Direct factor Xa inhibitors versus low molecular weight heparins or vitamin K antagonists for prevention of venous thromboembolism in elective primary hip or knee replacement or hip fracture repair

Abstract

Background

People undergoing major orthopaedic surgery are at increased risk of postoperative thromboembolic events. Low molecular weight heparins (LMWHs) are recommended for thromboprophylaxis in this population. New oral anticoagulants, including direct factor Xa inhibitors, are recommended as alternatives. They may have more advantages than disadvantages compared to LMWHs and vitamin K antagonists (VKAs, another type of anticoagulant).

Objectives

To assess the benefits and harms of prophylactic anticoagulation with direct factor Xa inhibitors compared with low molecular weight heparins and vitamin K antagonists in people undergoing major orthopaedic surgery for elective total hip or knee replacement or hip fracture surgery.

Search methods

We searched the Cochrane Vascular Specialised Register, CENTRAL, MEDLINE, Embase, two other databases, and two trial registers to 11 November 2023. We conducted reference checks to identify additional studies.

Selection criteria

We included randomised controlled trials (RCTs) comparing the effects of direct factor Xa inhibitors to LMWHs or VKAs in people undergoing major orthopaedic surgery.

Data collection and analysis

We used standard Cochrane methods. Our primary outcomes were all‐cause mortality, major venous thromboembolism (VTE), symptomatic VTE, major bleeding, and serious hepatic and non‐hepatic adverse events. We evaluated the risk of bias in the included studies using Cochrane's risk of bias 1 tool. We calculated estimates of treatment effects using risk ratios (RR) with 95% confidence intervals (CIs), and used GRADE criteria to assess the certainty of the evidence.

Main results

We included 53 RCTs (44,371 participants). Participants' average age was 64 years (range: 18 to 93 years). Only one RCT compared a VKA with direct factor Xa inhibitors. All 53 RCTs compared direct factor Xa inhibitors with LMWHs. Twenty‐three studies included participants undergoing total hip replacement; 21 studies, total knee replacement; and three studies included people having hip fracture surgery. The studies' average duration was approximately 42 days (range: two to 720 days).

Compared to LMWHs, direct factor Xa inhibitors may have little to no effect on all‐cause mortality, but the evidence is very uncertain (RR 0.83, 95% CI 0.52 to 1.31; I² = 0%; 28 studies, 29,698 participants; very low‐certainty evidence).

Direct factor Xa inhibitors may make little to no difference to major venous thromboembolic events compared to LMWHs, but the evidence is very uncertain (RR 0.51, 95% CI 0.37 to 0.71; absolute risk difference: 12 fewer major VTE events per 1000 participants, 95% CI 16 fewer to 7 fewer; I² = 48%; 28 studies, 24,574 participants; very low‐certainty evidence).

Compared to LMWHs, direct factor Xa inhibitors may reduce symptomatic VTE (RR 0.64, 95% CI 0.50 to 0.83; I² = 0%; 33 studies, 31,670 participants; low‐certainty evidence). The absolute benefit of substituting factor Xa inhibitors for LMWHs may be between two and five fewer symptomatic VTE episodes per 1000 patients.

In the meta‐analysis with all studies pooled, direct factor Xa inhibitors appeared to make little or no difference to major bleeding compared to LMWHs, but the evidence was very uncertain (RR 1.05, 95% CI 0.86 to 1.30; I² = 15%; 36 studies, 39,778 participants; very low certainty‐evidence). 
• In a subgroup analysis limited to studies comparing rivaroxaban to LMWHs, people given rivaroxaban may have had more major bleeding events (RR 1.94, 95% CI 1.26 to 2.98; I² = 0%; 17 studies, 17,630 participants; low‐certainty evidence). The absolute risk of substituting rivaroxaban for LMWH may be between one and seven more major bleeding events per 1000 patients. 
• In a subgroup analysis limited to studies comparing direct factor Xa inhibitors other than rivaroxaban to LMWHs, people given these other direct factor Xa inhibitors may have had fewer major bleeding events, but the evidence was very uncertain (RR 0.80, 95% CI 0.63 to 1.02; absolute risk difference: 3 fewer major bleeding events per 1000 participants, 95% CI 5 fewer to 0 fewer; I² = 0%; 19 studies, 22,148 participants; very low‐certainty evidence).

Direct factor Xa inhibitors may make little to no difference in serious hepatic adverse events compared to LMWHs, but the evidence is very uncertain (RR 3.01, 95% CI 0.12 to 73.93; 2 studies, 3169 participants; very low‐certainty evidence). Only two studies reported this outcome, with one death in the intervention group due to hepatitis reported in one study, and no events reported in the other study.

People given direct factor Xa inhibitors may have a lower risk of serious non‐hepatic adverse events than those given LMWHs (RR 0.89, 95% CI 0.81 to 0.97; I² = 18%; 15 studies, 26,246 participants; low‐certainty evidence). The absolute benefit of substituting factor Xa inhibitors for LMWH may be between three and 14 fewer serious non‐hepatic adverse events per 1000 patients.

Only one study compared a direct factor Xa inhibitor with a VKA. It reported outcome data with imprecise results due to the small number of events. It showed no difference in the effects of the study drugs.

Authors' conclusions

Oral direct factor Xa inhibitors may have little to no effect on all‐cause mortality, but the evidence is very uncertain.

Oral direct factor Xa inhibitors may slightly reduce symptomatic VTE events when compared with LMWH. They may make little or no difference to major VTE events, but the evidence is very uncertain.

In the evaluation of major bleeding, the evidence suggests rivaroxaban results in a slight increase in major bleeding events compared to LMWHs. The remaining oral direct factor Xa inhibitors may have little to no effect on major bleeding, but the evidence is very uncertain.

Oral direct factor Xa inhibitors may reduce serious non‐hepatic adverse events slightly compared to LMWHs. They may have little to no effect on serious hepatic adverse events, but the evidence is very uncertain.

Due to the high rates of missing participants and selective outcome reporting, the effect estimates may be biased.

Plain language summary

Direct factor Xa inhibitors or classic 'blood thinners': which leads to better outcomes for people undergoing major hip or knee surgery?

Key messages

• We do not know if classic anticoagulant medicines (commonly known as 'blood thinners') or newer blood‐thinning medicines called 'direct factor Xa inhibitors' (DFXa inhibitors) are better at preventing death or the development of blood clots in the deep veins of the legs or in the lungs of people who have had hip or knee replacement surgery.

• Compared to classic anticoagulants, DFXa inhibitors may slightly reduce the number of people who experience blood clot symptoms (such as breathing difficulties or pain). One type of DFXa inhibitor, rivaroxaban, may slightly increase the number of people who have serious uncontrolled bleeding.

Why are blood clots a concern for people having major hip or knee surgery?

'Venous thromboembolism' is when a blood clot forms in a vein, the vessels that carry blood back to the heart. Clots can narrow or block veins, leading to tissue damage, stroke, and death. People having hip or knee replacement surgery, or an operation to fix a broken hip, are at a higher risk of developing a blood clot.

How can blood clots be prevented and treated?

People undergoing major hip or knee surgery are typically given an anticoagulant medicine, commonly known as a 'blood thinner', to help prevent blood clots from forming.

There are two main types of 'classic' anticoagulants: (1) low molecular weight heparins (LMWHs), which are injected with a needle at a fixed dose; and (2) vitamin K antagonists (VKAs), given by mouth in variable amounts.

Newer anticoagulant medicines known as direct factor Xa inhibitors have been developed. They are given by mouth in fixed doses. Rivaroxaban and apixaban are types of DFXa inhibitors.

What did we want to find out?

We wanted to find out if DFXa inhibitors were better than classic anticoagulants at reducing the number of people who (1) died from any cause after major hip or knee surgery, and (2) developed blood clots or symptoms of blood clots (e.g. breathing difficulties and pain).

We also wanted to find out if DFXa inhibitors were associated with any unwanted effects, including uncontrolled bleeding, serious liver disease, and other serious unwanted events.

What did we do?

We searched for studies that compared DFXa inhibitors with classic anticoagulants in people undergoing hip or knee surgery. We compared and summarised the results of the studies and rated our confidence in the evidence, based on factors such as study methods and sizes.

What did we find?

We found 53 studies that involved 44,371 adults who had hip or knee replacement surgery, and had a high risk of developing blood clots in their lungs, legs, or pelvis. They received anticoagulant treatment for 6 to 39 days, and follow‐up lasted for an average of 42 days.

All 53 studies compared DFXa inhibitors with low molecular weight heparins (LMWHs). Just one study also investigated the VKA called warfarin.

The biggest study involved 5407 participants, and the smallest study, 50 participants. They were conducted in countries unevenly distributed around the world; most were done in high‐income regions. Roughly one‐third of participants (31%) were male. Participants' average age was 64 years. Pharmaceutical companies funded half of the studies (27 of 53).

Main results

Compared to LMWHs, we do not know if DFXa inhibitors reduce the number of people who:
• die from any cause after surgery, or
• develop blood clots in the lungs or deep veins of the leg or pelvis
because the evidence is very uncertain.

Compared to LMWHs, DFXa inhibitors may slightly reduce the number of people who have blood clot symptoms: between 2 and 5 fewer people out of every 1000 people given DFXa inhibitors would have blood clot symptoms compared to people given LMWHs.

The DFXa inhibitor rivaroxaban may slightly increase the number of people who have major uncontrolled bleeding compared to LMWHs. Between 1 and 7 more people out of every 1000 people given rivaroxaban may have uncontrolled bleeding compared to those given LMWHs.

The other DFXa inhibitors may have little to no effect on major bleeding, but the evidence is very uncertain.

Compared to LMWHs, DFXa inhibitors may have little to no effect on serious unwanted events affecting the liver, but the evidence is very uncertain.

They may slightly reduce other serious, non‐liver‐related events compared to LMWHs. Between 3 and 14 fewer people out of every 1000 people given DFXa inhibitors would have serious, non‐liver‐related events compared to people given LMWHs.

We did not find enough studies investigating VKAs to help us answer our questions.

What are the limitations of the evidence?

We have little confidence in the evidence because people in some studies were aware of which treatment they were getting. Most studies did not have a complete set of results for all participants. There were insufficient studies to be certain about the results of some outcomes.

How current is this evidence?

The evidence is current to November 2023.

Summary of findings

Summary of findings 1. Direct factor Xa inhibitors compared to LMWHs for prevention of venous thromboembolism in elective primary hip or knee replacement or hip fracture repair: primary outcomes.

Direct factor Xa inhibitors compared to LMWHs for prevention of VTE in elective primary hip or knee replacement or hip fracture repair
Patient or population: people undergoing elective primary hip or knee replacement or hip fracture repair Setting: hospital and outpatient settings Intervention: direct factor Xa inhibitors Comparison: LMWHs
Outcomes	Anticipated absolute effects* (95% CI)		Risk difference (95% CI)	Relative effect (95% CI)	№ of participants (RCTs)	Certainty of the evidence (GRADE)	Comments
	Risk with LMWHs	Risk with direct factor Xa inhibitors
All‐cause mortality Follow‐up: mean 41 days (range 10 to 84 days)a,b	Population:c Age 63 years (range 18 to 93) Male: 35%		0 fewer per 1000 (1 fewer to 1 more)	RR 0.83 (0.52 to 1.31)	29,698 (28 RCTs)	⊕⊝⊝⊝ Very lowd,e	A total of 74 all‐cause deaths were reported. OIS: 3,739,504 participantsf
	3 per 1000	2 per 1000 (1 to 3)
Major VTE Follow‐up: mean 39 days (range 10 to 180 days)a	Population: Age 63 years (range 18 to 93) Male: 36%		12 fewer per 1000 (16 fewer to 7 fewer)	RR 0.51 (0.37 to 0.71)	24,594 (28 RCTs)	⊕⊝⊝⊝ Very lowd,g	OIS: 6428 participants
	25 per 1000	13 per 1000 (9 to 17)
Symptomatic VTE Follow‐up: mean 40 days (range 6 to 84 days)a	Population: Age 64 years (range 18 to 93) Male: 33%		3 fewer per 1000 (5 fewer to 2 fewer)	RR 0.64 (0.50 to 0.83)	31,670 (33 RCTs)	⊕⊕⊝⊝ Lowd	OIS: 30,522 participants
	10 per 1000	6 per 1000 (5 to 8)
Major bleeding for rivaroxaban versus LMWHs Follow‐up: mean 38 days (range 10 to 168 days)a,b	Population: Age 63 years (range 18 to 93) Male: 39%		3 more per 1000 (1 more to 7 more)	RR 1.94 (1.26 to 2.98)	17,630 (17 RCTs)	⊕⊕⊝⊝ Lowd	OIS: 3710 participants
	3 per 1000	7 per 1000 (4 to 10)
Major bleeding for other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWHs Follow‐up: mean 44 days (range 7 to 70 days)a	Population: Age 63 years (range 19 to 93) Male: 32%		3 fewer per 1000 (5 fewer to 0 fewer)	RR 0.80 (0.63 to 1.02)	22,148 (19 RCTs)	⊕⊝⊝⊝ Very lowd,e	OIS: 36,170 participants
	13 per 1000	14 per 1000 (11 to 17)
Serious hepatic adverse events Follow‐up: mean 67 days (range 10 days to 22 months)a,b	Population: Age 64 years (range 25 to 73) Male: 20%		0 fewer per 1000 (0 fewer to 0 fewer)	RR 3.01 (0.12 to 73.93)	3169 (2 RCT)	⊕⊝⊝⊝ Very lowe,h	One single fatal serious hepatic adverse event was reported OIS: 49,332 participants
	0 per 1000	0 per 1000 (0 to 0)
Serious non‐hepatic adverse events Follow‐up: 41 days (range 10 to 70 days)a	Population: Age 64 years (range 18 to 93) Male: 33%		9 fewer per 1000 (14 fewer to 3 fewer)	RR 0.89 (0.81 to 0.97)	26,246 (15 RCTs)	⊕⊕⊝⊝ Lowi	None of the included studies defined 'serious adverse events', nor listed those adverse events judged to be serious. OIS: 18,868 participants
	77 per 1000	68 per 1000 (62 to 74)
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; LMWHs: low molecular weight heparins; OIS: optimal information size; RCT: randomised controlled trial; RR: risk ratio; RRR: risk rate reduction; RRI: risk rate increase; VTE: venous thromboembolism
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect

^aThe follow‐up mean value is the weighted average of time in days and the number of randomised participants in each study. 
^bOnly 1 study (Rahman 2020) reported a very extended follow‐up of 15 months (range 12 to 22 months).
^cPopulation: age is the weighted average of participants' ages in years and the number of randomised participants in each study (age range); gender is presented as the percentage of participants who were male.
^dDowngraded by two levels for very serious study limitations (overall high risk of bias in one or more domains, including attrition bias and selective reporting bias).
^eDowngraded by two levels for imprecision: insufficient number of events; overall sample size of the included studies for these outcomes was lower than the optimal information size (OIS); the 95% confidence interval of the overall pooled estimate for these outcomes overlapped 'no effect' (CI includes RR of 1.0); the CI failed to exclude important benefit or harm (RRR > 25% and/or RRI > 25%).
^fWe calculated the optimal information size using a web calculator (parameters: calculate sample size (for specified power); two‐sided test; alpha: 0.05; beta: 0.80. Input value for P1: (number of events / population) in the DFXa group; input value for P2: (number of events / population) in the LMWH group) for each outcome.
^gDowngraded by one level for moderate heterogeneity (I² = 48%) and P value less than 0.10 in the overall pooled estimate.
^hDowngraded by two levels for very serious study limitations (overall high risk of bias in one or more domains, including selection bias and selective reporting bias).
ⁱDowngraded by two levels for very serious study limitations (overall high risk of bias in one or more domains, including performance bias, attrition bias, and selective reporting bias).

Background

Description of the condition

Venous thromboembolism (VTE) is the presence of a blood clot that blocks a blood vessel within the venous system. People undergoing major orthopaedic surgery, such as hip or knee replacement, are at increased risk of blood clots for several reasons, including the following.

• Immobility: people often have reduced mobility after orthopaedic surgery, and prolonged immobility can lead to slowed blood flow, which raises the risk of clot formation.

• Surgical trauma, blood vessel damage, or both: damage to blood vessels and other tissues during surgery may activate blood clotting mechanisms and contribute to the formation of clots.

• Inflammation: surgery triggers an inflammatory response in the body, which may increase the production of clotting factors (and other substances that promote coagulation), and disrupt the normal balance between clotting and anticoagulation.

• Venous compression: hip and knee surgery may involve compressing the veins (as a result of surgical positioning or through use of devices such as tourniquets). This compression can impede blood flow and increase the risk of clot formation.

• Underlying conditions: people undergoing major orthopaedic surgery may have pre‐existing risk factors for clotting, including obesity, older age, or a history of venous thromboembolism.

Venous thromboembolism includes deep vein thrombosis (DVT) and pulmonary embolism (PE) (Bartholomew 2017). VTE occurs in 44% to 90% of people who undergo total hip or knee replacement and who do not receive anticoagulants, with an estimated 4.3% of patients presenting symptomatic VTE within 39 days after surgery (Falck‐Ytter 2012). Pulmonary embolism develops in up to 7% of people who undergo elective knee surgery (Stringer 1989), and is fatal in 0.7% of the people in whom it occurs (Ansari 1997). VTE has been observed three months postoperatively in 2.4% of people who have undergone hip replacement and in 1.7% of people who have undergone knee replacement, despite having received prophylaxis (White 1998). Symptomatic VTE occurs after people leave hospital, and the risk increases for at least two months after surgery (Douketis 2002; Leclerc 1998; Pellegrini 1996; White 1998), with 80% to 85% of cases being diagnosed after hospital discharge (Lapidus 2013).

'Thromboprophylaxis' refers to the treatments and measures available to help manage people's increased risk of VTE following major orthopaedic surgery. However, thromboprophylaxis approaches need to balance VTE risk reduction against the increased risk of bleeding complications associated with anticoagulants.

Description of the intervention

Anticoagulant medications, which reduce the blood's ability to clot, represent the mainstay of pharmacological thromboprophylaxis. Anticoagulant drugs are classified based on their mechanism of action and target within the coagulation cascade. There are six classes of anticoagulants, which are sometimes divided into 'older' or 'classic' anticoagulants (Ageno 2012; Garcia 2012; Weitz 2012):

• vitamin K antagonists (VKAs): oral warfarin or coumarin;

• indirect factor II (thrombin) and factor Xa inhibitors: unfractionated heparin;

• indirect factor Xa inhibitors: low molecular weight heparins (LMWHs), fondaparinux, danaparoid;

• direct parenteral thrombin inhibitors: hirudin, bivalirudin, argatroban;

and 'new' anticoagulants:

• indirect anticoagulants: AT mediated (idraparinux); protein C mediated (Solulin);

• direct oral anticoagulants (DOACs):

  • selective thrombin inhibitors;

  • selective factor Xa inhibitors;

  • factor VIIa inhibitors; factor VIIIa inhibitors;

  • factor IXa inhibitors.

This review focusses on direct factor Xa inhibitors, which are a type of direct oral anticoagulant (DOAC). For context, we also briefly describe LMWHs and VKAs.

Classic anticoagulants

VKAs are typically administered orally in the form of tablets. Warfarin is the most well‐known and commonly used VKA. It is associated with an increased risk of haemorrhage of 3% to 4% annually (Geerts 2001). Coagulation monitoring and dose adjustment are therefore routine for people taking VKAs (Ansell 2001; Hirsh 2001; Schulman 2003). Indirect and total costs of treatment are higher with the use of warfarin compared with other oral anticoagulants (Marcolino 2016).

LMWHs are given parenterally (i.e. by injection). Enoxaparin, dalteparin, and tinzaparin are types of LMWHs. LMWHs are more predictable than some other anticoagulants, and may have a lower risk of bleeding complications. Like VKAs, they are also associated with higher costs, especially in people at low risk of thromboembolic events (Türk 2016).

Direct factor Xa inhibitors

Factor Xa is an enzyme that plays a central role in the process of blood clot formation. It converts prothrombin, a plasma protein, into thrombin. Thrombin helps convert fibrinogen into fibrin, which forms the meshwork of a blood clot. Direct factor Xa inhibitors bind to and inhibit factor Xa without requiring antithrombin (in contrast to the indirect factor Xa inhibitors fondaparinux and idraparinux) (Rupprecht 2010). They have predictable pharmacodynamic and pharmacokinetic profiles which correlate well with inhibition of factor Xa activity and prolongation of prothrombin time. (These tests evaluate the time it takes for blood to clot: the first measures the activity of LMWHs and the second of VKAs. Regular monitoring of factor Xa inhibitors is unnecessary, unlike for VKAs.) Other advantages of direct factor Xa inhibitors include their availability in oral formulations and their rapid onset of action. Their effect also wanes quickly after discontinuation, with a half‐life of nine hours – an advantage in situations requiring rapid reversal or adjustment of anticoagulation (Kubitza 2005). Rivaroxaban and apixaban are types of direct factor Xa inhibitors.

Clinical guidelines

Evidence‐based clinical practice guidelines contain clinical recommendations to assist physicians, patients, and all decision‐makers to improve the quality of clinical care. The American College of Chest Physicians' (ACCP) guidelines (known as 'AT9') recommend thromboprophylaxis in people undergoing major orthopaedic surgery, using any of several agents (Falck‐Ytter 2012). These include LMWHs, direct thrombin inhibitors (e.g. dabigatran), low‐dose unfractionated heparin, adjusted‐dose VKAs, direct factor Xa inhibitors (including apixaban, fondaparinux, and rivaroxaban), aspirin, or an intermittent pneumatic compression device (IPCD), for a minimum of 10 to 14 days, with preference for LMWHs during an extended period of up to 35 days (Falck‐Ytter 2012). None of these recommendations have changed in the most recent version of the AT9 guidelines (Stevens 2021).

The American Society of Hematology (ASH) 2019 guidelines for orthopaedic surgery suggest using aspirin or anticoagulants (conditional recommendation based on very low certainty in the evidence of effects) (Anderson 2019). They suggest direct oral anticoagulants (DOACs) over LMWHs (conditional recommendation based on moderate certainty in the evidence of effects). The guideline panel suggests using any of the DOACs approved for use (conditional recommendation based on low certainty in the evidence of effects) (Anderson 2019). Conditional recommendations, also known as weak recommendations, suggest that the panel believes there is a close balance between the benefits and risks of the intervention, and there is uncertainty about the magnitude of the benefits or adverse effects. People's values and preferences are uncertain or variable, and the costs of the intervention may not always justify them.

While the ASH 2019 guidelines prefer DOACs to other classes, the AT9 ACCP 2012 guidelines do not.

How the intervention might work

As noted above, direct factor Xa inhibitors bind to and inhibit factor Xa without requiring antithrombin (in contrast to the indirect factor Xa inhibitors fondaparinux and idraparinux) (Rupprecht 2010). Direct factor Xa inhibitors have no effect on other coagulation factors, are of small size, and have the ability to inactivate the circulating factor Xa, and also the bound forms of factor Xa (e.g. factor Xa within the prothrombinase complex, clot‐bound factor Xa (Becker 2003; Comp 2003)). These latter two characteristics are an advantage because, theoretically, they increase efficacy. Regarding safety, direct factor Xa inhibitors' inhibition of factor Xa leads to decreased thrombin generation rather than inactivation of thrombin's catalytic activity. Thus, in theory, they should have no effect on platelet activation (Comp 2003; Ieko 2004), and therefore cause fewer bleeding events.

Therefore, the use of new oral anticoagulants might result in direct and indirect cost savings compared with traditional anticoagulants (Nutescu 2013). Agents of this group of importance for thromboprophylaxis in people undergoing major orthopaedic surgery include direct thrombin (factor IIa) inhibitors (such as dabigatran) and direct factor Xa inhibitors (such as apixaban, rivaroxaban, and edoxaban) (Bass 2015). As a general rule, 'new' oral anticoagulants which reduce the risk of venous thromboembolic events also increase the risk of bleeding events (Venker 2017). However, varying effects have been observed according to the type and dose of the drug. Although dabigatran appears to have an efficacy profile similar to LMWHs, increased risk of death has also been observed (Salazar 2010). Direct factor Xa inhibitors might represent a better target for thromboprophylaxis based on their mechanisms of action and the results of pre‐clinical and clinical studies (Ansell 2007).

Why it is important to do this review

The effectiveness of classic anticoagulants, such as heparin and warfarin, has been proved in numerous studies. However, data from clinical trials showed that, of people undergoing major orthopaedic surgery, 0.5% who received thromboprophylaxis with VKAs (Forster 2016), and 22.5% who received thromboprophylaxis with LMWHs (Salazar 2010), still experience venous thromboembolic events. Similarly, the incidence of venographic DVT and bleeding events has been shown to be 32.7% and 2.6%, respectively, for VKAs, and 17.9% and 3.7%, respectively, for LMHWs (Brown 2009). Therefore, better and safer anticoagulants are needed.

As mentioned above, of the new oral anticoagulants, direct factor Xa inhibitors have an interesting profile, with some studies showing increased efficacy compared with LMWHs (Augoustides 2012). Although at least 10 systematic reviews addressing the role of direct factor Xa inhibitors for thromboprophylaxis in people undergoing major orthopaedic surgery have been published (see Agreements and disagreements with other studies or reviews for details), some relevant outcomes have not been analysed, such as the incidence of fatal VTE, bleeding events, and serious adverse events. Most reviews have also not addressed heterogeneity regarding specific drugs, doses, timing and duration of prophylaxis. Therefore, a systematic analysis of relevant, high‐quality, and up‐to‐date evidence on this topic is needed.

Objectives

To assess the benefits and harms of prophylactic anticoagulation with direct factor Xa inhibitors compared with low molecular weight heparins and vitamin K antagonists in people undergoing major orthopaedic surgery for elective total hip or knee replacement or hip fracture surgery.

Methods

Criteria for considering studies for this review

Types of studies

We included randomised controlled trials (RCTs) that compared prophylactic anticoagulation with direct factor Xa inhibitors versus low molecular weight heparins (LMWHs) or vitamin K antagonists (VKAs) in the prevention of venous thromboembolism (VTE).

Types of participants

We included people who had elective major orthopaedic surgery, which includes total joint (hip or knee) replacement and hip fracture surgery (Kinov 2014). The terms 'replacement' and 'arthroplasty' are used interchangeably in the literature (Fawaz 2020). We have predominantly used the term 'replacement' in this review, and refer to the procedures as 'total hip replacement' (THR) and 'total knee replacement' (TKR).

Types of interventions

We included trials that evaluated prophylactic anticoagulation using direct factor Xa inhibitors, compared to LMWHs at standard doses, or VKAs. VKAs require dose adjustments to maintain therapeutic doses, which are achieved when the international normalised ratio (INR) is between 2 and 3. The INR is calculated with the prothrombin time, whose definition was previously explained (see Description of the intervention).

Types of outcome measures

Primary outcomes

• All‐cause mortality (during treatment or follow‐up): mortality associated with VTE (PE or DVT), bleeding, or treatment (not related to VTE or bleeding).

• Major VTE (during treatment or follow‐up): the incidence of proximal VTE, including DVT from the popliteal vein (proximal DVT), symptomatic DVT, PE‐related and VTE‐related death. PE could have been established through positive pulmonary angiography, high probability ventilation/perfusion scan, positive helicoidal tomography, postmortem evidence, or any other validated method. We accepted a diagnosis of DVT if made by positive venography or ultrasonography. We defined 'incidence' as the appearance of thrombosis in an area where it did not exist prior to the study.

• Symptomatic VTE (during treatment or follow‐up): including DVT or PE, fatal or non‐fatal.

• Major bleeding (during treatment or follow‐up): defined by a decrease in haemoglobin concentration of more than 2 g/dL, bleeding at a critical site (retroperitoneal, intracranial, or other critical site), or requiring a transfusion of two or more units of packed red blood cells (RBC).

• Serious hepatic adverse events (during treatment or follow‐up): the appearance of serious hepatopathy (liver disease), defined as fulminant hepatitis, symptoms of liver failure, or life‐threatening hepatopathy.

• Serious non‐hepatic adverse events (during treatment or follow‐up): the appearance of other serious adverse events associated with treatment, including heparin‐induced thrombocytopenia (HIT) or any life‐threatening event. Heparin‐induced thrombocytopenia is defined by the generation of HIT‐specific antibodies accompanied by an otherwise unexplained decrease in platelet count (usually > 50% fall, even if at its lowest point, the platelet count remained > 150 x 10⁹/L), or by skin lesions at heparin injection sites, or acute systemic reactions; for example, chills or cardiorespiratory distress after intravenous heparin bolus administration.

We decided to explore serious hepatic adverse effects separately from other serious adverse effects because a different but similar anticoagulant (ximelagatran, an oral direct thrombin inhibitor) was removed from the market in 2006 due to severe hepatotoxicity (EMEA 2006a; EMEA 2006b).

Secondary outcomes

• Fatal VTE (during treatment or follow‐up): deaths due to VTE events

• Asymptomatic distal DVT (during treatment or follow‐up): the incidence of asymptomatic distal DVT (below popliteal vein), as detected during a scheduled lower limb Doppler ultrasound or diagnosed by routine venogram

• Fatal bleeding (during treatment or follow‐up): deaths due to bleeding events

• Liver enzymes elevation (during treatment or follow‐up): temporary elevation of hepatic enzymes

• Minor adverse events (during treatment or follow‐up): morbidity associated with treatment (appearance of non‐fatal or significant haemorrhagic events, uncomplicated skin necrosis, or any other non‐life‐threatening event)

• Volume of blood loss (during treatment or follow‐up): postoperative drainage

All the review's primary and secondary outcomes are dichotomous outcomes, except for volume of blood loss, which is a continuous outcome.

Search methods for identification of studies

We applied no language, publication year, or publication status restrictions.

Electronic searches

The Cochrane Vascular Information Specialist first searched the following databases for relevant randomised controlled trials (see Appendix 1 for details of the search strategy used to search CENTRAL):

• the Cochrane Vascular Specialised Register via the Cochrane Register of Studies (CRS‐Web) (from 1 January 2017 to 1 September 2021);

• the Cochrane Central Register of Controlled Trials (CENTRAL) Cochrane Register of Studies Online (CRSO, 2021);

• MEDLINE (Ovid MEDLINE Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, Ovid MEDLINE Daily and Ovid MEDLINE) (1946 to 1 September 2021);

• Embase Ovid (1980 to 1 September 2021);

• CINAHL (Cumulative Index to Nursing and Allied Health Literature) EBSCO (from 1 January 2017 to 1 September 2021);

• AMED (Allied and Complementary Medicine) Ovid (from 1 January 2017 to 1 September 2021).

The Information Specialist also performed top‐up searches of the following trial registries on 1 September 2021 (see Appendix 2):

• ClinicalTrials.gov (www.clinicaltrials.gov);

• World Health Organization International Clinical Trials Registry Platform (www.who.int/trialsearch).

The Information Specialist modelled search strategies for the listed databases on the search strategy designed for CENTRAL. Where appropriate, they were combined with adaptations of the highly sensitive search strategy designed by Cochrane for identifying randomised controlled trials and controlled clinical trials (as described in Chapter 4 of the Cochrane Handbook for Systematic Reviews of Interventions; Lefebvre 2022). Search strategies for major databases are provided in Appendix 3.

The review authors subsequently conducted systematic top‐up searches of the following databases on 11 November 2023 using the same search strategies described in Appendix 1, Appendix 2, and Appendix 3:

• CENTRAL; 2023, Issue 10 (in CRSO, November 2023);

• MEDLINE (Ovid MEDLINE Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, Ovid MEDLINE Daily and Ovid MEDLINE) (from 1 September 2021 to 11 November 2023);

• Embase Ovid (from 1 September 2021 to 11 November 2023);

• ClinicalTrials.gov (www.clinicaltrials.gov);

• World Health Organization International Clinical Trials Registry Platform (www.who.int/trialsearch).

The review authors also performed an electronic search in the Latin American and Caribbean Health Science Literature (LILACS) database on 11 November 2023. The full search strategy is included in Appendix 4.

Searching other resources

We reviewed the reference lists of identified studies and previous systematic reviews in order to identify additional relevant RCTs.

Data collection and analysis

Our primary analyses present the results of all major orthopaedic surgery combined (i.e. total hip replacement and total knee replacement populations), all the oral direct factor Xa inhibitors as a group (instead of each drug separately), and all the reported events (in the treatment and follow‐up periods).

Selection of studies

Two review authors (JB, CS) independently evaluated the titles and abstracts of the reports of trials identified by electronic searches. We obtained copies of the full texts of those trials that met the selection criteria. We translated titles and abstracts of articles in languages other than English or Spanish into English and then evaluated their eligibility. We translated the entire text of the article if the title and abstract indicated that the study met the review's inclusion criteria. Two review authors (CS, JB) independently assessed the eligibility of all studies based on pre‐specified inclusion and exclusion criteria. We did not exclude studies on the basis of the journal in which reports were published, study author institutional affiliations, or study results. We resolved any disagreements through discussion to reach consensus. We listed studies excluded at the full‐text screening stage as formally excluded studies, with the reasons for exclusion (see Characteristics of excluded studies).

Data extraction and management

All review authors independently extracted data using pre‐designed data extraction sheets and verified the data. We compared the data extraction forms with each other to ensure consistency. We recorded important characteristics of the participants (inclusion and exclusion criteria; clinical and laboratory diagnosis criteria; number of participants; age; sex; duration of follow‐up; duration of the study); interventions (type of prophylactic anticoagulant; duration and time of initiation of the prophylactic anticoagulation treatment); and results.

Assessment of risk of bias in included studies

All review authors independently evaluated the methodological quality of each included study according to the Cochrane risk of bias approach – a domain‐based evaluation tool described in Higgins 2011b. The domains assessed are: (1) random sequence generation; (2) allocation concealment; (3) blinding of participants and personnel; (4) blinding of outcome assessment; (5) incomplete outcome data (attrition bias); (6) selective reporting; and (7) other potential biases.

Each review author issued a judgement of risk of bias (low, high, or unclear) for each domain. If it was not possible or prudent to arrive at a rating of high or low risk, we assigned a rating of unclear risk of bias. We resolved any disagreements by consensus. We obtained the information required for individual assessment of each study from its final report, published protocols, or by contacting the investigators.

We assessed incomplete outcome data separately for harm and benefit outcomes (previously known as 'safety' and 'efficacy' outcomes, respectively), due to previous experience regarding the prevalence of missing benefit results in this setting, where venography was the only diagnostic method for VTE (Salazar 2010).

In evaluating attrition bias, we reviewed all included studies and, where possible, completed all the CONSORT flow diagrams. If there were missing outcome data, we explored whether: (1) the reason for missing outcome data related to the studied outcomes (thrombosis, bleeding, death, or adverse events); (2) the missing data were imbalanced in numbers across intervention groups (imbalance defined as 5% or greater difference between intervention and control groups); and (3) there were different reasons for missing data across groups. We assessed a study as having a low risk of attrition bias if there were no missing outcome data, or if the missing data did not meet any of the above criteria. When the missing data met some or all of these criteria, we calculated whether the proportion of missing outcomes compared to the observed event risk was sufficient to have a clinically relevant impact on the intervention effect estimate, using these three hypothetical scenarios: (1) we assumed that missing participants had the same absolute risks as in the original study for both intervention and comparator groups; (2) we assumed that missing participants had the absolute risks derived from combining the results for all the studies included in the meta‐analysis for that outcome; and (3) we assumed that missing participants had the absolute risks derived from the study included in the meta‐analysis that showed the most positive effects for the comparator. If the new estimate of the absolute risks produced by any of these scenarios was more than 100% different from the original absolute risk of the intervention or comparator, we judged that the amount of missing outcome data might alter the estimate of the intervention effect, and we rated the study as having a high risk of attrition bias. See Appendix 5 and Appendix 6 for details of these calculations.

Measures of treatment effect

We summarised dichotomous outcomes for each study using the risk ratio (RR). For the lone continuous outcome, we used the mean difference (MD). We presented the results with 95% confidence intervals (CI) obtained with Review Manager (RevMan) software (RevMan 2024).

Once we had obtained the pooled effect estimate for each dichotomous outcome, we also presented the risk difference and its 95% CI, obtained with GRADEpro software (GRADEpro GDT 2024).

Unit of analysis issues

All the included RCTs had a parallel‐group design, where participants were individually randomised to either the intervention group (direct factor Xa inhibitors) or the control group (LMWHs or VKAs). We collected and analysed a single measurement of each outcome for each participant.

For phase II studies that included multiple treatment arms with different doses of direct factor Xa inhibitors, we initially planned to use only data from the treatment arm using the standard or intermediate dose. However, given that some non‐phase II studies (with multi‐arm designs) also assessed several doses of the experimental drugs, we decided to combine all doses of the direct factor Xa inhibitors in the main analysis. We then performed additional sensitivity analyses by dividing data into three daily dose subgroups (low dose, intermediate dose, and high dose) and, when necessary, an individual dose analysis.

Dealing with missing data

We attempted to contact the original investigators to request missing data. We re‐evaluated studies without complete information when additional information from the study authors became available. When we could not obtain missing data, we analysed the available data, and we noted any assumptions made. We performed imputation of missing data regarding the total numbers of participants (N) when data on the number of events were reported separately for the periods of treatment and post‐treatment follow‐up. Appendix 7 describes in detail the methodology we used for missing data imputation. We imputed data for the following outcomes and studies: all‐cause mortality (APROPOS 2007; ODIXa‐HIP2 2006; ODIXa‐KNEE 2005), major VTE (Fuji 2014‐THA; ODIXa‐HIP‐OD 2006), symptomatic VTE (Fuji 2014‐THA; Fuji 2014‐TKA; NCT01206972; ONYX‐1 2007), serious non‐hepatic adverse events (STARS J‐V 2015), fatal VTE (APROPOS 2007; ODIXa‐KNEE 2005; RECORD1 2008; RECORD2 2008; RECORD4 2009), and liver enzyme elevation (STARS J‐2 2014).

We complemented data from the final report of included studies with: (1) protocol descriptions of study rationale and study design articles; (2) clinical trial registry entries; (3) editorials and sub‐analyses of main studies; and (4) American Food and Drug Administration (FDA) briefing documents. Of note, we were specially careful not to add up reported events for components of the outcomes under study (e.g. pulmonary embolism (PE) plus proximal DVT to calculate major VTE), when reported information was insufficient to exclude the possibility of repetition of data from the same participant (e.g. one participant presenting PE and proximal DVT). We carefully considered missing data when evaluating potential biases and in the interpretation of results. When an intention‐to‐treat (ITT) analysis was not performed in the original studies, we included any possible events not included in the analyses reported, including the follow‐up periods.

Assessment of heterogeneity

We used the I² statistic instead of the Chi² test to ascertain homogeneity amongst studies, given that the former is more useful when there are few studies, and allows comparison amongst subgroups. The I² statistic is expressed as a percentage and describes the proportion of variability that is due to heterogeneity rather than sampling error. Based on Higgins 2003, we tentatively defined low heterogeneity as I² values of less than 25%, moderate heterogeneity as I² values between 25% and less than 50%, high heterogeneity as I² values between 50% and 75%, and very high heterogeneity as I² values above 75%.

Assessment of reporting biases

We performed a comprehensive search for published, unpublished, and ongoing studies that met the review's eligibility criteria. Our evaluation of the methodological quality of the included studies included our assessment of selective reporting bias. We assessed publication bias by visual inspection of funnel plots (Higgins 2011a), and using Peters' test for funnel plot asymmetry for dichotomous outcomes (Peters 2006), and Egger's test for continuous outcomes (Egger 1997). We chose Peters' test because it is recommended for dichotomous outcomes, regardless of the degree of heterogeneity and the magnitude of treatment effect (Jin 2015).

Data synthesis

When heterogeneity was low, we meta‐analysed data with a fixed‐effect model.

If we found statistical heterogeneity, we inspected the forest plot for the direction of effect of individual studies and overlap of confidence intervals, and we explored if a clinical explanation could be the cause of the heterogeneity by performing several subgroup analyses. If we could not explain the heterogeneity, we compared the fixed‐effects and a random‐effects meta‐analysis, searching for differences in direction or magnitude, with the intention of presenting the random‐effects result if there was no indication of funnel plot asymmetry. If there was an indication of funnel plot asymmetry, we performed a sensitivity analysis by excluding small studies.

We avoided presenting meta‐analysis results if: (1) we found very high unexplained statistical heterogeneity; or (2) the fixed‐effect and random‐effects meta‐analyses differed in direction or magnitude. Also, we did not perform meta‐analysis when we found high unexplained statistical heterogeneity with funnel plot asymmetry and different results in the sensitivity analysis excluding small studies.

Subgroup analysis and investigation of heterogeneity

Where there was heterogeneity amongst studies, we explored the reasons for such heterogeneity by performing subgroup analyses. Moreover, we subjectively evaluated heterogeneity amongst studies, using our clinical judgement to assess differences in participant populations, interventions, and measurement of outcomes. Bearing in mind that not all trials are designed to measure adverse events, we interpreted secondary outcomes with caution.

We considered the following subgroups relevant for this review:

• type of drug (rivaroxaban, apixaban, edoxaban, etc.);

• duration of the prophylactic anticoagulation, classified as short (up to 14 days) and extended (15 to 35 days) (Falck‐Ytter 2012);

• time of initiation of prophylactic anticoagulation (before surgery versus after surgery, applicable only for the comparator);

• frequency of administration (once versus twice daily) for those outcomes which showed association of the intervention drug with difference in the outcome;

• type of surgery: total hip replacement, total knee replacement, or hip fracture surgery.

Based on the results obtained, we considered it necessary to undertake some additional subgroup analyses:

• restriction of analysis to one of three daily dose groups (low, intermediate, and high, with limits established based on the highest daily dose studied in the included trials).

Differences between subgroups were established by a P value of less than 0.10 from the Chi² test for subgroup differences.

Sensitivity analysis

Since there are a variety of ways to perform a meta‐analysis, we tested the robustness of our results using these sensitivity analyses:

• restricting data to the active treatment period (i.e. excluding events that occurred after prophylaxis cessation);

• excluding studies with a high risk of selection bias (randomisation sequence generation and allocation concealment).

Summary of findings and assessment of the certainty of the evidence

We prepared a summary of findings table for the comparison of direct factor Xa inhibitors with LMWHs. We included the following primary outcomes: all‐cause mortality; major VTE; symptomatic VTE; major bleeding (disaggregated for rivaroxaban versus LMWHs, and all other direct factor Xa inhibitors versus LMWHs); serious hepatic adverse events; and serious non‐hepatic adverse events (Table 1). We have presented a summary of findings for our secondary outcomes in Table 2: fatal VTE; asymptomatic distal DVT; fatal bleeding; liver enzymes elevation; minor adverse events; and volume of blood loss. We graded the certainty of the evidence using criteria devised by the GRADE Working Group (Guyatt 2008), using the GRADEprofiler software (GRADEpro). We assigned one of four levels of certainty: high, moderate, low, or very low, based on the overall risk of bias in the included studies, the directness of the evidence, the (in)consistency of results, the precision of the estimates, and risk of publication bias (Higgins 2011a; Schünemann 2006; Schünemann 2023).

1. Direct factor Xa inhibitors compared to LMWHs for prevention of venous thromboembolism in elective primary hip or knee replacement or hip fracture repair: secondary outcomes.

Direct factor Xa inhibitors compared to LMWH for prevention of venous thromboembolism in elective primary hip or knee replacement or hip fracture repair: secondary outcomes
Patient or population: people undergoing elective primary hip or knee replacement or hip fracture repair Setting: hospital and outpatient settings Intervention: direct factor Xa inhibitors Comparison: LMWHs
Outcomes	Anticipated absolute effects* (95% CI)		Risk difference (95% CI)	Relative effect (95% CI)	№ of participants (RCTs)	Certainty of the evidence (GRADE)	Comments
	Risk with LMWHs	Risk with direct factor Xa inhibitors
Fatal VTE Follow‐up: 39 days (range 6 to 84 days)a,b	Population:c Age 62 (range 18 to 93) Male: 35%		0 fewer per 1000 (0 fewer to 1 more)	RR 1.19 (0.51 to 2.79)	27,183 (33 RCTs)	⊕⊝⊝⊝ Very lowd,e	A total of 15 fatal venous thromboembolic events were reported OIS:f 149,752 participants
	0 per 1000	0 per 1000 (0 to 1)
Asymptomatic distal DVT Follow‐up: 31 days (range 17 to 168 days)a	Population: Age 65 (range 30 to 88) Male: 26%		29 fewer per 1000 (43 fewer to 9 fewer)	RR 0.64 (0.46 to 0.89)	4818 (14 RCTs)	⊕⊕⊝⊝ Lowg	OIS: 3146 participants
	79 per 1000	51 per 1000 (36 to 71)
Fatal bleeding Follow‐up: 39 days (range 7 to 84 days)a,b	Population: Age 63 (range 18 to 93) Male: 35%		0 fewer per 1000 (0 fewer to 0 fewer)	RR 1.42 (0.33 to 6.04)	31,293 (26 RCTs)	⊕⊝⊝⊝ Very lowd,e	A total of 4 fatal bleeding events were reported OIS: 506,558 participants
	0 per 1000	0 per 1000 (0 to 0)
Liver enzymes elevation Follow‐up: 41 days (range 10 to 70 days)a	Population: Age 63 (range 18 to 93) Male: 34%		12 fewer per 1000 (16 fewer to 8 fewer)	RR 0.58 (0.46 to 0.73)	31,408 (21 RCTs)	⊕⊝⊝⊝ Very lowed,h,i	OIS: 11,742 participants
	29 per 1000	17 per 1000 (13 to 21)
Minor adverse events Follow‐up: 25 days (range 21 to 35 days)a	Population: Age 66 (range 50 to 81) Male: 13%		57 fewer per 1000 (108 fewer to 0 fewer)	RR 0.91 (0.83 to 1.00)	1011 (4 RCTs)	⊕⊕⊝⊝ Lowj	The most frequently reported minor adverse events included constipation, myalgia, and insomnia. OIS: 617,682 participants
	632 per 1000	575 per 1000 (525 to 632)
Volume of blood loss Follow‐up: 48 days (range 10 to 180 days)a	Population: Age 65 (range 26 to 93) Male: 38%			‐	7379 (14 RCTs)	⊕⊝⊝⊝ Very lowk,l,m	No meta‐analysis performed due to very high heterogeneity (I2 = 96%)
	‐	‐
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; DVT: deep vein thrombosis; LMWHs: low molecular weight heparins; OIS: optimal information size; RCT: randomised controlled trial; RR: risk ratio; ROB: risk of bias; RRR: risk rate reduction; RRI: risk rate increase; VTE: venous thromboembolism
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect

^aThe follow‐up mean value is the weighted average of time in days and the number of randomised participants in each study.
^bOnly Rahman 2020 reported a very extended follow‐up of 15 months (range 12 to 22 months).
^cPopulation: age is the weighted average of participants' ages in years and the number of randomised participants in each study (age range); gender is presented as the percentage of participants who were male
^dDowngraded by one level for very serious study limitations (overall high risk of bias in one or more domains, including selective reporting bias).
^eDowngraded by two levels for imprecision: insufficient number of events; overall sample size of the included studies for these outcomes was lower than the optimal information size (OIS); the 95% confidence interval of the overall pooled estimate for these outcomes overlapped 'no effect' (CI includes RR of 1.0); the CI failed to exclude important benefit or harm (RRR > 25% and/or RRI > 25%).
^fWe calculated the optimal information size using a web calculator (see footnote of Table 1 for more details).
^gDowngraded by two levels for very serious study limitations (overall high risk of bias in one or more domains, including attrition bias and selective reporting bias). 
^hDowngraded by one level for moderate heterogeneity (I² > 50%) and P value less than 0.10 in the overall pooled estimate.
ⁱCertainty of evidence was downgraded if the funnel plot was asymmetric by visual inspection and by statistically significant funnel plot asymmetry assessed by Peters' test for every analysed outcome. ^jDowngraded by two levels for very serious study limitations (overall high risk of bias in one or more domains, including performance bias and selective reporting bias).
^kDowngraded by two levels for very serious study limitations (overall high risk of bias in one or more domains, including selection, performance, attrition, and selective reporting bias). 
^lDowngraded by two levels for high or very high heterogeneity.
^mDowngraded by one level for imprecision: 95% confidence interval of the overall pooled estimate for this outcome overlapped no effect (CI includes RR of 1.0) and 95% CI failed to exclude important benefits or harm (RRR > 25% and/or RRI > 25%).

Regarding the overall risk of bias, we evaluated the certainty of the evidence for each outcome independently.

• If studies that contributed data to an outcome had a high risk of selection bias (due to inappropriate randomisation sequence or inappropriate allocation concealment), we downgraded the certainty of the evidence by one level. However, if those studies contributed 20% or more of the weight of the combined effect for the outcome, then we downgraded the certainty of the evidence by two levels for a high risk of selection bias because we considered this proportion of information from results at high risk of bias sufficient to affect the interpretation of results for the outcome.

• If studies that contributed data to an outcome had a high risk of performance bias, due to inappropriate blinding of participants and personnel, or detection bias, due to inappropriate blinding of outcome assessment, we downgraded the certainty of the evidence by one level. However, if the proportion of information from results at high risk of bias was sufficient to affect the interpretation of results for this outcome, then we downgraded the certainty of the evidence by two levels.

• If studies that contributed data to an outcome had missing participants in the intervention or control groups, we downgraded the certainty of the evidence by one level. However, if these missing data could seriously impact the overall effect estimate, then we downgraded the certainty of the evidence by two levels for high risk of attrition bias.

• If studies that contributed data to an outcome had a high risk of selective reporting bias, due to reported outcomes differing from those included in the protocol or included in the protocol after the study was published (amongst other issues), then we downgraded the certainty of the evidence by one level. However, if the other included studies that did not report the outcome in question also exhibited selective reporting issues (such as no protocol), uncertainty arises regarding whether there was selective omission of this outcome. In this case, we downgraded the certainty of the evidence by two levels for high risk of selective reporting bias.

If crucial study limitations in one or more of the above domains were sufficient to substantially lower our confidence in the estimate of effect, then we downgraded the certainty of the evidence by two levels for the overall risk of bias.

For the directness of the evidence, we downgraded the certainty of the evidence by one level if at least one of the components of the PICO question (participants, intervention, comparator, and outcomes) was different from the review question in more than 10% and up to 20% of all included studies for each outcome. We downgraded the certainty of the evidence by two levels if at least one of the components of the PICO question was different from the review question in more than 20% of the included studies for each outcome.

Regarding the inconsistency of results, we downgraded the certainty of the evidence by one level if there was high or very high heterogeneity (i.e. I² ≥ 50%), or if one of the two following conditions was present (or both were present and there was moderate heterogeneity (i.e. I² < 50%)): a P value of less than 0.10 in the overall pooled estimate for each outcome, or the absence of confidence interval overlap by visual inspection in one or more of the included studies for each outcome. We downgraded the certainty of the evidence by two levels if there was high or very high heterogeneity (i.e. I² ≥ 50%), the P value was less than 0.10 in the overall pooled estimate for each outcome, and the absence of confidence interval overlap by visual inspection in one or more of the included studies for each outcome.

For the precision of the estimates, we downgraded the certainty of the evidence by one level for imprecision if the 95% confidence interval of the overall pooled estimate for each outcome overlapped no effect (CI included RR of 1.0) and the CI failed to exclude important benefits or harm (risk rate reduction (RRR) > 25% and/or risk rate increase (RRI) > 25%) or if a study was terminated early with fewer than 300 events. We downgraded the certainty of the evidence by two levels for imprecision if there were an insufficient number of events, the sample size of the overall included studies for each outcome was lower than the optimal information size (OIS), the 95% confidence interval of the overall pooled estimate for each outcome overlapped no effect (CI included RR of 1.0), and the CI failed to exclude important benefit or harm (RRR > 25% and/or RRI > 25%), or if more than one study was terminated early with fewer than 300 events.

For each outcome, we calculated the optimal information size (OIS) using a web calculator (available at https://www.stat.ubc.ca/~rollin/stats/ssize/b2.html, parameters: calculate sample size (for specified power); two‐sided test; alpha: 0.05; beta: 0.80. Input value for P1: number of events or population in the DFXa group; input value for P2: number of events or population in the LMWH group). If the total number of participants for each outcome was less than the OIS, we did not downgrade the evidence certainty if the total number of participants exceeded 4000, and the 95% confidence interval of the overall pooled estimate excluded no effect, was narrow enough to exclude important benefit or important harm, or both (Guyatt 2011).

Regarding the risk of publication bias, we downgraded the certainty of the evidence if visual inspection of the funnel plot revealed asymmetry and there was statistically significant funnel plot asymmetry according to Peters' test (see Assessment of reporting biases) for each analysed outcome (see Table 3).

2. Publication bias analysis.

Outcome	Number of studies*	P value**
All‐cause mortality	16	0.7354*
Major VTE	24	0.5606*
Symptomatic VTE	25	0.7215*
Major bleeding	29	0.4488*
Major bleeding: rivaroxaban versus LMWHs	13	0.4898*
Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWHs	16	0.5463*
Serious hepatic adverse events	1	Not applicable
Serious non‐hepatic adverse events	16	0.7296*
Fatal VTE	10	0.3133*
Asymptomatic distal DVT	12	0.7972*
Fatal bleeding	4	Not applicable
Liver enzymes elevation	21	0.0101*
Minor adverse events	4	Not applicable
Volume of blood loss	14	0.1947**

*Excluding studies with 0 events
**Peters' test for dichotomous outcomes and Egger's test for continuous outcomes

Statistical test for funnel plot asymmetry was applied only to outcomes including 10 or more studies.

DVT: deep vein thrombosis; LMWHs: low molecular weight heparins; VTE: venous thromboembolism

In order to communicate the findings of our review using GRADE informative statements, we categorised the sizes of the summary effect estimates into four groups: large effect, moderate effect, small important effect, and trivial effect (Santesso 2020). We defined the criteria of the four effect size categories for each outcome, taking into consideration the small effect estimate for desirable anticipated effects and the trivial effect estimate for undesirable anticipated effects stated in the EtD (Evidence to Decision Framework) from recommendation 10 in the American Society of Hematology (ASH) 2019 guidelines (Anderson 2019; their EtD framework is available online at https://guidelines.gradepro.org/profile/9160FAA2-4F98-A3AA-9816-64DF796ABBC7). Our criteria for effect size categories are presented in Table 4.

3. Parameters of the four effect size categories for each outcome.

	Threshold of absolute effect size (risk difference) for review outcomes
Outcomes	Large effect	Moderate effect	Small important effect	Trivial, small, unimportant effect or no effect
All‐cause mortality	> 10 fewer per 1000 or > 10 more per 1000	5‐10 fewer per 1000 or 5‐10 more per 1000	2‐4 fewer per 1000 or 2‐4 more per 1000	Trivial: 1 fewer per 1000 or 1 more per 1000 (no effect 0 per 1000)
Major VTE	> 100 fewer per 1000 or > 100 more per 1000	50‐100 fewer per 1000 or 50‐100 more per 1000	10‐49 fewer per 1000 or 10‐49 more per 1000	Trivial: 1‐9 fewer per 1000 or 1‐9 more per 1000 (no effect 0 per 1000)
Symptomatic VTE	> 30 fewer per 1000 or > 30 more per 1000	10‐30 fewer per 1000 or 10‐30 more per 1000	3‐9 fewer per 1000 or 3‐9 more per 1000	Trivial: 1‐2 fewer per 1000 or 1‐2 more per 1000 (no effect 0 per 1000)
Major bleeding: rivaroxaban vs LMWHs	> 50 fewer per 1000 or > 50 more per 1000	20‐50 fewer per 1000 or 20‐50 more per 1000	3‐19 fewer per 1000 or 3‐19 more per 1000	Trivial: 1‐2 fewer per 1000 or 1‐2 more per 1000 (no effect 0 per 1000)
Major bleeding: other direct factor Xa inhibitors vs LMWHs	> 50 fewer per 1000 or > 50 more per 1000	20‐50 fewer per 1000 or 20‐50 more per 1000	3‐19 fewer per 1000 or 3‐19 more per 1000	Trivial: 1‐2 fewer per 1000 or 1‐2 more per 1000 (no effect 0 per 1000)
Serious hepatic adverse events	> 50 fewer per 1000 or > 50 more per 1000	20‐50 fewer per 1000 or 20‐50 more per 1000	3‐19 fewer per 1000 or 3‐19 more per 1000	Trivial: 1‐2 fewer per 1000 or 1‐2 more per 1000 (no effect 0 per 1000)
Serious non‐hepatic adverse events	> 50 fewer per 1000 or > 50 more per 1000	20‐50 fewer per 1000 or 20‐50 more per 1000	3‐19 fewer per 1000 or 3‐19 more per 1000	Trivial: 1‐2 fewer per 1000 or 1‐2 more per 1000 (no effect 0 per 1000)
Fatal VTE	> 10 fewer per 1000 or > 10 more per 1000	5‐10 fewer per 1000 or 5‐10 more per 1000	2‐4 fewer per 1000 or 2‐4 more per 1000	Trivial: 1 fewer per 1000 or 1 more per 1000 (no effect 0 per 1000)
Asymptomatic distal DVT	> 200 fewer per 1000 or > 200 more per 1000	100‐200 fewer per 1000 or 100‐200 more per 1000	50‐99 fewer per 1000 or 50‐99 more per 1000	Trivial: 1‐49 fewer per 1000 or 1‐49 more per 1000 (no effect 0 per 1000)
Fatal bleeding	> 10 fewer per 1000 or > 10 more per 1000	5‐10 fewer per 1000 or 5‐10 more per 1000	2‐4 fewer per 1000 or 2‐4 more per 1000	Trivial: 1 fewer per 1000 or 1 more per 1000 (no effect 0 per 1000)
Liver enzymes elevation	> 200 fewer per 1000 or > 200 more per 1000	100‐200 fewer per 1000 or 100‐200 more per 1000	50‐99 fewer per 1000 or 50‐99 more per 1000	Trivial: 1‐49 fewer per 1000 or 1‐49 more per 1000 (no effect 0 per 1000)
Minor adverse events	> 300 fewer per 1000 or > 300 more per 1000	200‐300 fewer per 1000 or 200‐300 more per 1000	100‐199 fewer per 1000 or 100‐199 more per 1000	Trivial: 1‐99 fewer per 1000 or 1‐99 more per 1000 (no effect 0 per 1000)
We have adhered to the suggested GRADE narrative statements for drawing conclusions based on the magnitude of the effect estimate from the meta‐analysis and the certainty of the evidence. These statements are summarised in Chapter 15, table 15.6.b, of the Cochrane Handbook (Schünemann 2023).

DVT: deep vein thrombosis; LMWHs: low molecular weight heparins; VTE: venous thromboembolism

For the estimation of the effect size, we used the risk difference presented in the summary of findings tables obtained by the GRADEproGDT software (Table 1; Table 2).

With the certainty of the evidence and the size of the absolute effect estimate obtained, the systematic review results could be summarised according to the GRADE informative statements. These statements are presented in the summary of main results in the Discussion and Authors' conclusions.

Results

Description of studies

See Characteristics of included studies, Characteristics of excluded studies, Characteristics of ongoing studies, and Characteristics of studies awaiting classification for detailed information.

Results of the search

We identified 2796 records through searching electronic databases and 138 additional records from other sources. After removing duplicates, 2582 records remained for screening of titles and abstracts. We excluded 2412 records at the title/abstract screening stage. We obtained the full texts of the remaining 170 records to assess eligibility for inclusion in the review. Of these, we listed three studies (three reports) as 'awaiting classification' (Hu 2022; Wang 2021; Wei 2008), and assessed 12 studies (13 reports) as ongoing (ChiCTR1800016829; IRCT20161121031003N4; IRCT20181128041784N1; IRCT20190325043107N16; JPRN‐UMIN000026819; JPRN‐UMIN000033422; NCT02085824; NCT02379663; NCT03088358; NCT05189002; NCT00408239; Pellegrini 2022). Two of the ongoing studies were in the recruiting phase at the time of completion of this review (ChiCTR1800016829; Pellegrini 2022), and the remaining 10 studies had not yet published results. We excluded 12 studies (12 reports) after retrieving and assessing full‐text reports (Çiçek 2021; CTRI/2011/07/001881; Deng 2020; DRIVE 2008; Eriksson 2000; Fuji 2010; Jiang 2018; Kwong 2007; Kwong 2008; Velik‐Salchner 2011; Verhamme 2013; Zhou 2019). We included 53 RCTs (142 reports), which met the eligibility criteria. See Figure 1.

1.

Study flow diagram

Included studies

See Characteristics of included studies.

We included 53 RCTs enroling a total of 44,371 randomised participants (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Bai 2021; Berezhnyak 2016; Changchun 2019; Chen 2016; Cohen 2013; DARINA 2021; EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Hoseinzadeh 2022; Hu 2015; Hui 2013; Jiang 2019; Kanan 2008; Karampinas 2019; Khalafallah 2018; Kim 2016; Kunal 2021; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Özler 2015; Rahman 2020; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Tang 2017; Wasko 2015; Weitz 2010; Wing 2020; Xie 2017; Yun‐Fei 2018; Zhou 2023; Zou 2014). The biggest study was performed in 5407 randomised people (ADVANCE‐3 2010), and the smallest study with outcome data was performed in 50 people (Wing 2020).

As expected, all included studies were parallel‐group in design, and 32 of them were stated to be double‐blinded for the comparisons of interest for this review (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Bai 2021; Cohen 2013; EXPERT 2009; Fuji 2014‐THA; Fuji 2014‐TKA; Kanan 2008; Kim 2016; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐2 2010; ONYX‐3 2014; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐V 2015; Tang 2017; Wasko 2015; Weitz 2010; Zhou 2023).

Most studies were performed in high‐income regions such as North America, Europe, Australia, and Japan. Twenty‐one studies included participants from India, China, Taiwan, or Thailand (ADVANCE‐3 2010; Bai 2021; Changchun 2019; Chen 2016; Fuji 2014‐THA; Fuji 2014‐TKA; Hu 2015; Hui 2013; Jiang 2019; Kunal 2021; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; Tang 2017; Wing 2020; Xie 2017; Yun‐Fei 2018; Zhou 2023; Zou 2014); eight studies included participants from Latin America (ADVANCE‐1 2009; ADVANCE‐3 2010; APROPOS 2007; Cohen 2013; Kanan 2008; RECORD1 2008; RECORD3 2008; RECORD4 2009); six studies included participants from South Africa, Egypt, or Sri Lanka (FOXTROT 2020; Rahman 2020; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009); five studies included participants from Russia and Ukraine (ADVANCE‐1 2009; ADVANCE‐3 2010; Berezhnyak 2016; FOXTROT 2020; Raskob 2010); and four studies included participants from Iran, Turkey, and Pakistan (Hoseinzadeh 2022; Özler 2015; RECORD1 2008; RECORD4 2009). The included studies were conducted in countries unevenly distributed around the world.

All the studies included mixed populations in terms of gender; the average percentage of males was 31%. Most participants were aged between 18 and 93 years; the average age was 64 years. Some but not all included studies provided details about participants' ethnicity. Of those that did, we established that 70.27% of participants were white, 22.43% Asian, 3.36% Black, 1.04% Latin, 0.82% Arab, 0.28% Hindu, and 1.8% were of 'other' ethnicities, meaning that some ethnicities were very under‐represented in the sample.

Regarding the type of surgery performed, 23 studies included participants with total hip replacement (ADVANCE‐3 2010; Bai 2021; Berezhnyak 2016; Fuji 2014‐THA; Hui 2013; Kanan 2008; Kim 2016; NCT01205932; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Rahman 2020; Raskob 2010; RECORD1 2008; RECORD2 2008; STARS J‐2 2014; STARS J‐V 2015; Wasko 2015; Wing 2020; Yun‐Fei 2018); 21 with total knee replacement (ADVANCE‐1 2009; ADVANCE‐2 2010; APROPOS 2007; Cohen 2013; DARINA 2021; EXPERT 2009; FOXTROT 2020; Fuji 2014‐TKA; Hu 2015; Jiang 2019; Karampinas 2019; Lassen 2003; NCT01206972; ODIXa‐KNEE 2005; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; Weitz 2010; Xie 2017; Zhou 2023; Zou 2014); six studies included participants undergoing either total hip replacement or total knee replacement (Agnelli 2007; Changchun 2019; Chen 2016; Khalafallah 2018; Kunal 2021; Özler 2015); and three studies included people undergoing hip fracture surgery (Hoseinzadeh 2022; STARS J‐4 2014; Tang 2017). Surgery was explicitly stated to have been elective in 28 studies (ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Berezhnyak 2016; DARINA 2021; EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Khalafallah 2018; Kim 2016; Kunal 2021; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Raskob 2010; RECORD1 2008; RECORD2 2008; STARS J‐V 2015; Weitz 2010). The underlying joint disease for surgery was osteoarthritis for most participants (70%) (ADVANCE‐2 2010; ADVANCE‐3 2010; Fuji 2014‐THA; Fuji 2014‐TKA; Kanan 2008; ONYX‐2 2010; ONYX‐3 2014; STARS E‐3 2014), with other underlying causes including degenerative joint disease, rheumatoid arthritis, and joint fracture.

In all 53 studies, the experimental intervention was a direct factor Xa inhibitor and was compared with LMWH. One study also compared a direct factor Xa inhibitor with the VKA warfarin (APROPOS 2007).

Study interventions included apixaban 5 to 20 mg/day in seven studies (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; APROPOS 2007; FOXTROT 2020; Jiang 2019; Kunal 2021), betrixaban 30 to 80 mg/day in one study (EXPERT 2009), darexaban 3 to 120 mg/day in five studies (Fuji 2014‐THA; Fuji 2014‐TKA; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014), edoxaban 15 to 90 mg/day in five studies (Raskob 2010; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015), eribaxaban 0.1 to 10 mg/day in one study (Cohen 2013), letaxaban 20 to 160 mg/day in one study (Weitz 2010), LY517717 25 to 150 mg/day in one study (Agnelli 2007), razaxaban 50 to 200 mg/day in one study (Lassen 2003), and rivaroxaban 5 to 60 mg/day in 31 studies (Bai 2021; Berezhnyak 2016; Changchun 2019; Chen 2016; DARINA 2021; Hoseinzadeh 2022; Hu 2015; Hui 2013; Kanan 2008; Karampinas 2019; Khalafallah 2018; Kim 2016; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; Özler 2015; Rahman 2020; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; Tang 2017; Wasko 2015; Wing 2020; Xie 2017; Yun‐Fei 2018; Zhou 2023; Zou 2014). Comparators included dalteparin 2500 to 5000 IU/day (Raskob 2010; Zou 2014), enoxaparin 40 to 80 mg/day (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Bai 2021; Chen 2016; Cohen 2013; EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Hoseinzadeh 2022; Hu 2015; Hui 2013; Kanan 2008; Khalafallah 2018; Kim 2016; Kunal 2021; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Özler 2015; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Tang 2017; Wasko 2015; Weitz 2010; Xie 2017; Zou 2014), nadroparin 0.3 mL/day (DARINA 2021), and titrated warfarin (APROPOS 2007).

In all studies, the experimental intervention was started between zero and 24 hours after surgery. The comparator was started between: six and 24 hours before surgery (ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; Berezhnyak 2016; FOXTROT 2020; Hu 2015; Kanan 2008; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; RECORD1 2008; RECORD2 2008; RECORD3 2008); zero to 36 hours after surgery (ADVANCE‐1 2009; APROPOS 2007; Bai 2021; Changchun 2019; Chen 2016; Cohen 2013; DARINA 2021; EXPERT 2009; Fuji 2014‐THA; Fuji 2014‐TKA; Hoseinzadeh 2022; Hui 2013; Jiang 2019; Karampinas 2019; Khalafallah 2018; Kim 2016; Kunal 2021; Lassen 2003; ODIXa‐KNEE 2005; Rahman 2020; Raskob 2010; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Tang 2017; Weitz 2010; Wing 2020; Xie 2017; Yun‐Fei 2018; Zhou 2023; Zou 2014); or after discharge (Özler 2015). The duration of the intervention ranged from six to 39 days, and post‐intervention follow‐up lasted between two and 720 days. The average follow‐up duration was 42 days.

Regarding the diagnostic instrument for measuring DVT, most (26/53) of the included studies evaluated DVT events with bilateral venography (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Cohen 2013; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Kanan 2008; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Weitz 2010), and three studies evaluated DVT events with unilateral venography (EXPERT 2009; Lassen 2003; STARS E‐3 2014). Regarding other non‐invasive instruments, 13 of 53 studies evaluated DVT events with Doppler ultrasound (Bai 2021; Hoseinzadeh 2022; Hu 2015; Hui 2013; Jiang 2019; Khalafallah 2018; Kim 2016; Kunal 2021; Wing 2020; Xie 2017; Yun‐Fei 2018; Zhou 2023; Zou 2014), and one study with computed tomography venography (Rahman 2020). Ten studies did not give any information regarding the diagnostic instrument for DVT (Berezhnyak 2016; Changchun 2019; Chen 2016; DARINA 2021; Karampinas 2019; NCT01205932; NCT01206972; Özler 2015; Tang 2017; Wasko 2015).

In seven of the 53 included studies, the duration of prophylaxis differed markedly between the two study arms (Berezhnyak 2016; Chen 2016; Hui 2013; NCT01205932; RECORD2 2008; Wing 2020; Yun‐Fei 2018). In four studies, the duration of prophylaxis was approximately five weeks for the factor Xa inhibitor groups and two weeks for the comparator groups, with outcome analysis taking place at these two different time points (Chen 2016; Hui 2013; RECORD2 2008; Wing 2020). In these studies, follow‐up was longer than two weeks for both groups. Two further studies also presented different durations of prophylaxis: in Berezhnyak 2016, 10 days for the control and 30 days for the intervention group, and in Yun‐Fei 2018, five weeks for the control and two weeks for the intervention group; however, they did not state the time point(s) analysed. The final study also had different prophylaxis durations for the two groups (five weeks for the intervention and two weeks for the comparison); however, investigators analysed the outcomes at the same time point (two weeks) (NCT01205932).

In nine included studies, participants were given co‐interventions which could affect study outcomes. Participants in the intervention and comparator groups in four studies received tranexamic acid (TXA) before surgery (Bai 2021; Karampinas 2019; Xie 2017; Zhou 2023). This co‐intervention could reduce postoperative bleeding. In three studies, participants in both study arms received intermittent pneumatic compression devices (IPC) after surgery (Khalafallah 2018; Kim 2016; Kunal 2021), and in Berezhnyak 2016, participants in both groups received elastic compression stockings. These co‐interventions could reduce postoperative VTE. In Özler 2015, the intervention and comparator groups received enoxaparin 0.4 mL twice daily during the hospital stay, initiated 12 hours after surgery, and were then switched to the study drugs for 10 days (in participants undergoing TKA) or 30 days (in those undergoing THA). This co‐intervention could reduce postoperative VTE. Two studies did not state if any co‐intervention was received (Changchun 2019; Jiang 2019).

We found 142 published data reports for the 53 included studies. Regarding unpublished data, we contacted authors of 38 studies to request missing data (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Chen 2016; Cohen 2013; EXPERT 2009; Fuji 2014‐THA; Fuji 2014‐TKA; Hui 2013; Kanan 2008; Kim 2016; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Özler 2015; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Tang 2017; Wasko 2015; Weitz 2010; Xie 2017; Zou 2014). We received responses from six authors (Fuji 2014‐THA; Fuji 2014‐TKA; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015). For details, see the 'Notes' section in the Characteristics of included studies table.

Ten of 53 studies reported no missing outcome data for benefit (i.e. efficacy) outcomes, and 11 of 53 studies reported no missing outcome data for harm (i.e. safety) outcomes.

Twenty‐two studies did not have a CONSORT flow diagram (Bai 2021; Berezhnyak 2016; Chen 2016; Fuji 2014‐THA; Fuji 2014‐TKA; Hoseinzadeh 2022; Hu 2015; Hui 2013; Jiang 2019; Kanan 2008; Karampinas 2019; Kunal 2021; Lassen 2003; NCT01205932; NCT01206972; Özler 2015; Rahman 2020; Wasko 2015; Wing 2020; Xie 2017; Yun‐Fei 2018; Zou 2014), and four studies presented tables with minimal information (Khalafallah 2018; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; Tang 2017), despite the fact that all were performed after the CONSORT statement was published (Begg 1996).

For benefit outcomes, 18 of 53 studies reported post‐randomisation exclusion rates lower than 10% (Bai 2021; Chen 2016; DARINA 2021; Hoseinzadeh 2022; Hu 2015; Hui 2013; Jiang 2019; Kanan 2008; Khalafallah 2018; Kim 2016; Kunal 2021; Özler 2015; Rahman 2020; Tang 2017; Wing 2020; Xie 2017; Yun‐Fei 2018; Zhou 2023). Thirty studies reported exclusion rates higher than 10% (up to 47%) (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Cohen 2013; EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Weitz 2010). Five studies did not report the exclusion rate for benefit outcomes (Berezhnyak 2016; Changchun 2019; Karampinas 2019; Wasko 2015; Zou 2014).

For harm outcomes, 43 of 53 studies reported post‐randomisation exclusion rates lower than 10% (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Bai 2021; Chen 2016; Cohen 2013; DARINA 2021; EXPERT 2009; Fuji 2014‐THA; Fuji 2014‐TKA; Hoseinzadeh 2022; Hu 2015; Hui 2013; Jiang 2019; Kanan 2008; Khalafallah 2018; Kim 2016; Kunal 2021; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Özler 2015; Rahman 2020; Raskob 2010; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Tang 2017; Wasko 2015; Weitz 2010; Wing 2020; Xie 2017; Yun‐Fei 2018; Zhou 2023). Four studies did not report the exclusion rate for harm outcomes (Berezhnyak 2016; Changchun 2019; Karampinas 2019; Zou 2014). The remaining six studies reported exclusion rates higher than 10% (up to 24%) for one or more of their harm outcomes (FOXTROT 2020; Lassen 2003; ODIXa‐HIP‐OD 2006; RECORD1 2008; RECORD2 2008; RECORD3 2008).

Most of the studies performed an ‘as‐treated’ or modified intention‐to‐treat (ITT) analysis without substantial departures in the intervention received from that assigned at randomisation.

Seven of 53 included studies did not report if authors registered any potential conflicts of interest (Jiang 2019; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP 2007; Özler 2015; Wasko 2015). The authors of 28 studies declared financial conflicts of interest (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Cohen 2013; DARINA 2021; EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Kanan 2008; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Weitz 2010). None of these 35 studies reported a management plan describing procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Authors of the remaining 18 studies declared no conflicts of interest (Bai 2021; Berezhnyak 2016; Changchun 2019; Chen 2016; Hoseinzadeh 2022; Hu 2015; Hui 2013; Karampinas 2019; Khalafallah 2018; Kim 2016; Kunal 2021; Rahman 2020; Tang 2017; Wing 2020; Xie 2017; Yun‐Fei 2018; Zhou 2023; Zou 2014).

Twenty‐seven of the 53 included studies reported external private sources of funding including Astellas Pharma (Fuji 2014‐THA; Fuji 2014‐TKA; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014), Bayer HealthCare (FOXTROT 2020; Kanan 2008; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009), Bayer Yakuhin, Ltd.(NCT01205932; NCT01206972), Bristol‐Myers Squibb‐Pfizer (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010), Daiichi Sankyo (Raskob 2010; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015), Johnson & Johnson Pharmaceutical Research & Development (RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009), Pfizer (Cohen 2013), Portola Pharmaceuticals (EXPERT 2009), and Takeda Global Research & Development (Weitz 2010). Only five of the 27 studies were performed in one country: Brazil (Kanan 2008: one centre; the private funding support was not specifically stated). The other four were conducted in Japan (NCT01205932: 42 centres; NCT01206972: 33 centres; STARS J‐4 2014: one centre; STARS J‐V 2015: one centre).

Thirteen of 53 studies did not report funding sources (Agnelli 2007; APROPOS 2007; Berezhnyak 2016; Chen 2016; DARINA 2021; Hu 2015; Karampinas 2019; Kim 2016; Kunal 2021; Lassen 2003; ODIXa‐HIP 2007; Özler 2015; Zou 2014). The remaining 13 studies reported national or local funds (Bai 2021; Changchun 2019; Hoseinzadeh 2022; Hui 2013; Jiang 2019; Khalafallah 2018; Rahman 2020; Tang 2017; Wasko 2015; Wing 2020; Xie 2017; Yun‐Fei 2018; Zhou 2023). Of these, 10 studies were conducted in China, one in Australia (Khalafallah 2018), one in Iran (Hoseinzadeh 2022), and one in Poland (Wasko 2015). All were single‐centre studies, except for Bai 2021, with three centres.

Excluded studies

We excluded 12 studies after retrieving and assessing full‐text reports (Çiçek 2021; CTRI/2011/07/001881; Deng 2020; DRIVE 2008; Eriksson 2000; Fuji 2010; Jiang 2018; Kwong 2007; Kwong 2008; Velik‐Salchner 2011; Verhamme 2013; Zhou 2019).

We excluded two studies because they were healthcare costs studies based on RCTs already included in this review (Kwong 2007; Kwong 2008). We excluded four studies because they had ineligible study designs: two were non‐RCTs (CTRI/2011/07/001881; Deng 2020), and two were observational in design (Çiçek 2021; Velik‐Salchner 2011). We excluded one study because it evaluated the intervention drug as treatment, not prophylaxis, for venous thromboembolism (Jiang 2018), and two further studies because they evaluated an ineligible intervention (DRIVE 2008; Eriksson 2000). We excluded two studies because they used an ineligible comparator (Fuji 2010; Verhamme 2013), and one study because enoxaparin was included in both treatment arms (Zhou 2019). See Characteristics of excluded studies for further details.

Studies awaiting classification

Three studies await classification because we were unable to obtain full‐text reports (Hu 2022; Wang 2021; Wei 2008). Available information in abstracts was insufficient to determine if these studies met the inclusion criteria for this review. See Characteristics of studies awaiting classification for further details.

Ongoing studies

We identified 12 ongoing trials that met the review eligibility criteria (ChiCTR1800016829; IRCT20161121031003N4; IRCT20181128041784N1; IRCT20190325043107N16; JPRN‐UMIN000026819; JPRN‐UMIN000033422; NCT02085824; NCT02379663; NCT03088358; NCT05189002; NCT00408239; Pellegrini 2022). Participants are undergoing total hip or knee replacement surgery or hip fracture surgery. All 12 have a parallel‐group study design. Half (six of 12 studies) are either double‐ or triple‐blinded, while five are explicitly not blinded (open‐label), and one study does not provide details on blinding (JPRN‐UMIN000026819). Six of 12 studies compare rivaroxaban to enoxaparin (with one study featuring a third study arm receiving dabigatran). The remaining comparisons are: edoxaban versus enoxaparin (two studies); three different doses of TeaRx, a factor Xa inhibitor, versus enoxaparin (one study); dimolegin versus dalteparin (one study); two doses of YM150 versus enoxaparin (one study); and rivaroxaban versus warfarin (one study). Countries where these studies are taking place are: China (one study); Iran (three studies); Japan (two studies); Poland (one study); and Russia (two studies). One study is being conducted in North America, and one in eight Asian countries. The final study did not provide location details.

At the time this review was being prepared, most of these trials had completed their participant recruitment phase, but we found no published results. See Characteristics of ongoing studies for further details.

Risk of bias in included studies

Our risk of bias assessment of the included studies is illustrated in Figure 2 and Figure 3. For details, see Characteristics of included studies.

2.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

3.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

We grouped the review outcomes into two categories for risk of attrition bias assessment: (1) benefit outcomes, including all‐cause mortality, major VTE, symptomatic VTE, fatal VTE, and asymptomatic distal DVT; and (2) harm outcomes, including major bleeding, serious adverse events (hepatic and non‐hepatic), fatal bleeding, liver enzymes elevation, minor adverse events, and volume of blood loss.

We gave special importance to selection bias (random sequence generation and concealment of allocation) and performed sensitivity analyses (Analysis 10.1 to Analysis 10.14) excluding those studies we deemed to have high risk of bias for these criteria (Agnelli 2007; Berezhnyak 2016; Cohen 2013; ODIXa‐HIP 2007; Rahman 2020).

10.1. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 1: All‐cause mortality

10.14. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 14: Volume of blood loss

Allocation

All included studies randomly assigned participants to treatment groups; however, adequate methodology was described only for 27 of 53 studies (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; APROPOS 2007; Bai 2021; DARINA 2021; FOXTROT 2020; Hoseinzadeh 2022; Hu 2015; Khalafallah 2018; Kim 2016; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Rahman 2020; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; Tang 2017; Xie 2017; Yun‐Fei 2018; Zhou 2023; Zou 2014). Four other studies described their methodology, reporting that participants were distributed sequentially for their intervention drug groups (Agnelli 2007; Berezhnyak 2016; Cohen 2013; ODIXa‐HIP 2007). In these studies, the probability of participants being allocated to each study group was different, so we considered their risk of bias to be high in this domain. We judged the risk of bias in the remaining 22 studies to be unclear (Changchun 2019; Chen 2016; EXPERT 2009; Fuji 2014‐THA; Fuji 2014‐TKA; Hui 2013; Jiang 2019; Kanan 2008; Karampinas 2019; Kunal 2021; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; Özler 2015; Raskob 2010; STARS J‐4 2014; STARS J‐V 2015; Wasko 2015; Weitz 2010; Wing 2020).

Adequate methods of allocation concealment were described in 17 of 53 studies (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; DARINA 2021; Hoseinzadeh 2022; Karampinas 2019; Khalafallah 2018; ODIXa‐KNEE 2005; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; Tang 2017; Weitz 2010; Xie 2017; Zhou 2023), generally consisting in allocation by a central office contacted by telephone or fax. We judged one study to be at high risk of bias (Rahman 2020), as the allocation sequence was not concealed until the moment of assignment from those assigning participants to the rivaroxaban group. We assessed the remaining 35 studies to be at unclear risk (Agnelli 2007; APROPOS 2007; Bai 2021; Berezhnyak 2016; Changchun 2019; Chen 2016; Cohen 2013; EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Hu 2015; Hui 2013; Jiang 2019; Kanan 2008; Kim 2016; Kunal 2021; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Özler 2015; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Wasko 2015; Wing 2020; Yun‐Fei 2018; Zou 2014).

Blinding

Regarding blinding of participants and personnel, 25 of 53 studies reported having used a double‐blind design for the comparisons of interest for this review (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Bai 2021; Kanan 2008; Kim 2016; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐3 2014; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐V 2015; Tang 2017; Wasko 2015; Zhou 2023), so we assessed these as being at low risk of performance bias. However, just 22 of them reported having also used a double‐dummy strategy (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Bai 2021; Kanan 2008; Kim 2016; NCT01205932; NCT01206972; ODIXa‐HIP‐OD 2006; ODIXa‐HIP2 2006; ODIXa‐KNEE 2005; ONYX‐3 2014; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐V 2015; Zou 2014), and only three studies explicitly indicated blinding of participants and personnel (ADVANCE‐2 2010; ONYX‐3 2014; Zhou 2023). Twenty‐one studies explicitly stated using an open‐label design for the comparison of interest (Berezhnyak 2016; Cohen 2013; DARINA 2021; EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Hu 2015; Jiang 2019; Karampinas 2019; Khalafallah 2018; ODIXa‐HIP 2007; ONYX‐1 2007; ONYX‐2 2010; Özler 2015; Rahman 2020; STARS J‐2 2014; STARS J‐4 2014; Weitz 2010; Xie 2017; Yun‐Fei 2018), and we assumed this design was also used in five additional studies based on the descriptions provided (Changchun 2019; Hoseinzadeh 2022; Hui 2013; Kunal 2021; Wing 2020). Thus, we assessed these 26 studies as being at high risk of performance bias. Two studies did not describe the blinding method (Chen 2016; Zou 2014).

Adequate blinding of outcome assessment was reported in six of 53 studies for benefit and harm outcomes, so we deemed these to have a low risk of detection bias (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; FOXTROT 2020; Khalafallah 2018; Zhou 2023). We judged the remaining 47 studies to be at an unclear risk of bias because blinding of data was not mentioned (Agnelli 2007; APROPOS 2007; Bai 2021; Berezhnyak 2016; Changchun 2019; Chen 2016; Cohen 2013; DARINA 2021; EXPERT 2009; Fuji 2014‐THA; Fuji 2014‐TKA; Hoseinzadeh 2022; Hu 2015; Hui 2013; Jiang 2019; Kanan 2008; Karampinas 2019; Kim 2016; Kunal 2021; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007;ONYX‐2 2010;ONYX‐3 2014; Özler 2015; Rahman 2020; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009;STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Tang 2017; Wasko 2015; Weitz 2010; Wing 2020; Xie 2017; Yun‐Fei 2018; Zou 2014).

Incomplete outcome data

Due to the important difference in exclusion rates between benefit and harm outcomes, we performed a separate analysis of these outcome types to evaluate the risk of attrition bias for incomplete outcome data.

Four studies did not evaluate any of the review outcomes (Berezhnyak 2016; Changchun 2019; Hoseinzadeh 2022; Kanan 2008) and we thus considered that this bias domain was not applicable to these studies.

Benefit outcomes

We judged nine of 53 studies to have a low risk of attrition bias: eight reported no missing outcome data (Bai 2021; Hui 2013; Kunal 2021; Özler 2015; Rahman 2020; Wing 2020; Yun‐Fei 2018; Zhou 2023). We assessed the remaining study as low risk due to at least one of the following reasons: (1) missing outcome data unlikely to be related to true outcome; (2) missing outcome data balanced in numbers across intervention groups, with similar reasons for missing data across groups; or (3) the proportion of missing outcomes compared with the observed event risk was not enough to have had a clinically relevant impact on the intervention effect estimate (STARS E‐3 2014).

We judged 29 of 53 studies to have a high risk of attrition bias: in 28 studies, the proportion of missing outcomes compared with the observed event risk was enough to have had a clinically relevant impact on the intervention effect estimate (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Cohen 2013; DARINA 2021; EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS J‐2 2014; STARS J‐4 2014; Xie 2017). See Appendix 5 for our calculations for each study. We judged the remaining one study to have a high risk of attrition bias because the reason for missing outcome data is likely to be related to one of the review outcomes (thrombosis, death), with either an imbalance in numbers or reasons for missing data across intervention groups (Weitz 2010). We also judged three other studies mentioned above to have a high risk of attrition bias for this reason (EXPERT 2009; ODIXa‐HIP 2007; ONYX‐1 2007).

We assessed 11 of 53 studies to have an unclear risk of attrition bias. Five did not have a CONSORT flow diagram, and it is unclear if there were exclusions or withdrawals (Hu 2015; Jiang 2019; Karampinas 2019; Wasko 2015; Zou 2014). The remaining six studies provided insufficient reporting of attrition/exclusions to permit a judgement of ‘low risk’ or ‘high risk’ (Chen 2016; Khalafallah 2018; Kim 2016; Lassen 2003; STARS J‐V 2015; Tang 2017).

Harm outcomes

We judged 22 of 53 studies to have a low risk of attrition bias; nine of these reported no missing outcome data (Bai 2021; Hui 2013; Kunal 2021; Özler 2015; Rahman 2020; Wasko 2015; Wing 2020; Yun‐Fei 2018; Zhou 2023). We judged the remaining 13 studies to have low risk of attrition bias due to at least one of the following reasons: (1) missing outcome data were unlikely to be related to the true outcome; (2) missing outcome data were balanced in numbers across intervention groups, with similar reasons for missing data across groups; or (3) the proportion of missing outcomes compared with the observed event risk is not enough to have had a clinically relevant impact on the intervention effect estimate (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; APROPOS 2007; Chen 2016; Cohen 2013; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ONYX‐1 2007; Raskob 2010; RECORD4 2009; STARS E‐3 2014; STARS J‐V 2015).

We assessed eight of 53 studies to have a high risk of attrition bias. In seven, the proportion of missing outcome data compared with the observed event risk is enough to have had a clinically relevant impact on the intervention effect estimate (Agnelli 2007; DARINA 2021; Fuji 2014‐TKA; ODIXa‐KNEE 2005; STARS J‐2 2014; STARS J‐4 2014; Xie 2017). See Appendix 6 for our calculations for each study. In the remaining study, we judged that the reason for missing outcome data is likely to be related to one of the review outcomes (bleeding, death, or adverse events), with either an imbalance in numbers or reasons for missing data across intervention groups (Weitz 2010).

We assessed 19 of 53 studies to have an unclear risk of attrition bias. Four did not have a CONSORT flow diagram, and it is not clear if there were exclusions or withdrawals (Hu 2015; Jiang 2019; Karampinas 2019; Zou 2014). The remaining 15 provided insufficient reporting of attrition/exclusions to permit a judgement of ‘low risk’ or ‘high risk’ (EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Khalafallah 2018; Kim 2016; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ONYX‐2 2010; ONYX‐3 2014; RECORD1 2008; RECORD2 2008; RECORD3 2008; Tang 2017).

Selective reporting

We assessed selective reporting by cross‐checking study outcomes between included reports and published protocols. We found protocols for 28 of 53 included studies (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; DARINA 2021; FOXTROT 2020; Khalafallah 2018; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐2 2010; ONYX‐3 2014; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Tang 2017; Wasko 2015; Weitz 2010). We deemed only three of these studies to be at low risk of reporting bias (FOXTROT 2020; Khalafallah 2018; Wasko 2015). In Wasko 2015, selective reporting did not involve any outcome included in this review, and in FOXTROT 2020 and Khalafallah 2018, the reported outcomes were the same as those specified in the protocol. The other studies reported outcomes not included in their protocols or did not report outcomes included in their protocols, with selective reporting involving outcomes considered in this review. Thus, we deemed 25 of 53 studies to be at high risk of selective reporting bias.

Changchun 2019 also has a protocol available, but this study did not evaluate any of the review outcomes. For that reason, we considered that the selective reporting domain was not applicable.

We assessed the remaining 24 studies to be at unclear risk of selective reporting (Bai 2021; Berezhnyak 2016; Chen 2016; Cohen 2013; EXPERT 2009; Fuji 2014‐THA; Fuji 2014‐TKA; Hoseinzadeh 2022; Hu 2015; Hui 2013; Jiang 2019; Kanan 2008; Karampinas 2019; Kim 2016; Kunal 2021; Lassen 2003; ONYX‐1 2007; Özler 2015; Rahman 2020; Wing 2020; Xie 2017; Yun‐Fei 2018; Zhou 2023; Zou 2014).

Other potential sources of bias

For this review, we considered the following potential sources of bias.

• Conflicts of interest

  • We assessed 35 studies as having a high risk of bias because they either did not report if conflicts of interest were registered, or did report conflicts of interest and did not describe a management plan to minimise the risk of bias to ensure research integrity (such as divestiture or independent data review).

  • We assessed the remaining 18 studies as having a low risk of bias as they explicitly reported that authors/investigators had no conflicts of interest to declare (see details in Characteristics of included studies).

• Sponsorship: we were unable to explore this in detail because we did not have access to the original study data in any of the included studies. Sponsorship bias could be the root cause of other types of bias we detected.

• Administration of an intervention prior to randomisation, that could enhance or diminish the effect of a subsequent, randomised, intervention. We found no evidence of this potential source of bias in the included studies.

• Additional participants recruited post randomisation: we found no evidence that any of the included studies recruited additional participants post randomisation.

• Contamination: we found no contamination.

• Studies stopped early for benefit: none of the included studies were stopped early for benefit.

• Inappropriate administration of an intervention

  • In seven of 53 studies, the duration of prophylaxis differed markedly for each intervention group. We considered four of these seven studies to be at high risk of bias because of this key difference and because the outcome analyses were at different time points (at five versus two weeks). Follow‐up was longer than two weeks.

  • We considered two of the seven studies to be at unclear risk of bias: they had different durations of prophylaxis (five weeks for the comparator group and two weeks for the intervention group); however, they did not state the time point(s) analysed.

  • The final study of the seven with different durations of prophylaxis (five weeks for the intervention group and two weeks for the comparator group) analysed the outcomes at the same time point (two weeks). We rated this study as having a low risk of bias for this reason. We judged the remaining 46 studies to be at low risk for this source of bias (more details in Included studies).

• Inappropriate administration of a co‐intervention

  • We judged 42 studies to be at low risk of bias as participants did not receive any inappropriate co‐intervention that could affect the outcomes addressed in this review.

  • We assessed the remaining 11 studies as having an unclear risk of bias: in nine studies, participants were given co‐interventions which could affect study outcomes, and two studies did not state if any co‐interventions were administered (see details in Included studies).

• Insensitive instrument for measuring outcomes, which can lead to underestimation of beneficial effects

  • We judged that all the studies (29 of 53) that measured DVT with venography (bilateral or unilateral) had a high risk of bias as venography is an insensitive instrument which can lead to an underestimation of beneficial effects (more details in Included studies and Discussion).

  • We assessed 14 studies as having a low risk of bias: 13 evaluated DVT events with Doppler ultrasound and one study with computed tomography venography.

  • We assessed the remaining 10 studies as having an unclear risk of bias as they provided no information regarding the diagnostic instrument for DVT.

• Studies with suspected fraud

  • We considered one study to be at high risk of bias because of suspected fraud (RECORD4 2009). This study is under investigation (Turpie 2022). We excluded this study from the adverse event outcomes (serious hepatic and non‐hepatic adverse events and minor adverse events) in our main analyses.

  • We considered the remaining 52 studies to be at low risk of bias (more details in Discussion).

Taking into account all these potential sources of bias, we judged that five of 53 included studies were at low risk of bias (Hoseinzadeh 2022; Hu 2015; Rahman 2020; Tang 2017; Zou 2014); 10 at unclear risk of bias (Bai 2021; Berezhnyak 2016; Changchun 2019; Karampinas 2019; Khalafallah 2018; Kim 2016; Kunal 2021; Xie 2017; Yun‐Fei 2018; Zhou 2023); and 38 at high risk of bias (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Chen 2016; Cohen 2013; DARINA 2021; EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Hui 2013; Jiang 2019; Kanan 2008; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Özler 2015; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Wasko 2015; Weitz 2010; Wing 2020).

Effects of interventions

See: Table 1

See Table 1 and Table 2. We performed quantitative synthesis only for the comparison between direct factor Xa inhibitors and LMWHs. We did not include five studies in the quantitative synthesis because they did not provide data amenable to inclusion in meta‐analysis (Berezhnyak 2016; Changchun 2019; Hoseinzadeh 2022; Kanan 2008; Lassen 2003). We were unable to perform meta‐analyses for the comparison of direct factor Xa inhibitors versus VKA as this was reported by only one study (APROPOS 2007).

If we found statistical heterogeneity, we inspected the forest plot for the direction of effect of individual studies and overlap of confidence intervals, and we explored if a clinical explanation could be the cause of the heterogeneity by performing several subgroup analyses. If we still could not explain the heterogeneity, we compared the fixed‐effect and random‐effects meta‐analyses, searching for differences in direction or magnitude, with the intention of presenting the random‐effects result if there was no indication of funnel plot asymmetry. If there was an indication of funnel plot asymmetry, we performed a sensitivity analysis in which we excluded small studies.

We avoided presenting meta‐analysis results if: (1) we found very high unexplained statistical heterogeneity; or (2) the fixed‐effect and random‐effects meta‐analytic results differed in direction or magnitude. In addition, we performed no meta‐analysis if we found high unexplained statistical heterogeneity with funnel plot asymmetry and different results in the sensitivity analysis excluding small studies.

The main analyses (Analysis 1.1 to Analysis 1.14) include all the outcomes with all the combined doses of direct factor Xa inhibitors, with the events of both surgeries (THA and TKA) combined, and include the events reported in the follow‐up period. We compared the events in the treatment and follow‐up periods with the population reported for the treatment period in each study.

1.1. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 1: All‐cause mortality

1.14. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 14: Volume of blood loss

Direct factor Xa inhibitors versus LMWHs

Primary outcomes

The main results are summarised in Analysis 1.1, Analysis 1.2, Analysis 1.3, Analysis 1.4, Analysis 1.5, Analysis 1.6, Analysis 1.7, Analysis 1.8, and Table 1. We carried out subgroup analyses to investigate if the type of drug, duration of prophylaxis, time of initiation of prophylaxis, frequency of administration, type of surgery, or drug dose had any effect on the outcomes. We report these subgroup analyses separately below.

1.2. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 2: Major VTE

1.3. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 3: Symptomatic VTE

1.4. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 4: Major bleeding

1.5. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 5: Major bleeding: rivaroxaban versus LMWH

1.6. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 6: Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH

1.7. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 7: Serious hepatic adverse events

1.8. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 8: Serious non‐hepatic adverse events

All‐cause mortality

Twenty‐eight studies reported all‐cause mortality for direct factor Xa inhibitors compared with LMWHs. Sixteen studies (26,108 participants) reported deaths (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; APROPOS 2007; Fuji 2014‐TKA; Kunal 2021; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐KNEE 2005; ONYX‐2 2010; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; Weitz 2010), and twelve studies (3590 participants) reported zero deaths (Chen 2016; FOXTROT 2020; Fuji 2014‐THA; Jiang 2019; Kim 2016; NCT01205932; NCT01206972; ODIXa‐HIP‐OD 2006; Rahman 2020; STARS J‐2 2014; STARS J‐4 2014; Xie 2017).

Direct factor Xa inhibitors may have little to no effect on all‐cause mortality compared to LMWHs, but the evidence is very uncertain (risk ratio (RR) 0.83, 95% confidence interval (CI) 0.52 to 1.31; risk difference: 0 fewer deaths per 1000 participants, 95% CI 1 fewer to 1 more; I² = 0%; 28 studies, 29,698 participants; very low‐certainty evidence; Analysis 1.1).

Major VTE

We defined major VTE as the incidence of proximal DVT (DVT from the popliteal vein), symptomatic DVT, pulmonary embolism (fatal or non‐fatal) and VTE‐related death. Twenty‐eight studies reported major VTEs for direct factor Xa inhibitors compared with LMWHs. Of these, 24 studies (23,983 participants) reported major VTE events (ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Kunal 2021; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; Wing 2020; Zhou 2023), and four studies (611 participants) reported zero events for this outcome (Özler 2015; STARS J‐2 2014; STARS J‐4 2014; Xie 2017).

There was moderate heterogeneity (I² = 48%), so we inspected the forest plot for the direction of effect of individual studies. The confidence intervals generally overlap, and the larger studies tend to show an effect favouring direct factor Xa inhibitors. We explored if a clinical explanation could be the cause of the heterogeneity by performing several subgroup analyses (Analysis 2.2; Analysis 3.2; Analysis 4.2; Analysis 5.2; Analysis 6.2; Analysis 7.2). We found statistically significant differences (P value < 0.10) between subgroups only in the drug type subgroup (Analysis 2.2) and daily dose subgroup (Analysis 7.2); see the subgroup analysis section below for more detailed information. We could not completely explain the heterogeneity, so we compared the fixed‐effect (RR 0.43, 95% CI 0.35 to 0.53) and the random‐effects meta‐analysis (RR 0.51, 95% CI 0.37 to 0.71), finding no major differences in direction or magnitude. Also, we found no funnel plot asymmetry (see the assessment of publication bias section below for more detailed information). For those reasons, we presented the combined results of the random‐effects model.

2.2. Analysis.

Comparison 2: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of drug, Outcome 2: Major VTE

3.2. Analysis.

Comparison 3: Direct factor Xa inhibitors versus LMWH: subgroup analysis by duration of prophylaxis, Outcome 2: Major VTE

4.2. Analysis.

Comparison 4: Direct factor Xa inhibitors versus LMWH: subgroup analysis by timing of start of prophylaxis (comparator), Outcome 2: Major VTE

5.2. Analysis.

Comparison 5: Direct factor Xa inhibitors versus LMWH: subgroup analysis by frequency of administration, Outcome 2: Major VTE

6.2. Analysis.

Comparison 6: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of surgery, Outcome 2: Major VTE

7.2. Analysis.

Comparison 7: Direct factor Xa inhibitors by type versus LMWH: subgroup analysis by daily all approved or low dose groups, Outcome 2: Major VTE

Direct factor Xa inhibitors may reduce major VTE events compared to LMWHs, but the evidence is very uncertain (RR 0.51, 95% CI 0.37 to 0.71; risk difference: 12 fewer major VTEs per 1000 participants; 95% CI 16 fewer to 7 fewer; I² = 48%; 28 studies, 24,594 participants; very low‐certainty evidence; Analysis 1.2).

Symptomatic VTE

Thirty‐three studies reported symptomatic VTEs for direct factor Xa inhibitors compared with LMWHs. Of these, 25 studies (29,981 participants) reported symptomatic VTE events (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Bai 2021; Chen 2016; EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Kunal 2021; NCT01205932; NCT01206972; ODIXa‐HIP‐OD 2006; ODIXa‐HIP 2007; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; Weitz 2010; Yun‐Fei 2018; Zou 2014), and eight studies (1689 participants) reported zero events for this outcome (Agnelli 2007; DARINA 2021; Karampinas 2019; Özler 2015; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Xie 2017).

Direct factor Xa inhibitors may reduce symptomatic VTE compared to LMWH (RR 0.64, 95% CI 0.50 to 0.83; risk difference: 3 fewer symptomatic VTEs per 1000; 95% CI 5 fewer to 2 fewer; I² = 0%; 33 studies, 31,670 participants; low‐certainty evidence; Analysis 1.3).

Major bleeding

Thirty‐six studies reported major bleeding events for direct factor Xa inhibitors compared with LMWHs (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Bai 2021; Cohen 2013; DARINA 2021; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Karampinas 2019; Khalafallah 2018; Kim 2016; Kunal 2021; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Özler 2015; Rahman 2020; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Weitz 2010; Xie 2017; Zhou 2023).

In the meta‐analysis with all studies pooled, direct factor Xa inhibitors may make little or no difference to major bleeding compared to LMWHs, but the evidence is very uncertain (RR 1.05, 95% CI 0.86 to 1.30; I² = 15%; 36 studies, 39,778 participants; very low certainty‐evidence; Analysis 1.4).

We observed a statistically significant difference (P value < 0.10) between subgroups based on drug type for this outcome, when we distinguished between studies that used rivaroxaban as the experimental intervention and those that used other direct factor Xa inhibitors. We could not perform individual comparisons for all experimental drug types due to the insufficient and unevenly distributed number of participants and studies across the different drug types. After a detailed subgroup analysis (see the type of drug and daily dose drug analyses in the subgroup analysis section below for more information: Analysis 2.4; Analysis 2.5; Analysis 2.6; Analysis 7.4), we decided to combine and present major bleeding events for studies comparing rivaroxaban with LMWHs (Analysis 1.5), and, separately, studies comparing other direct factor Xa inhibitors (not rivaroxaban) with LMWHs (Analysis 1.6).

2.4. Analysis.

Comparison 2: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of drug, Outcome 4: Major bleeding

2.5. Analysis.

Comparison 2: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of drug, Outcome 5: Major bleeding: rivaroxaban versus LMWH

2.6. Analysis.

Comparison 2: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of drug, Outcome 6: Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH

7.4. Analysis.

Comparison 7: Direct factor Xa inhibitors by type versus LMWH: subgroup analysis by daily all approved or low dose groups, Outcome 4: Major bleeding

Seventeen studies compared rivaroxaban to LMWHs, with 13 studies reporting major bleeding events (Bai 2021; DARINA 2021; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; Özler 2015; Rahman 2020; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; Zhou 2023), and four studies (1739 participants) reporting zero events for this outcome (Karampinas 2019; Khalafallah 2018; Kim 2016, Xie 2017). Participants given rivaroxaban may have a higher risk of major bleeding events than those given LMWHs (RR 1.94, 95% CI 1.26 to 2.98; risk difference: 3 more major bleeding events per 1000 participants, 95% CI 1 more to 7 more; I² = 0%; 17 studies, 17,630 participants; low‐certainty evidence; Analysis 1.5).

Nineteen studies compared other direct factor Xa inhibitors to LMWHs, with 16 studies reporting major bleeding events (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Cohen 2013; Fuji 2014‐TKA; Kunal 2021; ONYX‐2 2010; ONYX‐3 2014; Raskob 2010; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Weitz 2010), and three studies (823 participants) reporting zero events for this outcome (FOXTROT 2020; Fuji 2014‐THA; ONYX‐1 2007). Participants given direct factor Xa inhibitors (other than rivaroxaban) may have a lower risk of major bleeding events than those given LMWHs, but the evidence is very uncertain (RR 0.80, 95% CI 0.63 to 1.02; risk difference: 3 fewer major bleeding events per 1000 participants, 95% CI 5 fewer to 0 fewer; I² = 0%; 19 studies, 22,148 participants; very low‐certainty evidence; Analysis 1.6).

For the purposes of this review, major bleeding events included any clinically overt bleeding accompanied by one or more of the following: (1) haemoglobin decreased by 2 g/dL or more within 24 hours; (2) transfusion of two or more units of packed red blood cells (RBC); (3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); (4) bleeding into the operated joint, requiring an additional operation or intervention; or (5) fatal bleeding. For the first two criteria, some studies adopted alternative definitions, including (1) a bleeding index score of 2.0 or higher (bleeding index = units of red blood cells transfused plus pre‐bleed haemoglobin minus post‐bleed haemoglobin) (ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014); (2) transfusion of four or more units (approximately 800 mL) of blood (Özler 2015; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Weitz 2010); and (3) postoperative blood transfusion (Rahman 2020). Some studies adopted additional criteria, including (1) intramuscular bleeding with compartment syndrome (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010); and (2) the need to discontinue study medication (APROPOS 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐3 2014; Raskob 2010; Xie 2017). Four studies did not report the definition used for major bleeding events (Fuji 2014‐THA; Fuji 2014‐TKA; NCT01205932; NCT01206972), and one study restricted major bleeding events to bleeding at a critical site (Kim 2016).

Serious hepatic adverse events

Only two studies reported data on serious liver‐related adverse events, consisting of one single death due to hepatitis in the apixaban group (ADVANCE‐2 2010); the other study had no events in any group (Rahman 2020). Direct factor Xa inhibitors may make little to no difference in serious hepatic adverse events compared to LMWHs, but the evidence is very uncertain (RR 3.01, 95% CI 0.12 to 73.93; risk difference: 0 fewer serious hepatic adverse events per 1000 participants, 95% CI 0 fewer to 0 fewer; 2 studies, 3169 participants; very low‐certainty evidence; Analysis 1.7). We did not evaluate heterogeneity for this outcome because there was only one study with events.

Serious non‐hepatic adverse events

We included 15 studies reporting serious non‐hepatic adverse events in the pooled meta‐analysis (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; APROPOS 2007; Fuji 2014‐THA; Fuji 2014‐TKA; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Weitz 2010). Although the RECORD4 2009 study reported this outcome, as noted above (Other potential sources of bias), we removed it from the analysis for risk of bias. Direct factor Xa inhibitors may reduce the risk of serious non‐hepatic adverse events slightly compared to LMWHs (RR 0.89, 95% CI 0.81 to 0.97; risk difference: 9 fewer serious non‐hepatic adverse events per 1000 participants, 95% CI 14 fewer to 3 fewer; I² = 18%; 15 studies, 26,246 participants; low‐certainty evidence; Analysis 1.8).

We included all adverse events reported as "serious" in the primary studies in this meta‐analysis. Notably, none of the included studies defined "serious adverse events" or listed those adverse events judged to be serious. Eight additional studies also reported data on serious adverse events; however, they could not be included in the meta‐analysis because the reported data were insufficient (Cohen 2013; EXPERT 2009; NCT01205932; NCT01206972; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐KNEE 2005; ONYX‐1 2007).

Secondary outcomes

The main results are summarised in Analysis 1.9, Analysis 1.10, Analysis 1.11, Analysis 1.12, Analysis 1.13, Analysis 1.14, and Table 2. We carried out subgroup analyses to investigate if the type of drug, duration of prophylaxis, time of initiation of prophylaxis, frequency of administration, type of surgery, or drug dose had any effect on the outcomes, and we report these separately below.

1.9. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 9: Fatal VTE

1.10. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 10: Asymptomatic distal DVT

1.11. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 11: Fatal bleeding

1.12. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 12: Liver enzymes elevation

1.13. Analysis.

Comparison 1: Direct factor Xa inhibitors versus LMWHs, Outcome 13: Minor adverse events

Fatal VTE

We included 33 studies reporting on the outcome fatal VTE for direct factor Xa inhibitors compared with LMWHs. Ten studies (18,372 participants) reported fatal VTE events (ADVANCE‐2 2010; ADVANCE‐3 2010; APROPOS 2007; Fuji 2014‐TKA; ODIXa‐HIP 2007; ODIXa‐KNEE 2005; RECORD1 2008; RECORD2 2008; RECORD4 2009; Tang 2017), and 23 studies (8811 participants) reported zero events for this outcome (Chen 2016; Cohen 2013; DARINA 2021; EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Jiang 2019; Kim 2016; Kunal 2021; NCT01205932; NCT01206972; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Özler 2015; Rahman 2020; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Weitz 2010; Xie 2017). Compared to LMWHs, direct factor Xa inhibitors may make little or no difference to the risk of death from venous thromboembolic events, but the evidence is very uncertain (RR 1.19, 95% CI 0.51 to 2.79; risk difference: 0 fewer fatal VTEs per 1000 participants, 95% CI 0 fewer to 1 more; I² = 0%; 33 studies, 27,183 participants; very low‐certainty evidence; Analysis 1.9).

Asymptomatic distal DVT

Fourteen studies reported on asymptomatic distal DVTs for direct factor Xa inhibitors compared with LMWHs. Twelve studies (4602 participants) reported asymptomatic distal DVT events (EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Hu 2015; Hui 2013; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Xie 2017), and two studies (216 participants) reported zero events for this outcome (Kunal 2021; Özler 2015).

There was moderate heterogeneity (I² = 29%), so we inspected the forest plot for the direction of effect of individual studies. The confidence intervals generally overlap, and the larger studies tend to show an effect favouring direct factor Xa inhibitors. We performed subgroup analyses to see if a clinical explanation could explain the heterogeneity (Analysis 2.9; Analysis 3.8; Analysis 4.9; Analysis 5.9; Analysis 6.9; Analysis 7.7), but we found no subgroup differences. We then compared the fixed‐effect (RR 0.65, 95% CI 0.52 to 0.81) and the random‐effects meta‐analysis (RR 0.64, 95% CI 0.46 to 0.89), and found no major differences in direction or magnitude. In addition, we found no funnel plot asymmetry (see the assessment of publication bias section below for more detailed information). For those reasons, we presented the combined results of the random‐effects model.

2.9. Analysis.

Comparison 2: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of drug, Outcome 9: Asymptomatic distal DVT

3.8. Analysis.

Comparison 3: Direct factor Xa inhibitors versus LMWH: subgroup analysis by duration of prophylaxis, Outcome 8: Asymptomatic distal DVT

4.9. Analysis.

Comparison 4: Direct factor Xa inhibitors versus LMWH: subgroup analysis by timing of start of prophylaxis (comparator), Outcome 9: Asymptomatic distal DVT

5.9. Analysis.

Comparison 5: Direct factor Xa inhibitors versus LMWH: subgroup analysis by frequency of administration, Outcome 9: Asymptomatic distal DVT

6.9. Analysis.

Comparison 6: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of surgery, Outcome 9: Asymptomatic distal DVT

7.7. Analysis.

Comparison 7: Direct factor Xa inhibitors by type versus LMWH: subgroup analysis by daily all approved or low dose groups, Outcome 7: Asymptomatic distal DVT

The overall pooled estimate across studies revealed that direct factor Xa inhibitors may reduce the risk of asymptomatic distal DVT events compared to LMWHs (RR 0.64, 95% CI 0.46 to 0.89; risk difference: 29 fewer asymptomatic distal DVTs per 1000 participants, 95% CI 43 fewer to 9 fewer; I² = 29%; 14 studies, 4818 participants; low‐certainty evidence; Analysis 1.10).

Fatal bleeding

Twenty‐six studies reported on fatal bleeding, with four studies (10,794 participants) reporting fatal bleeding events (ADVANCE‐2 2010; Fuji 2014‐TKA; RECORD1 2008; RECORD4 2009), and 22 studies (20,499 participants) reporting zero events for this outcome (ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Cohen 2013; EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Jiang 2019; Kim 2016; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; Rahman 2020; RECORD2 2008; RECORD3 2008; STARS J‐2 2014; STARS J‐4 2014; Weitz 2010; Xie 2017). Direct factor Xa inhibitors may make little or no difference to death from bleeding compared to LMWHs, but the evidence is very uncertain (RR 1.42, 95% CI 0.33 to 6.04; risk difference: 0 fewer fatal bleeding events per 1000 participants, 95% CI 0 fewer to 0 fewer; I² = 0%; 26 studies, 31,293 participants; very low‐certainty evidence; Analysis 1.11).

Liver enzymes elevation

Twenty‐one studies reported on liver enzymes elevation (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; EXPERT 2009; Fuji 2014‐THA; Fuji 2014‐TKA; Kim 2016; ODIXa‐HIP 2007; ODIXa‐HIP‐OD 2006; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Weitz 2010).

There was moderate heterogeneity (I² = 49%), so we inspected the forest plot for the direction of the effect of individual studies and the overlap of confidence intervals, and observed no important differences between them. We performed several subgroup analyses to explore if the heterogeneity could be explained by clinical differences (Analysis 2.10; Analysis 3.9; Analysis 4.10; Analysis 5.10; Analysis 6.10; Analysis 7.8). We found a statistically significant difference (P value < 0.10) only in the subgroup analysis for timing of the start of control drug prophylaxis (i.e. LMWHs started before versus after surgery) (Analysis 4.10). This did not fully explain the heterogeneity, so we compared the fixed‐effect (RR 0.60, 95% CI 0.52 to 0.70) and the random‐effects meta‐analysis (RR 0.58, 95% CI 0.46 to 0.73), and observed no major differences in direction or magnitude. However, we did detect funnel plot asymmetry (see the assessment of publication bias section below for more detailed information), so we performed a sensitivity analysis excluding small studies, and observed no major differences in direction or magnitude. For these reasons, we presented the combined results of the random‐effects model.

2.10. Analysis.

Comparison 2: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of drug, Outcome 10: Liver enzymes elevation

3.9. Analysis.

Comparison 3: Direct factor Xa inhibitors versus LMWH: subgroup analysis by duration of prophylaxis, Outcome 9: Liver enzymes elevation

4.10. Analysis.

Comparison 4: Direct factor Xa inhibitors versus LMWH: subgroup analysis by timing of start of prophylaxis (comparator), Outcome 10: Liver enzymes elevation

5.10. Analysis.

Comparison 5: Direct factor Xa inhibitors versus LMWH: subgroup analysis by frequency of administration, Outcome 10: Liver enzymes elevation

6.10. Analysis.

Comparison 6: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of surgery, Outcome 10: Liver enzymes elevation

7.8. Analysis.

Comparison 7: Direct factor Xa inhibitors by type versus LMWH: subgroup analysis by daily all approved or low dose groups, Outcome 8: Liver enzymes elevation

The overall pooled estimate for liver enzymes elevation showed that direct factor Xa inhibitors may limit liver enzymes elevation compared to LMWH, but the evidence is very uncertain (RR 0.58, 95% CI 0.46 to 0.73; risk difference: 12 fewer liver enzymes elevation events per 1000 participants, 95% CI 16 fewer to 8 fewer; I² = 49%; 21 studies, 31,408 participants; very low‐certainty evidence; Analysis 1.12).

For the purposes of this review, we defined liver enzymes elevation as an increase in ALT (alanine transaminase) levels above three times the upper limit of normal. We also included data from studies with different criteria for this outcome in the meta‐analysis: liver enzymes elevation defined as an increase in ALT and/or AST (aspartate transaminase) levels greater than three times the upper limit of normal (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Raskob 2010; STARS J‐V 2015); and as total bilirubin levels three times higher than at baseline plus ALT or AST levels five times higher than at baseline (Kim 2016). Two additional studies also reported data on liver enzymes elevation; however, we could not include these in the meta‐analysis because the reported data were insufficient (NCT01205932; NCT01206972).

Minor adverse events

Only four studies reported minor adverse events for direct factor Xa inhibitors compared with LMWHs (DARINA 2021; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA). Direct factor Xa inhibitors may make little to no difference to minor adverse events compared to LMWHs (RR 0.91, 95% CI 0.83 to 1.00; risk difference: 57 fewer minor adverse events per 1000 participants, 95% CI 108 fewer to 0 fewer; I² = 0%; 4 studies, 1011 participants; low‐certainty evidence; Analysis 1.13).

We included all adverse events reported as "minor" in primary studies in this meta‐analysis. The most frequently reported minor adverse events included constipation, myalgia, and insomnia.

Volume of blood loss

Fourteen studies reported the volume of blood loss for direct factor Xa inhibitors compared with LMWHs (Hu 2015; Hui 2013; Jiang 2019; Khalafallah 2018; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; RECORD4 2009; Tang 2017; Wasko 2015; Weitz 2010; Wing 2020; Zhou 2023). There was very high heterogeneity (I² = 96%). We explored if a clinical explanation could be the cause of the heterogeneity by performing several subgroup analyses (Analysis 2.11; Analysis 3.10; Analysis 4.11; Analysis 5.11; Analysis 6.11; Analysis 7.9). We found statistically significant differences between subgroups (P value < 0.10) in four subgroups (type of drug, frequency of administration, type of surgery, and daily drug dose; see the subgroup analysis section below for more detailed information). However, clinical differences did not explain the very high heterogeneity. We also found no funnel plot asymmetry (see the assessment of publication bias section below for more detailed information). Due to the very high unexplained heterogeneity, we did not perform a meta‐analysis for this outcome.

2.11. Analysis.

Comparison 2: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of drug, Outcome 11: Volume of blood loss

3.10. Analysis.

Comparison 3: Direct factor Xa inhibitors versus LMWH: subgroup analysis by duration of prophylaxis, Outcome 10: Volume of blood loss

4.11. Analysis.

Comparison 4: Direct factor Xa inhibitors versus LMWH: subgroup analysis by timing of start of prophylaxis (comparator), Outcome 11: Volume of blood loss

5.11. Analysis.

Comparison 5: Direct factor Xa inhibitors versus LMWH: subgroup analysis by frequency of administration, Outcome 11: Volume of blood loss

6.11. Analysis.

Comparison 6: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of surgery, Outcome 11: Volume of blood loss

7.9. Analysis.

Comparison 7: Direct factor Xa inhibitors by type versus LMWH: subgroup analysis by daily all approved or low dose groups, Outcome 9: Volume of blood loss

The subgroup and sensitivity analyses performed are reported in this review (see below). The volume of blood loss consisted of the volume of post‐operative drainage in most studies, but other definitions were also employed (Hui 2013; Wasko 2015), which could be one of the sources of the heterogeneity. Three additional studies also reported data on volume of blood loss (RECORD1 2008; RECORD2 2008; RECORD3 2008). However, we could not include these in the meta‐analysis because the studies did not report standard deviations of the mean.

Subgroup analyses

We performed subgroup analyses in this review for two purposes: (1) to investigate sources of heterogeneity in the meta‐analysis; and (2) to provide more accurate estimates of treatment effects than the primary pooled analysis effects (main analyses Analysis 1.1 to Analysis 1.14) in pre‐defined clinically relevant subgroups of participants.

In the main meta‐analyses, we found moderate heterogeneity in one primary benefit outcome, major VTE (I² = 48%: Analysis 1.2), and in two secondary benefit outcomes: asymptomatic distal VTE (I² = 29%; Analysis 1.10) and liver enzymes elevation (I² = 49%; Analysis 1.12). We found very high statistical heterogeneity only in the secondary harm outcome volume of blood loss (I² = 96%; Analysis 1.14). Bearing in mind that not all trials are designed to measure adverse events, we interpreted secondary harm outcomes with caution.

We explored the reasons for heterogeneity in these four outcomes in the subgroups considered relevant for this review: type of drug; duration of the prophylactic anticoagulation, classified as 'short' and 'extended'; time of initiation of prophylactic anticoagulation (before surgery versus after surgery, applicable only for the comparator drug, LMWHs); frequency of administration (once versus twice daily); type of surgery, which included total hip replacement (THR), total knee replacement (TKR), both THR and TKR, and hip fracture surgery (HFS); and restriction of analysis to all approved doses or, if there is no approved dose, one of three daily dose groups (low, intermediate, and high, with limits established based on the lowest and highest daily dose studied in included trials).

Due to the design of the included studies, the data in all of these subgroups are independent. This means that the study participants contribute to just one of the subgroups in the forest plot, except for frequency of administration and the daily dose subgroups, because some studies (Phase II) compared different doses or frequencies of a specific direct factor Xa inhibitor with the same comparator (LMWH).

In each subgroup, we did not include studies with zero events in the intervention and control groups. We excluded comparisons with insufficient studies (fewer than 10 studies) amongst all the subgroups from the subgroup analysis.

Regarding the exploration of a more accurate estimate of effect in each pre‐defined subgroup, we explored all the subgroups by visual inspection. In those where we detected an interaction (meaning that significant statistical differences between subgroups existed (P value < 0.10 from the Chi² test for subgroup differences)), we analysed the covariate distribution effect (taking into account the number of trials and participants contributing to each subgroup). In the analysis of subgroups, we examined whether the covariate distribution was proportional among the subgroups, whether the total number of subgroups included ten or more trials, whether a subgroup reached the optimal information size (OIS), or whether the number of participants was large enough to find a precise estimate of the effect. We also explored the heterogeneity between results from the trials within each subgroup.

The descriptions of the subgroups presented in this review are consistent with recommendations in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2023) and Richardson 2019.

Type of drug

The categories within this subgroup were: apixaban, betrixaban, darexaban, edoxaban, erixaban, letaxaban, LY517717, and rivaroxaban. When a category (or drug type) is not reported, it means there were no data to include it. For all the analysed outcomes, the data in the categories are independent, meaning that the study participants contributed to no more than one of the drug types in the forest plot.

We evaluated three outcomes (major VTE, major bleeding, and volume of blood loss) because the test for subgroup differences suggested that there was a statistically significant subgroup effect (P < 0.10) (Analysis 2.2; Analysis 2.4; Analysis 2.11). In the remaining outcomes, the test for subgroup differences indicated that there was no statistically significant subgroup effect (P > 0.10), suggesting that the type of drug does not modify the effect of direct factor X inhibitors in comparison to LMWHs (Analysis 2.1; Analysis 2.3; Analysis 2.7; Analysis 2.8; Analysis 2.9; Analysis 2.10). Due to this, we present in all of these outcomes the combined effect of all direct factor Xa inhibitors instead of each drug separately.

2.1. Analysis.

Comparison 2: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of drug, Outcome 1: All‐cause mortality

2.3. Analysis.

Comparison 2: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of drug, Outcome 3: Symptomatic VTE

2.7. Analysis.

Comparison 2: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of drug, Outcome 7: Serious non‐hepatic adverse events

2.8. Analysis.

Comparison 2: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of drug, Outcome 8: Fatal VTE

For major VTE, the test for subgroup differences suggests that there is a statistically significant subgroup effect (P = 0.01), meaning that the type of drug statistically significantly modifies the effect of direct factor X inhibitors in comparison to LMWHs for the outcome major VTE (Analysis 2.2). The treatment effect favours direct factor X inhibitors over LMWHs for apixaban, edoxaban, and rivaroxaban, while there is no difference for darexaban and LY517717; therefore, the subgroup effect is qualitative. For apixaban, a sufficient number of trials (6) and participants (7910; optimal information size (OIS) 6950) were included. There is no heterogeneity between results from the trials within this drug type (I² = 0%) that requires further exploration. However, in edoxaban and LY517717, there were a smaller number of trials (just one study per drug), and fewer participants (< 800 for each drug) contributed data to them than to the other drug types, meaning that the analysis may not be able to detect subgroup differences. Darexaban probably has a sufficient number of trials (5). However, an insufficient number of participants (2992; OIS 111,760) were included, meaning that the analysis may not be able to detect subgroup differences, despite the fact that there is no heterogeneity between the results from the trials within this drug (I² = 0%). Regarding rivaroxaban, a large number of trials (11) and participants (11,929; OIS 2780) were included; however, there is substantial unexplained heterogeneity between the trials exploring this drug (I² = 63%). A visual inspection of the forest plot confirms that heterogeneity is higher within this drug than across all trials (I² = 48%). In addition, the confidence intervals generally overlap, and the larger studies tend to show an effect estimate of the beneficial effect of rivaroxaban over LMWH. Therefore, the validity of the treatment effect estimate for this drug is uncertain, as individual trial results are inconsistent.

The covariate distribution is a matter of concern for this subgroup analysis because the number of participants and trials is not evenly distributed across the different drug types, which could affect the reliability of the subgroup analysis to produce valid results regarding which drug has better efficacy due to fewer major VTE events (Analysis 2.2). Also, the direct factor Xa inhibitors that differ from the main analysis of all drug combined effects (darexaban and LY517717) for major VTE events present an insufficient number of participants to consider the difference important.

The heterogeneity in the overall pooled estimate for major VTE across studies (Analysis 1.2) is partially explained by the type of drug analysed.

For those reasons, for major VTE events, we present the combined effect of all direct factor Xa inhibitors instead of each drug separately in the main analysis.

For major bleeding, the test for subgroup differences suggests that there is a statistically significant subgroup effect (P = 0.02), meaning that the type of direct factor Xa inhibitor statistically significantly modifies the effect of direct factor X inhibitors in comparison to LMWHs for the outcome of major bleeding (Analysis 2.4). The treatment effect does not show any difference between direct factor X inhibitors over LMWHs for apixaban, darexaban, edoxaban, erixaban, letaxaban, and LY517717. Only in rivaroxaban are there more bleeding events than in LMWH; therefore, the subgroup effect is qualitative.

For rivaroxaban (the only drug that presents a different combined effect than the main analysis, which includes all direct factor Xa inhibitors together with all doses), a sufficient number of trials (13) and participants (15,891; OIS 3462) were included. There is no heterogeneity between results from the trials assessing this drug (I² = 0%) that requires further exploration. Also, the 95% confidence interval of the overall pooled estimate for this drug is precise enough because it excludes no effect (CI around RR excludes 1) and important harm (relative risk increase (RRI) > 25%) (Analysis 2.4). The evaluation of whether these findings are affected by the dose in the direct factor Xa inhibitors is described in the daily drug dose subgroups (see subgroup analysis by drug type: daily approved dose groups, below, for a detailed explanation: Analysis 7.4).

Regarding the direct factor X inhibitor drugs with no difference in major bleeding events compared to LMWH, only apixaban probably has a sufficient number of trials (5). However, an insufficient number of participants (12,687; OIS 149,356) were included, meaning that the analysis may not be able to detect subgroup differences. Also, there is unexplained heterogeneity between the trials within this drug (I² = 30%). A visual inspection of the forest plot confirms that heterogeneity is higher within the apixaban studies than across all trials (I² = 15%); also, the confidence intervals generally overlap, and the larger studies tend to show an effect estimate of the beneficial effect of apixaban over LMWH. Therefore, the validity of the treatment effect estimate for this drug is uncertain, as individual trial results are inconsistent.

In summary, rivaroxaban is the only drug type with more major bleeding events in comparison with LMWHs: RR 1.94 (95% CI 1.26 to 2.98), with no heterogeneity (I² = 0%). However, the covariate distribution is a matter of concern for this subgroup analysis due to the fact that the number of participants and trials is not evenly distributed across the different drug types, which could affect the reliability of the subgroup analysis to produce valid results regarding which of the other drugs besides rivaroxaban has better safety due to fewer major bleeding events (Analysis 2.4).

The heterogeneity in the overall pooled estimate for major bleeding across studies (Analysis 1.4) is explained by the type of drug analysed.

For those reasons, for major bleeding events, we present the results of the combined effect of rivaroxaban separately from the combined effect of the other direct factor Xa inhibitors, instead of all drugs combined (Analysis 1.5; Analysis 1.6).

For volume of blood loss, the test for subgroup differences suggests that there is a statistically significant subgroup effect (P < 0.00001), meaning that the type of drug statistically significantly modifies the effect of direct factor X inhibitors in comparison to LMWH for the outcome volume of blood loss. The subgroup effect is qualitative; however, the drugs that differ with respect to the main analysis have an insufficient number of participants. Also, the number of participants and trials is not evenly distributed across the different drug types, which affects the reliability of the subgroup analysis to produce valid results (Analysis 2.11).

There were insufficient studies for reliable subgroup analyses, and they did not explain the very high heterogeneity in the overall pooled estimate for volume of blood loss across studies. (Analysis 1.14).

Duration of prophylaxis

The categories within this subgroup were the following: short prophylaxis (≤ 14 days); and extended prophylaxis (> 14 days). When a category is not reported, it means there were no data to include it. For all the analysed outcomes, the data in the subgroups are independent, meaning that the study participants contribute to no more than one of the categories in the forest plot.

In all the analysed outcomes, the test for subgroup differences indicates that there is no statistically significant subgroup effect (P > 0.10), suggesting that the duration of prophylaxis does not modify the effect of direct factor X inhibitors in comparison to LMWH. (Analysis 3.1; Analysis 3.2; Analysis 3.3; Analysis 3.4; Analysis 3.5; Analysis 3.6; Analysis 3.7; Analysis 3.8; Analysis 3.9; Analysis 3.10).

3.1. Analysis.

Comparison 3: Direct factor Xa inhibitors versus LMWH: subgroup analysis by duration of prophylaxis, Outcome 1: All‐cause mortality

3.3. Analysis.

Comparison 3: Direct factor Xa inhibitors versus LMWH: subgroup analysis by duration of prophylaxis, Outcome 3: Symptomatic VTE

3.4. Analysis.

Comparison 3: Direct factor Xa inhibitors versus LMWH: subgroup analysis by duration of prophylaxis, Outcome 4: Major bleeding

3.5. Analysis.

Comparison 3: Direct factor Xa inhibitors versus LMWH: subgroup analysis by duration of prophylaxis, Outcome 5: Major bleeding: rivaroxaban versus LMWH

3.6. Analysis.

Comparison 3: Direct factor Xa inhibitors versus LMWH: subgroup analysis by duration of prophylaxis, Outcome 6: Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH

3.7. Analysis.

Comparison 3: Direct factor Xa inhibitors versus LMWH: subgroup analysis by duration of prophylaxis, Outcome 7: Serious non‐hepatic adverse events

Also, for the outcomes with heterogeneity in the primary analysis (major VTE, asymptomatic distal DVT, liver enzymes elevation, and volume of blood loss), the duration of prophylaxis categories did not explain the heterogeneity (Analysis 3.2; Analysis 3.8; Analysis 3.9; Analysis 3.10).

Timing of start of prophylaxis (comparator)

The categories within this subgroup were the following: start of comparator before surgery; and start of comparator after surgery. When a category is not reported, it means there were no data to include it. For all the analysed outcomes, the data in the categories are independent, meaning that the study participants contribute to no more than one of the categories in the forest plot.

In two outcomes (symptomatic VTE and liver enzymes elevation), the test for subgroup differences suggests that there is a statistically significant subgroup effect (P < 0.10). For symptomatic VTE, the subgroup effect is qualitative and for liver enzymes elevation, the subgroup effect is quantitative. However, for both outcomes, the category that differs with respect to the main analysis has an insufficient number of participants. Also, the covariate distribution is a matter of concern, which could affect the reliability of the subgroup analysis (Analysis 4.3; Analysis 4.10).

4.3. Analysis.

Comparison 4: Direct factor Xa inhibitors versus LMWH: subgroup analysis by timing of start of prophylaxis (comparator), Outcome 3: Symptomatic VTE

In the remaining outcomes analysed, the test for subgroup differences indicates that there is no statistically significant subgroup effect (P > 0.10), suggesting that the timing of the start of prophylaxis in the LMWH group does not modify the effect of direct factor X inhibitors in comparison to LMWHs (Analysis 4.1; Analysis 4.2; Analysis 4.4; Analysis 4.5; Analysis 4.6; Analysis 4.7; Analysis 4.8; Analysis 4.9; Analysis 4.11).

4.1. Analysis.

Comparison 4: Direct factor Xa inhibitors versus LMWH: subgroup analysis by timing of start of prophylaxis (comparator), Outcome 1: All‐cause mortality

4.4. Analysis.

Comparison 4: Direct factor Xa inhibitors versus LMWH: subgroup analysis by timing of start of prophylaxis (comparator), Outcome 4: Major bleeding

4.5. Analysis.

Comparison 4: Direct factor Xa inhibitors versus LMWH: subgroup analysis by timing of start of prophylaxis (comparator), Outcome 5: Major bleeding: rivaroxaban versus LMWH

4.6. Analysis.

Comparison 4: Direct factor Xa inhibitors versus LMWH: subgroup analysis by timing of start of prophylaxis (comparator), Outcome 6: Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH

4.7. Analysis.

Comparison 4: Direct factor Xa inhibitors versus LMWH: subgroup analysis by timing of start of prophylaxis (comparator), Outcome 7: Serious non‐hepatic adverse events

4.8. Analysis.

Comparison 4: Direct factor Xa inhibitors versus LMWH: subgroup analysis by timing of start of prophylaxis (comparator), Outcome 8: Fatal VTE

Also, for the outcomes with heterogeneity in the primary analysis (major VTE, asymptomatic distal DVT, liver enzymes elevation, and volume of blood loss), the timing of the start of prophylaxis in the LMWH categories (before or after surgery) did not explain the heterogeneity (Analysis 4.2; Analysis 4.9; Analysis 4.10; Analysis 4.11).

Frequency of administration

The categories within the subgroup were the following: once daily administration; and twice daily administration. When a category is not reported, it means there were no data to include it.

In order to avoid dependent data in the categories within a subgroup (due to some phase II studies that use the same comparator for different frequencies of administration doses for the intervention), we presented the analysis by evenly splitting the comparator events and the participants for those studies in each category. There was a concern that this could modify the combined effect of the results in the categories. Thus, we performed a sensitivity analysis, and found no difference in the category combined effect when we compared the split and un‐split data for each category (data not presented).

In three outcomes (symptomatic VTE, serious non‐hepatic adverse events, and volume of blood loss), the test for subgroup differences suggests that there is a statistically significant subgroup effect (P < 0.10). For the three outcomes, the subgroup effect is qualitative. However, the category that differs with respect to the main analysis has an insufficient number of participants. Also, the covariate distribution is a matter of concern, which could affect the reliability of the subgroup analysis (Analysis 5.3; Analysis 5.7; Analysis 5.11).

5.3. Analysis.

Comparison 5: Direct factor Xa inhibitors versus LMWH: subgroup analysis by frequency of administration, Outcome 3: Symptomatic VTE

5.7. Analysis.

Comparison 5: Direct factor Xa inhibitors versus LMWH: subgroup analysis by frequency of administration, Outcome 7: Serious non‐hepatic adverse events

In the remaining analysed outcomes, the test for subgroup differences indicates that there is no statistically significant subgroup effect (P > 0.10), suggesting that the frequency of administration does not modify the effect of direct factor X inhibitors in comparison to LMWH (Analysis 5.1; Analysis 5.2; Analysis 5.4; Analysis 5.5; Analysis 5.6 ; Analysis 5.8; Analysis 5.9; Analysis 5.10).

5.1. Analysis.

Comparison 5: Direct factor Xa inhibitors versus LMWH: subgroup analysis by frequency of administration, Outcome 1: All‐cause mortality

5.4. Analysis.

Comparison 5: Direct factor Xa inhibitors versus LMWH: subgroup analysis by frequency of administration, Outcome 4: Major bleeding

5.5. Analysis.

Comparison 5: Direct factor Xa inhibitors versus LMWH: subgroup analysis by frequency of administration, Outcome 5: Major bleeding: rivaroxaban versus LMWH

5.6. Analysis.

Comparison 5: Direct factor Xa inhibitors versus LMWH: subgroup analysis by frequency of administration, Outcome 6: Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH

5.8. Analysis.

Comparison 5: Direct factor Xa inhibitors versus LMWH: subgroup analysis by frequency of administration, Outcome 8: Fatal VTE

Also, for the outcomes with heterogeneity in the primary analysis (major VTE, asymptomatic distal DVT, liver enzymes elevation, and volume of blood loss), the frequency of administration categories (once or twice daily) were unable to explain the heterogeneity.

Type of surgery

The categories within this subgroup were the following: total hip or knee replacement surgery; total hip replacement surgery only; total knee replacement surgery only; and hip fracture surgery only. When a category is not reported, it means there were no data to include it. For all the analysed outcomes, the data in the categories are independent, meaning that the study participants contribute to no more than one of the categories in the forest plot.

In three outcomes (symptomatic VTE, major bleeding, and volume of blood loss), the test for subgroup differences suggests that there is a statistically significant subgroup effect (P < 0.10). For major bleeding, the subgroup effect is quantitative, and for symptomatic VTE and volume of blood loss, the subgroup effect is qualitative. However, for the three outcomes, the categories that differ with respect to the main analysis have an insufficient number of participants. Also, the covariate distribution is a matter of concern, which could affect the reliability of the subgroup analysis (Analysis 6.3; Analysis 6.4; Analysis 6.11).

6.3. Analysis.

Comparison 6: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of surgery, Outcome 3: Symptomatic VTE

6.4. Analysis.

Comparison 6: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of surgery, Outcome 4: Major bleeding

In the remaining outcomes analysed, the test for subgroup differences indicates that there is no statistically significant subgroup effect (P > 0.10), suggesting that the type of surgery does not modify the effect of direct factor X inhibitors in comparison to LMWHs (Analysis 6.1; Analysis 6.2; Analysis 6.5; Analysis 6.6; Analysis 6.7; Analysis 6.8; Analysis 6.9; Analysis 6.10).

6.1. Analysis.

Comparison 6: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of surgery, Outcome 1: All‐cause mortality

6.5. Analysis.

Comparison 6: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of surgery, Outcome 5: Major bleeding: rivaroxaban versus LMWH

6.6. Analysis.

Comparison 6: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of surgery, Outcome 6: Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH

6.7. Analysis.

Comparison 6: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of surgery, Outcome 7: Serious non‐hepatic adverse events

6.8. Analysis.

Comparison 6: Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of surgery, Outcome 8: Fatal VTE

Also, for the outcomes with heterogeneity in the primary analysis (major VTE, asymptomatic distal DVT, liver enzymes elevation, and volume of blood loss), the type of surgery categories were unable to explain the heterogeneity (Analysis 6.4; Analysis 6.9; Analysis 6.10; Analysis 6.11).

Daily dose groups

There are four direct factor Xa inhibitors that are still on the market and have approved doses for one or several thrombotic conditions:

• apixaban: 2.5 to 20 mg/day;

• betrixaban: 40 to 160 mg/day;

• edoxaban: 30 to 60 mg/day;

• rivaroxaban: 5 to 30 mg/day.

By visual inspection, in none of the outcomes did the pooled estimate of these drug categories by drug type present a more accurate result than the original analysis, which included all drug doses by drug type.

For the remaining direct factor Xa inhibitors with no approved dose, we established three daily dose groups based on the maximum daily dose of each intervention drug throughout the included studies. Daily dose groups were established as follows:

• darexaban: ≤ 40 mg/day (low dose), 40.1–80 mg/day (medium dose), and > 80 mg/day (high dose);

• eribaxaban: ≤ 2.5 mg/day (low dose), 2.6–6.6 mg/day (medium dose), and > 6.6 mg/day (high dose);

• letaxaban: ≤ 40.0 mg/day (low dose), 40.1–106.6 mg/day (medium dose), and > 106.6 mg/day (high dose);

• LY517717: ≤ 50 mg/day (low dose), 50.1–100 mg/day (medium dose), and > 100 mg/day (high dose).

The categories within the subgroup were the following: low dose, medium dose, and high dose. When a category is not reported, it means there were no data to include it. If a comparison were performed including low, medium, and high‐dose categories in one subgroup analysis, the data would not be independent for all the analysed outcomes, meaning that the study participants contributed to more than one of the subgroup categories in the forest plot (in the comparator group: LMWH). This prevents us from making inferences about this subgroup. For this reason and according to the protocol, we intended to evaluate just one of the three categories in a subgroup analysis. We chose to evaluate the low‐dose category because it has more studies and weight in all the evaluated outcomes (the medium and high doses have fewer than 10 studies in almost all the outcomes). By visual inspection, in none of the outcomes did the pooled estimate of this low‐dose category by drug type present a more accurate result than the original analysis, which included all drug doses.

In three outcomes (major VTE, major bleeding, and volume of blood loss), the test for subgroup differences suggests that there is a statistically significant subgroup effect (P < 0.10) (Analysis 7.2; Analysis 7.4; Analysis 7.9).

In the remaining outcomes analysed (all‐cause mortality, serious non‐hepatic adverse events, fatal VTE, asymptomatic distal DVT, and liver enzymes elevation), the test for subgroup differences indicates that there is no statistically significant subgroup effect (P > 0.10), suggesting that the low drug dose does not modify the effect of direct factor X inhibitors in comparison to LMWHs (Analysis 7.1; Analysis 7.3; Analysis 7.5; Analysis 7.6; Analysis 7.7; Analysis 7.8). Also, the same effect as in the main analysis with all the combined doses was observed for these outcomes. For those reasons, for these outcomes, we present all the combined doses for the combined effect of all direct factor Xa inhibitors instead of each drug dose separately in the main analysis.

7.1. Analysis.

Comparison 7: Direct factor Xa inhibitors by type versus LMWH: subgroup analysis by daily all approved or low dose groups, Outcome 1: All‐cause mortality

7.3. Analysis.

Comparison 7: Direct factor Xa inhibitors by type versus LMWH: subgroup analysis by daily all approved or low dose groups, Outcome 3: Symptomatic VTE

7.5. Analysis.

Comparison 7: Direct factor Xa inhibitors by type versus LMWH: subgroup analysis by daily all approved or low dose groups, Outcome 5: Serious non‐hepatic adverse events

7.6. Analysis.

Comparison 7: Direct factor Xa inhibitors by type versus LMWH: subgroup analysis by daily all approved or low dose groups, Outcome 6: Fatal VTE

For major VTE, the combined treatment effect shows fewer major VTE events in the direct factor X inhibitors over LMWHs, as happens in the main analysis (Analysis 1.2). As in the analysis of drug type with all the combined doses (Analysis 2.2), the treatment effect favours direct factor X inhibitors over LMWHs for apixaban, edoxaban, and rivaroxaban, while there is no difference for darexaban and LY517717; therefore, the subgroup effect is qualitative. For this outcome, most of the categories that agree with the main analysis have a sufficient number of studies and participants. However, the categories that differ with respect to the main analysis have an insufficient number of participants or studies. Also, the covariate distribution is a matter of concern, which could affect the reliability of the subgroup analysis (Analysis 7.2). The heterogeneity in the overall pooled estimate for major VTE across studies is not explained by the doses of drugs analysed (Analysis 1.2). Also, the same effect as in the main analysis with all the combined doses was observed for this outcome.

For those reasons, for major VTE events, we present all the combined doses for the combined effect of all direct factor Xa inhibitors instead of each drug separately in the main analysis.

For major bleeding, the test for subgroup differences suggests that there is a statistically significant subgroup effect (P = 0.02) (Analysis 2.4), meaning that the type of direct factor Xa inhibitor statistically significantly modifies the effect of direct factor X inhibitors in comparison to LMWH for the outcome of major bleeding for the analysis for each drug with all the combined doses. In the subgroup analysis, which includes the therapeutic doses approved (or small doses) (Analysis 7.4), the same is observed. The treatment effect does not show any difference between direct factor Xa inhibitors over LMWHs for apixaban, darexaban, edoxaban, erixaban, letaxaban, and LY517717. Only rivaroxaban shows more bleeding events than in LMWHs; therefore, the subgroup effect is qualitative.

As opposed to all the other outcomes, for major bleeding events, all the direct factor Xa inhibitors that agree with the main analysis have an insufficient number of studies and participants. However, rivaroxaban is the only one that differs with respect to the main analysis due to the higher number of major bleeding events and the sufficient number of participants and studies. There is no heterogeneity amongst them (I² = 0) that requires further analysis. The covariate distribution is a matter of concern, which could affect the reliability of the low dose effect in the other direct factor Xa inhibitors (Analysis 7.4).

In order to explore if the higher major bleeding events observed in rivaroxaban (all doses) in comparison with the other direct factor Xa inhibitors (all doses) (Analysis 2.4), and also observed in this approved dose subgroup analysis (Analysis 7.4), in which rivaroxaban doses ranged from 5 to 30 mg/day, could be related to the higher doses in this subgroup and not related to the standard dosis for orthopaedic surgery (rivaroxaban 10 mg/day, apixaban 5 mg/day), we performed a sensitivity analysis excluding rivaroxaban and apixaban (other doses) for major bleeding events. Despite no statistical significance being reached, probably due to the insufficient number of participants, a tendency for more major bleeding events was observed only in the rivaroxaban 10 mg/day group and not in the other direct factor Xa inhibitors. (See sensitivity analysis by daily approved orthopaedic surgery dose groups for major bleeding below for a detailed explanation in Analysis 8.1). Also, in analyses 2.4, 7.4, and 8.1, the test for subgroup effect is statistically significant. In all of these analyses, rivaroxaban (independently of its dose) is the only direct factor Xa inhibitor that explains this subgroup difference. Therefore, drug dose is not the cause of the observed higher major bleeding events for rivaroxaban. Also, drug dose does not modify the effect of all the other direct factor Xa inhibitors in comparison with LMWHs.

8.1. Analysis.

Comparison 8: Direct factor Xa inhibitors by type versus LMWH: sensitivity analysis by daily approved orthopedic surgery dose groups for major bleeding, Outcome 1: Major bleeding: rivaroxaban 10 mg/day, apixaban 5 mg/day, and other direct factor Xa drugs at low dose vs LMWH

The heterogeneity in the overall pooled estimate for major bleeding across studies is explained by rivaroxaban (Analysis 7.4), irrespective of the dose group.

For those reasons, for major bleeding events, we present the results of the combined effect of all doses of rivaroxaban separately from the combined effect of all doses of the other direct factor Xa inhibitors instead of all drugs combined.

For volume of blood loss, the combined treatment effect shows no difference in the direct factor X inhibitors in comparison with LMWHs, as happens in the main analysis (Analysis 1.14). As in the analysis of drug type with all the combined doses (Analysis 2.11), the subgroup effect is qualitative. The category that agrees with the main analysis has a sufficient number of studies and participants, and the heterogeneity is less than the main analysis. However, the other two categories that differ with respect to the main analysis have an insufficient number of participants and studies. Also, the covariate distribution is a matter of concern, which could affect the reliability of the subgroup analysis (Analysis 7.9).

Regarding the outcomes with heterogeneity in the primary analysis (major VTE, asymptomatic distal DVT, liver enzymes elevation, and volume of blood loss), we evaluated if the daily dose subgroups could explain the heterogeneity. Only for the volume of blood loss is the very high heterogeneity partially explained by the daily drug dose.

Sensitivity analysis

Since there are a variety of ways to perform a meta‐analysis, we tested the robustness of our results using sensitivity analysis. In each sensitivity analysis, we did not include studies with zero events in the intervention and control groups.

Effect of daily approved orthopaedic surgery dose groups for major bleeding

Results are summarised in Analysis 8.1 (sensitivity analysis by daily approved orthopaedic surgery dose groups for major bleeding).

To evaluate if the higher major bleeding events observed in rivaroxaban groups (all doses) (Analysis 1.5) and in the subgroup with all the approved doses (Analysis 7.4) differ from other direct factor Xa inhibitors, we performed a sensitivity analysis including only the approved dose for rivaroxaban for VTE prevention on orthopaedic surgery (10 mg/day) compared with LMWHs and the other direct factor Xa inhibitors including only the approved dose (or small dose if there is no specific, approved dose for VTE prevention on orthopaedic surgery) compared with LMWHs. In this analysis, the test for subgroup differences suggests that there is a statistically significant subgroup effect (P < 0.10), meaning that in this drug dose, the type of drug statistically significantly modifies the effect of direct factor X inhibitors in comparison to LMWHs for the outcome of major bleeding. The data in the categories are independent, meaning that the study participants contribute to no more than one category in the forest plot. The treatment effect does not show any difference between direct factor X inhibitors over LMWHs for apixaban and the other direct factor Xa inhibitors at low doses.

Only in rivaroxaban 10 mg/day is there a tendency for more bleeding events than in LMWHs; therefore, the subgroup effect is qualitative. Also, more than five trials in each drug type category were included (17, 6, and 12 respectively), with over 4000 participants in each category (15,858, 12,287, and 5714, respectively; however, in none of the drug type categories was the OIS achieved), so the covariate distribution is probably a matter of concern for this subgroup analysis. There is low heterogeneity (I² = 20%) between the results of the trials investigating apixaban 5 mg/day. A visual inspection of the forest plot confirms that heterogeneity is higher within apixaban than across all trials (I² = 9%); also, the confidence intervals generally overlap, and the larger studies tend to show an effect estimate of the beneficial effect of apixaban over LMWHs. Therefore, the validity of the treatment effect estimate for the apixaban category is uncertain, as individual trial results are inconsistent. There is no heterogeneity between the results of the trials of rivaroxaban (10 mg/day) and the other direct factor Xa inhibitors at low doses (I² = 0%) that requires further exploration (Analysis 8.1). Despite no statistical significance being reached for rivaroxaban 10 mg/day versus LMWHs, probably due to the insufficient number of participants (less than the OIS), a tendency for more bleeding events was observed.

More major bleeding events are observed in rivaroxaban‐approved doses (Analysis 7.4) and rivaroxaban (all doses combined) in comparison with LMWHs (Analysis 1.5). Upon visual inspection, we found that the drug type subclassification (rivaroxaban versus other direct factor Xa inhibitors), not the drug dose, was responsible for the statistical significance of the difference between subgroups in these drug dose analyses. Therefore, drug dose is not the cause of the higher major bleeding events observed in the rivaroxaban comparison, as opposed to the comparison between other direct factor Xa inhibitors and LMWHs, in which no difference in major bleeding events was found. The results of this sensitivity analysis strengthened our confidence to present the effect of rivaroxaban versus LMWHs for major bleeding events separately from the other direct factor Xa inhibitors.

Restricted to treatment period results

Results are summarised in Analysis 9: sensitivity analysis by restriction to treatment period results.

We performed sensitivity analyses by restricting outcome data to the period of administration of the intervention, so post‐treatment follow‐up events were excluded (Analysis 9.1 to Analysis 9.14). Effect estimates remained unchanged or with similar directions and magnitudes for all outcomes in comparison with the original analysis (Analysis 1.1 to Analysis 1.14).

9.1. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 1: All‐cause mortality

9.14. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 14: Volume of blood loss

Exclusion of studies with a high risk of selection bias (random sequence generation and allocation concealment)

Results are summarized in Analysis 10. We performed sensitivity analysis (Analysis 10.1 to Analysis 10.14) by excluding those studies deemed to have a high risk of selection bias (Agnelli 2007; Berezhnyak 2016; Cohen 2013; ODIXa‐HIP 2007; Rahman 2020).

Effect estimates remained unchanged or with similar directions and magnitudes for all outcomes in comparison with the original analysis, except for one harm outcome: major bleeding with rivaroxaban. The overall effect without the studies at high risk for selection bias (RR 1.60, 95% CI 1.00 to 2.57; I = 0%; 11 studies; 15,106 participants; Analysis 10.5) is different from the original effect (RR 1.94, 95% CI 1.26 to 2.98; I = 0%; 17 studies, 17,630 participants; Analysis 1.5). The magnitude of the effect remains similar, but the direction of the effect changes slightly; the test for subgroup differences shows no statistical difference (P = 0.83; data not available in the analysis).

10.5. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 5: Major bleeding: rivaroxaban versus LMWH

Exclusion of studies with a high risk of 'other' bias

We performed sensitivity analyses by excluding studies deemed to have a high risk of bias for not reporting conflicts of interest or for reporting conflicts of interest without describing a management plan to minimise the risk of bias to ensure research integrity, such as divestitures or independent data reviews.

Seven studies did not report that conflicts of interest were registered (Jiang 2019; Lassen 2003; NCT01205932; NCT01206972; ODIXa‐HIP 2007; Özler 2015; Wasko 2015). The authors of 28 studies declared financial conflicts of interest (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Cohen 2013; DARINA 2021; EXPERT 2009; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Kanan 2008; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Weitz 2010). None of these 35 studies reported a management plan describing procedures or additional steps for minimising the risk of bias. Due to the high number of excluded studies, the magnitude of the effect size differed and the 95% confidence interval became wider than in the original analysis. This analysis is not presented because no useful information was provided.

Regarding inappropriate administration of an intervention, we performed a sensitivity analysis (Analysis 11.1 to Analysis 11.14), excluding six studies: four studies with high risk for this bias in which the duration of prophylaxis differed markedly for each intervention group and the outcome analysis was at different time points (Chen 2016; Hui 2013; RECORD2 2008; Wing 2020); and two studies with an unclear risk for this bias due to the different durations of prophylaxis for the study groups and for not stating the time point analysed (Berezhnyak 2016; Yun‐Fei 2018). Effect estimates remained unchanged, with similar directions and magnitudes for all outcomes in comparison with the original analysis (see Analysis 11.1 to Analysis 11.14).

11.1. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 1: All‐cause mortality

11.14. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 14: Volume of blood loss

Regarding the use of an insensitive instrument to measure outcomes, which can lead to underestimation of beneficial effects, we performed sensitivity analysis on the VTE‐related outcomes, excluding the 29 studies that evaluated DVT events with venography. Of these, 26 studies used bilateral venography (ADVANCE‐1 2009; ADVANCE‐2 2010; ADVANCE‐3 2010; Agnelli 2007; APROPOS 2007; Cohen 2013; FOXTROT 2020; Fuji 2014‐THA; Fuji 2014‐TKA; Kanan 2008; ODIXa‐HIP 2007; ODIXa‐HIP2 2006; ODIXa‐HIP‐OD 2006; ODIXa‐KNEE 2005; ONYX‐1 2007; ONYX‐2 2010; ONYX‐3 2014; Raskob 2010; RECORD1 2008; RECORD2 2008; RECORD3 2008; RECORD4 2009; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015; Weitz 2010), and three studies evaluated DVT events with unilateral venography (EXPERT 2009; Lassen 2003; STARS E‐3 2014). We deemed these studies to have a high risk of bias. We performed sensitivity analysis for all outcomes by excluding those studies deemed to have a high risk of bias for these criteria. However, due to the high number of excluded studies, the magnitude of the effect size differed and the 95% confidence interval became wider than the original analysis. This analysis is not presented because no useful information was provided.

Exclusion of RECORD4 2009 study

Currently, the RECORD4 2009 study is under investigation. During a recent audit of the study, the FDA identified numerous data integrity deficiencies at eight of the 16 clinical trial sites (Demasi 2022; Turpie 2022), raising concerns about the accuracy of the reported results. For that reason, we performed a sensitivity analysis excluding RECORD4 2009. (Analysis not presented.) Effect estimates remained unchanged and with similar directions and magnitudes for all outcomes in comparison with the original analysis. We excluded this study from our main analysis of adverse events (serious hepatic and non‐hepatic adverse events and minor adverse events). This study presented data for just one of these three outcomes (serious non‐hepatic adverse events: 114 events for 1526 participants in the direct factor Xa inhibitors group and 134 events for 1508 participants in the LMWH group).

Assessment of publication bias

Results are summarised in Table 3.

We assessed publication bias by visual inspection of funnel plots for all meta‐analyses. Asymmetry was found by visual inspection for major VTE, serious non‐hepatic adverse events, asymptomatic distal DVT, liver enzymes elevation and volume of blood loss. We could not assess serious hepatic adverse events because only one study with events was found. Additionally, we performed tests for funnel plot asymmetry for those meta‐analyses including 10 or more studies (Sterne 2011). Peters' test was used for dichotomous outcomes and Egger's test for continuous outcomes. We used the package "meta" (Schwarzer 2007) for R software version 3.3.2 (R‐project 2009). We established funnel plot asymmetry with a P value greater than 0.10. Overall, we found evidence of publication bias only for liver enzymes elevation (P = 0.010) (Figure 4, Table 3). Visual inspection of funnel plots showed that the asymmetry observed for liver enzymes elevation was in the expected direction for publication bias (favouring the treatment groups). The "missing" studies would have favoured LMWHs. The summary effects obtained by fixed‐effect and random‐effects meta‐analyses for this outcome were similar: 0.60 (95% CI 0.53 to 0.70) and 0.58 (95% CI 0.46 to 0.73), respectively. Nevertheless, we performed a sensitivity analysis excluding the seven small studies that, according to Figure 4, favoured the intervention (EXPERT 2009; Fuji 2014‐TKA; Kim 2016; STARS E‐3 2014; STARS J‐2 2014; STARS J‐4 2014; STARS J‐V 2015), and obtained a result with a similar direction and effect estimate magnitude as the original analysis. Also, the fixed‐effect and the random‐effects meta‐analytic results were similar to those of the original analysis: fixed‐effect 0.69 (95% CI 0.59 to 0.81) and random‐effect 0.69 (95% CI 0.58 to 0.83) (Analysis 12.1).

4.

Funnel plot of comparison: 1 Direct factor Xa inhibitors versus LMWH, outcome: 1.10 Liver enzymes elevation.

12.1. Analysis.

Comparison 12: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding small studies for liver enzymes elevation outcome, Outcome 1: Liver enzymes elevation

Direct factor Xa inhibitors versus vitamin K antagonists (VKAs)

One included study reported data for the comparison between direct factor Xa inhibitors and vitamin K antagonists (warfarin) (APROPOS 2007). This study compared apixaban (dose 5 to 20 mg/day, given once or twice daily) with warfarin started after surgery and administered for a duration of 10 to 14 days (short prophylaxis), in people undergoing elective TKR surgery. The study reported summary data for the treatment period for several review outcomes. It reported summary data for the follow‐up period (at 30 days postoperatively) only for all‐cause mortality, fatal VTE, and fatal bleeding. This study also compared apixaban with LMWHs (see Characteristics of included studies).

Primary outcomes

All‐cause mortality

During the treatment period, one death was reported (1/638 in the apixaban group and 0/109 in the warfarin group), corresponding to a fatal pulmonary embolism. During the follow‐up period, two additional deaths were reported (two in the apixaban group and zero in the warfarin group, total number of participants under follow‐up was not reported), corresponding to a myocardial infarction and end‐stage heart failure with cachexia.

Major VTE

During the treatment period, 13 major VTE events were reported (11/638 in the apixaban group and 2/109 in the warfarin group). The difference was reported by the study authors as not statistically significant.

Symptomatic VTE

The study reported no data for this outcome.

Major bleeding

During the treatment period, 18 major bleeding events were reported (18/917 in the apixaban group and 0/151 in the warfarin group). The study did not evaluate the statistical significance of this difference.

Serious hepatic adverse events

The study reported no data for this outcome.

Serious non‐hepatic adverse events

During the treatment period, 89 serious non‐hepatic adverse events were reported (80/917 in the apixaban group and 9/151 in the warfarin group). The study did not evaluate the statistical significance of this difference, nor did it provide a definition of serious adverse events.

Secondary outcomes

Fatal VTE

As mentioned above, only one VTE‐related death in the apixaban group (fatal pulmonary embolism) was reported during the study treatment and follow‐up periods.

Asymptomatic distal DVT

The study reported no data for this outcome.

Fatal bleeding

The study reported no bleeding‐related deaths during the study treatment and follow‐up periods.

Liver enzymes elevation

During the treatment period, 15 liver enzymes elevation events were reported (12/899 in the apixaban group and 3/150 in the warfarin group). The study did not evaluate the statistical significance of this difference.

Minor adverse events

The study reported no data for this outcome.

Volume of blood loss

The study reported no data for this outcome.

Discussion

We included 53 randomised controlled trials (RCTs), enroling 44,371 participants. We assessed the evidence on the effects of using direct factor Xa inhibitors as prophylaxis for venous thromboembolism (VTE) after elective major orthopaedic surgery compared with using low molecular weight heparins (LMWHs, 53 studies) and vitamin K antagonists (VKAs, 1 study). Five studies did not provide data for quantitative synthesis.

The comparison between direct factor Xa inhibitors and VKAs was addressed in only one included study, which reported data on several review outcomes. Given the small number of events for most of these outcomes, the results are not precise enough to show a difference in the effects of the study drugs. We did not include these data in the quantitative synthesis; therefore, the results of this review are only applicable to the comparison of direct factor Xa inhibitors with LMWHs.

Summary of main results

For every analysed outcome, when it was feasible, we presented the combined effect of all major orthopaedic surgeries (the THR and TKR populations), all the oral direct factor Xa inhibitors as a group (instead of each drug separately), all the combined doses, and all the reported events (in the treatment and follow‐up period).

After a primary systematic meta‐analysis of the data (Analysis 1.1 to Analysis 1.14), we performed several pre‐specified subgroup and sensitivity analyses, exploring heterogeneity and providing more accurate estimates of treatment effects (Analysis 2.1 to Analysis 12.1). After a meticulous evaluation of these subgroups and sensitivity analyses, we decided to report all the outcomes (except major bleeding events) as we had planned; that is, as all direct factor Xa inhibitors combined, with all studied doses combined, and with both major orthopaedic surgeries combined, instead of each of them independently, because, after reviewing the evidence, we believe this is the best way to summarise it. In terms of major bleeding events, analysis revealed that drug type was the only subgroup that modified the combined effect. Rivaroxaban is the only direct factor Xa inhibitor that may have more major bleeding events than LMWHs. There may have been little or no difference in major bleeding events experienced by participants given the other direct factor Xa inhibitors and those given LMWHs.

We assessed the overall risk of bias for each outcome. We conducted a thorough analysis of each potential bias individually. 'Unclear risk of bias' was our most common assessment (41% of all risk assessments) across the eight risk of bias domains for all 53 included studies. We deemed that there was a high risk of bias in 31% of all risk assessments. However, despite this, we deemed the overall risk of bias as high in all outcomes. Nearly all the analysed outcomes did not raise concerns about selection and performance bias. Across all outcomes, most studies raised serious concerns about attrition, reporting, and other forms of bias.

We rated the certainty of the evidence as low for five of 12 outcomes (symptomatic VTE; major bleeding for rivaroxaban versus LMWHs; serious non‐hepatic adverse events; asymptomatic distal DVT; and minor adverse events). We rated the remaining outcomes as having very low‐certainty evidence (all‐cause mortality; major VTE; major bleeding for the non‐rivaroxaban direct factor Xa inhibitors; serious hepatic adverse events; fatal VTE; fatal bleeding; liver enzymes elevation; volume of blood loss).

We did not synthesise in a meta‐analysis the evidence for volume of blood loss due to very high unexplained heterogeneity.

Regarding the primary outcomes, we found that oral direct factor Xa inhibitors may:

• have little to no effect on all‐cause mortality, but the evidence is very uncertain;

• reduce major VTE events, but the evidence is very uncertain;

• slightly reduce symptomatic VTE events when compared with LMWHs.

Also, in the evaluation of major bleeding, the evidence suggests rivaroxaban results in a slight increase in major bleeding events when compared with LMWHs. However, the remaining oral direct factor Xa inhibitors may have little to no effect on major bleeding compared to LMWHs, but the evidence is very uncertain. Direct factor Xa inhibitors may have little to no effect on serious hepatic adverse events, but the evidence is very uncertain. Direct factor Xa inhibitors may reduce serious non‐hepatic adverse events slightly when compared with LMWHs.

Regarding secondary outcomes, we found that oral direct factor Xa inhibitors may have little to no effect on fatal VTE, fatal bleeding, and liver enzymes elevations compared to LMWHs, but the evidence is very uncertain for these three outcomes. We found that direct factor Xa inhibitors may result in a slight reduction in asymptomatic distal VTE events, and may make little to no difference in minor adverse events when compared with LMWHs.

The absolute benefit of substituting factor Xa inhibitors for LMWHs may be between two and five fewer symptomatic venous thromboembolism episodes per 1000 patients, and between three and 14 fewer serious non‐hepatic adverse events per 1000 patients. The absolute risk of substituting the factor Xa inhibitor rivaroxaban for LMWHs may be between one and seven more major bleeding events per 1000 patients.

Overall completeness and applicability of evidence

Although all the studies included in this review enroled people undergoing major orthopaedic surgery, the procedure was explicitly stated to have been elective in only 61% of the studies. Surgery was explicitly stated to have been primary (i.e. did not include revision surgery) and unilateral in 39% and 32% of included studies, respectively. Any underlying joint disease was varied in most studies, with osteoarthritis being the main aetiology. These characteristics were not consistently described throughout the included studies. Therefore, the results of this review cannot be generalised to specific groups of patients. Only three studies examined people undergoing hip fracture surgery (Hoseinzadeh 2022; STARS J‐4 2014; Tang 2017), and only two of these reported data on all‐cause mortality, fatal VTE, major VTE, symptomatic VTE, asymptomatic distal DVT, fatal bleeding, major bleeding, serious non‐hepatic adverse events, liver enzyme elevation, and volume of blood loss (STARS J‐4 2014; Tang 2017). Therefore, the results of this review cannot be generalised to people undergoing hip fracture surgery, given their under‐representation.

Despite the absence of statistical heterogeneity in most of the outcomes, there is relevant clinical heterogeneity between people who undergo elective hip or knee replacement and those who undergo urgent surgery for a hip fracture. The latter are much older and have a very high short‐term mortality rate.

We performed quantitative synthesis for all direct factor Xa inhibitors studied in RCTs, except for razaxaban, as the only study evaluating it provided no information about any of the review's outcomes (Lassen 2003). We addressed additional considerations regarding the duration, timing of start, and daily frequency of administration of prophylaxis in this review.

Although most studies reported adverse events, detailed descriptions of the events under surveillance were not available. Therefore, serious hepatic adverse events might have been reported under the category of serious adverse events, leading to underestimation of the incidence of this outcome in our review. This possibility also implies that the 'serious non‐hepatic adverse events' outcome might have been heterogeneously reported throughout the studies. However, inconsistency for this outcome was low (I² < 25%) in the main review analysis.

We had concerns about the diagnostic instrument – venography – that nearly 55% (29 of 53) of the included studies used to confirm a diagnosis of deep vein thrombosis (DVT). In 17 studies with a total of 21,288 participants, venography either could not be performed or was performed with uninterpretable results in 6078 (29%) participants. The remaining 12 studies did not report the number of venograms not performed. Other non‐invasive instruments, such as ultrasound and computed tomography venography, are frequently used in clinical practice to diagnose DVT events (Tovey 2003). In this review, 13 studies evaluated DVT with Doppler ultrasound and one study with computed tomography venography. Of these, three studies reported that there were no participant exclusions due to instrument deficiency (Bai 2021; Hoseinzadeh 2022; Zhou 2023). In two further studies, we were able to estimate participant exclusions (Khalafallah 2018; Kim 2016). In these five studies, with a total of 1944 participants, only 42 (2.2%) were not analysed (although we do not know if these 42 were excluded because ultrasonography could not be performed, the ultrasonogram was uninterpretable, or for some other reason). We consider venography to be an insensitive instrument to measure outcomes, which can lead to underestimation of beneficial effects.

We also had concerns about investigators' potential conflicts of interest. As conflicts of interest can lead to biases that seriously affect research, whether consciously or not (AAMC 2008; Dana 2003; Warner 2004), mechanisms are required to avoid this. These mechanisms include transparent reporting of the existence of conflicts of interest, and taking steps to reduce bias and ensure research integrity, such as instituting a conflict of interest management committee, investigator divestment, and independent data review (Bero 2017; Ioannidis 2005). Of the studies included in this review, most had a high risk of bias due to failing to report conflicts of interest or lack thereof.

We considered the included study, RECORD4 2009, to be at high risk of bias because of suspected fraud. This study is under investigation. During an audit years after the publication of the study, the US Food and Drug Administration (FDA) identified numerous data integrity deficiencies at eight of the 16 clinical trial sites (Demasi 2022; Turpie 2022), raising concerns about the accuracy of the reported results.

Finally, although we had concerns about the potential for sponsorship bias, we were unable to explore it as we did not have access to the included studies' original data. Sponsorship bias is an important bias but difficult to identify without full disclosure and access to study data. Secrecy and closeness between those who manufacture drugs, those who test them, and those who advertise them need to be avoided (Jefferson 2020).

Certainty of the evidence

This review included 53 RCTs. Most of this body of evidence addresses the comparison between direct factor Xa inhibitors and LMWHs. Although studies differed in individual methodological assessments, most shared important characteristics such as a parallel‐group design, inclusion of people scheduled for primary unilateral joint surgery with similar inclusion and exclusion criteria, use of enoxaparin as a comparator, low post‐randomisation exclusion rates for safety outcomes, and similar definitions for most outcomes. There was a low risk of selection bias in most studies. The risk for performance, attrition (benefit outcomes), reporting, and other biases was high in most studies, and the risk for selection and detection bias was mostly unclear.

Using the GRADE approach, we rated the certainty of the evidence as very low to low for both primary and secondary outcomes.

We downgraded the certainty of the evidence by two levels for 10 of the 13 outcomes, and by one level for the remaining three outcomes due to the high overall risk of bias in the studies (particularly in the attrition and reporting bias domains). We downgraded the evidence certainty for selection bias in two outcomes (serious hepatic adverse events; volume of blood loss) and for performance bias in three outcomes only (serious non‐hepatic adverse events, minor adverse events; volume of blood loss). We considered that the remaining outcomes were unlikely to be affected by an unblinded design.

Regarding inconsistency, we found zero or low heterogeneity in nine of 13 outcomes. We found serious inconsistency due to moderate heterogeneity in two outcomes (major VTE and liver enzymes elevation), and downgraded the certainty of the evidence by one level for this reason. We found very serious inconsistency due to high unexplained heterogeneity (I² = 96%) for the final outcome, volume of blood loss, and for this reason, we did not pool studies in a meta‐analysis.

We had no concerns regarding the indirectness of evidence in any analysed outcome.

Regarding imprecision, even if the total number of participants for an outcome did not reach the optimal information size, we nevertheless did not downgrade for imprecision if we included more than 4000 participants in the overall pooled estimate for each outcome and the 95% confidence interval excluded no effect, or was narrow enough to exclude important benefit or important harm, or both (Guyatt 2011). We had no concerns about imprecision in seven outcomes. In five of six outcomes (all‐cause mortality; major bleeding for the comparison of other direct factor Xa inhibitors (besides rivaroxaban) versus LMWHs; serious hepatic adverse events; fatal VTE; fatal bleeding), we downgraded the evidence by two levels for imprecision because an insufficient number of participants were included in the overall pooled estimate to exclude no effect or exclude important benefit or harm. In the remaining outcome (volume of blood loss), we downgraded by one level for imprecision because the overall pooled estimate failed to exclude no effect or exclude important benefit or harm.

In one outcome (liver enzymes elevation), we had concerns about the likelihood of publication bias, based on our visual inspection of the funnel plot and using Peters' test for funnel plot asymmetry (Peters 2006). Consequently, we downgraded the certainty of the evidence for this outcome.

Potential biases in the review process

We endeavoured to limit the risk of introducing bias throughout the review process by having at least two review authors, working independently, perform study screening and selection, data extraction, and risk of bias assessment. We are confident that we included all relevant published studies in this review, given that electronic and other sources were comprehensively searched; however, relevant unpublished studies might have been missed. Our analysis of publication bias showed that our results for liver enzymes elevation might have been biased in favour of direct factor Xa inhibitors.

There are a number of potential biases in the review process, given the presence of the following limitations: (1) we obtained limited additional data from study authors; (2) incomplete reporting of methods limited the assessment of the risk of bias in the included studies; and (3) there was inconsistent use of the different study drugs throughout the studies. We also made some deviations from the published protocol during the review process (see Differences between protocol and review), which we considered appropriate and necessary in order to obtain applicable results.

We carried out data imputation mostly for studies with larger sample sizes, outcome measures, and longer follow‐up periods, given that these studies often have different analysis sets and losses to follow‐up. We imputed only sample size data. We assumed these missing data corresponded to data that were not explicitly reported. We imputed missing data only for the total numbers of participants (N) when data on the numbers of events were reported separately for the periods of treatment and post‐treatment follow‐up. (See Appendix 7 for a detailed description of the cases in which these assumptions were made.)

Agreements and disagreements with other studies or reviews

Several systematic reviews have assessed the role of direct factor Xa inhibitors as prophylactic agents for prevention of VTE in people undergoing major orthopaedic surgery (Caldeira 2017; Cao 2010; Gómez‐Outes 2012; Kapoor 2017; Ma 2015; Neumann 2012; Ning 2016; Turun 2011; Venker 2017; Yoshida Rde 2013). Our results regarding all direct factor Xa inhibitors compared with LMWHs were derived from a larger number of studies and participants than the results in previous reviews, and most outcomes we analysed had low heterogeneity. In addition, we included a wider array of direct factor Xa inhibitors than in other systematic reviews.

Previous systematic reviews have assessed only five of the 12 outcomes considered in this review. These reviews also included outcomes such as total VTE (Caldeira 2017; Turun 2011; Venker 2017), pulmonary embolism (Neumann 2012; Ning 2016; Yoshida Rde 2013), total DVT (Kapoor 2017; Ma 2015), symptomatic DVT (Neumann 2012; Yoshida Rde 2013), proximal DVT (Yoshida Rde 2013), asymptomatic DVT (Ning 2016), total bleeding (Cao 2010; Neumann 2012; Yoshida Rde 2013), clinically relevant non‐major bleeding (Cao 2010; Ning 2016; Yoshida Rde 2013), and postoperative wound infection (Ning 2016). Although differing in numbers of included studies, these reviews showed that rivaroxaban was associated with a reduced risk of major VTE and symptomatic VTE, without causing a difference in the effect on major bleeding and liver enzymes elevation, as compared with LMWHs. Similarly, no difference in the risk of major bleeding was seen when all direct factor Xa inhibitors were analysed together, or for apixaban and edoxaban individually. No difference in the risk of major VTE was seen for apixaban individually. Of note, people undergoing hip fracture surgery were not included in any of the above‐mentioned reviews.

Our results agree with some of the findings mentioned above but differ regarding the effects of rivaroxaban on major bleeding and liver enzymes elevation. When comparing the relative estimates of direct factor Xa inhibitors versus LMWHs (risk ratios and their 95% confidence intervals) in those reviews with this one, we found in our subgroup analysis that rivaroxaban is associated with more major bleeding events than LMWHs. All but one previous review found that rivaroxaban was not associated with a difference in the risk of major bleeding (Caldeira 2017; Cao 2010; Gómez‐Outes 2012; Turun 2011; Venker 2017; Yoshida Rde 2013). However, we included more studies and participants in our review, and detected no heterogeneity in the subgroup analysis restricted to the rivaroxaban versus LMWHs comparison. In contrast to previous reviews, we found a risk difference, with more major bleeding events for rivaroxaban. Re‐expressing these results as an absolute effect estimate, if it is assumed that three out of 1000 people given LMWHs would have major bleeding events, then seven out of 1000 (95% CI 4 to 10) people given rivaroxaban would.

The only previous review that analysed the effect of rivaroxaban on liver enzymes elevation did not find any association with relative estimates (Yoshida Rde 2013). In our review, when evaluating the relative effect estimates, we found that people given direct factor Xa inhibitors were less likely to have elevated liver enzymes than those given LMWHs, and the same was true for a comparison restricted to rivaroxaban versus LMWHs. Analysing the absolute effect estimate range, our conclusion was that oral direct factor Xa inhibitors may have little to no effect on liver enzymes elevation, but the evidence was very uncertain.

The Neumann 2012 systematic review was the cornerstone for developing the American Society of Hematology (ASH) 2019 guideline recommendation regarding the use of direct oral anticoagulants (DOACs) in people undergoing major orthopaedic surgery. The DOACs evaluated were the direct factor Xa inhibitors and dabigatran, an oral thrombin inhibitor (Anderson 2019; recommendation 10). The Neumann 2012 review was very well‐designed with robust methodology. Amongst other interesting statistical analyses, the authors used meta‐regression, which is not something we did in our review. They also assessed the primary studies' risk of bias and evaluated the certainty of evidence according to GRADE methodology. They excluded studies that reported composite outcomes without providing information on the components of the outcome, or that presented confounding comparisons. We preferred to include all studies whose PICO met our review criteria and to evaluate the confounding comparisons in a sensitivity analysis, by excluding studies with a high risk of bias due to the inappropriate administration of the intervention (different duration of the intervention) (Analysis 11.1 to Analysis 11.14; see the Effects of interventions, sensitivity analysis section, for more information). Our review and Neumann 2012 had some differences in the risk of bias assessment. Neumann and colleagues rated most of the included studies as having a low risk of bias for selective reporting, whereas we rated the same studies as high risk in this domain because we detected selective reporting for several of this review's outcomes (the two reviews evaluated different outcomes). Both reviews assessed most of the studies as having a high risk of bias for incomplete outcome data in three outcomes (mortality, symptomatic DVT and PE, and bleeding). However, Neumann and colleagues reported a low risk of bias in all studies in the 'other bias' domain.

Both reviews analysed the impact of the high number of missing participants. Neumann and colleagues performed a sensitivity analysis that included all participants in the denominators, using extreme assumptions about outcomes in participants with missing data: that is, they assumed that the event risk in people with missing data in the intervention group was two or three times the event risk for those with available data, and in control groups, they assumed the same event risk for those with missing and available data. Ultimately, the results of their main and sensitivity analyses did not differ. We performed three scenario analyses and a worst‐case‐scenario analysis that led us to rate most outcomes as having a high risk of attrition bias (see Incomplete outcome data (attrition bias) above for a detailed explanation; see also Appendix 5, and Appendix 6). Another difference between Neumann 2012 and our review is the approach taken to evaluating dose effect: Neumann and colleagues subgrouped doses (into high, intermediate, and low), and compared each subgroup with the main analysis, whereas we adopted a different approach (see Effects of interventions, daily dose subgroup analysis, for further details). Finally, Neumann and colleagues rated the certainty of evidence as high in three of the four analysed outcomes and moderate in one (due to inconsistency) (Neumann 2012).

The 2019 ASH Guidelines used the Neumann 2012 review and included two more outcomes (Anderson 2019). Authors rated four of six of these outcomes as having moderate certainty of evidence and the remaining two as having high certainty. In all the outcomes, the authors did not find any serious limitations regarding the risk of bias (Anderson 2019, recommendation 10, supplemental material: EtD framework is available online at https://guidelines.gradepro.org/profile/9160FAA2‐4F98‐A3AA‐9816‐64DF796ABBC7). By contrast, we downgraded evidence certainty by two levels for serious limitations regarding the high risk of bias in the included studies, rating the certainty of the evidence as low or very low in all outcomes included in our review. The 2019 ASH Guidelines address four efficacy outcomes (mortality, symptomatic pulmonary embolism, symptomatic proximal DVT, and symptomatic distal DVT) and two safety outcomes (major bleeding and re‐operation). Like us, the ASH Guidelines considered differences of less than three events per 1000 people as unlikely to be important to patients for both efficacy (benefit) and safety (harm) outcomes. We calculated the effect estimate for each outcome in our review to constitute either a 'small important effect', 'trivial effect', or no effect, according to the GRADE approach and terminology (Schünemann 2006). For all‐cause mortality and symptomatic VTE, our findings are similar to those of the 2019 ASH Guidelines, in terms of the direction and magnitude of the absolute effect estimate for these outcomes.

Regarding major bleeding, the 2019 ASH Guidelines did not find a difference in effect between DOACs, with moderate certainty of evidence (Anderson 2019; recommendation 10). The authors calculated an absolute risk difference of zero fewer people with a major bleeding event per 1000 people (95% CI from 2 fewer to 3 more). We found more major bleeding events in the rivaroxaban groups than the LMWH groups (with low certainty of evidence). Rivaroxaban resulted in a slight increase in major bleeding events when compared with LMWHs. We calculated an absolute risk difference of three more people with a major bleeding event per 1000 people (95% CI from 1 more to 7 more). This effect remains even when higher dose groups are not included. For the other direct factor Xa inhibitors, we found they may have little to no effect on major bleeding, but the evidence was very uncertain (an absolute risk difference of three fewer people with a major bleeding event per 1000; 95% CI from 5 fewer to 0 fewer).

Authors' conclusions

Implications for practice.

Direct factor Xa inhibitors may have little to no effect on all‐cause mortality and major venous thromboembolic events compared to low molecular weight heparins (LMWHs), but the evidence is very uncertain. Direct factor Xa inhibitors may result in a slight reduction in symptomatic venous thromboembolism compared to LMWHs.

In terms of harms, the direct factor Xa inhibitor rivaroxaban may result in a slight increase in major bleeding events compared to LMWHs. Other direct factor Xa inhibitors may have little to no effect on major bleeding, but the evidence is very uncertain.

Compared to LMWHs, direct factor Xa inhibitors may have little to no effect on serious hepatic adverse events, but the evidence is very uncertain. They may reduce serious non‐hepatic adverse events slightly compared to LMWHs.

Due to the high rates of missing participants and selective outcome reporting, the summary effect estimates may be biased.

Implications for research.

We would encourage investigators conducting future randomised controlled trials that compare direct factor Xa inhibitors to LMWHs, vitamin K antagonists, or both, for prevention of thromboembolism in people undergoing major orthopaedic surgery to:

• improve the design and reporting of studies by always declaring investigators' conflicts of interest and by presenting a management plan to prevent bias for this reason (such as independent outcome analysis);

• use optimal methods for diagnosing deep vein thrombosis (DVT). Conventional venography has serious limitations. We were unable to include a sizeable number of participants in the efficacy population due to venography not being performed or being unreadable in the primary studies. More recent studies used other diagnostic instruments, such as Doppler ultrasound or computer venography, with minimal missing participants.

• avoid reporting bias by publishing results for all the outcomes as they were planned in the protocol;

• analyse critical outcomes besides major venous thromboembolism, such as mortality, symptomatic pulmonary embolism, symptomatic proximal DVT, and clinically relevant bleeding (the latter was reported in a minority of the included studies).

• employ an authoritative, widely‐used definition of 'major bleeding', such as the one adopted by the Scientific and Standardization Committee (Schulman 2010);

• increase the follow‐up period to no less than 60 days in this setting;

• include more participants undergoing hip fracture surgery;

• evaluate older people (more than 70 years of age) and people from under‐represented ethnic groups;

• aim to determine the effects of different doses of edoxaban on the risk of thromboembolic and bleeding events in larger populations;

• aim to determine the differences in the effects of rivaroxaban and other direct factor Xa inhibitors on the risk of thromboembolic and bleeding events by head‐to‐head comparisons;

• aim to determine the differences in the effects of aspirin, the new oral anticoagulants (direct factor VII and IX) with direct factor Xa inhibitors on the risk of thromboembolic and bleeding events by head‐to‐head comparisons;

• describe the effects of direct factor Xa inhibitors on the risk of serious adverse events, especially those related to hepatotoxicity.

History

Protocol first published: Issue 6, 2015

Appendices

Appendix 1. CENTRAL search November 2023

#1	MESH DESCRIPTOR Thrombosis	0
#2	MESH DESCRIPTOR Thromboembolism	0
#3	MESH DESCRIPTOR Venous Thromboembolism	0
#4	MESH DESCRIPTOR Venous Thrombosis EXPLODE ALL TREES	0
#5	(thrombus* or thrombopro* or thrombotic* or thrombolic* or thromboemboli* or thrombos* or embol*):ti,ab,kw	4201
#6	MESH DESCRIPTOR Pulmonary Embolism EXPLODE ALL TREES	0
#7	(PE or DVT or VTE):ti,ab,kw	1293
#8	((vein* or ven*) near thromb*):ti,ab,kw	1392
#9	(blood near3 clot*):ti,ab,kw	850
#10	(pulmonary near3 clot*):ti,ab,kw	2
#11	(lung near3 clot*):ti,ab,kw	1
#12	#1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9 OR #10 OR #11	5576
#13	MESH DESCRIPTOR Arthroplasty, Replacement, Hip EXPLODE ALL TREES	0
#14	MESH DESCRIPTOR Arthroplasty, Replacement, Knee EXPLODE ALL TREES	0
#15	MESH DESCRIPTOR Hip Fractures EXPLODE ALL TREES	0
#16	((hip near3 (surg* or replac* or arthroplast* or prosthe* or hemiarthroplas*))):ti,ab,kw	1423
#17	((knee near3 (surg* or replac* or arthroplast* or prosthe*))):ti,ab,kw	2092
#18	(((hip or femur or femoral) near3 fracture)):ti,ab,kw	863
#19	#13 OR #14 OR #15 OR #16 OR #17 OR #18	3752
#20	MESH DESCRIPTOR Factor Xa Inhibitors EXPLODE ALL TREES	0
#21	MESH DESCRIPTOR Factor Xa EXPLODE ALL TREES	0
#22	(*x?ban):ti,ab,kw	268384
#23	(Xarelto or Bay‐597939 or Bay597939 ):ti,ab,kw	34
#24	(BMS‐562247 or BMS‐562247):ti,ab,kw	0
#25	(Factor X* near4 (antag* or inhib* or block*)):ti,ab,kw	129
#26	(FX* near4 (antag* or inhib* or block*)):ti,ab,kw	25
#27	(10* near4 (antag* or inhib* or block*) ):ti,ab,kw	255
#28	(DU‐176b or DU176b):ti,ab,kw	0
#29	(LY517717 or LY‐517717):ti,ab,kw	0
#30	(PRT054021 or PRT‐054021):ti,ab,kw	0
#31	(YM150 or YM‐150 or DU‐176b or DU176*):ti,ab,kw	0
#32	(GW813893 or Tak 442 or PD0348292 or GSK‐813893 or GSK813893):ti,ab,kw	0
#33	#20 OR #21 OR #22 OR #23 OR #24 OR #25 OR #26 OR #27 OR #28 OR #29 OR #30 OR #31 OR #32	268384
#34	#12 AND #19 AND #33	235

with dates in CRSO from 01/09/2021 to 12/11/2023.

Appendix 2. Trial registries searches from February 2017 to November 2023

ClinicalTrials.gov

18 studies found for: (hip OR knee) AND thromboprophylaxis until February 2017

3 studies found for: (hip OR knee) AND thromboprophylaxis until November 2023

World Health Organization International Clinical Trials Registry Platform

189 records for 59 trials found for (Hip or knee)/Title AND (thrombosis or DVT or embolism)/Condition OR thromboprophylaxis/Intervention until February 2017

46 records for 0 trials found for (Hip or knee)/Title AND (thrombosis or DVT or embolism)/Condition OR thromboprophylaxis/Intervention until November 2023

ISRCTN Register

5 results (hip OR knee) AND thromboprophylaxis until February 2017

9 results (hip OR knee) AND thromboprophylaxis until November 2023

Cochrane Vascular Specialised Register

151 records until February 2017

112 records until September 2021

Appendix 3. Sources searched and search strategies (last searched November 2023)

Source	Search strategy	Hits retrieved
1. VASCULAR REGISTER IN CRSW (Date of most recent search: 1 September 2021)	#1 Xa AND INREGISTER	May 2018: 3 May 2020: 81 Sep 2021: 28
2. CENTRAL via CRSO (Date of most recent search: 11 November 2023)	#1 MESH DESCRIPTOR Thrombosis #2 MESH DESCRIPTOR Thromboembolism #3 MESH DESCRIPTOR Venous Thromboembolism #4 MESH DESCRIPTOR Venous Thrombosis EXPLODE ALL TREES #5 (thrombus* or thrombopro* or thrombotic* or thrombolic* or thromboemboli* or thrombos* or embol*):TI,AB,KY #6 MESH DESCRIPTOR Pulmonary Embolism EXPLODE ALL TREES #7 (PE or DVT or VTE):TI,AB,KY #8 (((vein* or ven*) near thromb*)):TI,AB,KY #9 ((blood near3 clot*)):TI,AB,KY #10 (pulmonary near3 clot*):TI,AB,KY #11 ((lung near3 clot*)):TI,AB,KY #12 #1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9 OR #10 OR #11 #13 MESH DESCRIPTOR Arthroplasty, Replacement, Hip EXPLODE ALL TREES #14 MESH DESCRIPTOR Arthroplasty, Replacement, Knee EXPLODE ALL TREES #15 MESH DESCRIPTOR Hip Fractures EXPLODE ALL TREES #16 (hip near3 (surg* or replac* or arthroplast* or prosthe* or hemiarthroplas*)):TI,AB,KY #17 (knee near3 (surg* or replac* or arthroplast* or prosthe*)):TI,AB,KY #18 (((hip or femur or femoral) near3 fracture)):TI,AB,KY #19 #13 OR #14 OR #15 OR #16 OR #17 OR #18 #20 MESH DESCRIPTOR Factor Xa Inhibitors EXPLODE ALL TREES #21 MESH DESCRIPTOR Factor Xa EXPLODE ALL TREES #22 *x?ban:TI,AB,KY #23 (Xarelto or Bay‐597939 or Bay597939 ):TI,AB,KY #24 (BMS‐562247 or BMS‐562247):TI,AB,KY #25 (Factor X* near4 (antag* or inhib* or block*)):TI,AB,KY #26 (FX* near4 (antag* or inhib* or block*)):TI,AB,KY #27 (10* near4 (antag* or inhib* or block*) ):TI,AB,KY #28 (DU‐176b or DU176b):TI,AB,KY #29 (LY517717 or LY‐517717):TI,AB,KY #30 (PRT054021 or PRT‐054021):TI,AB,KY #31 YM150 or YM‐150 or DU‐176b or DU176* #32 GW813893 or Tak 442 or PD0348292 or GSK‐813893 or GSK813893 #33 #20 OR #21 OR #22 OR #23 OR #24 OR #25 OR #26 OR #27 OR #28 OR #29 OR #30 OR #31 OR #32 #34 #12 AND #19 AND #33	Feb 2017: 168 May 2018: 70 May 2020: 724 Sep 2021: 293 Nov 2023: 235
3. ClinicalTrials.gov (Date of most recent search: 11 November 2023)	hip OR knee | thromboprophylaxis Hip OR knee Arthropathy | thromboprophylaxis / Interventional studies	May 2018: 1 May 2020: 2 Sep 2021: 0 Nov 2023: 4
4. ICTRP Search Portal (Date of most recent search: 11 November 2023)	thromboprophylaxis thromboembolism	May 2018: 25 May 2020: N/A Sep 2021: 13 Nov 2023: 10
5. MEDLINE Ovid MEDLINE Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, Ovid MEDLINE Daily and Ovid MEDLINE 1946 to Present (Date of most recent search: 11 November 2023)	1 THROMBOSIS/ 2 THROMBOEMBOLISM/ 3 Venous Thromboembolism/ 4 exp Venous Thrombosis/ 5 (thrombus* or thrombopro* or thrombotic* or thrombolic* or thromboemboli* or thrombos* or embol*).ti,ab. 6 exp Pulmonary Embolism/ 7 (PE or DVT or VTE).ti,ab. 8 (blood adj3 clot*).ti,ab. 9 (pulmonary adj3 clot*).ti,ab. 10 (lung adj3 clot*).ti,ab. 11 ((vein* or ven*) adj thromb*).ti,ab. 12 or/1‐11 13 exp Arthroplasty, Replacement, Hip/ 14 exp Arthroplasty, Replacement, Knee/ 15 exp Hip Fractures/ 16 (hip adj3 (surg* or replac* or arthroplast* or prosthe* or hemiarthroplas*)).ti,ab. 17 (knee adj3 (surg* or replac* or arthroplast* or prosthe*)).ti,ab. 18 ((hip or femur or femoral) adj3 fracture).ti,ab. 19 or/13‐18 20 exp Factor Xa Inhibitors/ 21 exp Factor Xa/ 22 (Xarelto or Bay‐597939 or Bay597939).ti,ab. 23 (BMS‐562247 or BMS‐562247).ti,ab. 24 (Factor X* adj4 (antag* or inhib* or block)).ti,ab. 25 (FX* adj4 (antag* or inhib* or block*)).ti,ab. 26 (10* adj4 (antag* or inhib* or block*)).ti,ab. 27 (DU‐176b or DU176b).ti,ab. 28 (LY517717 or LY‐517717).ti,ab. 29 (PRT054021 or PRT‐054021).ti,ab. 30 (YM150 or YM‐150 or DU‐176b or DU176*).ti,ab. 31 (GW813893 or Tak 442 or PD0348292 or GSK‐813893 or GSK813893).ti,ab. 32 or/20‐31 33 12 and 19 and 32 34 randomized controlled trial.pt. 35 controlled clinical trial.pt. 36 randomized.ab. 37 placebo.ab. 38 drug therapy.fs. 39 randomly.ab. 40 trial.ab. 41 groups.ab. 42 or/34‐41 43 exp animals/ not humans.sh. 44 42 not 43 45 33 and 44	May 2018: 33 May 2020: 40 Sep 2021: 21 Nov 2023: 27
6. Embase (Date of most recent search: 11 November 2023)	1 thrombosis/ 2 thromboembolism/ 3 venous thromboembolism/ 4 exp vein thrombosis/ 5 (thrombus* or thrombopro* or thrombotic* or thrombolic* or thromboemboli* or thrombos* or embol*).ti,ab. 6 exp lung embolism/ 7 (PE or DVT or VTE).ti,ab. 8 ((vein* or ven*) adj thromb*).ti,ab. 9 (blood adj3 clot*).ti,ab. 10 (pulmonary adj3 clot*).ti,ab. 11 (lung adj3 clot*).ti,ab. 12 or/1‐11 13 exp hip replacement/ 14 exp knee replacement/ 15 exp hip fracture/ 16 (hip adj3 (surg* or replac* or arthroplast* or prosthe* or hemiarthroplas*)).ti,ab. 17 (knee adj3 (surg* or replac* or arthroplast* or prosthe*)).ti,ab. 18 ((hip or femur or femoral) adj3 fracture).ti,ab. 19 or/13‐18 20 exp blood clotting factor 10a inhibitor/ 21 exp blood clotting factor 10a/ 22 (Xarelto or Bay‐597939 or Bay597939).ti,ab. 23 (BMS‐562247 or BMS‐562247).ti,ab. 24 (Factor X* adj4 (antag* or inhib* or block*)).ti,ab. 25 (FX* adj4 (antag* or inhib* or block*)).ti,ab. 26 (10* adj4 (antag* or inhib* or block*)).ti,ab. 27 (DU‐176b or DU176b).ti,ab. 28 (LY517717 or LY‐517717).ti,ab. 29 (PRT054021 or PRT‐054021).ti,ab. 30 (YM150 or YM‐150 or DU‐176b or DU176*).ti,ab. 31 (GW813893 or Tak 442 or PD0348292 or GSK‐813893 or GSK813893).ti,ab. 32 or/20‐31 33 12 and 19 and 32 34 randomized controlled trial/ 35 controlled clinical trial/ 36 random$.ti,ab. 37 randomization/ 38 intermethod comparison/ 39 placebo.ti,ab. 40 (compare or compared or comparison).ti. 41 ((evaluated or evaluate or evaluating or assessed or assess) and (compare or compared or comparing or comparison)).ab. 42 (open adj label).ti,ab. 43 ((double or single or doubly or singly) adj (blind or blinded or blindly)).ti,ab. 44 double blind procedure/ 45 parallel group$1.ti,ab. 46 (crossover or cross over).ti,ab. 47 ((assign$ or match or matched or allocation) adj5 (alternate or group$1 or intervention$1 or patient$1 or subject$1 or participant$1)).ti,ab. 48 (assigned or allocated).ti,ab. 49 (controlled adj7 (study or design or trial)).ti,ab. 50 (volunteer or volunteers).ti,ab. 51 trial.ti. 52 or/34‐51 53 33 and 52	May 2018: 98 May 2020: 128 Sep 2021: 140 Nov 2023: 276
7. CINAHL via EBSCO (Date of most recent search: 1 September 2021)	S45 S31 AND S44 S44 S32 OR S33 OR S34 OR S35 OR S36 OR S37 OR S38 OR S39 OR S40 OR S41 OR S42 OR S43 S43 (MH "Crossover Design") S42 TX random* S41 TX trial* S40 TX "latin square" S39 (MH "Random Assignment") S38 (MH "Single‐Blind Studies") or (MH "Double‐Blind Studies") or (MH "Triple‐Blind Studies") S37 (MH "Factorial Design") 0 S36 (MH "Placebos") S35 (MH "Clinical Trials") S34 TX "multi‐centre study" OR "multi‐center study" OR "multicentre study" OR "multicenter study" OR "multi‐site study" S33 TX crossover OR "cross‐over" S32 AB placebo* 28,095 S31 S12 AND S19 AND S30 S30 S20 OR S21 OR S22 OR S23 OR S24 OR S25 OR S26 OR S27 OR S28 OR S29 S29 TX GW813893 or Tak 442 or PD0348292 or GSK‐813893 or GSK813893 S28 TX YM150 or YM‐150 or DU‐176b or DU176* S27 TX PRT054021 or PRT‐054021 S26 TX LY517717 or LY‐517717 S25 TX DU‐176b or DU176b S24 TX 10* n4 (antag* or inhib* or block*) S23 TX FX* n4 (antag* or inhib* or block*) S22 TX BMS‐562247 or BMS‐562247 1 S21 TX Xarelto or Bay‐597939 or Bay597939 S20 TX Factor Xa S19 S13 OR S14 OR S15 OR S16 OR S17 OR S18 S18 TX ((hip or femur or femoral) n3 fracture) S17 TX knee n3 (surg* or replac* or arthroplast* or prosthe*) S16 TX hip n3 (surg* or replac* or arthroplast* or prosthe* or hemiarthroplas*)) S15 (MH "Hip Fractures+") S14 (MH "Arthroplasty, Replacement, Knee") S13 (MH "Arthroplasty, Replacement, Hip") S12 S1 OR S2 OR S3 OR S4 OR S5 OR S6 OR S7 OR S8 OR S9 OR S10 OR S11 S11 TX lung n3 clot* S10 TX pulmonary 3 clot* S9 TX blood n3 clot* S8 TX ((vein* or ven*) n thromb*) S7 TX PE or DVT or VTE S6 (MH "Pulmonary Embolism") S5 TX thrombus* or thrombopro* or thrombotic* or thrombolic* or thromboemboli* or thrombos* or embol* S4 (MH "Venous Thrombosis+") S3 (MH "Venous Thromboembolism") S2 (MH "Thromboembolism") S1 (MH "Thrombosis")	May 2018: 0 May 2020: 6 Sep 2021: 1
8. AMED via OVID (Date of most recent search: 1 September 2021)	1 THROMBOSIS/ 2 THROMBOEMBOLISM/ 3 (thrombus* or thrombopro* or thrombotic* or thrombolic* or thromboemboli* or thrombos* or embol*).ti,ab. 4 exp Pulmonary Embolism/ 5 (PE or DVT or VTE).ti,ab. 6 (blood adj3 clot*).ti,ab. 7 (pulmonary adj3 clot*).ti,ab. 8 (lung adj3 clot*).ti,ab. 9 ((vein* or ven*) adj thromb*).ti,ab. 10 or/1‐9 11 exp Arthroplasty replacement hip/ 12 exp Arthroplasty replacement knee/ 13 exp Hip fractures/ 14 (hip adj3 (surg* or replac* or arthroplast* or prosthe* or hemiarthroplas*)).ti,ab. 15 (knee adj3 (surg* or replac* or arthroplast* or prosthe*)).ti,ab. 16 ((hip or femur or femoral) adj3 fracture).ti,ab. 12 17 or/11‐16 18 Factor Xa Inhibitors.ti,ab. 19 (Xarelto or Bay‐597939 or Bay597939).ti,ab. 20 (BMS‐562247 or BMS‐562247).ti,ab. 21 (Factor X* adj4 (antag* or inhib* or block*)).ti,ab. 22 (FX* adj4 (antag* or inhib* or block*)).ti,ab. 23 (10* adj4 (antag* or inhib* or block*)).ti,ab. 24 (DU‐176b or DU176b).ti,ab. 25 (LY517717 or LY‐517717).ti,ab. 26 (PRT054021 or PRT‐054021).ti,ab. 27 (YM150 or YM‐150 or DU‐176b or DU176*).ti,ab. 28 (GW813893 or Tak 442 or PD0348292 or GSK‐813893 or GSK813893).ti,ab. 29 or/18‐28 30 10 and 17 and 29	May 2018: 0 May 2020: 0 Sep 2021: 0
TOTAL before deduplication		May 2018: 230 May 2020: 981 Sep 2021: 496 Nov 2023: 552
TOTAL after deduplication		May 2018: 170 May 2020: 886 Sep 2021: 492 Nov 2023: 491

Appendix 4. LILACS (last searched November 2023)

#1 Rivaroxaban (palabras del título) AND prophylaxis (palabras del título)

#2 Apixaban (palabras del título) AND prophylaxis (palabras del título)

#3 Edoxaban (palabras del título) AND prophylaxis (palabras del título)

#4 Darexaban (palabras del título) AND prophylaxis (palabras del título)

#5 Eribaxaban (palabras del título) AND prophylaxis (palabras del título)

#6 Betrixaban (palabras del título) AND prophylaxis (palabras del título)

#7 Letaxaban (palabras del título) AND prophylaxis (palabras del título)

#8 Razaxaban (palabras del título) AND prophylaxis (palabras del título)

#9 #1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8

Appendix 5. Analyses of benefit outcome (major VTE or, in case the study did not address this outcome, another benefit VTE‐related outcome) for every study with missing participants

Supplementary materials contain additional data and information that support or enhance the article. Supplementary materials may not be subject to the same editorial scrutiny as the content of the article, and Cochrane has not copy‐edited, typeset or proofread these materials. The material in these sections has been supplied by the author(s) for publication under a Licence for Publication and the author(s) are solely responsible for the material. Cochrane accordingly gives no representations or warranties of any kind in relation to, and accepts no liability for any reliance on or use of, such material.

If included studies are not mentioned here, this indicates that they did not have any missing participants or that they clearly reported missing participants.

For ADVANCE‐1 2009, for outcome major VTE, there were 26 events over 1269 participants (total randomised 1599; missing outcome data 330) in the direct FXa inhibitor group and 20 events over 1216 participants (total randomised 1596; missing outcome data 380) in the LMWH group. The calculated absolute risk (baseline AR) is 2,05 for the intervention and 1,64 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 7 new events for the direct FXa inhibitor group and 7 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 2,06 for the direct FXa inhibitor group and 1,69 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 5 new events for the direct FXa inhibitor group and 10 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,94 for the direct FXa inhibitor group and 1,88 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 29 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 3,44 for the direct FXa inhibitor group and 1,25 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 1% and 5% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be ‐11% and 24% for the LMWH group. And the difference for approach 3 would be 139% for the direct FXa inhibitor group and ‐39,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ADVANCE 2 2010 study, for outcome Major VTE, there were 13 events over 1195 participants (total randomized 1528; missing outcome data 333) in the direct FXa inhibitor group and 26 events over 1199 participants (total randomized 1529; missing outcome data 330) in the LMWH group. The calculated absolute risk (baseline AR) is 1,09 for the intervention and 2,17 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 4 new events for the direct FXa inhibitor group and 8 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 1,11 for the direct FXa inhibitor group and 2,22 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 5 new events for the direct FXa inhibitor group and 9 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,18 for the direct FXa inhibitor group and 2,29 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 29 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 2,75 for the direct FXa inhibitor group and 1,7 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 2% and 5% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 8,99999999999999% and 12,% for the LMWH group. And the difference for approach 3 would be 166% for the direct FXa inhibitor group and ‐47% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ADVANCE 3 2010 study, for outcome Major VTE, there were 10 events over 2199 participants (total randomized 2708; missing outcome data 509) in the direct FXa inhibitor group and 25 events over 2195 participants (total randomized 2699; missing outcome data 504) in the LMWH group. The calculated absolute risk (baseline AR) is 0,45 for the intervention and 1,14 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 3 new events for the direct FXa inhibitor group and 6 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,48 for the direct FXa inhibitor group and 1,15 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 8 new events for the direct FXa inhibitor group and 13 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,66 for the direct FXa inhibitor group and 1,41 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 44 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 1,99 for the direct FXa inhibitor group and 0,93 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 3% and 1% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 21,% and 27% for the LMWH group. And the difference for approach 3 would be 154% for the direct FXa inhibitor group and ‐21,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Agnelli 2007 study, for outcome Major VTE, there were 18 events over 313 participants (total randomized 490; missing outcome data 177) in the direct FXa inhibitor group and 10 events over 65 participants (total randomized 91; missing outcome data 26) in the LMWH group. The calculated absolute risk (baseline AR) is 5,75 for the intervention and 15,38 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 11 new events for the direct FXa inhibitor group and 4 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 5,92 for the direct FXa inhibitor group and 15,38 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 3 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 4,29 for the direct FXa inhibitor group and 12,09 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 16 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 6,94 for the direct FXa inhibitor group and 10,99 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 17% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be ‐146% and ‐329,% for the LMWH group. And the difference for approach 3 would be 119,% for the direct FXa inhibitor group and ‐439,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the APROPOS 2007 study, for outcome Major VTE, there were 11 events over 638 participants (total randomized 933; missing outcome data 295) in the direct FXa inhibitor group and 5 events over 109 participants (total randomized 152; missing outcome data 43) in the LMWH group. The calculated absolute risk (baseline AR) is 1,72 for the intervention and 4,59 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 6 new events for the direct FXa inhibitor group and 2 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 1,82 for the direct FXa inhibitor group and 4,61 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 5 new events for the direct FXa inhibitor group and 6 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,71 for the direct FXa inhibitor group and 7,24 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 26 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 3,97 for the direct FXa inhibitor group and 3,29 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 10% and 2% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be ‐1% and 265,% for the LMWH group. And the difference for approach 3 would be 225% for the direct FXa inhibitor group and ‐130,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Chen 2016 study, for outcome Major VTE, there were 2 events over 135 participants (total randomized 136; missing outcome data 1) in the direct FXa inhibitor group and 3 events over 133 participants (total randomized 134; missing outcome data 1) in the LMWH group. The calculated absolute risk (baseline AR) is 1,48 for the intervention and 2,26 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 2,21 for the direct FXa inhibitor group and 2,99 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 2,21 for the direct FXa inhibitor group and 2,99 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 2,21 for the direct FXa inhibitor group and 2,24 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 73% and 73% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 73% and 73,% for the LMWH group. And the difference for approach 3 would be 73% for the direct FXa inhibitor group and ‐1,99999999999996% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Cohen 2013 study, for outcome TOTAL VTE, there were 119 events over 561 participants (total randomized 1009; missing outcome data 448) in the direct FXa inhibitor group and 34 events over 188 participants (total randomized 402; missing outcome data 214) in the LMWH group. The calculated absolute risk (baseline AR) is 21,21 for the intervention and 18,09 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 96 new events for the direct FXa inhibitor group and 39 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 21,31 for the direct FXa inhibitor group and 18,16 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 9.1), there would be 63 new events for the direct FXa inhibitor group and 35 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 18,04 for the direct FXa inhibitor group and 17,16 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is STARS J4‐2014), there would be 30 new events for the direct FXa inhibitor group and 8 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 14,77 for the direct FXa inhibitor group and 10,45 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 10% and 7% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be ‐317,% and ‐93,% for the LMWH group. And the difference for approach 3 would be ‐644,% for the direct FXa inhibitor group and ‐764% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the DARINA 2021 study, for outcome Major VTE, there were 0 events over 48 participants (total randomized 51; missing outcome data 3) in the direct FXa inhibitor group and 0 events over 45 participants (total randomized 47; missing outcome data 2) in the LMWH group. The calculated absolute risk (baseline AR) is 0 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,96 for the direct FXa inhibitor group and 2,13 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 1,96 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 0% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 196% and 213% for the LMWH group. And the difference for approach 3 would be 196% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the EXPERT 2009 study, for outcome SYMPT VTE, there were 3 events over 135 participants (total randomized 172; missing outcome data 37) in the direct FXa inhibitor group and 1 events over 40 participants (total randomized 43; missing outcome data 3) in the LMWH group. The calculated absolute risk (baseline AR) is 2,22 for the intervention and 2,5 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 2,33 for the direct FXa inhibitor group and 4,65 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.3), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 2,33 for the direct FXa inhibitor group and 4,65 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is STARS E3 2014), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 2,33 for the direct FXa inhibitor group and 4,65 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 11% and 215% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 11,% and 215,% for the LMWH group. And the difference for approach 3 would be 11,% for the direct FXa inhibitor group and 215,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the FOXTROT 2020 study, for outcome Major VTE, there were 2 events over 83 participants (total randomized 100; missing outcome data 17) in the direct FXa inhibitor group and 3 events over 76 participants (total randomized 102; missing outcome data 26) in the LMWH group. The calculated absolute risk (baseline AR) is 2,41 for the intervention and 3,95 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 2 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 3 for the direct FXa inhibitor group and 4,9 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 3 for the direct FXa inhibitor group and 3,92 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 2 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 4 for the direct FXa inhibitor group and 2,94 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 59% and 95% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 59,% and ‐3,00000000000003% for the LMWH group. And the difference for approach 3 would be 159% for the direct FXa inhibitor group and ‐101,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Fuji 2014‐THA study, for outcome Major VTE, there were 2 events over 268 participants (total randomized 343; missing outcome data 75) in the direct FXa inhibitor group and 0 events over 82 participants (total randomized 104; missing outcome data 22) in the LMWH group. The calculated absolute risk (baseline AR) is 0,75 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,87 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 2 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,17 for the direct FXa inhibitor group and 0,96 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 7 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 2,62 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 12% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 42,% and 96% for the LMWH group. And the difference for approach 3 would be 187% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Fuji 2014‐TKA study, for outcome Major VTE, there were 6 events over 157 participants (total randomized 181; missing outcome data 24) in the direct FXa inhibitor group and 2 events over 74 participants (total randomized 91; missing outcome data 17) in the LMWH group. The calculated absolute risk (baseline AR) is 3,82 for the intervention and 2,7 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 3,87 for the direct FXa inhibitor group and 3,3 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 3,87 for the direct FXa inhibitor group and 3,3 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 3 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 4,97 for the direct FXa inhibitor group and 2,2 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 5% and 60% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 5,00000000000003% and 60,% for the LMWH group. And the difference for approach 3 would be 115,% for the direct FXa inhibitor group and ‐50% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Kanan 2008 study, for outcome Major VTE, there were 3 events over 33 participants (total randomized 35; missing outcome data 2) in the direct FXa inhibitor group and 2 events over 32 participants (total randomized 32; missing outcome data 0) in the LMWH group. The calculated absolute risk (baseline AR) is 9,09 for the intervention and 6,25 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 11,43 for the direct FXa inhibitor group and 6,25 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 11,43 for the direct FXa inhibitor group and 6,25 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 11,43 for the direct FXa inhibitor group and 6,25 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 234% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 234% and 0% for the LMWH group. And the difference for approach 3 would be 234% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Khalafallah 2018 study, for outcome TOTAL VTE, there were 8 events over 356 participants (total randomized 375; missing outcome data 19) in the direct FXa inhibitor group and 13 events over 366 participants (total randomized 375; missing outcome data 9) in the LMWH group. The calculated absolute risk (baseline AR) is 2,25 for the intervention and 3,55 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 2,4 for the direct FXa inhibitor group and 3,73 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 9.1), there would be 3 new events for the direct FXa inhibitor group and 2 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 2,93 for the direct FXa inhibitor group and 4 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is STARS J4‐2014), there would be 2 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 2,67 for the direct FXa inhibitor group and 3,73 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 15% and 18% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 68,% and 45,% for the LMWH group. And the difference for approach 3 would be 42,% for the direct FXa inhibitor group and 18,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Kim 2016 study, for outcome all‐cause mortality, there were 0 events over 350 participants (total randomized 379; missing outcome data 29) in the direct FXa inhibitor group and 0 events over 351 participants (total randomized 379; missing outcome data 28) in the LMWH group. The calculated absolute risk (baseline AR) is 0 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.1), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,26 for the direct FXa inhibitor group and 0,26 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ADVANCE 2 2010), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,26 for the direct FXa inhibitor group and 0,26 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 0% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 26% and 26% for the LMWH group. And the difference for approach 3 would be 26% for the direct FXa inhibitor group and 26% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the NCT01205932 study, for outcome Major VTE, there were 1 events over 235 participants (total randomized 303; missing outcome data 68) in the direct FXa inhibitor group and 0 events over 79 participants (total randomized 99; missing outcome data 20) in the LMWH group. The calculated absolute risk (baseline AR) is 0,43 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,66 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 2 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,99 for the direct FXa inhibitor group and 1,01 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 6 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 2,31 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 23% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 56,% and 101% for the LMWH group. And the difference for approach 3 would be 188% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the NCT01206972 study, for outcome Major VTE, there were 3 events over 158 participants (total randomized 226; missing outcome data 68) in the direct FXa inhibitor group and 0 events over 55 participants (total randomized 76; missing outcome data 21) in the LMWH group. The calculated absolute risk (baseline AR) is 1,9 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 2 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 2,21 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 2 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 2,21 for the direct FXa inhibitor group and 1,32 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 6 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 3,98 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 31% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 31,% and 132% for the LMWH group. And the difference for approach 3 would be 208% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ODIXa‐HIP 2007 study, for outcome Major VTE, there were 17 events over 359 participants (total randomized 479; missing outcome data 120) in the direct FXa inhibitor group and 5 events over 107 participants (total randomized 162; missing outcome data 55) in the LMWH group. The calculated absolute risk (baseline AR) is 4,74 for the intervention and 4,67 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 6 new events for the direct FXa inhibitor group and 3 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 4,8 for the direct FXa inhibitor group and 4,94 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 2 new events for the direct FXa inhibitor group and 2 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 3,97 for the direct FXa inhibitor group and 4,32 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 11 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 5,85 for the direct FXa inhibitor group and 3,09 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 6% and 27% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be ‐77% and ‐35,% for the LMWH group. And the difference for approach 3 would be 111,% for the direct FXa inhibitor group and ‐158% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ODIXa‐HIP2 2006 study, for outcome Major VTE, there were 9 events over 442 participants (total randomized 586; missing outcome data 144) in the direct FXa inhibitor group and 5 events over 106 participants (total randomized 136; missing outcome data 30) in the LMWH group. The calculated absolute risk (baseline AR) is 2,04 for the intervention and 4,72 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 3 new events for the direct FXa inhibitor group and 2 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 2,05 for the direct FXa inhibitor group and 5,15 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 3 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 2,05 for the direct FXa inhibitor group and 4,41 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 13 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 3,75 for the direct FXa inhibitor group and 3,68 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 1% and 43% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 0,999999999999979% and ‐31,% for the LMWH group. And the difference for approach 3 would be 171% for the direct FXa inhibitor group and ‐104,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ODIXa‐HIP‐OD 2006 study, for outcome Major VTE, there were 18 events over 511 participants (total randomized 713; missing outcome data 202) in the direct FXa inhibitor group and 3 events over 107 participants (total randomized 160; missing outcome data 53) in the LMWH group. The calculated absolute risk (baseline AR) is 3,52 for the intervention and 2,8 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 8 new events for the direct FXa inhibitor group and 2 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 3,65 for the direct FXa inhibitor group and 3,13 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 4 new events for the direct FXa inhibitor group and 2 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 3,09 for the direct FXa inhibitor group and 3,13 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 18 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 5,05 for the direct FXa inhibitor group and 1,88 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 13% and 33% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be ‐43,% and 33,% for the LMWH group. And the difference for approach 3 would be 153,% for the direct FXa inhibitor group and ‐92% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ODIXa‐KNEE 2005 study, for outcome Major VTE, there were 11 events over 296 participants (total randomized 506; missing outcome data 210) in the direct FXa inhibitor group and 3 events over 70 participants (total randomized 105; missing outcome data 35) in the LMWH group. The calculated absolute risk (baseline AR) is 3,72 for the intervention and 4,29 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 8 new events for the direct FXa inhibitor group and 2 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 3,75 for the direct FXa inhibitor group and 4,76 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 4 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 2,96 for the direct FXa inhibitor group and 3,81 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 19 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 5,93 for the direct FXa inhibitor group and 2,86 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 3% and 47% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be ‐76,% and ‐48% for the LMWH group. And the difference for approach 3 would be 221,% for the direct FXa inhibitor group and ‐143,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ONYX‐1 2007 study, for outcome Major VTE, there were 10 events over 116 participants (total randomized 142; missing outcome data 26) in the direct FXa inhibitor group and 0 events over 31 participants (total randomized 36; missing outcome data 5) in the LMWH group. The calculated absolute risk (baseline AR) is 8,62 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 3 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 9,15 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 7,75 for the direct FXa inhibitor group and 2,78 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 3 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 9,15 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 53% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be ‐86,9999999999999% and 278% for the LMWH group. And the difference for approach 3 would be 53,0000000000001% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ONYX‐2 2010 study, for outcome Major VTE, there were 5 events over 581 participants (total randomized 849; missing outcome data 268) in the direct FXa inhibitor group and 0 events over 127 participants (total randomized 168; missing outcome data 41) in the LMWH group. The calculated absolute risk (baseline AR) is 0,86 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 3 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,94 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 5 new events for the direct FXa inhibitor group and 2 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,18 for the direct FXa inhibitor group and 1,19 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 24 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 3,42 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 8% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 32,% and 119% for the LMWH group. And the difference for approach 3 would be 256% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ONYX‐3 2014 study, for outcome Major VTE, there were 26 events over 1208 participants (total randomized 1587; missing outcome data 379) in the direct FXa inhibitor group and 10 events over 328 participants (total randomized 405; missing outcome data 77) in the LMWH group. The calculated absolute risk (baseline AR) is 2,15 for the intervention and 3,05 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 9 new events for the direct FXa inhibitor group and 3 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 2,21 for the direct FXa inhibitor group and 3,21 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 6 new events for the direct FXa inhibitor group and 2 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 2,02 for the direct FXa inhibitor group and 2,96 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 33 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 3,72 for the direct FXa inhibitor group and 2,47 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 6% and 16% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be ‐13,% and ‐8,99999999999999% for the LMWH group. And the difference for approach 3 would be 157,% for the direct FXa inhibitor group and ‐58,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Raskob 2010 study, for outcome Major VTE, there were 21 events over 630 participants (total randomized 728; missing outcome data 98) in the direct FXa inhibitor group and 20 events over 144 participants (total randomized 175; missing outcome data 31) in the LMWH group. The calculated absolute risk (baseline AR) is 3,33 for the intervention and 13,89 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 4 new events for the direct FXa inhibitor group and 5 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 3,43 for the direct FXa inhibitor group and 14,29 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 2 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 3,16 for the direct FXa inhibitor group and 12 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 9 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 4,12 for the direct FXa inhibitor group and 11,43 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 10% and 40% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be ‐17,% and ‐189,% for the LMWH group. And the difference for approach 3 would be 79% for the direct FXa inhibitor group and ‐246,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the RECORD1 2008 study, for outcome Major VTE, there were 4 events over 1686 participants (total randomized 2266; missing outcome data 580) in the direct FXa inhibitor group and 33 events over 1678 participants (total randomized 2275; missing outcome data 597) in the LMWH group. The calculated absolute risk (baseline AR) is 0,24 for the intervention and 1,97 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 2 new events for the direct FXa inhibitor group and 12 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,26 for the direct FXa inhibitor group and 1,98 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 9 new events for the direct FXa inhibitor group and 15 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,57 for the direct FXa inhibitor group and 2,11 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 50 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 2,38 for the direct FXa inhibitor group and 1,45 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 2% and 1% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 33,% and 14,% for the LMWH group. And the difference for approach 3 would be 214,% for the direct FXa inhibitor group and ‐52% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the RECORD2 2008 study, for outcome Major VTE, there were 6 events over 961 participants (total randomized 1252; missing outcome data 291) in the direct FXa inhibitor group and 49 events over 962 participants (total randomized 1257; missing outcome data 295) in the LMWH group. The calculated absolute risk (baseline AR) is 0,62 for the intervention and 5,09 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 2 new events for the direct FXa inhibitor group and 16 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,64 for the direct FXa inhibitor group and 5,17 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 5 new events for the direct FXa inhibitor group and 8 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,88 for the direct FXa inhibitor group and 4,53 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 26 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 2,56 for the direct FXa inhibitor group and 3,9 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 2% and 8% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 26% and ‐56,% for the LMWH group. And the difference for approach 3 would be 194% for the direct FXa inhibitor group and ‐119% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the RECORD3 2008 study, for outcome Major VTE, there were 9 events over 908 participants (total randomized 1254; missing outcome data 346) in the direct FXa inhibitor group and 24 events over 925 participants (total randomized 1277; missing outcome data 352) in the LMWH group. The calculated absolute risk (baseline AR) is 0,99 for the intervention and 2,59 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 4 new events for the direct FXa inhibitor group and 10 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 1,04 for the direct FXa inhibitor group and 2,66 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 6 new events for the direct FXa inhibitor group and 9 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,2 for the direct FXa inhibitor group and 2,58 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 30 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 3,11 for the direct FXa inhibitor group and 1,88 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 5% and 7% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 21,% and ‐0,999999999999979% for the LMWH group. And the difference for approach 3 would be 212% for the direct FXa inhibitor group and ‐71% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the RECORD4 2009 study, for outcome Major VTE, there were 13 events over 1122 participants (total randomized 1584; missing outcome data 462) in the direct FXa inhibitor group and 22 events over 1112 participants (total randomized 1564; missing outcome data 452) in the LMWH group. The calculated absolute risk (baseline AR) is 1,16 for the intervention and 1,98 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 6 new events for the direct FXa inhibitor group and 9 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 1,2 for the direct FXa inhibitor group and 1,98 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 7 new events for the direct FXa inhibitor group and 12 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,26 for the direct FXa inhibitor group and 2,17 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 40 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 3,35 for the direct FXa inhibitor group and 1,41 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 4% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 10,% and 19,% for the LMWH group. And the difference for approach 3 would be 219,% for the direct FXa inhibitor group and ‐57,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the STARS E‐3 2014 study, for outcome symp VTE, there were 4 events over 299 participants (total randomized 360; missing outcome data 61) in the direct FXa inhibitor group and 1 events over 295 participants (total randomized 356; missing outcome data 61) in the LMWH group. The calculated absolute risk (baseline AR) is 1,34 for the intervention and 0,34 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 1,39 for the direct FXa inhibitor group and 0,56 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.3), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,39 for the direct FXa inhibitor group and 0,56 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is STARS E‐3 2014), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 1,39 for the direct FXa inhibitor group and 0,56 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 5% and 22% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 4,99999999999998% and 22,% for the LMWH group. And the difference for approach 3 would be 4,99999999999998% for the direct FXa inhibitor group and 22,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the STARS J‐2 2014 study, for outcome Major VTE, there were 0 events over 150 participants (total randomized 175; missing outcome data 25) in the direct FXa inhibitor group and 0 events over 74 participants (total randomized 89; missing outcome data 15) in the LMWH group. The calculated absolute risk (baseline AR) is 0 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,57 for the direct FXa inhibitor group and 1,12 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 3 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 1,71 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 0% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 57,% and 112,% for the LMWH group. And the difference for approach 3 would be 171% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the STARS J‐4 2014 study, for outcome Major VTE, there were 0 events over 46 participants (total randomized 62; missing outcome data 16) in the direct FXa inhibitor group and 0 events over 27 participants (total randomized 30; missing outcome data 3) in the LMWH group. The calculated absolute risk (baseline AR) is 0 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,61 for the direct FXa inhibitor group and 3,33 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 2 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 3,23 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 0% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 161% and 333% for the LMWH group. And the difference for approach 3 would be 323% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the STARS J‐V 2015 study, for outcome symp VTE, there were 0 events over 255 participants (total randomized 307; missing outcome data 52) in the direct FXa inhibitor group and 0 events over 248 participants (total randomized 303; missing outcome data 55) in the LMWH group. The calculated absolute risk (baseline AR) is 0 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.3), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,33 for the direct FXa inhibitor group and 0,33 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is STARS E‐3 2014), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,33 for the direct FXa inhibitor group and 0,33 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 0% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 33% and 33% for the LMWH group. And the difference for approach 3 would be 33% for the direct FXa inhibitor group and 33% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Weitz 2010 study, for outcome symp VTE, there were 14 events over 621 participants (total randomized 885; missing outcome data 264) in the direct FXa inhibitor group and 4 events over 109 participants (total randomized 163; missing outcome data 54) in the LMWH group. The calculated absolute risk (baseline AR) is 2,25 for the intervention and 3,67 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 6 new events for the direct FXa inhibitor group and 2 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 2,26 for the direct FXa inhibitor group and 3,68 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.3), there would be 2 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,81 for the direct FXa inhibitor group and 3,07 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is STARS E‐3 2014), there would be 4 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 2,03 for the direct FXa inhibitor group and 3,07 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 1% and 1% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be ‐44,% and ‐60,% for the LMWH group. And the difference for approach 3 would be ‐22,% for the direct FXa inhibitor group and ‐60,% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Xie 2017 study, for outcome Major VTE, there were 0 events over 96 participants (total randomized 98; missing outcome data 2) in the direct FXa inhibitor group and 0 events over 98 participants (total randomized 98; missing outcome data 0) in the LMWH group. The calculated absolute risk (baseline AR) is 0 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an efficacy outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR1 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the efficacy outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.2), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,02 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ONIX1 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the corresponding AR3 including these new events plus the previous one and divided by all the randomized patients would be 1,02 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 0% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 102% and 0% for the LMWH group. And the difference for approach 3 would be 102% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

Appendix 6. Analyses of harm outcome (major bleeding or, in case the study did not address this outcome, another harm‐related outcome) for every study with missing participants

Supplementary materials contain additional data and information that support or enhance the article. Supplementary materials may not be subject to the same editorial scrutiny as the content of the article, and Cochrane has not copy‐edited, typeset or proofread these materials. The material in these sections has been supplied by the author(s) for publication under a Licence for Publication and the author(s) are solely responsible for the material. Cochrane accordingly gives no representations or warranties of any kind in relation to, and accepts no liability for any reliance on or use of, such material.

If included studies are not mentioned here, this indicates that they did not have any missing participants or that they clearly reported missing participants.

For the ADVANCE 1 2009 study, for outcome Major Bleeding, there were 11 events over 1596 participants (total randomized 1599; missing outcome data 3) in the direct FXa inhibitor group and 22 events over 1588 participants (total randomized 1596; missing outcome data 8) in the LMWH group. The calculated absolute risk (baseline AR) is 0,69 for the intervention and 1,39 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,75 for the direct FXa inhibitor group and 1,44 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,75 for the direct FXa inhibitor group and 1,44 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,75 for the direct FXa inhibitor group and 1,38 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 6% and 5% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 6% and 5% for the LMWH group. And the difference for approach 3 would be 6% for the direct FXa inhibitor group and ‐1% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ADVANCE 2 2010 study, for outcome Major Bleeding, there were 9 events over 1501 participants (total randomized 1528; missing outcome data 27) in the direct FXa inhibitor group and 14 events over 1508 participants (total randomized 1529; missing outcome data 21) in the LMWH group. The calculated absolute risk (baseline AR) is 0,6 for the intervention and 0,93 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,65 for the direct FXa inhibitor group and 0,98 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,65 for the direct FXa inhibitor group and 0,98 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 2 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,72 for the direct FXa inhibitor group and 0,92 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 5% and 5% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 5% and 5% for the LMWH group. And the difference for approach 3 would be 12% for the direct FXa inhibitor group and ‐1% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ADVANCE 3 2010 study, for outcome Major Bleeding, there were 22 events over 2673 participants (total randomized 2708; missing outcome data 35) in the direct FXa inhibitor group and 18 events over 2659 participants (total randomized 2699; missing outcome data 40) in the LMWH group. The calculated absolute risk (baseline AR) is 0,82 for the intervention and 0,68 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,85 for the direct FXa inhibitor group and 0,7 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,85 for the direct FXa inhibitor group and 0,7 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 2 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,89 for the direct FXa inhibitor group and 0,67 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 3% and 2% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 3% and 2% for the LMWH group. And the difference for approach 3 would be 7% for the direct FXa inhibitor group and ‐1% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Agnelli 2007 study, for outcome Major Bleeding, there were 2 events over 417 participants (total randomized 490; missing outcome data 73) in the direct FXa inhibitor group and 1 events over 90 participants (total randomized 91; missing outcome data 1) in the LMWH group. The calculated absolute risk (baseline AR) is 0,48 for the intervention and 1,11 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,61 for the direct FXa inhibitor group and 2,2 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,61 for the direct FXa inhibitor group and 2,2 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 4 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 1,22 for the direct FXa inhibitor group and 1,1 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 13% and 109,% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 13% and 109,% for the LMWH group. And the difference for approach 3 would be 74% for the direct FXa inhibitor group and ‐1% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the APROPOS 2007 study, for outcome Major Bleeding, there were 18 events over 917 participants (total randomized 933; missing outcome data 16) in the direct FXa inhibitor group and 0 events over 149 participants (total randomized 152; missing outcome data 3) in the LMWH group. The calculated absolute risk (baseline AR) is 1,96 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 2,04 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 2,04 for the direct FXa inhibitor group and 0,66 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 2,04 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 8% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 8% and 66% for the LMWH group. And the difference for approach 3 would be 8% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Chen 2016 study, for outcome Major Bleeding, there were 3 events over 135 participants (total randomized 136; missing outcome data 1) in the direct FXa inhibitor group and 8 events over 133 participants (total randomized 134; missing outcome data 1) in the LMWH group. The calculated absolute risk (baseline AR) is 2,22 for the intervention and 6,02 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 2,94 for the direct FXa inhibitor group and 6,72 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 2,94 for the direct FXa inhibitor group and 6,72 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 2,94 for the direct FXa inhibitor group and 5,97 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 72% and 70% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 72% and 70% for the LMWH group. And the difference for approach 3 would be 72% for the direct FXa inhibitor group and ‐5% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Cohen 2013 study, for outcome Major Bleeding, there were 5 events over 992 participants (total randomized 1009; missing outcome data 17) in the direct FXa inhibitor group and 2 events over 397 participants (total randomized 402; missing outcome data 5) in the LMWH group. The calculated absolute risk (baseline AR) is 0,5 for the intervention and 0,5 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,59 for the direct FXa inhibitor group and 0,75 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,59 for the direct FXa inhibitor group and 0,75 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,59 for the direct FXa inhibitor group and 0,5 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 9% and 25% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 9% and 25% for the LMWH group. And the difference for approach 3 would be 9% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the DARINA 2021 study, for outcome Major Bleeding, there were 0 events over 48 participants (total randomized 51; missing outcome data 3) in the direct FXa inhibitor group and 2 events over 45 participants (total randomized 47; missing outcome data 2) in the LMWH group. The calculated absolute risk (baseline AR) is 0 for the intervention and 4,44 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 0 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 6,38 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,96 for the direct FXa inhibitor group and 6,38 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 1,96 for the direct FXa inhibitor group and 4,26 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 0% and 194% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 196% and 194% for the LMWH group. And the difference for approach 3 would be 196% for the direct FXa inhibitor group and ‐18% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the EXPERT 2009 study, for outcome Liver enzymes elevation, there were 1 events over 171 participants (total randomized 172; missing outcome data 1) in the direct FXa inhibitor group and 4 events over 43 participants (total randomized 43; missing outcome data 0) in the LMWH group. The calculated absolute risk (baseline AR) is 0,58 for the intervention and 9,3 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 1,16 for the direct FXa inhibitor group and 9,3 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.10), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,16 for the direct FXa inhibitor group and 9,3 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is RASKOB 2010), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 1,16 for the direct FXa inhibitor group and 9,3 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 58,% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 58,% and 0% for the LMWH group. And the difference for approach 3 would be 58,% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the FOXTROT 2020 study, for outcome Major Bleeding, there were 0 events over 100 participants (total randomized 100; missing outcome data 0) in the direct FXa inhibitor group and 0 events over 102 participants (total randomized 102; missing outcome data 0) in the LMWH group. The calculated absolute risk (baseline AR) is 0 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 0% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 0% and 0% for the LMWH group. And the difference for approach 3 would be 0% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Fuji 2014‐THA study, for outcome Major Bleeding, there were 0 events over 343 participants (total randomized 343; missing outcome data 0) in the direct FXa inhibitor group and 0 events over 103 participants (total randomized 104; missing outcome data 1) in the LMWH group. The calculated absolute risk (baseline AR) is 0 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 0 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0,96 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 0% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 0% and 96% for the LMWH group. And the difference for approach 3 would be 0% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Fuji 2014‐TKA study, for outcome Major Bleeding, there were 1 events over 180 participants (total randomized 181; missing outcome data 1) in the direct FXa inhibitor group and 0 events over 90 participants (total randomized 91; missing outcome data 1) in the LMWH group. The calculated absolute risk (baseline AR) is 0,56 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 1,1 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,1 for the direct FXa inhibitor group and 1,1 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 1,1 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 54% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 54% and 110,% for the LMWH group. And the difference for approach 3 would be 54% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Khalafallah 2018 study, for outcome Major Bleeding, there were 0 events over 369 participants (total randomized 375; missing outcome data 6) in the direct FXa inhibitor group and 0 events over 371 participants (total randomized 375; missing outcome data 4) in the LMWH group. The calculated absolute risk (baseline AR) is 0 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,27 for the direct FXa inhibitor group and 0,27 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,27 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 0% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 27% and 27% for the LMWH group. And the difference for approach 3 would be 27% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Kim 2016 study, for outcome Major Bleeding, there were 0 events over 350 participants (total randomized 379; missing outcome data 29) in the direct FXa inhibitor group and 0 events over 351 participants (total randomized 379; missing outcome data 28) in the LMWH group. The calculated absolute risk (baseline AR) is 0 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,26 for the direct FXa inhibitor group and 0,26 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 2 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,53 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 0% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 26% and 26% for the LMWH group. And the difference for approach 3 would be 53% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ODIXa‐HIP 2007 study, for outcome Major Bleeding, there were 21 events over 463 participants (total randomized 479; missing outcome data 16) in the direct FXa inhibitor group and 0 events over 162 participants (total randomized 162; missing outcome data 0) in the LMWH group. The calculated absolute risk (baseline AR) is 4,54 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 4,59 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 4,59 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 4,59 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 5% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 5% and 0% for the LMWH group. And the difference for approach 3 would be 5% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ODIXa‐HIP2 2006 study, for outcome Major Bleeding, there were 15 events over 572 participants (total randomized 586; missing outcome data 14) in the direct FXa inhibitor group and 2 events over 132 participants (total randomized 136; missing outcome data 4) in the LMWH group. The calculated absolute risk (baseline AR) is 2,62 for the intervention and 1,52 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 2,73 for the direct FXa inhibitor group and 2,21 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 2,73 for the direct FXa inhibitor group and 2,21 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 2,73 for the direct FXa inhibitor group and 1,47 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 11% and 69% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 11% and 69% for the LMWH group. And the difference for approach 3 would be 11% for the direct FXa inhibitor group and ‐5% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ODIXa‐HIP‐OD 2006 study, for outcome Major Bleeding, there were 24 events over 668 participants (total randomized 713; missing outcome data 45) in the direct FXa inhibitor group and 3 events over 157 participants (total randomized 160; missing outcome data 3) in the LMWH group. The calculated absolute risk (baseline AR) is 3,59 for the intervention and 1,91 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 2 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 3,65 for the direct FXa inhibitor group and 2,5 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 3,51 for the direct FXa inhibitor group and 2,5 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 3 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 3,79 for the direct FXa inhibitor group and 1,88 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 6% and 59% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be ‐8% and 59% for the LMWH group. And the difference for approach 3 would be 20% for the direct FXa inhibitor group and ‐3% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ODIXa‐KNEE 2005 study, for outcome Major Bleeding, there were 14 events over 509 participants (total randomized 506; missing outcome data ‐3) in the direct FXa inhibitor group and 2 events over 104 participants (total randomized 105; missing outcome data 1) in the LMWH group. The calculated absolute risk (baseline AR) is 2,75 for the intervention and 1,92 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be ‐1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 2,57 for the direct FXa inhibitor group and 2,86 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be ‐1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 2,57 for the direct FXa inhibitor group and 2,86 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be ‐1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 2,57 for the direct FXa inhibitor group and 1,9 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be ‐18% and 94% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be ‐18% and 94% for the LMWH group. And the difference for approach 3 would be ‐18% for the direct FXa inhibitor group and ‐2% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ONYX‐1 2007 study, for outcome Major Bleeding, there were 0 events over 139 participants (total randomized 142; missing outcome data 3) in the direct FXa inhibitor group and 0 events over 36 participants (total randomized 36; missing outcome data 0) in the LMWH group. The calculated absolute risk (baseline AR) is 0 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,7 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,7 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 0% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 70% and 0% for the LMWH group. And the difference for approach 3 would be 70% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ONYX‐2 2010 study, for outcome Major Bleeding, there were 1 events over 794 participants (total randomized 849; missing outcome data 55) in the direct FXa inhibitor group and 1 events over 166 participants (total randomized 168; missing outcome data 2) in the LMWH group. The calculated absolute risk (baseline AR) is 0,13 for the intervention and 0,6 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,24 for the direct FXa inhibitor group and 1,19 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,24 for the direct FXa inhibitor group and 1,19 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 3 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,47 for the direct FXa inhibitor group and 0,6 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 11% and 59% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 11% and 59% for the LMWH group. And the difference for approach 3 would be 34% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the ONYX‐3 2014 study, for outcome Major Bleeding, there were 24 events over 1529 participants (total randomized 1587; missing outcome data 58) in the direct FXa inhibitor group and 8 events over 393 participants (total randomized 405; missing outcome data 12) in the LMWH group. The calculated absolute risk (baseline AR) is 1,57 for the intervention and 2,04 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 1,58 for the direct FXa inhibitor group and 2,22 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,58 for the direct FXa inhibitor group and 2,22 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 3 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 1,7 for the direct FXa inhibitor group and 1,98 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 1% and 18% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 1% and 18% for the LMWH group. And the difference for approach 3 would be 13% for the direct FXa inhibitor group and ‐6% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Raskob 2010 study, for outcome Major Bleeding, there were 5 events over 724 participants (total randomized 728; missing outcome data 4) in the direct FXa inhibitor group and 0 events over 172 participants (total randomized 175; missing outcome data 3) in the LMWH group. The calculated absolute risk (baseline AR) is 0,69 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,82 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,82 for the direct FXa inhibitor group and 0,57 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,82 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 13% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 13% and 57,% for the LMWH group. And the difference for approach 3 would be 13% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the RECORD1 2008 study, for outcome Major Bleeding, there were 6 events over 2209 participants (total randomized 2266; missing outcome data 57) in the direct FXa inhibitor group and 2 events over 2224 participants (total randomized 2275; missing outcome data 51) in the LMWH group. The calculated absolute risk (baseline AR) is 0,27 for the intervention and 0,09 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,31 for the direct FXa inhibitor group and 0,13 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,31 for the direct FXa inhibitor group and 0,13 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 3 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,4 for the direct FXa inhibitor group and 0,09 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 4% and 4% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 4% and 4% for the LMWH group. And the difference for approach 3 would be 13% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the RECORD2 2008 study, for outcome Major Bleeding, there were 1 events over 1228 participants (total randomized 1252; missing outcome data 24) in the direct FXa inhibitor group and 1 events over 1229 participants (total randomized 1257; missing outcome data 28) in the LMWH group. The calculated absolute risk (baseline AR) is 0,08 for the intervention and 0,08 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,16 for the direct FXa inhibitor group and 0,16 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,16 for the direct FXa inhibitor group and 0,16 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 2 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,24 for the direct FXa inhibitor group and 0,08 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 8% and 8% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 8% and 8% for the LMWH group. And the difference for approach 3 would be 16% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the RECORD3 2008 study, for outcome Major Bleeding, there were 7 events over 1220 participants (total randomized 1254; missing outcome data 34) in the direct FXa inhibitor group and 6 events over 1239 participants (total randomized 1277; missing outcome data 38) in the LMWH group. The calculated absolute risk (baseline AR) is 0,57 for the intervention and 0,48 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,64 for the direct FXa inhibitor group and 0,55 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,64 for the direct FXa inhibitor group and 0,55 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 2 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,72 for the direct FXa inhibitor group and 0,47 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 7% and 7% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 7% and 7% for the LMWH group. And the difference for approach 3 would be 15% for the direct FXa inhibitor group and ‐1% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the RECORD4 2009 study, for outcome Major Bleeding, there were 10 events over 1526 participants (total randomized 1584; missing outcome data 58) in the direct FXa inhibitor group and 4 events over 1508 participants (total randomized 1564; missing outcome data 56) in the LMWH group. The calculated absolute risk (baseline AR) is 0,66 for the intervention and 0,27 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,69 for the direct FXa inhibitor group and 0,32 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,69 for the direct FXa inhibitor group and 0,32 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 3 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,82 for the direct FXa inhibitor group and 0,26 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 3% and 5% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 3% and 5% for the LMWH group. And the difference for approach 3 would be 16% for the direct FXa inhibitor group and ‐1% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the STARS E‐3 2014 study, for outcome Major Bleeding, there were 4 events over 354 participants (total randomized 360; missing outcome data 6) in the direct FXa inhibitor group and 1 events over 349 participants (total randomized 356; missing outcome data 7) in the LMWH group. The calculated absolute risk (baseline AR) is 1,13 for the intervention and 0,29 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 1,39 for the direct FXa inhibitor group and 0,56 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,39 for the direct FXa inhibitor group and 0,56 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 1,39 for the direct FXa inhibitor group and 0,28 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 26% and 27% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 26% and 27% for the LMWH group. And the difference for approach 3 would be 26% for the direct FXa inhibitor group and ‐1% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the STARS J‐2 2014 study, for outcome Major Bleeding, there were 1 events over 174 participants (total randomized 175; missing outcome data 1) in the direct FXa inhibitor group and 0 events over 87 participants (total randomized 89; missing outcome data 2) in the LMWH group. The calculated absolute risk (baseline AR) is 0,57 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 1,14 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,14 for the direct FXa inhibitor group and 1,12 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 1,14 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 57,% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 57,% and 112,% for the LMWH group. And the difference for approach 3 would be 57,% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the STARS J‐4 2014 study, for outcome Major Bleeding, there were 1 events over 59 participants (total randomized 62; missing outcome data 3) in the direct FXa inhibitor group and 1 events over 29 participants (total randomized 30; missing outcome data 1) in the LMWH group. The calculated absolute risk (baseline AR) is 1,69 for the intervention and 3,45 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 3,23 for the direct FXa inhibitor group and 6,67 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 3,23 for the direct FXa inhibitor group and 6,67 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 3,23 for the direct FXa inhibitor group and 3,33 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 154% and 322% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 154% and 322% for the LMWH group. And the difference for approach 3 would be 154% for the direct FXa inhibitor group and ‐12% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the STARS J‐V 2015 study, for outcome Major Bleeding, there were 2 events over 303 participants (total randomized 307; missing outcome data 4) in the direct FXa inhibitor group and 6 events over 301 participants (total randomized 303; missing outcome data 2) in the LMWH group. The calculated absolute risk (baseline AR) is 0,66 for the intervention and 1,99 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,98 for the direct FXa inhibitor group and 2,31 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,98 for the direct FXa inhibitor group and 2,31 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,98 for the direct FXa inhibitor group and 1,98 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 32% and 32% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 32% and 32% for the LMWH group. And the difference for approach 3 would be 32% for the direct FXa inhibitor group and ‐1% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Weitz 2010 study, for outcome Major Bleeding, there were 4 events over 877 participants (total randomized 885; missing outcome data 8) in the direct FXa inhibitor group and 3 events over 161 participants (total randomized 163; missing outcome data 2) in the LMWH group. The calculated absolute risk (baseline AR) is 0,46 for the intervention and 1,86 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0,56 for the direct FXa inhibitor group and 2,45 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 1 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 0,56 for the direct FXa inhibitor group and 2,45 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 0,56 for the direct FXa inhibitor group and 1,84 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 10% and 59% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 10% and 59% for the LMWH group. And the difference for approach 3 would be 10% for the direct FXa inhibitor group and ‐2% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

For the Xie 2017 study, for outcome Major Bleeding, there were 0 events over 96 participants (total randomized 98; missing outcome data 2) in the direct FXa inhibitor group and 0 events over 98 participants (total randomized 98; missing outcome data 0) in the LMWH group. The calculated absolute risk (baseline AR) is 0 for the intervention and 0 for the comparator. The events in missing participants for approach 1 (assuming the missing outcome data for an safety outcome* have the same absolute risk (AR) of the original study for the intervention and the comparator) would be 0 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR1 including these new events plus the previous one and divided by all the randomized patients would be 0 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 2 (assuming the missing outcome data for the safety outcome, have the combined absolute risk of all the studies included in the meta‐analysis of this outcome for the intervention and the comparator) (see Analysis 1.4), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR2, including these new events plus the previous one and divided by all the randomized patients would be 1,02 for the direct FXa inhibitor group and 0 for the LMWH group. For approach 3 (the most extreme study within the meta‐analysis is ODIXa‐HIP 2007), there would be 1 new events for the direct FXa inhibitor group and 0 for the LMWH group, and the correspondig AR3 including these new events plus the previous one and divided by all the randomized patients would be 1,02 for the direct FXa inhibitor group and 0 for the LMWH group. The difference in comparison with the baseline AR for approach 1 for the direct FXa inhibitor group would be 0% and 0% for the LMWH group. The difference for approach 2 for the direct FXa inhibitor group would be 102% and 0% for the LMWH group. And the difference for approach 3 would be 102% for the direct FXa inhibitor group and 0% for the LMWH group. All the differences over 100% are considered large enough to modify the effect in the estimate for this outcome.

Appendix 7. Imputation of missing data

Imputation of missing data was carried out only for the total numbers of participants (N) when data on the number of events were reported separately for the periods of treatment and post‐treatment follow‐up. We established methods of data imputation for three different settings:

Case 1: N differs for the treatment and follow‐up periods (losses to follow‐up). N was imputed as the total number of participants for the treatment period.

Case 2: N is not reported for any period (but N is reported for efficacy or safety sets for other outcomes). N was imputed as the total number of participants reported for the efficacy or safety set according to the type of outcome (efficacy or safety outcome). When different total numbers of participants were reported for different outcomes for the corresponding efficacy/safety set, N was imputed as the bigger total number of participants.

Case 3: N is not reported for the same efficacy/safety set for treatment and follow‐up period. N was imputed as the total number of participants of the efficacy/safety set reported for the treatment period.

Data and analyses

Comparison 1. Direct factor Xa inhibitors versus LMWHs.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 All‐cause mortality	28	29698	Risk Ratio (M‐H, Fixed, 95% CI)	0.83 [0.52, 1.31]
1.2 Major VTE	28	24594	Risk Ratio (M‐H, Random, 95% CI)	0.51 [0.37, 0.71]
1.3 Symptomatic VTE	33	31670	Risk Ratio (M‐H, Fixed, 95% CI)	0.64 [0.50, 0.83]
1.4 Major bleeding	36	39778	Risk Ratio (M‐H, Fixed, 95% CI)	1.05 [0.86, 1.30]
1.5 Major bleeding: rivaroxaban versus LMWH	17	17630	Risk Ratio (M‐H, Fixed, 95% CI)	1.94 [1.26, 2.98]
1.6 Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH	19	22148	Risk Ratio (M‐H, Fixed, 95% CI)	0.80 [0.63, 1.02]
1.7 Serious hepatic adverse events	2	3169	Risk Ratio (M‐H, Fixed, 95% CI)	3.01 [0.12, 73.93]
1.8 Serious non‐hepatic adverse events	15	26246	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.81, 0.97]
1.9 Fatal VTE	33	27183	Risk Ratio (M‐H, Fixed, 95% CI)	1.19 [0.51, 2.79]
1.10 Asymptomatic distal DVT	14	4818	Risk Ratio (M‐H, Random, 95% CI)	0.64 [0.46, 0.89]
1.11 Fatal bleeding	26	31293	Risk Ratio (M‐H, Fixed, 95% CI)	1.42 [0.33, 6.04]
1.12 Liver enzymes elevation	21	31408	Risk Ratio (M‐H, Random, 95% CI)	0.58 [0.46, 0.73]
1.13 Minor adverse events	4	1011	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.83, 1.00]
1.14 Volume of blood loss	14	7379	Mean Difference (IV, Random, 95% CI)	0.57 [‐29.58, 30.73]

Comparison 2. Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of drug.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 All‐cause mortality	16	26118	Risk Ratio (M‐H, Fixed, 95% CI)	0.83 [0.52, 1.31]
2.1.1 Apixaban	5	12502	Risk Ratio (M‐H, Fixed, 95% CI)	1.20 [0.54, 2.68]
2.1.2 Darexaban	2	978	Risk Ratio (M‐H, Fixed, 95% CI)	1.04 [0.11, 9.59]
2.1.3 Edoxaban	1	896	Risk Ratio (M‐H, Fixed, 95% CI)	2.15 [0.12, 39.70]
2.1.4 Letaxaban	1	730	Risk Ratio (M‐H, Fixed, 95% CI)	0.53 [0.02, 12.94]
2.1.5 Rivaroxaban	7	11012	Risk Ratio (M‐H, Fixed, 95% CI)	0.64 [0.34, 1.17]
2.2 Major VTE	24	23983	Risk Ratio (M‐H, Random, 95% CI)	0.51 [0.37, 0.71]
2.2.1 Apixaban	6	7910	Risk Ratio (M‐H, Random, 95% CI)	0.47 [0.31, 0.71]
2.2.2 Darexaban	5	2992	Risk Ratio (M‐H, Random, 95% CI)	0.94 [0.51, 1.73]
2.2.3 Edoxaban	1	774	Risk Ratio (M‐H, Random, 95% CI)	0.24 [0.13, 0.43]
2.2.4 LY517717	1	378	Risk Ratio (M‐H, Random, 95% CI)	1.25 [0.38, 4.11]
2.2.5 Rivaroxaban	11	11929	Risk Ratio (M‐H, Random, 95% CI)	0.44 [0.25, 0.78]
2.3 Symptomatic VTE	25	29981	Risk Ratio (M‐H, Fixed, 95% CI)	0.64 [0.50, 0.82]
2.3.1 Apixaban	5	11914	Risk Ratio (M‐H, Fixed, 95% CI)	1.00 [0.61, 1.62]
2.3.2 Betrixaban	1	175	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.10, 8.31]
2.3.3 Darexaban	4	1415	Risk Ratio (M‐H, Fixed, 95% CI)	0.78 [0.22, 2.81]
2.3.4 Edoxaban	1	594	Risk Ratio (M‐H, Fixed, 95% CI)	3.95 [0.44, 35.10]
2.3.5 Letaxaban	1	730	Risk Ratio (M‐H, Fixed, 95% CI)	0.61 [0.21, 1.83]
2.3.6 Rivaroxaban	13	15153	Risk Ratio (M‐H, Fixed, 95% CI)	0.49 [0.35, 0.68]
2.4 Major bleeding	29	37216	Risk Ratio (M‐H, Fixed, 95% CI)	1.05 [0.86, 1.30]
2.4.1 Apixaban	5	12687	Risk Ratio (M‐H, Fixed, 95% CI)	0.84 [0.64, 1.11]
2.4.2 Darexaban	3	3152	Risk Ratio (M‐H, Fixed, 95% CI)	0.74 [0.36, 1.54]
2.4.3 Edoxaban	5	2552	Risk Ratio (M‐H, Fixed, 95% CI)	0.99 [0.40, 2.43]
2.4.4 Erixaban	1	1389	Risk Ratio (M‐H, Fixed, 95% CI)	0.67 [0.16, 2.78]
2.4.5 Letaxaban	1	1038	Risk Ratio (M‐H, Fixed, 95% CI)	0.24 [0.06, 1.08]
2.4.6 LY517717	1	507	Risk Ratio (M‐H, Fixed, 95% CI)	0.43 [0.04, 4.71]
2.4.7 Rivaroxaban	13	15891	Risk Ratio (M‐H, Fixed, 95% CI)	1.94 [1.26, 2.98]
2.5 Major bleeding: rivaroxaban versus LMWH	13	15891	Risk Ratio (M‐H, Fixed, 95% CI)	1.94 [1.26, 2.98]
2.6 Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH	16	21325	Risk Ratio (M‐H, Fixed, 95% CI)	0.80 [0.63, 1.02]
2.7 Serious non‐hepatic adverse events	15	26246	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.81, 0.97]
2.7.1 Apixaban	4	12591	Risk Ratio (M‐H, Fixed, 95% CI)	0.99 [0.87, 1.13]
2.7.2 Darexaban	2	716	Risk Ratio (M‐H, Fixed, 95% CI)	0.56 [0.18, 1.68]
2.7.3 Edoxaban	5	2552	Risk Ratio (M‐H, Fixed, 95% CI)	1.04 [0.63, 1.69]
2.7.4 Letaxaban	1	1038	Risk Ratio (M‐H, Fixed, 95% CI)	0.76 [0.45, 1.27]
2.7.5 Rivaroxaban	3	9349	Risk Ratio (M‐H, Fixed, 95% CI)	0.78 [0.68, 0.89]
2.8 Fatal VTE	10	18372	Risk Ratio (M‐H, Fixed, 95% CI)	1.19 [0.51, 2.79]
2.8.1 Apixaban	3	9211	Risk Ratio (M‐H, Fixed, 95% CI)	2.39 [0.45, 12.76]
2.8.2 Darexaban	1	218	Risk Ratio (M‐H, Fixed, 95% CI)	1.31 [0.05, 31.84]
2.8.3 Rivaroxaban	6	8943	Risk Ratio (M‐H, Fixed, 95% CI)	0.86 [0.29, 2.55]
2.9 Asymptomatic distal DVT	12	4602	Risk Ratio (M‐H, Random, 95% CI)	0.64 [0.46, 0.89]
2.9.1 Apixaban	1	159	Risk Ratio (M‐H, Random, 95% CI)	0.50 [0.26, 0.98]
2.9.2 Betrixaban	1	175	Risk Ratio (M‐H, Random, 95% CI)	2.67 [0.65, 11.01]
2.9.3 Darexaban	2	560	Risk Ratio (M‐H, Random, 95% CI)	0.86 [0.47, 1.54]
2.9.4 Edoxaban	4	1394	Risk Ratio (M‐H, Random, 95% CI)	0.47 [0.31, 0.73]
2.9.5 Rivaroxaban	4	2314	Risk Ratio (M‐H, Random, 95% CI)	0.38 [0.12, 1.23]
2.10 Liver enzymes elevation	21	31408	Risk Ratio (M‐H, Random, 95% CI)	0.58 [0.46, 0.73]
2.10.1 Apixaban	4	12484	Risk Ratio (M‐H, Random, 95% CI)	0.81 [0.61, 1.07]
2.10.2 Betrixaban	1	214	Risk Ratio (M‐H, Random, 95% CI)	0.06 [0.01, 0.55]
2.10.3 Darexaban	2	710	Risk Ratio (M‐H, Random, 95% CI)	0.47 [0.21, 1.04]
2.10.4 Edoxaban	5	2552	Risk Ratio (M‐H, Random, 95% CI)	0.29 [0.11, 0.79]
2.10.5 Letaxaban	1	1038	Risk Ratio (M‐H, Random, 95% CI)	0.73 [0.30, 1.77]
2.10.6 Rivaroxaban	7	13903	Risk Ratio (M‐H, Random, 95% CI)	0.62 [0.47, 0.81]
2.10.7 LY517717	1	507	Risk Ratio (M‐H, Random, 95% CI)	0.55 [0.28, 1.06]
2.11 Volume of blood loss	14	7379	Mean Difference (IV, Random, 95% CI)	0.57 [‐29.58, 30.73]
2.11.1 Apixaban	1	220	Mean Difference (IV, Random, 95% CI)	‐82.50 [‐88.72, ‐76.28]
2.11.2 Letaxaban	1	1038	Mean Difference (IV, Random, 95% CI)	107.10 [58.08, 156.12]
2.11.3 Rivaroxaban	12	6121	Mean Difference (IV, Random, 95% CI)	‐5.08 [‐20.15, 9.99]

Comparison 3. Direct factor Xa inhibitors versus LMWH: subgroup analysis by duration of prophylaxis.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 All‐cause mortality	15	24375	Risk Ratio (M‐H, Fixed, 95% CI)	0.97 [0.59, 1.58]
3.1.1 Short prophylaxis (≤ 14 days)	11	15011	Risk Ratio (M‐H, Fixed, 95% CI)	0.81 [0.45, 1.47]
3.1.2 Extended prophylaxis (> 14 days)	4	9364	Risk Ratio (M‐H, Fixed, 95% CI)	1.46 [0.59, 3.59]
3.2 Major VTE	24	23983	Risk Ratio (M‐H, Random, 95% CI)	0.51 [0.37, 0.71]
3.2.1 Short prophylaxis (≤ 14 days)	16	11792	Risk Ratio (M‐H, Random, 95% CI)	0.59 [0.42, 0.82]
3.2.2 Extended prophylaxis (> 14 days)	8	12191	Risk Ratio (M‐H, Random, 95% CI)	0.33 [0.15, 0.71]
3.3 Symptomatic VTE	22	27035	Risk Ratio (M‐H, Fixed, 95% CI)	0.69 [0.53, 0.89]
3.3.1 Short prophylaxis (≤ 14 days)	17	16157	Risk Ratio (M‐H, Fixed, 95% CI)	0.72 [0.54, 0.97]
3.3.2 Extended prophylaxis (> 14 days)	5	10878	Risk Ratio (M‐H, Fixed, 95% CI)	0.56 [0.31, 1.02]
3.4 Major bleeding	27	34639	Risk Ratio (M‐H, Fixed, 95% CI)	1.06 [0.86, 1.31]
3.4.1 Short prophylaxis (≤ 14 days)	20	21683	Risk Ratio (M‐H, Fixed, 95% CI)	1.15 [0.85, 1.55]
3.4.2 Extended prophylaxis (> 14 days)	7	12956	Risk Ratio (M‐H, Fixed, 95% CI)	0.95 [0.71, 1.27]
3.5 Major bleeding: rivaroxaban versus LMWH	11	13314	Risk Ratio (M‐H, Fixed, 95% CI)	2.05 [1.32, 3.18]
3.5.1 Short prophylaxis (≤ 14 days) rivaroxaban	8	8668	Risk Ratio (M‐H, Fixed, 95% CI)	2.14 [1.32, 3.46]
3.5.2 Extended prophylaxis (> 14 days) rivaroxaban	3	4646	Risk Ratio (M‐H, Fixed, 95% CI)	1.58 [0.52, 4.77]
3.6 Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH	19	34281	Risk Ratio (M‐H, Fixed, 95% CI)	0.86 [0.71, 1.03]
3.6.1 Short prophylaxis (≤ 14 days) Other DFXa	12	13015	Risk Ratio (M‐H, Fixed, 95% CI)	0.70 [0.47, 1.04]
3.6.2 Extended prophylaxis (> 14 days) Other DFXa	7	21266	Risk Ratio (M‐H, Fixed, 95% CI)	0.93 [0.76, 1.14]
3.7 Serious non‐hepatic adverse events	14	23789	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.83, 1.01]
3.7.1 Short prophylaxis (≤ 14 days)	12	14024	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.79, 1.03]
3.7.2 Extended prophylaxis (> 14 days)	2	9765	Risk Ratio (M‐H, Fixed, 95% CI)	0.94 [0.81, 1.08]
3.8 Asymptomatic distal DVT	11	4496	Risk Ratio (M‐H, Random, 95% CI)	0.66 [0.48, 0.90]
3.8.1 Short prophylaxis (≤ 14 days)	10	4302	Risk Ratio (M‐H, Random, 95% CI)	0.67 [0.49, 0.91]
3.8.2 Extended prophylaxis (> 14 days)	1	194	Risk Ratio (M‐H, Random, 95% CI)	0.11 [0.01, 2.08]
3.9 Liver enzymes elevation	20	29077	Risk Ratio (M‐H, Random, 95% CI)	0.60 [0.47, 0.76]
3.9.1 Short prophylaxis (≤ 14 days)	18	19575	Risk Ratio (M‐H, Random, 95% CI)	0.55 [0.41, 0.74]
3.9.2 Extended prophylaxis (> 14 days)	2	9502	Risk Ratio (M‐H, Random, 95% CI)	0.78 [0.58, 1.04]
3.10 Volume of blood loss	11	6493	Mean Difference (IV, Random, 95% CI)	4.17 [‐32.39, 40.74]
3.10.1 Short prophylaxis (≤ 14 days)	7	5912	Mean Difference (IV, Random, 95% CI)	25.75 [‐16.18, 67.68]
3.10.2 Extended prophylaxis (> 14 days)	4	581	Mean Difference (IV, Random, 95% CI)	‐24.22 [‐76.83, 28.39]

Comparison 4. Direct factor Xa inhibitors versus LMWH: subgroup analysis by timing of start of prophylaxis (comparator).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
4.1 All‐cause mortality	16	26108	Risk Ratio (M‐H, Fixed, 95% CI)	0.83 [0.52, 1.31]
4.1.1 Start of comparator before surgery	8	16774	Risk Ratio (M‐H, Fixed, 95% CI)	0.76 [0.41, 1.41]
4.1.2 Start of comparator after surgery	8	9334	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.46, 1.83]
4.2 Major VTE	21	23297	Risk Ratio (M‐H, Random, 95% CI)	0.50 [0.35, 0.71]
4.2.1 Start of comparator before surgery	12	18309	Risk Ratio (M‐H, Random, 95% CI)	0.50 [0.31, 0.82]
4.2.2 Start of comparator after surgery	9	4988	Risk Ratio (M‐H, Random, 95% CI)	0.49 [0.29, 0.83]
4.3 Symptomatic VTE	22	29295	Risk Ratio (M‐H, Fixed, 95% CI)	0.63 [0.49, 0.82]
4.3.1 Start of comparator before surgery	9	19639	Risk Ratio (M‐H, Fixed, 95% CI)	0.50 [0.34, 0.72]
4.3.2 Start of comparator after surgery	13	9656	Risk Ratio (M‐H, Fixed, 95% CI)	0.80 [0.56, 1.13]
4.4 Major bleeding	29	37216	Risk Ratio (M‐H, Fixed, 95% CI)	1.05 [0.86, 1.30]
4.4.1 Start of comparator before surgery	11	23253	Risk Ratio (M‐H, Fixed, 95% CI)	1.24 [0.89, 1.72]
4.4.2 Start of comparator after surgery	18	13963	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.70, 1.20]
4.5 Major bleeding: rivaroxaban versus LMWH	13	15891	Risk Ratio (M‐H, Fixed, 95% CI)	1.94 [1.26, 2.98]
4.5.1 Start of comparator before surgery	6	11523	Risk Ratio (M‐H, Fixed, 95% CI)	2.23 [1.23, 4.06]
4.5.2 Start of comparator after surgery	7	4368	Risk Ratio (M‐H, Fixed, 95% CI)	1.62 [0.88, 3.00]
4.6 Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH	16	21325	Risk Ratio (M‐H, Fixed, 95% CI)	0.80 [0.63, 1.02]
4.6.1 Start of comparator before surgery	5	11730	Risk Ratio (M‐H, Fixed, 95% CI)	0.87 [0.58, 1.31]
4.6.2 Start of comparator after surgery	11	9595	Risk Ratio (M‐H, Fixed, 95% CI)	0.76 [0.57, 1.03]
4.7 Serious non‐hepatic adverse events	15	26246	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.81, 0.97]
4.7.1 Start of comparator before surgery	5	17690	Risk Ratio (M‐H, Fixed, 95% CI)	0.86 [0.77, 0.95]
4.7.2 Start of comparator after surgery	10	8556	Risk Ratio (M‐H, Fixed, 95% CI)	0.98 [0.82, 1.17]
4.8 Fatal VTE	10	18372	Risk Ratio (M‐H, Fixed, 95% CI)	1.19 [0.51, 2.79]
4.8.1 Start of comparator before surgery	5	13816	Risk Ratio (M‐H, Fixed, 95% CI)	1.36 [0.44, 4.18]
4.8.2 Start of comparator after surgery	5	4556	Risk Ratio (M‐H, Fixed, 95% CI)	0.99 [0.26, 3.69]
4.9 Asymptomatic distal DVT	11	4438	Risk Ratio (M‐H, Random, 95% CI)	0.67 [0.46, 0.96]
4.9.1 Start of comparator before surgery	1	90	Risk Ratio (M‐H, Random, 95% CI)	0.25 [0.03, 2.15]
4.9.2 Start of comparator after surgery	10	4348	Risk Ratio (M‐H, Random, 95% CI)	0.69 [0.47, 1.00]
4.10 Liver enzymes elevation	21	31408	Risk Ratio (M‐H, Random, 95% CI)	0.58 [0.46, 0.73]
4.10.1 Start of comparator before surgery	8	19041	Risk Ratio (M‐H, Random, 95% CI)	0.72 [0.56, 0.92]
4.10.2 Start of comparator after surgery	13	12367	Risk Ratio (M‐H, Random, 95% CI)	0.43 [0.29, 0.64]
4.11 Volume of blood loss	14	7379	Mean Difference (IV, Random, 95% CI)	0.57 [‐29.58, 30.73]
4.11.1 Start of comparator before surgery	5	2276	Mean Difference (IV, Random, 95% CI)	15.51 [‐33.95, 64.97]
4.11.2 Start of comparator after surgery	9	5103	Mean Difference (IV, Random, 95% CI)	‐4.33 [‐40.84, 32.17]

Comparison 5. Direct factor Xa inhibitors versus LMWH: subgroup analysis by frequency of administration.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
5.1 All‐cause mortality	17	26323	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.57, 1.38]
5.1.1 Once daily administration	9	12163	Risk Ratio (M‐H, Fixed, 95% CI)	0.69 [0.39, 1.24]
5.1.2 Twice daily administration	10	14160	Risk Ratio (M‐H, Fixed, 95% CI)	1.26 [0.63, 2.53]
5.2 Major VTE	24	24347	Risk Ratio (M‐H, Random, 95% CI)	0.52 [0.38, 0.70]
5.2.1 Once daily administration	15	14031	Risk Ratio (M‐H, Random, 95% CI)	0.44 [0.27, 0.71]
5.2.2 Twice daily administration	11	10316	Risk Ratio (M‐H, Random, 95% CI)	0.62 [0.45, 0.85]
5.3 Symptomatic VTE	25	30017	Risk Ratio (M‐H, Fixed, 95% CI)	0.64 [0.50, 0.82]
5.3.1 Once daily administration	15	16106	Risk Ratio (M‐H, Fixed, 95% CI)	0.51 [0.37, 0.70]
5.3.2 Twice daily administration	11	13911	Risk Ratio (M‐H, Fixed, 95% CI)	0.93 [0.62, 1.40]
5.4 Major bleeding	29	38081	Risk Ratio (M‐H, Fixed, 95% CI)	1.07 [0.88, 1.31]
5.4.1 Once daily administration	22	21857	Risk Ratio (M‐H, Fixed, 95% CI)	1.29 [0.93, 1.81]
5.4.2 Twice daily administration	11	16224	Risk Ratio (M‐H, Fixed, 95% CI)	0.95 [0.74, 1.22]
5.5 Major bleeding: rivaroxaban versus LMWH	13	16053	Risk Ratio (M‐H, Fixed, 95% CI)	2.07 [1.37, 3.15]
5.5.1 Once daily administration	11	14199	Risk Ratio (M‐H, Fixed, 95% CI)	1.86 [1.17, 2.96]
5.5.2 Twice daily administration	3	1854	Risk Ratio (M‐H, Fixed, 95% CI)	2.88 [1.14, 7.29]
5.6 Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH	16	22028	Risk Ratio (M‐H, Fixed, 95% CI)	0.81 [0.65, 1.03]
5.6.1 Once daily administration	11	7658	Risk Ratio (M‐H, Fixed, 95% CI)	0.82 [0.50, 1.36]
5.6.2 Twice daily administration	8	14370	Risk Ratio (M‐H, Fixed, 95% CI)	0.81 [0.63, 1.05]
5.7 Serious non‐hepatic adverse events	15	26556	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.81, 0.97]
5.7.1 Once daily administration	10	12990	Risk Ratio (M‐H, Fixed, 95% CI)	0.82 [0.72, 0.93]
5.7.2 Twice daily administration	7	13566	Risk Ratio (M‐H, Fixed, 95% CI)	0.96 [0.85, 1.09]
5.8 Fatal VTE	10	17987	Risk Ratio (M‐H, Fixed, 95% CI)	1.26 [0.53, 2.96]
5.8.1 Once daily administration	4	8111	Risk Ratio (M‐H, Fixed, 95% CI)	0.81 [0.23, 2.82]
5.8.2 Twice daily administration	6	9876	Risk Ratio (M‐H, Fixed, 95% CI)	1.87 [0.55, 6.39]
5.9 Asymptomatic distal DVT	12	4602	Risk Ratio (M‐H, Random, 95% CI)	0.64 [0.46, 0.89]
5.9.1 Once daily administration	8	3708	Risk Ratio (M‐H, Random, 95% CI)	0.55 [0.36, 0.84]
5.9.2 Twice daily administration	4	894	Risk Ratio (M‐H, Random, 95% CI)	0.85 [0.46, 1.54]
5.10 Liver enzymes elevation	21	31630	Risk Ratio (M‐H, Random, 95% CI)	0.58 [0.46, 0.72]
5.10.1 Once daily administration	14	17433	Risk Ratio (M‐H, Random, 95% CI)	0.52 [0.38, 0.71]
5.10.2 Twice daily administration	9	14197	Risk Ratio (M‐H, Random, 95% CI)	0.71 [0.53, 0.95]
5.11 Volume of blood loss	12	5744	Mean Difference (IV, Random, 95% CI)	‐14.33 [‐44.60, 15.93]
5.11.1 Once daily administration	9	4207	Mean Difference (IV, Random, 95% CI)	‐8.48 [‐23.08, 6.12]
5.11.2 Twice daily administration	3	1537	Mean Difference (IV, Random, 95% CI)	‐63.08 [‐117.35, ‐8.80]

Comparison 6. Direct factor Xa inhibitors versus LMWH: subgroup analysis by type of surgery.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
6.1 All‐cause mortality	16	26108	Risk Ratio (M‐H, Fixed, 95% CI)	0.83 [0.52, 1.31]
6.1.1 Total hip replacement surgery only	7	12911	Risk Ratio (M‐H, Fixed, 95% CI)	0.95 [0.48, 1.88]
6.1.2 Total knee replacement surgery only	8	13101	Risk Ratio (M‐H, Fixed, 95% CI)	0.72 [0.38, 1.37]
6.1.3 Total hip or knee replacement surgery	1	96	Risk Ratio (M‐H, Fixed, 95% CI)	1.00 [0.06, 15.53]
6.2 Major VTE	24	23983	Risk Ratio (M‐H, Random, 95% CI)	0.51 [0.36, 0.71]
6.2.1 Total hip replacement surgery only	13	15212	Risk Ratio (M‐H, Random, 95% CI)	0.44 [0.25, 0.75]
6.2.2 Total knee replacement surgery only	9	8297	Risk Ratio (M‐H, Random, 95% CI)	0.53 [0.38, 0.75]
6.2.3 Total hip or knee replacement surgery	2	474	Risk Ratio (M‐H, Random, 95% CI)	1.37 [0.47, 3.98]
6.3 Symptomatic VTE	25	29981	Risk Ratio (M‐H, Fixed, 95% CI)	0.64 [0.50, 0.83]
6.3.1 Total hip replacement surgery only	11	15252	Risk Ratio (M‐H, Fixed, 95% CI)	0.43 [0.27, 0.67]
6.3.2 Total knee replacement surgery only	12	14365	Risk Ratio (M‐H, Fixed, 95% CI)	0.78 [0.57, 1.07]
6.3.3 Total hip or knee replacement surgery	2	364	Risk Ratio (M‐H, Fixed, 95% CI)	0.99 [0.25, 3.90]
6.4 Major bleeding	29	37216	Risk Ratio (M‐H, Fixed, 95% CI)	1.05 [0.86, 1.30]
6.4.1 Total hip replacement surgery only	13	19427	Risk Ratio (M‐H, Fixed, 95% CI)	1.41 [1.00, 2.00]
6.4.2 Total knee replacement surgery only	12	16978	Risk Ratio (M‐H, Fixed, 95% CI)	0.90 [0.63, 1.29]
6.4.3 Hip fracture surgery only	1	88	Risk Ratio (M‐H, Fixed, 95% CI)	0.49 [0.03, 7.58]
6.4.4 Total hip or knee replacement surgery	3	723	Risk Ratio (M‐H, Fixed, 95% CI)	0.77 [0.56, 1.08]
6.5 Major bleeding: rivaroxaban versus LMWH	13	15891	Risk Ratio (M‐H, Fixed, 95% CI)	1.94 [1.26, 2.98]
6.5.1 Total hip replacement surgery only	7	9452	Risk Ratio (M‐H, Fixed, 95% CI)	2.58 [1.41, 4.73]
6.5.2 Total knee replacement surgery only	5	6319	Risk Ratio (M‐H, Fixed, 95% CI)	1.45 [0.77, 2.73]
6.5.3 Total hip or knee replacement surgery	1	120	Risk Ratio (M‐H, Fixed, 95% CI)	0.33 [0.01, 8.02]
6.6 Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH	16	21325	Risk Ratio (M‐H, Fixed, 95% CI)	0.80 [0.63, 1.02]
6.6.1 Total hip replacement surgery only	6	9975	Risk Ratio (M‐H, Fixed, 95% CI)	0.93 [0.60, 1.45]
6.6.2 Hip fracture surgery only	1	88	Risk Ratio (M‐H, Fixed, 95% CI)	0.49 [0.03, 7.58]
6.6.3 Total knee replacement surgery only	7	10659	Risk Ratio (M‐H, Fixed, 95% CI)	0.72 [0.46, 1.11]
6.6.4 Total hip or knee replacement surgery	2	603	Risk Ratio (M‐H, Fixed, 95% CI)	0.79 [0.57, 1.10]
6.7 Serious non‐hepatic adverse events	15	26246	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.81, 0.97]
6.7.1 Total hip replacement surgery only	7	14429	Risk Ratio (M‐H, Fixed, 95% CI)	0.88 [0.77, 0.99]
6.7.2 Total knee replacement surgery only	7	11729	Risk Ratio (M‐H, Fixed, 95% CI)	0.90 [0.79, 1.03]
6.7.3 Hip fracture surgery only	1	88	Risk Ratio (M‐H, Fixed, 95% CI)	0.49 [0.11, 2.29]
6.8 Fatal VTE	10	18372	Risk Ratio (M‐H, Fixed, 95% CI)	1.19 [0.51, 2.79]
6.8.1 Total hip replacement surgery only	4	10759	Risk Ratio (M‐H, Fixed, 95% CI)	0.98 [0.27, 3.49]
6.8.2 Total knee replacement surgery only	5	7422	Risk Ratio (M‐H, Fixed, 95% CI)	1.88 [0.50, 6.99]
6.8.3 Hip fracture surgery only	1	191	Risk Ratio (M‐H, Fixed, 95% CI)	0.33 [0.01, 8.00]
6.9 Asymptomatic distal DVT	12	4602	Risk Ratio (M‐H, Random, 95% CI)	0.64 [0.46, 0.89]
6.9.1 Total hip replacement surgery only	4	1183	Risk Ratio (M‐H, Random, 95% CI)	0.53 [0.24, 1.15]
6.9.2 Total knee replacement surgery only	7	3346	Risk Ratio (M‐H, Random, 95% CI)	0.65 [0.45, 0.96]
6.9.3 Hip fracture surgery only	1	73	Risk Ratio (M‐H, Random, 95% CI)	1.76 [0.19, 16.10]
6.10 Liver enzymes elevation	21	31408	Risk Ratio (M‐H, Random, 95% CI)	0.58 [0.46, 0.73]
6.10.1 Total hip replacement surgery only	10	16122	Risk Ratio (M‐H, Random, 95% CI)	0.59 [0.44, 0.80]
6.10.2 Total knee replacement surgery only	9	14691	Risk Ratio (M‐H, Random, 95% CI)	0.55 [0.35, 0.86]
6.10.3 Hip fracture surgery only	1	88	Risk Ratio (M‐H, Random, 95% CI)	0.06 [0.00, 1.00]
6.10.4 Total hip or knee replacement surgery	1	507	Risk Ratio (M‐H, Random, 95% CI)	0.55 [0.28, 1.06]
6.11 Volume of blood loss	14	7379	Mean Difference (IV, Random, 95% CI)	0.57 [‐29.58, 30.73]
6.11.1 Total hip replacement surgery only	6	2342	Mean Difference (IV, Random, 95% CI)	1.18 [‐19.12, 21.49]
6.11.2 Total knee replacement surgery only	6	4106	Mean Difference (IV, Random, 95% CI)	‐13.92 [‐59.21, 31.38]
6.11.3 Hip fracture surgery only	1	191	Mean Difference (IV, Random, 95% CI)	55.50 [12.20, 98.80]
6.11.4 Total hip or knee replacement surgery	1	740	Mean Difference (IV, Random, 95% CI)	‐5.00 [‐33.25, 23.25]

Comparison 7. Direct factor Xa inhibitors by type versus LMWH: subgroup analysis by daily all approved or low dose groups.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
7.1 All‐cause mortality	13	23682	Risk Ratio (M‐H, Fixed, 95% CI)	0.88 [0.55, 1.41]
7.1.1 Apixaban dose (2.5 to 20 mg/day)	5	12502	Risk Ratio (M‐H, Fixed, 95% CI)	1.20 [0.54, 2.68]
7.1.2 Edoxaban dose (30 to 60 mg/day)	1	527	Risk Ratio (M‐H, Fixed, 95% CI)	3.40 [0.18, 65.49]
7.1.3 Rivaroxaban dose (5 to 30 mg/day)	7	10653	Risk Ratio (M‐H, Fixed, 95% CI)	0.68 [0.37, 1.25]
7.2 Major VTE	24	21871	Risk Ratio (M‐H, Random, 95% CI)	0.54 [0.37, 0.78]
7.2.1 Apixaban dose (2.5 to 20 mg/day)	5	7790	Risk Ratio (M‐H, Random, 95% CI)	0.47 [0.31, 0.71]
7.2.2 Darexaban low dose (≤ 40 mg/day)	5	1908	Risk Ratio (M‐H, Random, 95% CI)	1.32 [0.61, 2.85]
7.2.3 Edoxaban dose (30 to 60 mg/day)	1	453	Risk Ratio (M‐H, Random, 95% CI)	0.19 [0.08, 0.41]
7.2.4 LY517717 low dose (≤ 50 mg/day)	1	114	Risk Ratio (M‐H, Random, 95% CI)	1.33 [0.28, 6.29]
7.2.5 Rivaroxaban dose (5 to 30 mg/day)	12	11606	Risk Ratio (M‐H, Random, 95% CI)	0.44 [0.25, 0.80]
7.3 Symptomatic VTE	24	28635	Risk Ratio (M‐H, Fixed, 95% CI)	0.65 [0.50, 0.84]
7.3.1 Apixaban dose (2.5 to 20 mg/day)	5	11914	Risk Ratio (M‐H, Fixed, 95% CI)	1.03 [0.63, 1.67]
7.3.2 Betrixaban dose (40 to 160 mg/day)	1	105	Risk Ratio (M‐H, Fixed, 95% CI)	0.62 [0.04, 9.57]
7.3.3 Darexaban low dose (≤ 40 mg/day)	3	740	Risk Ratio (M‐H, Fixed, 95% CI)	0.82 [0.18, 3.81]
7.3.4 Edoxaban dose (30 to 60 mg/day)	1	594	Risk Ratio (M‐H, Fixed, 95% CI)	3.95 [0.44, 35.10]
7.3.5 Letaxaban low dose (≤ 40 mg/day)	1	387	Risk Ratio (M‐H, Fixed, 95% CI)	0.49 [0.13, 1.79]
7.3.6 Rivaroxaban dose (5 to 30 mg/day)	13	14895	Risk Ratio (M‐H, Fixed, 95% CI)	0.50 [0.36, 0.70]
7.4 Major bleeding	28	33695	Risk Ratio (M‐H, Fixed, 95% CI)	0.97 [0.78, 1.19]
7.4.1 Apixaban dose (2.5 to 20 mg/day)	5	12687	Risk Ratio (M‐H, Fixed, 95% CI)	0.84 [0.64, 1.11]
7.4.2 Darexaban low dose (≤ 40 mg/day)	2	1791	Risk Ratio (M‐H, Fixed, 95% CI)	0.45 [0.18, 1.11]
7.4.3 Edoxaban dose (30 to 60 mg/day)	5	2094	Risk Ratio (M‐H, Fixed, 95% CI)	1.03 [0.42, 2.50]
7.4.4 Erixaban low dose (≤ 2.5 mg/day)	1	1184	Risk Ratio (M‐H, Fixed, 95% CI)	0.67 [0.15, 2.99]
7.4.5 Letaxaban low dose (≤ 40 mg/day)	1	555	Risk Ratio (M‐H, Fixed, 95% CI)	0.14 [0.01, 1.30]
7.4.6 LY517717 low dose (≤ 50 mg/day)	1	156	Risk Ratio (M‐H, Fixed, 95% CI)	0.45 [0.02, 10.94]
7.4.7 Rivaroxaban dose (5 to 30 mg/day)	13	15228	Risk Ratio (M‐H, Fixed, 95% CI)	1.63 [1.06, 2.51]
7.5 Serious non‐hepatic adverse events	15	25043	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.81, 0.97]
7.5.1 Apixaban dose (2.5 to 20 mg/day)	4	12591	Risk Ratio (M‐H, Fixed, 95% CI)	0.99 [0.87, 1.13]
7.5.2 Darexaban low dose (≤ 40 mg/day)	2	454	Risk Ratio (M‐H, Fixed, 95% CI)	0.44 [0.11, 1.80]
7.5.3 Edoxaban dose (30 to 60 mg/day)	5	2094	Risk Ratio (M‐H, Fixed, 95% CI)	0.97 [0.59, 1.61]
7.5.4 Letaxaban low dose (≤ 40 mg/day)	1	555	Risk Ratio (M‐H, Fixed, 95% CI)	0.84 [0.48, 1.49]
7.5.5 Rivaroxaban dose (5 to 30 mg/day)	3	9349	Risk Ratio (M‐H, Fixed, 95% CI)	0.78 [0.68, 0.89]
7.6 Fatal VTE	9	17933	Risk Ratio (M‐H, Fixed, 95% CI)	1.27 [0.53, 3.09]
7.6.1 Apixaban dose (2.5 to 20 mg/day)	3	9211	Risk Ratio (M‐H, Fixed, 95% CI)	2.39 [0.45, 12.76]
7.6.2 Rivaroxaban dose (5 to 30 mg/day)	6	8722	Risk Ratio (M‐H, Fixed, 95% CI)	0.98 [0.33, 2.85]
7.7 Asymptomatic distal DVT	12	4254	Risk Ratio (M‐H, Random, 95% CI)	0.63 [0.46, 0.87]
7.7.1 Apixaban dose (2.5 to 20 mg/day)	1	159	Risk Ratio (M‐H, Random, 95% CI)	0.50 [0.26, 0.98]
7.7.2 Betrixaban dose (40 to 160 mg/day)	1	105	Risk Ratio (M‐H, Random, 95% CI)	2.46 [0.55, 11.02]
7.7.3 Darexaban low dose (≤ 40 mg/day)	2	360	Risk Ratio (M‐H, Random, 95% CI)	0.95 [0.50, 1.82]
7.7.4 Edoxaban dose (30 to 60 mg/day)	4	1316	Risk Ratio (M‐H, Random, 95% CI)	0.46 [0.30, 0.71]
7.7.5 Rivaroxaban dose (5 to 30 mg/day)	4	2314	Risk Ratio (M‐H, Random, 95% CI)	0.38 [0.12, 1.23]
7.8 Liver enzymes elevation	21	29496	Risk Ratio (M‐H, Random, 95% CI)	0.60 [0.47, 0.77]
7.8.1 Apixaban dose (2.5 to 20 mg/day)	4	12484	Risk Ratio (M‐H, Random, 95% CI)	0.81 [0.61, 1.07]
7.8.2 Betrixaban dose (40 to 160 mg/day)	1	127	Risk Ratio (M‐H, Random, 95% CI)	0.06 [0.00, 1.04]
7.8.3 Darexaban low dose (≤ 40 mg/day)	2	451	Risk Ratio (M‐H, Random, 95% CI)	0.42 [0.13, 1.31]
7.8.4 Edoxaban dose (30 to 60 mg/day)	5	2094	Risk Ratio (M‐H, Random, 95% CI)	0.28 [0.09, 0.81]
7.8.5 Letaxaban low dose (≤ 40 mg/day)	1	555	Risk Ratio (M‐H, Random, 95% CI)	0.68 [0.25, 1.84]
7.8.6 LY517717 low dose (≤ 50 mg/day)	1	156	Risk Ratio (M‐H, Random, 95% CI)	1.09 [0.30, 3.91]
7.8.7 Rivaroxaban dose (5 to 30 mg/day)	7	13629	Risk Ratio (M‐H, Random, 95% CI)	0.63 [0.48, 0.84]
7.9 Volume of blood loss	14	6237	Mean Difference (IV, Random, 95% CI)	‐6.66 [‐35.87, 22.55]
7.9.1 Apixaban dose (2.5 to 20 mg/day)	1	220	Mean Difference (IV, Random, 95% CI)	‐82.50 [‐88.72, ‐76.28]
7.9.2 Letaxaban low dose (≤ 40 mg/day)	1	555	Mean Difference (IV, Random, 95% CI)	97.50 [41.27, 153.73]
7.9.3 Rivaroxaban dose (5 to 30 mg/day)	12	5462	Mean Difference (IV, Random, 95% CI)	‐9.31 [‐22.50, 3.87]

Comparison 8. Direct factor Xa inhibitors by type versus LMWH: sensitivity analysis by daily approved orthopedic surgery dose groups for major bleeding.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
8.1 Major bleeding: rivaroxaban 10 mg/day, apixaban 5 mg/day, and other direct factor Xa drugs at low dose vs LMWH	27	31382	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.74, 1.14]
8.1.1 Rivaroxaban 10 mg/day	13	14119	Risk Ratio (M‐H, Fixed, 95% CI)	1.47 [0.93, 2.33]
8.1.2 Apixaban 5 mg/day	5	12075	Risk Ratio (M‐H, Fixed, 95% CI)	0.82 [0.62, 1.08]
8.1.3 Other direct factor Xa inhibitors at low dose	9	5188	Risk Ratio (M‐H, Fixed, 95% CI)	0.64 [0.37, 1.11]

Comparison 9. Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
9.1 All‐cause mortality	12	24828	Risk Ratio (M‐H, Fixed, 95% CI)	0.88 [0.49, 1.58]
9.2 Major VTE	22	23837	Risk Ratio (M‐H, Random, 95% CI)	0.50 [0.36, 0.69]
9.3 Symptomatic VTE	22	29542	Risk Ratio (M‐H, Fixed, 95% CI)	0.63 [0.48, 0.83]
9.4 Major bleeding	27	36960	Risk Ratio (M‐H, Fixed, 95% CI)	1.09 [0.85, 1.40]
9.5 Major bleeding: rivaroxaban versus LMWH	12	15731	Risk Ratio (M‐H, Fixed, 95% CI)	1.93 [1.22, 3.06]
9.5.1 Rivaroxaban	12	15731	Risk Ratio (M‐H, Fixed, 95% CI)	1.93 [1.22, 3.06]
9.6 Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH	15	21229	Risk Ratio (M‐H, Fixed, 95% CI)	0.81 [0.59, 1.10]
9.6.1 Other direct factor Xa inhibitors	15	21229	Risk Ratio (M‐H, Fixed, 95% CI)	0.81 [0.59, 1.10]
9.7 Serious hepatic adverse events	1	3009	Risk Ratio (M‐H, Fixed, 95% CI)	3.01 [0.12, 73.93]
9.8 Serious non‐hepatic adverse events	15	26246	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.81, 0.97]
9.9 Fatal VTE	8	17788	Risk Ratio (M‐H, Fixed, 95% CI)	1.20 [0.45, 3.21]
9.10 Asymptomatic distal DVT	11	4496	Risk Ratio (M‐H, Random, 95% CI)	0.66 [0.48, 0.90]
9.11 Fatal bleeding	2	7467	Risk Ratio (M‐H, Fixed, 95% CI)	2.99 [0.31, 28.75]
9.12 Liver enzymes elevation	20	30707	Risk Ratio (M‐H, Random, 95% CI)	0.57 [0.44, 0.74]
9.13 Minor adverse events	4	1011	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.83, 1.00]
9.14 Volume of blood loss	12	7109	Mean Difference (IV, Random, 95% CI)	8.07 [‐13.79, 29.93]

9.2. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 2: Major VTE

9.3. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 3: Symptomatic VTE

9.4. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 4: Major bleeding

9.5. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 5: Major bleeding: rivaroxaban versus LMWH

9.6. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 6: Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH

9.7. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 7: Serious hepatic adverse events

9.8. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 8: Serious non‐hepatic adverse events

9.9. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 9: Fatal VTE

9.10. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 10: Asymptomatic distal DVT

9.11. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 11: Fatal bleeding

9.12. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 12: Liver enzymes elevation

9.13. Analysis.

Comparison 9: Direct factor Xa inhibitors versus LMWH: sensitivity analysis restricted to treatment period results, Outcome 13: Minor adverse events

Comparison 10. Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
10.1 All‐cause mortality	15	25642	Risk Ratio (M‐H, Fixed, 95% CI)	0.81 [0.51, 1.29]
10.2 Major VTE	22	23139	Risk Ratio (M‐H, Random, 95% CI)	0.46 [0.33, 0.65]
10.3 Symptomatic VTE	24	29515	Risk Ratio (M‐H, Fixed, 95% CI)	0.63 [0.49, 0.81]
10.4 Major bleeding	25	34535	Risk Ratio (M‐H, Fixed, 95% CI)	0.97 [0.78, 1.21]
10.5 Major bleeding: rivaroxaban versus LMWH	11	15106	Risk Ratio (M‐H, Fixed, 95% CI)	1.60 [1.00, 2.57]
10.5.1 Rivaroxaban	11	15106	Risk Ratio (M‐H, Fixed, 95% CI)	1.60 [1.00, 2.57]
10.6 Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH	14	19429	Risk Ratio (M‐H, Fixed, 95% CI)	0.82 [0.64, 1.04]
10.6.1 Other direct factor Xa inhibitors	14	19429	Risk Ratio (M‐H, Fixed, 95% CI)	0.82 [0.64, 1.04]
10.7 Serious hepatic adverse events	1	3009	Risk Ratio (M‐H, Fixed, 95% CI)	3.01 [0.12, 73.93]
10.8 Serious non‐hepatic adverse events	15	26246	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.81, 0.97]
10.9 Fatal VTE	9	17906	Risk Ratio (M‐H, Fixed, 95% CI)	1.22 [0.50, 2.95]
10.10 Asymptomatic distal DVT	12	4602	Risk Ratio (M‐H, Random, 95% CI)	0.64 [0.46, 0.89]
10.11 Fatal bleeding	4	10794	Risk Ratio (M‐H, Fixed, 95% CI)	1.42 [0.33, 6.04]
10.12 Liver enzymes elevation	19	30296	Risk Ratio (M‐H, Random, 95% CI)	0.56 [0.43, 0.73]
10.13 Minor adverse events	4	1011	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.83, 1.00]
10.14 Volume of blood loss	11	5922	Mean Difference (IV, Random, 95% CI)	1.11 [‐34.50, 36.72]

10.2. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 2: Major VTE

10.3. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 3: Symptomatic VTE

10.4. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 4: Major bleeding

10.6. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 6: Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH

10.7. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 7: Serious hepatic adverse events

10.8. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 8: Serious non‐hepatic adverse events

10.9. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 9: Fatal VTE

10.10. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 10: Asymptomatic distal DVT

10.11. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 11: Fatal bleeding

10.12. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 12: Liver enzymes elevation

10.13. Analysis.

Comparison 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias in random sequence generation or allocation concealment, Outcome 13: Minor adverse events

Comparison 11. Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
11.1 All‐cause mortality	15	24375	Risk Ratio (M‐H, Fixed, 95% CI)	0.97 [0.59, 1.58]
11.2 Major VTE	22	22010	Risk Ratio (M‐H, Random, 95% CI)	0.55 [0.41, 0.75]
11.3 Symptomatic VTE	22	27098	Risk Ratio (M‐H, Fixed, 95% CI)	0.73 [0.55, 0.95]
11.4 Major bleeding	28	34759	Risk Ratio (M‐H, Fixed, 95% CI)	1.05 [0.85, 1.30]
11.5 Major bleeding: rivaroxaban versus LMWH	12	13434	Risk Ratio (M‐H, Fixed, 95% CI)	1.97 [1.28, 3.04]
11.5.1 Rivaroxaban versus LMWH	12	13434	Risk Ratio (M‐H, Fixed, 95% CI)	1.97 [1.28, 3.04]
11.6 Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH	16	21325	Risk Ratio (M‐H, Fixed, 95% CI)	0.80 [0.63, 1.02]
11.6.1 Other direct factor Xa inhibitors versus LMWH	16	21325	Risk Ratio (M‐H, Fixed, 95% CI)	0.80 [0.63, 1.02]
11.7 Serious hepatic adverse events	1	3009	Risk Ratio (M‐H, Fixed, 95% CI)	3.01 [0.12, 73.93]
11.8 Serious non‐hepatic adverse events	14	23789	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.83, 1.01]
11.9 Fatal VTE	9	16639	Risk Ratio (M‐H, Fixed, 95% CI)	1.54 [0.60, 3.92]
11.10 Asymptomatic distal DVT	11	4496	Risk Ratio (M‐H, Random, 95% CI)	0.66 [0.48, 0.90]
11.11 Fatal bleeding	4	10794	Risk Ratio (M‐H, Fixed, 95% CI)	1.42 [0.33, 6.04]
11.12 Liver enzymes elevation	20	29077	Risk Ratio (M‐H, Random, 95% CI)	0.60 [0.47, 0.76]
11.13 Minor adverse events	4	1011	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.83, 1.00]
11.14 Volume of blood loss	12	7223	Mean Difference (IV, Random, 95% CI)	2.77 [‐33.97, 39.51]

11.2. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 2: Major VTE

11.3. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 3: Symptomatic VTE

11.4. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 4: Major bleeding

11.5. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 5: Major bleeding: rivaroxaban versus LMWH

11.6. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 6: Major bleeding: other direct factor Xa inhibitors (excluding rivaroxaban) versus LMWH

11.7. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 7: Serious hepatic adverse events

11.8. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 8: Serious non‐hepatic adverse events

11.9. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 9: Fatal VTE

11.10. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 10: Asymptomatic distal DVT

11.11. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 11: Fatal bleeding

11.12. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 12: Liver enzymes elevation

11.13. Analysis.

Comparison 11: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with high risk of bias for inappropriate intervention administration (different intervention durations), Outcome 13: Minor adverse events

Comparison 12. Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding small studies for liver enzymes elevation outcome.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
12.1 Liver enzymes elevation	14	28568	Risk Ratio (M‐H, Random, 95% CI)	0.69 [0.58, 0.83]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

ADVANCE‐1 2009.

Study characteristics
Methods	Aim: non‐inferiority and superiority, efficacy and safety Design: parallel‐group Blinding: double‐blind, double‐dummy; groups blinded: participants, outcome adjudicators, data analysts Study duration: not reported Duration of intervention: 10 to 14 days Duration of follow‐up: 60 days Method of randomisation: interactive central telephone system, with block sizes of 4; stratified by site and type of joint replacement (unilateral/bilateral) Method of concealment of allocation: use of an interactive central telephone system for allocation of participants Primary efficacy analysis set: participants who had an efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: not reported Exclusions: Apixaban group (445): 3 did not receive study drug, 169 did not undergo venography, 273 had uninterpretable venography (91 unilateral, 70 proximal segments not readable, 112 distal segments not readable) Enoxaparin group (474): 8 did not receive study drug, 192 did not undergo venography, 274 had uninterpretable venography (98 unilateral, 90 proximal segments not readable, 86 distal segments not readable)
Participants	Diagnosis: scheduled to TKR (unilateral/bilateral) or revision of knee prosthesis Location of participants: 14 countries (Canada, USA, Mexico, Hungary, Denmark, Australia, Israel, Norway, Argentina, Sweden, Brazil, Russia, Poland, Turkey) Number of participants randomised: 3195 Age (years; mean and range): (primary efficacy analysis population) apixaban 66.1 (26 to 93); enoxaparin 65.5 (33 to 87) Gender (male %): (primary efficacy analysis population) apixaban 468 (40.4%); enoxaparin 452 (40%) Baseline imbalances: none Inclusion criteria: scheduled to undergo TKR surgery for one or both knees, including revision of a previously inserted artificial joint Exclusion criteria: active bleeding or a contraindication to anticoagulant prophylaxis, or requiring ongoing anticoagulant or antiplatelet treatment uncontrolled hypertension, active hepatobiliary disease, clinically significant impairment of renal function, thrombocytopaenia, anaemia, allergy to heparin, and allergy to radiographic contrast dye or another contraindication to bilateral venography
Interventions	Number of intervention groups: 2 Concomitant interventions: none Excluded interventions: none Apixaban group: Intervention: apixaban (oral) 2.5 mg twice daily Starting time: after surgery: 12 to 24 hours Duration: 10 to 14 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 30 mg twice daily Starting time: after surgery: 12 to 24 hours Duration: 10 to 14 days
Outcomes	Number of outcomes: according to protocol: 39 reported: 26. Reported additionally: symptomatic VTE and VTE‐related death, major VTE and death, symptomatic DVT (treatment and follow‐up periods), PE (follow‐up period), all‐cause death (follow‐up period), minor bleeding, liver laboratory abnormalities (follow‐up period), thrombocytopaenia (treatment period), serious adverse events (follow‐up period) not reported: major bleeding, CRNM bleeding, major and CRNM bleeding, any bleeding (follow‐up period), PE, VTE, VTE‐related death, symptomatic VTE, proximal DVT (total, asymptomatic), distal DVT (total, asymptomatic), bleeding AEs, AEs leading to discontinuation, mean change from baseline in SBP, DBP and heart rate, laboratory abnormalities in haematology, kidney function, electrolyte and other chemistry OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 10 to 14 days (treatment), 60 days (additionally, follow‐up) Time points reported: 10 to 14 days (treatment), 60 days (additionally, follow‐up) Symptomatic VTE: Definition: incidence of symptomatic VTE or VTE‐related death Time points measured: 10 to 14 days (treatment), 60 days (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2 g/dL within 24 hours; 2) transfusion of ≥ 2 U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding; 6) intramuscular bleeding with compartment syndrome Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake) Serious AEs: non‐hepatic Definition: number of participants with at least one serious adverse event (definition not provided), excluding those liver‐related Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Liver enzymes elevation: Definition: incidence of ALT/AST > 3 times upper limit of normal (ULN) on same date Time points measured: 10 to 14 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Time points reported: 10 to 14 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) RESULTS All‐cause mortality: Summary data (10 to 14 days): see Analysis 9.1 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (60 day follow‐up): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: not specified, ITT analysis, 79 missing participants (not specified if correspond to those lost to follow‐up) ‐Subgroups reported: none Symptomatic VTE: Summary data (10 to 14 days): see Analysis 9.3 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (60 day follow‐up): not reported Major bleeding: Summary data (12 to 16 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 11 missing participants (did not receive study drug) ‐Subgroups reported: none Summary data (60 day follow‐up): not reported Serious AEs: non‐hepatic: Summary data (12 to 16 days): see Analysis 9.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 11 missing participants (did not receive study drug) ‐Subgroups reported: none Summary data (60 day follow‐up): see Analysis 1.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 11 missing participants (did not receive study drug) ‐Subgroups reported: none Liver enzymes elevation: Summary data (10 to 14 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 11 missing participants (did not receive study drug) ‐Subgroups reported: none Summary data (60 day follow‐up): see Analysis 1.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 11 missing participants (did not receive study drug) ‐Subgroups reported: none
Notes	Source of funding: Bristol‐Myers Squibb, Pfizer Conflicts of interest: authors (6) had financial relationships with several pharmaceutical companies, including Bristol‐Myers Squibb and Pfizer Published protocol: registered in ClinicalTrials.gov as NCT00371683, and in ICTRP as EUCTR2006‐002161‐39‐DK Observations: during follow‐up period, participants received additional prophylaxis at the discretion of the investigator Missing data requested from authors: specific groups blinded, date of start and end of study, description of method of randomisation, details about withdrawals, definition of serious AEs, follow‐up summary data for major bleeding, follow‐up missing participants for all‐cause mortality Missing data obtained from authors: none For the efficacy set (per‐protocol analysis), 28.8% of the randomised participants were excluded (27.8% in the apixaban group and 29.7% in the enoxaparin group. They estimated 30% before before sample size calculation). Reasons for exclusion were reported; however, no follow‐up of these participants was performed and the incidence of VTE among them remains unknown. (They reported the symptomatic VTE events in all the randomized participants.) The sample size estimation was insufficient to detect differences even with inferiority analysis, due to the unexpected lower rates of events. For the safety set, 0.3% of the randomised participants were excluded (0.2% in the apixaban group and 0.5% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The use of a computer‐generated random sequence was assumed given the reported use of an interactive central telephone system with block sizes of 4; stratified by site and type of joint replacement (unilateral/bilateral)
Allocation concealment (selection bias)	Low risk	Participants were allocated by the use of an interactive central telephone system
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Low risk	Outcome assessment was performed by independent, blinded central adjudication committees. Data analysis was performed by the study sponsor, with a statistical analysis established before unblinding of the databases
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcomes. Proportion of missing outcomes compared with the observed event risk is enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcomes. Missing outcome data balanced in reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk is not enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 8 outcomes not included in the protocol were reported, 21 outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan that explains the procedures or additional steps to be taken to minimize the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

ADVANCE‐2 2010.

Study characteristics
Methods	Aim: non‐inferiority and superiority, efficacy and safety Design: parallel‐group (phase III) Blinding: double‐blind, double‐dummy; groups blinded: participants, outcome adjudicators, data analysts Study duration: June 2007 to November 2008 Duration of intervention: 10 to 14 days Duration of follow‐up: 60 days Method of randomisation: Bristol‐Myers Squibb randomisation centre with SAS, with block sizes of 4; stratified by site and type of joint replacement (unilateral/bilateral) Method of concealment of allocation: use of an interactive central telephone system for allocation of participants Primary efficacy analysis set: participants who had an efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: not reported Exclusions: Apixaban group (525): 27 did not receive study drug, 215 did not undergo venography, 337 had uninterpretable venography (43 unilateral, 75 proximal segments not readable, 291 distal segments not readable) Enoxaparin group (511): 21 did not receive study drug, 209 did not undergo venography, 323 had uninterpretable venography (38 unilateral, 83 proximal segments not readable, 202 distal segments not readable)
Participants	Diagnosis: scheduled to elective TKR (unilateral/bilateral), including revision Location of participants: 27 countries (including countries from South Africa, Europe, Latin America and Asia/Pacific) Number of participants randomised: 3057 Age (years; mean and range): (primary efficacy analysis population) apixaban 65.1 (59 to 72); enoxaparin 66.0 (60 to 73) Gender (male %): (primary efficacy analysis population) apixaban 289 (30%); enoxaparin 267 (27%) Baseline imbalances: none Inclusion criteria: scheduled to have unilateral elective TKR or same‐day bilateral knee replacement, including revision Exclusion criteria: active bleeding or a contraindication to anticoagulant prophylaxis, or need for continuing anticoagulant or antiplatelet treatment uncontrolled hypertension, active hepatobiliary disease, impaired renal function, thrombocytopaenia, anaemia, heparin allergy, allergy to radiographic contrast dye, or other disorders preventing bilateral venography
Interventions	Number of intervention groups: 2 Concomitant interventions: none Excluded interventions: none Apixaban group: Intervention: apixaban (oral) 2.5 mg twice daily Starting time: after surgery: 12 to 24 hours Duration: 10 to 14 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: 12 hours Duration: 10 to 14 days
Outcomes	Number of outcomes: according to protocol: 18 reported: 22. Reported additionally: major VTE, symptomatic (including fatal) VTE, DVT, all‐cause death, total bleeding, minor bleeding, serious AEs, drug‐related serious AEs, drug‐related AEs, total AEs (treatment period), fatal VTE, symptomatic non‐fatal VTE (follow‐up period), symptomatic DVT, PE and fatal bleeding (treatment and follow‐up periods) not reported: non‐fatal PE, non‐serious AEs (treatment period), laboratory marked abnormalities on urinalysis, kidney function, electrolyte and other clinical tests (treatment and follow‐up periods) OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Major VTE Definition: incidence of proximal DVT, symptomatic DVT, PE and VTE‐related death Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment period) Symptomatic VTE Definition: incidence of symptomatic VTE Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment period) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2g/dL within 24 hours; 2) transfusion of ≥ 2U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding; 6) intramuscular bleeding with compartment syndrome Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake) Serious AEs: hepatic Definition: not included as an outcome, reported as death from hepatopathy (infectious versus drug‐related) Time points measured: not measured Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake) Serious AEs: non‐hepatic Definition: number of participants with at least one serious AE (definition not provided), excluding those liver‐related Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Fatal VTE Definition: incidence of death from VTE Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Fatal bleeding Definition: incidence of death from bleeding Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Liver enzymes elevation: Definition: incidence of ALT/AST > 3 times upper limit of normal (ULN) on same date Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake), 60 days (additionally, follow‐up) RESULTS All‐cause mortality: Summary data (12 to 16 days): see Analysis 9.1 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (60‐day follow‐up): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: not specified, not specified, 130 missing participants (not specified if correspond to those lost to follow‐up) ‐Subgroups reported: none Major VTE: Summary data (12 to 16 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, 63 missing participants ‐Subgroups reported: none Summary data (60‐day follow‐up): not reported Symptomatic VTE: Summary data (12 to 16 days): see Analysis 9.3 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (60‐day follow‐up): not reported Major bleeding: Summary data (12 to 16 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 48 missing participants (did not receive study drug) ‐Subgroups reported: none Summary data (60‐day): not reported Serious AEs: hepatopathy: Summary data (12 to 16 days): apixaban 1/NR; enoxaparin NR/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Summary data (60‐day): not reported Serious AEs: non‐hepatic: Summary data (12 to 16 days): see Analysis 9.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 48 missing participants (did not receive study drug) ‐Subgroups reported: none Summary data (60‐day follow‐up): see Analysis 1.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 48 missing participants (did not receive study drug) ‐Subgroups reported: none Fatal VTE: Summary data (12 to 16 days): see Analysis 9.9 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (60‐day follow‐up): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: not specified, not specified, 130 missing participants (not specified if correspond to those lost to follow‐up) ‐Subgroups reported: none Fatal bleeding: Summary data (12 to 16 days): apixaban 0/1528; enoxaparin 0/1529 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (60‐day follow‐up): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: not specified, not specified, 130 missing participants (not specified if correspond to those lost to follow‐up) ‐Subgroups reported: none Liver enzymes elevation: Summary data (12 to 16 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 48 missing participants (did not receive study drug) ‐Subgroups reported: none Summary data (60‐day follow‐up): see Analysis 1.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 48 missing participants (did not receive study drug) ‐Subgroups reported: none
Notes	Source of funding: Bristol‐Myers Squibb, Pfizer Conflicts of interest: authors (6) had financial relationships with several pharmaceutical companies, including Bristol‐Myers Squibb and Pfizer Published protocol: registered in ClinicalTrials.gov as NCT00452530, and in ICTRP as EUCTR2006‐006896‐19‐AT Observations: during follow‐up period, participants with subclinical thrombosis received continued prophylaxis or treatment according to local practice Missing data requested from authors: specific groups blinded, details about withdrawals, definition of serious AEs, treatment period summary data for symptomatic VTE, follow‐up summary data for major bleeding and serious AEs, treatment period number of participants for serious AEs, follow‐up missing participants for fatal VTE, symptomatic VTE and fatal bleeding Missing data obtained from authors: none For the efficacy set (per‐protocol analysis), 33.9% of the randomised participants were excluded (34.4% in the apixaban group and 33.4% in the enoxaparin group. They estimated 30% before sample size calculation). Reasons for exclusion were reported. However, no follow‐up of these participants was performed and the incidence of VTE among them remains unknown. (They reported the symptomatic VTE events in all the randomized participants.) For the safety set, 1.6% of the randomised participants were excluded (1.8% in the apixaban group and 1.4% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated, in the Bristol‐Myers Squibb randomisation centre, with SAS, block sizes of 4
Allocation concealment (selection bias)	Low risk	Participants were allocated by the use of an interactive central telephone system
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. "Investigators" were also reported to have been blinded. We assumed this to mean the personnel in direct contact with participants (clinicians/providers)
Blinding of outcome assessment (detection bias)	Low risk	Outcome assessment was performed by an independent, blinded central adjudication committee. Data analysis was performed by statisticians masked to treatment allocation, with a statistical analysis established before locking of the databases
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome. Proportion of missing outcomes compared with the observed event risk is enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome. Missing outcome data balanced in reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk is not enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 15 outcomes not included in the protocol were reported, six outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan that explains the procedures or additional steps to be taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

ADVANCE‐3 2010.

Study characteristics
Methods	Aim: non‐inferiority and superiority, efficacy and safety Design: parallel‐group (phase III) Blinding: double‐blind, double‐dummy; groups blinded: participants, outcome adjudicators, data analysts Study duration: March 2007 to September 2009 Duration of intervention: 35 ± 3 days Duration of follow‐up: 65 ± 5 days Method of randomisation: Bristol‐Myers Squibb randomisation centre with SAS, with block sizes of 4; stratified by site Method of concealment of allocation: use of an interactive telephone system for allocation of participants Primary efficacy analysis set: participants who had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: not reported Exclusions: Apixaban group: 363 did not undergo venography, 396 had uninterpretable venography (83 unilateral, 63 proximal segments not readable, 250 distal segments not readable); 35 did not receive study drug Enoxaparin group: 364 did not undergo venography, 418 had uninterpretable venography (66 unilateral, 74 proximal segments not readable, 278 distal segments not readable); 40 did not receive study drug
Participants	Diagnosis: scheduled to THR, elective, or revision of hip prosthesis Location of participants: 21 countries (regions included North and South America, Europe, Asia and Australia) Number of participants randomised: 5407 Age (years; mean and range): (primary efficacy analysis population) apixaban 60.9 (19 to 90); enoxaparin 60.0 (19 to 91) Gender (male %): (primary efficacy analysis population) apixaban 925 (47.5%); enoxaparin 912 (47.6%) Baseline imbalances: none Inclusion criteria: scheduled to undergo elective THR or revision of a previously inserted hip prosthesis Exclusion criteria: active bleeding or a contraindication to anticoagulant prophylaxis, or requiring ongoing anticoagulant or antiplatelet treatment
Interventions	Number of intervention groups: 2 Concomitant interventions: none Excluded interventions: none Apixaban group: Intervention: apixaban (oral) 2.5 mg twice daily Starting time: after surgery: 12 to 24 hours Duration: 35 ± 3 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: 12 ± 3 hours Duration: 35 ± 3 days
Outcomes	Number of outcomes: according to protocol: 23 reported: 23. Reported additionally: symptomatic + fatal VTE, fatal VTE, non‐fatal PE, minor bleeding, serious AEs (treatment period), symptomatic DVT, fatal PE, fatal bleeding (treatment and follow‐up periods) not reported: distal DVT, marked abnormalities in clinical (haematology, electrolyte, kidney, urinalysis) laboratory tests, myocardial infarction and stroke, neurologic AEs (treatment period), all‐cause death, serious AEs (follow‐up period), bleeding‐related AEs (treatment and follow‐up periods) OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 35 ± 3 days (treatment), 60 ± 5 days (additionally, follow‐up) Time points reported: 35 ± 3 days (treatment), 60 ± 5 days (additionally, follow‐up) Major VTE: Definition: incidence of proximal DVT, symptomatic DVT, PE and VTE‐related death Time points measured: 35 ± 3 days (treatment), 60 ± 5 days (additionally, follow‐up) Time points reported: 35 ± 3 days (treatment) Symptomatic VTE: Definition: incidence of symptomatic VTE or VTE‐related death Time points measured: 35 ± 3 days (treatment), 60 ± 5 days (additionally, follow‐up) Time points reported: 35 ± 3 days (treatment) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2 g/dL within 24 hours; 2) transfusion of ≥ 2 U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding; 6) intramuscular bleeding with compartment syndrome Time points measured: 37 ± 3 days (treatment, 2 days after the last study drug intake), 60 ± 5 days (additionally, follow‐up) Time points reported: 37 ± 3 days (treatment, 2 days after the last study drug intake) Serious AEs: non‐hepatic Definition: number of patients with at least one serious AE (definition not provided), excluding those liver‐related Time points measured: 37 ± 3 days (treatment, 2 days after the last study drug intake), 60 ± 5 days (additionally, follow‐up) Time points reported: 37 ± 3 days (treatment, 2 days after the last study drug intake) Fatal VTE: Definition: incidence of fatal and nonfatal PE Time points measured: 35 ± 3 days (treatment), 60 ± 5 days (additionally, follow‐up) Time points reported: 35 ± 3 days (treatment), 60 ± 5 days (additionally, follow‐up) Fatal bleeding: Definition: incidence of any bleeding Time points measured: 37 ± 3 days (treatment, 2 days after the last study drug intake), 60 ± 5 days (additionally, follow‐up) Time points reported: 37 ± 3 days (treatment, 2 days after the last study drug intake), 60 ± 5 days (additionally, follow‐up) Liver enzymes elevation: Definition: incidence of ALT/AST > 3 times upper limit of normal (ULN) on same date Time points measured: 37 ± 3 days (treatment, 2 days after the last study drug intake), 60 ± 5 days (additionally, follow‐up) Time points reported: 37 ± 3 days (treatment, 2 days after the last study drug intake), 60 ± 5 days (additionally, follow‐up) RESULTS All‐cause mortality: Summary data (35 ± 3 days): see Analysis 9.1 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (60 ± 5 days follow‐up): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: not specified, not specified, 232 missing participants (not specified if correspond to those lost to follow‐up) ‐Subgroups reported: none Major VTE: Summary data (35 ± 3 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, 1013 missing participants ‐Subgroups reported: none Summary data (60 ± 5 days follow‐up): not reported Symptomatic VTE: Summary data (35 ± 3 days): see Analysis 9.3 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (60 ± 5 days follow‐up): not reported Major bleeding: Summary data (37 ± 3 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 75 missing participants (did not receive study drug) ‐Subgroups reported: none Summary data (60 ± 5 days follow‐up): not reported Serious AEs: non‐hepatic: Summary data (37 ± 3 days): Analysis 9.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 75 missing participants (did not receive study drug) ‐Subgroups reported: none Summary data (60 ± 5 days follow‐up): not reported Fatal VTE: Summary data (35 ± 3 days): see Analysis 9.9 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (60 ± 5 days follow‐up): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: not specified, not specified, 232 missing participants (not specified if correspond to those lost to follow‐up) ‐Subgroups reported: none Fatal bleeding: Summary data (37 ± 3 days): see Analysis 9.11 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (60 ± 5 days follow‐up): not reported Liver enzymes elevation: Summary data (37 ± 3 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 162 missing participants (reasons not fully explained) ‐Subgroups reported: none Summary data (60 ± 5 days follow‐up): see Analysis 1.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 575 missing participants (reasons not fully explained) ‐Subgroups reported: none
Notes	Source of funding: Bristol‐Myers Squibb, Pfizer Conflicts of interest: authors (6) had financial relationships with several pharmaceutical companies, including Bristol‐Myers Squibb and Pfizer Published protocol: as Supplementary Material to the study report. Also registered in ClinicalTrials.gov as NCT00423319, and in ICTRP as EUCTR2006‐002402‐60‐SE Observations: none Missing data requested from authors: specific groups blinded, details about withdrawals, definition of serious AEs, follow‐up summary data for major bleeding and serious AEs, treatment period missing participants for liver enzymes elevation, follow‐up missing participants for all‐cause mortality, fatal VTE, fatal bleeding and liver enzymes elevation Missing data obtained from authors: none For the efficacy set (per‐protocol analysis), 13.4% of the randomised participants were excluded (28%/13.4% in the apixaban group and 29%/13.5% in the enoxaparin group for the treatment period/follow‐up period, respectively). Reasons for exclusion were reported. For the safety set, 3.0% of the randomised participants were excluded (2.9% in the apixaban group and 3.1% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated, in the Bristol‐Myers Squibb randomisation centre, with SAS, block sizes of 4
Allocation concealment (selection bias)	Low risk	Participants were allocated by the use of an interactive central telephone system
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Low risk	Outcome assessment was performed by an independent, blinded central adjudication committee. Data analysis was performed by statisticians masked to treatment allocation, with a statistical analysis established before locking of the databases
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome. Proportion of missing outcomes compared with the observed event risk is enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome. Missing outcome data balanced in reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk is not enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov and as supplementary material with the included report. Reported outcomes differed from those included in the protocol: 8 outcomes not included in the protocol were reported, 7 outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

Agnelli 2007.

Study characteristics
Methods	Aim: non‐inferiority, efficacy and safety Design: parallel‐group/dose‐escalation (phase II) Blinding: double‐blind, double‐dummy; groups blinded: participants, outcome adjudicators Study duration: December 2003 to April 2005 Duration of intervention: 6 to 10 days Duration of follow‐up: 22 ± 7 days Method of randomisation: stratified by type of surgery (THR/TKR); no additional description provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) underwent planned surgery, 3) had a primary efficacy outcome that could be evaluated, 4) had no protocol violations that could potentially impact on efficacy evaluation (per‐protocol population) Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: not reported Exclusions: LY517717 groups (not detailed according to dose group): 3 did not receive study drug; 8 did not undergo planned surgery; 91 were not evaluable for any VTE event, 41 did not undergo venography, 50 had uninterpretable venography (not further detailed); 24 had major protocol violations (3 received < 5 doses of study drug, 9 used excluded intervention, 3 violated study drug tasting requirement, 9 other). Enoxaparin group: 1 did not receive study drug; 2 did not undergo planned surgery; 16 were not evaluable for any VTE event, 12 did not undergo venography, 4 had uninterpretable venography (not further detailed); 6 had major protocol violations (0 received < 5 doses of study drug, 2 used excluded intervention, 3 violated study drug tasting requirement, 1 other)
Participants	Diagnosis: scheduled to THR/TKR, elective, primary, unilateral Location of participants: 8 countries (Australia, Austria, Belgium, Czech Republic, Germany, Hungary, Italy and Poland) Number of participants randomised: 511 Age (years; mean and range): (safety analysis population) LY517717 25 mg / 50 mg / 75 mg / 100 mg / 125 mg / 150 mg (once daily) 63.4 (44‐75) / 62.4 (33‐75) / 64.4 (45‐74) / 62.4 (31‐75) / 61.6 (32‐75) / 62.1 (33‐75); enoxaparin 63.5 (42‐75) Gender (male %): (safety analysis population) LY517717 25 mg / 50 mg / 75 mg / 100 mg / 125 mg / 150 mg (once daily) 10 (31.2%) / 15 (44.1%) / 11 (34.4%) / 50 (47.2%) / 57 (51.8%) / 49 (47.6%); enoxaparin 38 (42.2%) Baseline imbalances: slight differences amongst groups regarding gender and proportion of hip surgeries; notable differences regarding the number of participants amongst groups Inclusion criteria: scheduled for elective primary unilateral THR or TKR between 18 and 75 years of age weighing between 50 kg and 120 kg Exclusion criteria: clinically significant bleeding disorder; baseline International Normalised Ratio > 1.5 or platelet count < 100x109/L active peptic ulcer disease; recent stroke; uncontrolled hypertension; intracranial tumour; brain or spinal cord surgery anticoagulant or thrombolytic therapy in the previous 7 days; treatment with aspirin or any non‐steroidal antiinflammatory drug (other than COX‐2 inhibitors) within 48 hours of surgery serum creatinine levels ≥ 2.0 mg/dL (177 mmol/L); baseline aspartate aminotransferase (AST) or alanine aminotransferase (ALT) ≥ 2 times the upper limit of normal or total bilirubin outside of the normal range Hb < 10 g/dL female gender and childbearing potential failure or inability to provide informed consent
Interventions	Number of intervention groups: 7 Concomitant interventions: none Excluded interventions: none LY517717 25 mg group: Intervention: LY517717 (oral) 25 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 6 to 10 days (doses) LY517717 50 mg group: Intervention: LY517717 (oral) 50 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 6 to 10 days (doses) LY517717 75 mg group: Intervention: LY517717 (oral) 75 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 6 to 10 days (doses) LY517717 100 mg group: Intervention: LY517717 (oral) 100 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 6 to 10 days (doses) LY517717 125 mg group: Intervention: LY517717 (oral) 125 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 6 to 10 days (doses) LY517717 150 mg group: Intervention: LY517717 (oral) 150 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 6 to 10 days (doses) Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: evening Duration: 6 to 10 days (doses)
Outcomes	Number of outcomes: According to protocol: no protocol available OUTCOMES Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT Time points measured: 6 to 10 days (treatment), 22 ± 7 days (additionally, follow‐up) Time points reported: 6 to 10 days (treatment), 22 ± 7 days (additionally, follow‐up) Symptomatic VTE: Definition: incidence of symptomatic VTE Time points measured: 6 to 10 days (treatment), 22 ± 7 days (additionally, follow‐up) Time points reported: 6 to 10 days (treatment) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) a bleeding index ≥ 2.0 (bleeding index = units of red blood cells transfused + pre‐bleed Hb − post‐bleed Hb); 2) bleeding at a critical site (intracranial, intraocular, pericardial, or retroperitoneal); 3) bleeding into the operated joint, requiring an additional operation or intervention; 4) fatal bleeding Time points measured: 6 to 10 days (treatment), 22 ± 7 days (additionally, follow‐up) Time points reported: 6 to 10 days (treatment), 22 ± 7 days (additionally, follow‐up) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 6 to 10 days (treatment), 22 ± 7 days (additionally, follow‐up) Time points reported: 6 to 10 days (treatment), 22 ± 7 days (additionally, follow‐up) Liver enzymes elevation: Definition: incidence of ALT >3 times ULN Time points measured: 6 to 10 days (treatment) Time points reported: 6 to 10 days (treatment) RESULTS Major VTE: Summary data (6 to 10 days): LY517717 25 mg once daily ‐ 50 mg once daily ‐ 75 mg once daily ‐ 100 mg once daily ‐ 125 mg once daily ‐ 150 mg once daily ‐ enoxaparin 40 mg once daily: 0/28 ‐ 3/21 ‐ 6/22 ‐ 5/77 ‐ 2/81 ‐ 0/84 ‐ 3/65 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 133 missing participants ‐Subgroups reported: none Summary data (22 ± 7 days follow‐up): LY517717 25 mg once daily ‐ 50 mg once daily ‐ 75 mg once daily ‐ 100 mg once daily ‐ 125 mg once daily ‐ 150 mg once daily ‐ enoxaparin 40 mg once daily: 0/NR ‐ 0/NR ‐ 0/NR ‐ 1/NR ‐ 0/NR ‐ 1/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Symptomatic VTE: Summary data (6 to 10 days): LY517717 25 mg once daily ‐ 50 mg once daily ‐ 75 mg once daily ‐ 100 mg once daily ‐ 125 mg once daily ‐ 150 mg once daily ‐ enoxaparin 40 mg once daily: 0/28 ‐ 0/21 ‐ 0/22 ‐ 0/77 ‐ 0/81 ‐ 0/84 ‐ 0/65 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 133 missing participants ‐Subgroups reported: none Summary data (22 ± 7 days follow‐up): not reported Major bleeding: Summary data (6 to 10 days): LY517717 25 mg once daily ‐ 50 mg once daily ‐ 75 mg once daily ‐ 100 mg once daily ‐ 125 mg once daily ‐ 150 mg once daily ‐ enoxaparin 40 mg once daily: 0/32 ‐ 0/34 ‐ 0/32 ‐ 1/106 ‐ 0/110 ‐ 0/103 ‐ 0/90 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 4 missing participants (did not receive study drug) ‐Subgroups reported: none Summary data (22 ± 7 days follow‐up): LY517717 25 mg once daily ‐ 50 mg once daily ‐ 75 mg once daily ‐ 100 mg once daily ‐ 125 mg once daily ‐ 150 mg once daily ‐ enoxaparin 40 mg once daily: 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 1/NR ‐ 0/NR ‐ 1/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Fatal bleeding: Summary data (6 to 10 days): LY517717 25 mg once daily ‐ 50 mg once daily ‐ 75 mg once daily ‐ 100 mg once daily ‐ 125 mg once daily ‐ 150 mg once daily ‐ enoxaparin 40 mg once daily: 0/32 ‐ 0/34 ‐ 0/32 ‐ 0/106 ‐ 0/110 ‐ 0/103 ‐ 0/90 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 4 missing participants (did not receive study drug) ‐Subgroups reported: none Summary data (22 ± 7 days follow‐up): not reported Liver enzymes elevation: Summary data (6 to 10 days): LY517717 25 mg once daily ‐ 50 mg once daily ‐ 75 mg once daily ‐ 100 mg once daily ‐ 125 mg once daily ‐ 150 mg once daily ‐ enoxaparin 40 mg once daily: 2/32 ‐ 2/34 ‐ 1/32 ‐ 4/106 ‐ 1/110 ‐ 3/103 ‐ 5/90 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 4 missing participants (did not receive study drug) ‐Subgroups reported: none
Notes	Source of funding: not explicitly stated, likely to be Eli Lilly and Company Conflicts of interest: authors (3) had financial relationships with several pharmaceutical companies, including Eli Lilly and Company Published protocol: registered in ClinicalTrials.gov as NCT00074828 Observations: randomisation to some groups was performed according to the results of two interim analyses; thus the probability of being allocated to each intervention group was not the same throughout the study Missing data requested from authors: specific groups blinded, description of method of randomisation and concealment of allocation, follow‐up summary data for symptomatic VTE and fatal bleeding, treatment period number of participants for major VTE and major bleeding, source of funding, protocol of study Missing data obtained from authors: none For the efficacy set (per‐protocol analysis), 26.0% of the randomised participants were excluded (25.5% in the LY517717 groups and 28.6% in the enoxaparin group). Reasons for exclusion were reported. Protocol violations were reported without deeper explanations. Imbalance in reasons for missing data across intervention groups. Symptomatic VTE events in all the randomised participants not reported. None of the arms were prematurely stopped due to bleeding; however, 3 arms were stopped due to low doses. For the safety set, 0.8% of the randomised participants were excluded (0.7% in the LY517717 groups and 1.1% in the enoxaparin group). Reasons for exclusion were reported. Imbalance in reasons for missing data across intervention groups.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Randomisation was stated, but the method is not described. Of note, under the dose‐escalation design, the probability of participants being allocated to each LY577717 group was different
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded central adjudication committee. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome. Missing outcome data unbalanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk is enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome. Missing outcome data unbalanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk is enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in 2005 without outcomes in ClinicalTrials.gov. Reported outcomes were included in 2008 after the study was finished.
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

APROPOS 2007.

Study characteristics
Methods	Aim: efficacy and safety Design: parallel‐group (phase II) Blinding: double‐blind, double‐dummy (for the apixaban versus enoxaparin comparisons); groups blinded: participants, outcome adjudicators Study duration: not reported Duration of intervention: 12 ± 2 days Duration of follow‐up: 30 days Method of randomisation: computer‐generated random sequence Method of concealment of allocation: not mentioned Primary efficacy analysis set: participants who had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: 108 participants discontinued study drug, reasons and allocated groups were not completely reported Exclusions: apixaban 2.5 mg twice daily: 42 (15 had no venography, 27 had uninterpretable venography); 5 mg once daily: 60 (6 did not receive study drug, 25 had no venography, 29 had uninterpretable venography); 5 mg twice daily: 52 (4 did not receive study drug, 15 had no venography, 33 had uninterpretable venography); 10 mg once daily 52 (1 did not receive study drug, 20 had no venography, 31 had uninterpretable venography); 10 mg twice daily: 44 (1 did not receive study drug, 19 had no venography, 24 had uninterpretable venography); 20 mg once daily: 46 (5 did not receive study drug, 12 had no venography, 29 had uninterpretable venography) enoxaparin 30 mg twice daily: 43 (3 did not receive study drug, 18 had no venography, 22 had uninterpretable venography) warfarin, titrated: 44 (2 did not receive study drug, 9 had no venography, 33 had uninterpretable venography)
Participants	Diagnosis: scheduled to TKR Location of participants: 8 countries (Argentina, Australia, Canada, Mexico, Denmark, Israel, Poland, USA) Number of participants randomised: 1238 Age (years; mean and range): (primary efficacy analysis population) apixaban 2.5 mg once daily / 2.5 mg twice daily / 5 mg once daily / 5 mg twice daily / 10 mg once daily / 10 mg twice daily / 20 mg once daily 67.6 (46‐88) / 66.9 (31‐87) / 66.4 (46‐84) / 67.2 (28‐86) / 66.4 (37‐87) / 65.8 (35‐90); enoxaparin 66.5 (36‐88); warfarin 66.8 (43‐85) Gender (male %): (primary efficacy analysis population) apixaban 2.5 mg once daily / 2.5 mg twice daily / 5 mg once daily / 5 mg twice daily / 10 mg once daily / 10 mg twice daily / 20 mg once daily 32% / 35% / 34.4% / 41% / 38.3% / 35.3%; enoxaparin 38.2%; warfarin 39.2% Baseline imbalances: none Inclusion criteria: scheduled to have a TKR Exclusion criteria: child‐bearing potential if a woman presence of bleeding/coagulation disorders history of heparin‐induced thrombocytopenia; intracranial/intraocular haemorrhage within the past 5 years; GI bleeding within 90 days of surgery or ulcer disease within 30 days before surgery; brain, spinal, ophthalmologic or major surgery/trauma within 90 days prior to surgery known VTE disease within the past 12 months uncontrolled hypertension; malignant disease; active hepatobiliary disease known or suspected GI disease that may affect absorption of study medication ALT, AST or bilirubin (direct or total) > 1.5 x ULN; INR >1.4 or activated partial thromboplastin time >1.4·control value hypersensitivity to UFH, LMWH, warfarin or other vitamin K antagonists, porcine products or iodinated contrast medium (for venography) treatment with medications affecting coagulation/platelet function (aspirin, clopidogrel, ticlopidine, dipyridamole or sulfinpyrazone) within 7 days prior to surgery
Interventions	Number of intervention groups: 8 Concomitant interventions: none Excluded interventions: none Apixaban 2.5 mg twice daily group: Intervention: apixaban (oral) 2.5 mg twice daily Starting time: after surgery: 12 to 24 hours Duration: 12 ± 2 days Apixaban 5 mg once daily group: Intervention: apixaban (oral) 5 mg once daily Starting time: after surgery: 12 to 24 hours Duration: 12 ± 2 days Apixaban 5 mg twice daily group: Intervention: apixaban (oral) 5 mg twice daily Starting time: after surgery: 12 to 24 hours Duration: 12 ± 2 days Apixaban 10 mg once daily group: Intervention: apixaban (oral) 10 mg once daily Starting time: after surgery: 12 to 24 hours Duration: 12 ± 2 days Apixaban 10 mg twice daily group: Intervention: apixaban (oral) 10 mg twice daily Starting time: after surgery: 12 to 24 hours Duration: 12 ± 2 days Apixaban 20 mg once daily group: Intervention: apixaban (oral) 20 mg once daily Starting time: after surgery: 12 to 24 hours Duration: 12 ± 2 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 30 mg twice daily Starting time: after surgery: 12 to 24 hours Duration: 12 ± 2 days Warfarin group: Intervention: warfarin (oral) start dose 5 mg once daily, adjusted to maintain INR 1.8‐3.0 Starting time: after surgery: evening Duration: 12 ± 2 days
Outcomes	Number of outcomes: according to protocol: 4 reported: 23. Reported additionally: total VTE, major VTE, PE, proximal DVT, proximal DVT and PE + all‐cause death, symptomatic DVT, asymptomatic DVT, total bleeding, CRNM bleeding, bleeding with surgical intervention, total AEs, serious AEs, myocardial infarction, stroke, neurological events, wound‐related infections, liver enzymes elevation (treatment period), all‐cause death, fatal VTE, fatal bleeding (treatment and follow‐up periods) not reported: pharmacokinetic profile (treatment period) OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 14 ± 2 days (treatment), 30 days (additionally, follow‐up) Time points reported: 14 ± 2 days (treatment), 30 days (additionally, follow‐up) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT Time points measured: 14 ± 2 days (treatment), 30 days (additionally, follow‐up) Time points reported: 14 ± 2 days (treatment) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2 g/dL relative to postsurgical value; 2) transfusion of ≥ 2 U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding; 6) need to discontinue study medication Time points measured: 14 ± 2 days (treatment), 30 days (additionally, follow‐up) Time points reported: 14 ± 2 days (treatment) Serious AEs: non‐hepatic: Definition: number of patients with at least one serious AE (definition not provided), excluding those liver‐related Time points measured: 14 ± 2 days (treatment), 30 days (additionally, follow‐up) Time points reported: 14 ± 2 days (treatment) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 14 ± 2 days (treatment), 30 days (additionally, follow‐up) Time points reported: 14 ± 2 days (treatment), 30 days (additionally, follow‐up) Fatal bleeding: Definition: incidence of any bleeding Time points measured: 14 ± 2 days (treatment), 30 days (additionally, follow‐up) Time points reported: 14 ± 2 days (treatment), 30 days (additionally, follow‐up) Liver enzymes elevation: Definition: incidence of ALT > 3 times ULN Time points measured: 14 ± 2 days (treatment), 30 days (additionally, follow‐up) Time points reported: 14 ± 2 days (treatment) RESULTS All‐cause mortality: Summary data (14 ± 2 days): apixaban 2.5 mg twice daily ‐ 5 mg once daily ‐ 5 mg twice daily ‐ 10 mg once daily ‐ 10 mg twice daily ‐ 20 mg once daily ‐ enoxaparin 30 mg twice daily ‐ titrated warfarin: 1/111 ‐ 0/97 ‐ 0/105 ‐ 0/105 ‐ 0/110 ‐ 0/110 ‐ 0/109 ‐ 0/109 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 382 missing participants ‐Subgroups reported: none Summary data (30 days follow‐up): apixaban 2.5 mg twice daily ‐ 5 mg once daily ‐ 5 mg twice daily ‐ 10 mg once daily ‐ 10 mg twice daily ‐ 20 mg once daily ‐ enoxaparin 30 mg twice daily ‐ titrated warfarin: 1/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 1/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Major VTE: Summary data (14 ± 2 days): see Analysis 9.2, warfarin 2/109 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 382 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): not reported Major bleeding: Summary data (14 ± 2 days): see Analysis 1.4, warfarin 0/151 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 21 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): not reported Serious AEs: non‐hepatic Summary data (14 ± 2 days): see Analysis 9.8, warfarin 9/151 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 21 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): not reported Fatal VTE: Summary data (14 ± 2 days): apixaban 2.5 mg twice daily ‐ 5 mg once daily ‐ 5 mg twice daily ‐ 10 mg once daily ‐ 10 mg twice daily ‐ 20 mg once daily ‐ enoxaparin 30 mg twice daily ‐ titrated warfarin: 1/111 ‐ 0/97 ‐ 0/105 ‐ 0/105 ‐ 0/110 ‐ 0/110 ‐ 0/109 ‐ 0/109 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 382 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): apixaban 2.5 mg twice daily ‐ 5 mg once daily ‐ 5 mg twice daily ‐ 10 mg once daily ‐ 10 mg twice daily ‐ 20 mg once daily ‐ enoxaparin 30 mg twice daily ‐ titrated warfarin: 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Fatal bleeding: Summary data (14 ± 2 days): see Analysis 1.11, warfarin 0/171 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 21 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): apixaban 2.5 mg twice daily ‐ 5 mg once daily ‐ 5 mg twice daily ‐ 10 mg once daily ‐ 10 mg twice daily ‐ 20 mg once daily ‐ enoxaparin 30 mg twice daily ‐ titrated warfarin: 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Liver enzymes elevation: Summary data (14 ± 2 days): see Analysis 9.12, warfarin 3/150 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 42 missing participants (21 unexplained) ‐Subgroups reported: none Summary data (30‐day follow‐up): not reported
Notes	Source of funding: not explicitly stated, likely to be Bristol‐Myers Squibb Conflicts of interest: authors (6) had financial relationships with several pharmaceutical companies, including Brystol‐Myers Squibb Published protocol: registered in clinicaltrials.gov as NCT00097357, and in ICTRP as EUCTR2004‐001128‐19‐DK Observations: none Missing data requested from authors: specific groups blinded, date of start and end of study, description of method for concealment of allocation, details about withdrawals, definition of serious AEs, follow‐up summary data for major VTE, major bleeding, serious AEs and liver enzymes elevation, treatment period number of participants for all‐cause mortality, fatal VTE and fatal bleeding Missing data obtained from authors: none For the efficacy set, 30.9% of the randomised participants were excluded post‐randomisation (31.6% in the apixaban groups, 28.8% in the enoxaparin group and 28.3% in the warfarin group). Reasons for exclusion were reported. For the safety set, 1.7% of the randomised participants were excluded (1.7% in the apixaban groups, 2.0% in the enoxaparin group and 1.3% in the warfarin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random sequence was computer‐generated
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion for the comparison with LMWH, but not for the comparison with warfarin. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent central adjudication committee, but it was not explicitly stated as blinded to treatment allocation. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome. Missing outcome data unbalanced in reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk is enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome. Missing outcome data balanced in reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk is not enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 20 outcomes not included in the protocol were reported, 1 outcome included in the protocol was not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

Bai 2021.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: double‐blind, double‐dummy; groups blinded: not detailed Study duration: 7 days Duration of intervention: 7 days Duration of follow‐up: not mentioned Method of randomisation: computer‐generated random sequence Method of concealment of allocation: computer‐generated random sequence Primary efficacy analysis set: coagulation difference with lab values Primary safety analysis set: bleeding events Withdrawals: not reported Exclusions: not reported
Participants	Diagnosis: scheduled to cementless unilateral primary THA Location of participants: 1 country (China) Number of participants randomised: 228 Age (years): Rivaroxaban group mean age of 61.2 years (range 46 to 83) Enoxaparin group mean age of 61.9 years (43 to 82) Gender (male %): Rivaroxaban group: 114 participants (male 37%) Enoxaparin group: 114 participants (male 45%) Baseline imbalances: not found Inclusion criteria: patients undergoing cementless unilateral primary THA, aged 18 years or older and diagnosed with osteoarthritis or osteonecrosis of the femoral head Exclusion criteria: (1) had haemorrhagic disease or a contraindication to anticoagulant prophylaxis, (2) had lower extremity DVT that was confirmed by preoperative colour Doppler ultrasonography, (3) had a history of thromboembolic disease, anticoagulant therapy, or antiplatelet treatment, or (4) had a history of severe liver diseases, uncontrolled hypertension, clinically significant impairment of renal function, or any other organ insufficiencies.
Interventions	Number of intervention groups: 2 Concomitant interventions: intravenous infusion of 1 g of tranexamic acid 30 minutes before incision and 1 g of tranexamic acid 3 hours later. Drainage tube prior to arthrotomy closure and removed 24 hours after surgery. Blood transfusions if the haemoglobin level was < 7.0 g/dL or if anaemia symptoms. Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 14 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 4000 AxaIU (0.4 mL) once daily Starting time: after surgery: 6 to 8 hours Duration: 14 days
Outcomes	Number of outcomes According to protocol: no protocol available Reported: thromboelastography and routine coagulation tests performed on the day before the operation, 1 day and 7 days after the operation Primary safety outcomes: major bleeding, minor bleeding Secondary safety outcomes: wound complications Other outcomes reported: DVT and PE Reported outcomes relevant to this review: Primary safety outcomes Major bleeding: definition: acute, clinically overt bleeding that was fatal, or occurred in critical organs (e.g. retroperitoneal, pericardial, intracranial, intraocular, and spinal cord bleeding), or required re‐operation or significant bleeding in extra‐surgical sites and associated with a fall in the haemoglobin level of at least 20 g/L. Time points measured: 7 to 14 days (treatment) Time points reported: 7 to 14 days (treatment) Minor bleeding: definition: bruising or ecchymosis, incision haematomas, gastrointestinal bleeding, and other bleeding events that were not major events. Time points measured: 7 to 14 days (treatment) Time points reported: 7 to 14 days (treatment) Secondary safety outcomes Wound complications: definition: composite of excessive wound haematoma, reported surgical‐site bleeding, and postoperative wound infection Time points measured: 7 to 14 days (treatment) Time points reported: 7 to 14 days (treatment) Other outcomes reported: DVT (colour Doppler ultrasonography of both lower limbs) and PE (not defined) Time points measured: 7 to 14 days (treatment) Time points reported: 7 to 14 days (treatment) RESULTS Rivaroxaban group: Total VTE: (2+0)/114 Symptomatic VTE: (0+0)/114 postoperative DVT: 2 /114 participants (1.8%) all asymptomatic pulmonary embolism: 0 Major bleeding: 1/114 (bleeding of digestive tract) Minor bleeding: 4/114 (3.4%) Enoxaparin group: Total VTE: (4+0)/114 Symptomatic VTE: (1+0)/114 postoperative DVT: 4/114 participants (3.5%) 3 asymptomatic, 1 symptomatic pulmonary embolism: 0 Major bleeding: 0/114 (bleeding of digestive tract) Minor bleeding: 3/114 (2.6%)
Notes	Source of funding: National Natural Science Foundation of China (81902244), Research Project of Jiangsu Provincial Health Department (H201528),Jiangsu Provincial Medical Youth Talent (QNRC2016801), and Xuzhou Science and Technology Innovation Project (KC19063). Conflicts of interest: author(s) declared no potential conflicts of interest Published protocol: not found Observations: none Missing data requested from authors: none Missing data obtained from authors: none No exclusion or withdrawal was reported for any study group. No CONSORT flow diagram. Quote: "only patients with length of hospital stay longer than 7 days were enrolled in the study, which may have introduced selection bias (patients discharged before postoperative day 7 were excluded from analysis)."
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated random sequence
Allocation concealment (selection bias)	Unclear risk	Computer‐generated random sequence. However, "only patients with length of hospital stay longer than 7 days were enrolled in the study, which may have introduced selection bias (patients discharged before postoperative day 7 were excluded from analysis)."
Blinding of participants and personnel (performance bias)	Low risk	Double‐blind, double‐dummy; groups blinded: not detailed
Blinding of outcome assessment (detection bias)	Unclear risk	Not stated
Incomplete outcome data (attrition bias) Benefit outcomes	Low risk	No exclusion or withdrawal was reported for any study group.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	No exclusion or withdrawal was reported for any study group.
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	Unclear risk	There may be a risk of bias for an inappropriate co‐intervention, but there is either insufficient information to assess whether an important risk of bias exists, or insufficient rationale or evidence that an identified problem will introduce bias.

Berezhnyak 2016.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Blinding: open‐label Exclusions: none Method of concealment of allocation: not mentioned Method of randomisation: not described Primary efficacy analysis set: all treated participants Primary safety analysis set: all treated participants Study duration: 2010 to 2013 Withdrawals: none
Participants	Baseline characteristics Rivaroxaban group Age; mean (range): not reported Gender; F/M: not reported LMWH group Age; mean (range): not reported Gender; F/M: not reported Overall Age; mean (range): 57.3 (30 to 74) Gender; F/M: 195/156 Inclusion criteria: same as in the RECORD 2 and RE‐NOVATE studies: participants ≥ 18 years and > 40 kg, scheduled for primary elective unilateral total hip replacement who provided signed informed consent Exclusion criteria: same as in the RECORD 2 and RE‐NOVATE studies: "Any bleeding diathesis; History of acute intracranial disease or hemorrhagic stroke, major surgery, trauma, uncontrolled hypertension or myocardial infarction within the past 3 months; Gastrointestinal or urogenital bleeding or ulcer disease within the past 6 months; Severe liver disease; Aspartate aminotransferase or Alanine Aminotransferase (ALT) levels more than two times the Upper Limit of the Normal Range (ULN) within the past month; Severe renal insufficiency (creatinine clearance < 30 mL min−1); Concomitant long‐acting non‐steroidal anti‐inflammatory drug therapy (also contraindicated during study treatment); Active malignant disease; Pregnancy" Pretreatment: not reported Diagnosis: coxarthrosis of 2–3 degrees scheduled for hip replacement Location of participants: single centre, Russia Number of participants randomised: 351
Interventions	Rivaroxaban group Intervention: 10 mg once daily Starting time: 8 hours after surgery Duration: 30 days Concomitant interventions: elastic compression of the lower extremities and early activation Excluded interventions: long‐acting NSAIDs LMWH group Intervention: 40 mg once daily Starting time: 12 hours before surgery Duration: 10 days Concomitant interventions: elastic compression of the lower extremities and early activation Excluded interventions: long‐acting NSAIDs
Outcomes	None of the review outcomes were analysed.
Notes	Missing data requested from authors: none Missing data obtained from authors: none No exclusion or withdrawal was reported for any study group. No data for any review outcome was reported. No CONSORT flow diagram. The source of funding was not stated. Authors declared no conflicts of interest. The administration of the intervention for a longer period of time than that for the active control (5 weeks versus 2 weeks) implies that the control may have been favoured in the design of the study, since follow‐up was longer than 2 weeks and the time analysis of the outcomes was not stated.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Randomisation was stated, but the method was not described. Of note, the number of participants in each study group differed markedly.
Allocation concealment (selection bias)	Unclear risk	Not reported. Quote: "Material and Methods the period from 2010 to 2013, a prospective randomized study was conducted based on the department of orthopedics and anesthesiology and the department of reanimation of the Regional clinical hospital in Barnaul. The Committee on Biomedical Ethics approved the work. The study included 351 patients:"
Blinding of participants and personnel (performance bias)	High risk	The study was open‐label for the comparison of rivaroxaban to LMWHs.
Blinding of outcome assessment (detection bias)	Unclear risk	No blinding of any group was reported
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	Unclear risk	There may be a risk of bias for an inappropriate intervention and a co‐intervention, but there was either insufficient information to assess whether an important risk of bias exists, or insufficient rationale or evidence that an identified problem would introduce bias. The administration of the intervention for a longer period of time than that for the active control (5 weeks versus 2 weeks) implies that the control may have been favoured in the design of the study, since follow‐up was longer than 2 weeks and the time analysis of the outcomes was not stated.

Changchun 2019.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Blinding: not described Exclusions: none Method of concealment of allocation: not described Method of randomisation: not described Primary efficacy analysis set: all treated participants Primary safety analysis set: all treated participants Study duration: from 2014 to 2017 Withdrawals: none
Participants	Baseline characteristics Rivaroxaban group Age; mean (range): 68.8 ± 6.5 Gender; F/M: 102/100 LMWH group Age; mean (range): 66.3 ± 14.0 Gender; F/M: 99/96 Overall Age; mean (range): 67.4 ± 11.3 Gender; F/M: 201/196 Inclusion criteria: scheduled to unilateral hip or knee arthroplasty. Preoperative thromboelastogram Exclusion criteria: "congenital coagulation disorder; Liver or kidney failure (creatinine clearance <30mL/min); Acute cardiovascular or cerebrovascular diseases; Hypertension; Malignancy; History of thrombosis; Preoperative use of anticoagulants, antiplatelet therapy" Pretreatment: no statistically significant differences were observed Diagnosis: arthrosis, rheumatoid arthritis, necrosis of the femur head scheduled to THR, or TKR Location of participants: single centre (China) Number of participants randomised: 397 Missing data: 0
Interventions	Rivaroxaban group Intervention: rivaroxaban (oral) 10 mg once daily Starting time: 12 hours postoperatively Duration: not reported Concomitant interventions: not reported Excluded interventions: not reported LMWH group Intervention: LMWH (subcutaneous) 4000 IU once daily Starting time: 12 hours postoperatively Duration: not reported Concomitant interventions: not reported Excluded interventions: not reported
Outcomes	None of the review outcomes were analysed.
Notes	Source of funding: China National High‐Tech Research and Technology Program (863 Program), No. 2013AA032203; Northern Sichuan Scientific Research Development Program, No. CBY17‐A‐YB12. Conflict of interest: "The therapeutic intervention in the article uses artificial hip and knee prostheses. All authors of the article declare that they have not received funding for this device. There is no conflict of interest during the research and writing of the article." Missing data requested from authors: none Missing data obtained from authors: none No missing data reported. None of the reported outcomes were included in the review.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding of participants and personnel (performance bias)	High risk	Not described. Open‐label design can be assumed
Blinding of outcome assessment (detection bias)	Unclear risk	Not described. Open‐label design can be assumed
Other bias	Unclear risk	There may be a risk of bias for an inappropriate co‐intervention, but there is either insufficient information to assess whether an important risk of bias exists, or insufficient rationale or evidence that an identified problem will introduce bias.

Chen 2016.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: not described Study duration: October 2011 to July 2014 Duration of intervention: 35 days Duration of follow‐up: 35 days Method of randomisation: not described Method of concealment of allocation: not mentioned Primary efficacy analysis set: all randomised participants who completed the study Primary safety analysis set: same as efficacy set Withdrawals: rivaroxaban group: 1 lost to follow‐up, enoxaparin group: 1 lost to follow‐up Exclusions: no participant was excluded
Participants	Diagnosis: scheduled to THA or TKA Location of participants: single centre (China) Number of participants randomised: 270 Age (years; mean ± SD): rivaroxaban group 72.7 ± 8.5; enoxaparin group 69.9 ± 10.4 Gender (male %): rivaroxaban group 43%; enoxaparin group 38% Baseline imbalances: statistically significant difference in age Inclusion criteria Hip‐knee osteoarthritis or rheumatoid arthritis Peri‐hip fracture‐caused total hip replacement or internal fixation Exclusion criteria Elderly participants Those who showed multiple complications Those who could not tolerate the surgery Known hypercoagulable state Active bleeding or bleeding trend Obvious liver diseases (such as acute hepatitis, chronic active hepatitis, cirrhosis) Coagulation disorder before or after the surgery Perioperative infections
Interventions	Number of intervention groups: 2 Concomitant interventions: none Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 12 hours Duration: 35 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: after surgery: 12 hours Duration: until discharge
Outcomes	Number of outcomes: According to protocol: no protocol is available OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 35 days Time points reported: 35 days Symptomatic VTE: Definition: incidence of suspected DVT confirmed through colour Doppler ultrasonography Time points measured: 35 days Time points reported: 35 days Fatal VTE: Definition: incidence of death due to VTE Time points measured: 35 days Time points reported: 35 days RESULTS All‐cause mortality: Summary data (35 days): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 2 missing participants ‐Subgroups reported: none Symptomatic VTE: Summary data (35 days): see Analysis 1.3 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 2 missing participants ‐Subgroups reported: none Fatal VTE: Summary data (35 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 2 missing participants ‐Subgroups reported: none
Notes	Source of funding: not declared Conflicts of interest: none Published protocol: none Observations: none Missing data requested from authors: specific groups blinded, description of method for randomisation and concealment of allocation, duration (mean and SD) of treatment for enoxaparin group, source of funding, protocol of study Missing data obtained from authors: none Duration of prophylaxis differed markedly for each intervention group and the outcome analysis were at different time points; 5 weeks versus +‐2 weeks. For the analysis set, 0.3% of the randomised participants were excluded (0.7% in the rivaroxaban group and 0.8% in the enoxaparin group). Reasons for exclusion were reported. One participant was lost in each group. Reasons not provided. No CONSORT flow diagram.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Unclear risk	Blinding of the study was not stated. No blinding of any group was reported
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of the study was not stated. No blinding of any group was reported
Incomplete outcome data (attrition bias) Benefit outcomes	Unclear risk	Missing outcome data reported. Insufficient information about reasons and numbers with missing outcome data. Proportion of missing outcomes compared with the observed event risk is too small to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome. Proportion of missing outcomes compared with the observed event risk is too small to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	High risk	There is at least one important risk of bias. The study had a potential source of bias related to the specific study design used (inappropriate intervention as duration of prophylaxis differed markedly for each intervention group, and the outcome analysis was at different time points; 5 weeks versus 2 weeks). The intervention could be favoured in the design of the study, since follow‐up was longer than 2 weeks and the analysis of the outcomes were at different times.

Cohen 2013.

Study characteristics
Methods	Aim: dose‐response study, efficacy and safety Design: parallel‐group/dose‐escalation (phase II) Blinding: open‐label (for the eribaxaban versus enoxaparin comparisons); groups blinded: outcome adjudicators Study duration: not reported Duration of intervention: 6 to 14 days Duration of follow‐up: 32 ± 4 days Method of randomisation: stratified by region (North America/other), no additional description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: not reported Exclusions: eribaxaban groups (not detailed according to dose group): 17 did not receive study drug, 274 did not undergo venography, 157 had uninterpretable venography; enoxaparin 30 mg twice daily: 5 did not receive study drug, 165 did not undergo venography, 44 had uninterpretable venography
Participants	Diagnosis: scheduled to TKR, unilateral Location of participants: Australia, Europe, America and South Africa Number of participants randomised: 1411 Age (years; mean and range): (safety analysis population) eribaxaban 0.1 mg / 0.3 mg / 0.5 mg / 1 mg / 2.5 mg / 4 mg / 10 mg (once daily) 67.3 (47‐86) / 67.0 (43‐86) / 68.4 (49‐85) / 66.6 (38‐87) / 67.2 (42‐87) / 65.8 (45‐84) / 65.8 (28‐87); enoxaparin 66.8 (28‐88) Gender (male %): (safety analysis population) eribaxaban 0.1 mg / 0.3 mg / 0.5 mg / 1 mg / 2.5 mg / 4 mg / 10 mg (once daily) 25 (41.0%) / 44 (31.2%) / 59 (32.2%) / 73 (36.1%) / 75 (37.5%) / 51 (36.4%) / 28 (43.1%); enoxaparin 140 (35.3%) Baseline imbalances: slight differences regarding gender among the groups Inclusion criteria At least 18 years of age Scheduled to undergo unilateral TKR Women were either postmenopausal or surgically sterilised Signed an informed consent document Exclusion criteria Low body weight (< 45 kg for women, < 57 kg for men) DVT or PE within the past year, or suspected post‐thrombotic state Significant bleeding Anaemia (haemoglobin level < 10.0 g/dL) Thrombocytopenia (platelet count < 100x103/µL) Uncontrolled hypertension Creatinine clearance < 30 mL/min (based on the Cockcroft‐Gault formula) Active hepatic disease Cancer Immunocompromised status Ischaemic stroke or myocardial infarction within 3 months Surgery/trauma within 6 months Persistent immobilisation Current use of anticoagulants, platelet aggregation inhibitors, or any other drug influencing coagulation History of significant adverse reaction to heparin
Interventions	Number of intervention groups: 8 Concomitant interventions: none as part of the study; elastic compression stockings were allowed Excluded interventions: none Eribaxaban 0.1 mg once daily group: Intervention: eribaxaban (oral) 0.1 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 6 to 14 days Eribaxaban 0.3 mg once daily group: Intervention: eribaxaban (oral) 0.3 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 6 to 14 days Eribaxaban 0.5 mg once daily group: Intervention: eribaxaban (oral) 0.5 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 6 to 14 days Eribaxaban 1 mg once daily group: Intervention: eribaxaban (oral) 1 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 6 to 14 days Eribaxaban 2.5 mg once daily group: Intervention: eribaxaban (oral) 2 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 6 to 14 days Eribaxaban 4 mg once daily group: Intervention: eribaxaban (oral) 4 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 6 to 14 days Eribaxaban 10 mg once daily group: Intervention: eribaxaban (oral) 10 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 6 to 14 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 30 mg twice daily Starting time: after surgery: 12‐24h Duration: 6 to 14 days
Outcomes	Number of outcomes: According to protocol: no protocol is available. OUTCOMES Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2 g/dL within 24h; 2) transfusion of ≥ 2 U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) fatal bleeding; 5) intramuscular bleeding with compartment syndrome Time points measured: 6 to 14 days (treatment), 32 ± 4 days (additionally, follow‐up) Time points measured: 6 to 14 days (treatment) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 6 to 14 days (treatment), 32 ± 4 days (additionally, follow‐up) Time points reported: 6 to 14 days (treatment), 32 ± 4 days (additionally, follow‐up) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 6 to 14 days (treatment), 32 ± 4 days (additionally, follow‐up) Time points reported: 6 to 14 days (treatment), 32 ± 4 days (additionally, follow‐up) RESULTS Major bleeding: Summary data (6 to 14 days): eribaxaban 0.1 mg once daily ‐ 0.3 mg once daily ‐ 0.5 mg once daily ‐ 1 mg once daily ‐ 2.5 mg once daily ‐ 4 mg once daily ‐ 10 mg once daily ‐ enoxaparin 30 mg twice daily: 0/61 ‐ 0/141 ‐ 2/183 ‐ 1/202 ‐ 1/200 ‐ 0/140 ‐ 1/65 ‐ 3/397 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 22 missing participants ‐Subgroups reported: none Summary data (32 ± 4 days follow‐up): not reported Fatal VTE: Summary data (6 to 14 days): eribaxaban 0.1 mg once daily ‐ 0.3 mg once daily ‐ 0.5 mg once daily ‐ 1 mg once daily ‐ 2.5 mg once daily ‐ 4 mg once daily ‐ 10 mg once daily ‐ enoxaparin 30 mg twice daily: 0/35 ‐ 0/89 ‐ 0/104 ‐ 0/120 ‐ 0/112 ‐ 0/74 ‐ 0/27 ‐ 0/188 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 662 missing participants ‐Subgroups reported: none Summary data (32 ± 4 days follow‐up): eribaxaban 0.1 mg once daily ‐ 0.3 mg once daily ‐ 0.5 mg once daily ‐ 1 mg once daily ‐ 2.5 mg once daily ‐ 4 mg once daily ‐ 10 mg once daily ‐ enoxaparin 30 mg twice daily: 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Fatal bleeding: Summary data (6 to 14 days): eribaxaban 0.1 mg once daily ‐ 0.3 mg once daily ‐ 0.5 mg once daily ‐ 1 mg once daily ‐ 2.5 mg once daily ‐ 4 mg once daily ‐ 10 mg once daily ‐ enoxaparin 30 mg twice daily: 0/61 ‐ 0/141 ‐ 0/183 ‐ 0/202 ‐ 0/200 ‐ 0/140 ‐ 0/65 ‐ 0/397 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 22 missing participants ‐Subgroups reported: none Summary data (32 ± 4 days follow‐up): eribaxaban 0.1 mg once daily ‐ 0.3 mg once daily ‐ 0.5 mg once daily ‐ 1 mg once daily ‐ 2.5 mg once daily ‐ 4 mg once daily ‐ 10 mg once daily ‐ enoxaparin 30 mg twice daily: 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none
Notes	Source of funding: Pfizer, Inc. Conflicts of interest: authors (5) had financial relationships with several pharmaceutical companies, including Pfizer Published protocol: registered in clinicaltrials.gov as NCT00306254, and in ICTRP as EUCTR2005‐005179‐14‐IT Observations: none Missing data requested from authors: specific groups blinded, date of start and end of study, description of the method of randomisation and concealment of allocation, details about withdrawals and exclusions, definition of serious AEs, treatment period summary data for serious AEs, follow‐up summary data for all‐cause mortality, major bleeding and serious AEs, treatment period number of participants for all‐cause mortality, follow‐up number of participants for fatal VTE and fatal bleeding Missing data obtained from authors: none For the efficacy set, 46.9% of the randomised participants were excluded (44.4% in the eribaxaban groups and 53.2% in the enoxaparin group). Reasons for exclusion were reported. Imbalance in reasons for missing data across intervention groups. Symptomatic VTE events in all the randomised patients not reported. For the safety set, 1.6% of the randomised participants were excluded (1.7% in the eribaxaban groups and 1.2% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Randomisation was stated, but the method not described. Of note, under the dose‐escalation design, the probability of participants being allocated to each eribaxaban group was different
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	High risk	Study intervention was open‐label for the eribaxaban versus enoxaparin comparisons
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment (except for all‐cause mortality) was performed by an independent, blinded imaging core laboratory. Blinding of outcome assessment was not mentioned for safety outcomes. Blinding of the data analysts was not mentioned.
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data balanced in reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk is insufficient to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	Unclear risk	The protocol was published as a conference abstract only
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan describing procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

DARINA 2021.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Blinding: open‐label Exclusions: 5 participants were excluded from the analysis set Method of concealment of allocation: sealed envelopes Method of randomisation: computer‐generated randomised list Primary efficacy analysis set: per‐protocol analysis set Primary safety analysis set: per‐protocol analysis set Study duration: from July 2012 to November 2013 Withdrawals: 5 participants withdrew from the study
Participants	Baseline characteristics Rivaroxaban group Age; mean (range): 65 ± 7.5 Gender; F/M: 30/18 Nadroparin group Age; mean (range): 67 ± 10 Gender; F/M: 31/14 Overall Age; mean (range): 66 ± 8.7 Gender; F/M: 87/51 Inclusion criteria: people ≥ 18 years old; weighing more than 40 kg; scheduled for primary elective TKA; provided signed informed consent Exclusion criteria: "Known inherited or acquired clinically significant active high risk of bleeding or bleeding disorder; Major surgery, trauma, uncontrolled severe arterial hypertension or myocardial infarction within the last 3 months; History of acute intracranial disease or haemorrhagic stroke; Gastrointestinal or urogenital bleeding or ulcer disease within the last 6 months; Cirrhotic patients with moderate hepatic impairment (aspartate or alanine aminotransferase levels higher than twice the upper limit of the normal range within the last month); Severe renal insufficiency (creatinine clearance <30mL/min); Other indication for treatment with anticoagulants; Active malignant disease; Pregnancy or breastfeeding" Pretreatment: baseline characteristic differences were not tested for statistical significance Diagnosis: scheduled to TKA Location of participants: single centre (the Netherlands) Number of participants randomised: 148 (98 for nadroparin and rivaroxaban groups) Exclusions/lost to follow‐up: 2 for nadroparin and 3 for rivaroxaban
Interventions	Rivaroxaban group Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery Duration: 6 weeks Concomitant interventions: standard orthopaedic care and physiotherapy Excluded interventions: none Nadroparin group Intervention: nadroparin 0.3 mL (subcutaneous) once daily Starting time: after surgery Duration: 6 weeks Concomitant interventions: standard orthopaedic care and physiotherapy Excluded interventions: none Dabigatran group (direct thrombin inhibitor: not included in the analysis)
Outcomes	Symptomatic VTE (instrument for diagnosis not stated) Major bleeding Fatal VTE Minor adverse events Clinically relevant non‐major bleeding (outcome not included in this review)
Notes	Source of funding: "The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not‐for‐profit sectors." Conflicts of interest: "MvH reports grants from Bayer and personal fees from Boehringer Ingelheim during the conduct of the study. Other authors have nothing to report." Missing data requested from authors: none Missing data obtained from authors: none For the analysis sets, 5.1% of the randomised participants were excluded (4.3% in the nadroparin group and 5.9% in the rivaroxaban group). Reasons for exclusion were reported. 2 participants in rivaroxaban group were lost to follow‐up with no explanation provided. Imbalance in reasons for missing data across intervention groups. Different types of knee surgery between groups. "(80%) patients had fast‐track TKA. Compared with patients in the other treatment groups, a larger share of patients in the dabigatran group had fast‐track TKA. Thirty‐three (24%) patients received a patella prosthesis during TKA. Compared with the nadroparin group, a larger share of patients in the rivaroxaban group and smaller share of patients in the dabigatran group received a new patella."
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomised list
Allocation concealment (selection bias)	Low risk	Sealed envelopes
Blinding of participants and personnel (performance bias)	High risk	Open‐label for the comparison between rivaroxaban and nadroparin
Blinding of outcome assessment (detection bias)	Unclear risk	Open‐label for the comparison of interest
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	High risk	Missing outcome data reported. Not provided enough information regarding reasons for missing outcome data. Missing outcome data unbalanced in reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: all but one reported outcome were included in the protocol. Selective reporting involved several outcomes included in the review
Other bias	High risk	There is at least one important risk of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review.

EXPERT 2009.

Study characteristics
Methods	Aim: exploratory study, efficacy and safety Design: parallel‐group Blinding: open‐label (for the betrixaban versus enoxaparin comparisons); groups blinded: outcome adjudicators Study duration: May 2006 to November 2006 Duration of intervention: 10 to 14 days Duration of follow‐up: 4 ± 2 weeks Method of randomisation: in a 2:2:1 ratio, no additional description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: participants who had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: betrixaban 15 mg twice daily: 2 (1 withdrew consent, 1 refused); betrixaban 40 mg twice daily: 7 (4 withdrew consent, 3 had adverse event); enoxaparin 30 mg twice daily: 0 Exclusions: betrixaban 15 mg twice daily: 15 (9 had no venography, 2 had uninterpretable venography (not in time window), 2 had abnormal laboratory values, 1 was withdrawn because of physician's decision, 1 was lost to follow‐up); betrixaban 40 mg twice daily: 8 (5 had no venography, 1 had uninterpretable venography (not in time window), 2 had abnormal laboratory values); enoxaparin 30 mg twice daily: 2 (1 had no venography, 1 was withdrawn because of investigator's decision)
Participants	Diagnosis: scheduled to TKR, elective, unilateral Location of participants: 2 countries (USA and Canada) Number of participants randomised: 215 Age (years; mean and range): (all randomised population) betrixaban 15 mg twice daily: 63 (47‐75); betrixaban 40 mg twice daily: 65 (47‐75); enoxaparin 30 mg twice daily: 62 (43‐75) Gender (male %): (all randomised population) betrixaban 15 mg twice daily: 32 (36.8%); betrixaban 40 mg twice daily: 32 (38.1%); enoxaparin 30 mg twice daily: 21 (48.8%) Baseline imbalances: slight differences regarding gender among the groups Inclusion criteria: Men and women between 18 and 75 years of age and weighing between 50 and 120 kg Scheduled to undergo elective primary unilateral TKR Female participants were required to be without reproductive potential, i.e. postmenopausal for ≥ 2 years or after hysterectomy Exclusion criteria: Patients with bleeding disorders, a recent episode of internal bleeding, or at high risk for bleeding Patients with platelet count < 100,000/mm3, haemoglobin < 10 g/dL or haematocrit < 30%
Interventions	Number of intervention groups: 3 Concomitant interventions: none Excluded interventions: thrombolytics agents and anticoagulants (within 7 days prior to surgery and throughout the treatment period) Betrixaban 15 mg twice daily group: Intervention: betrixaban (oral) 15 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 10 to 14 days Betrixaban 40 mg twice daily group: Intervention: betrixaban (oral) 40 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 10 to 14 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 30 mg twice daily Starting time: after surgery: 12‐24h Duration: 10 to 14 days
Outcomes	Number of outcomes: According to protocol: no protocol is available. OUTCOMES Symptomatic VTE: Definition: incidence of symptomatic VTE Time points measured: 10 to 14 days (treatment), 4 ± 2 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 10 to 14 days (treatment), 4 ± 2 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Asymptomatic distal DVT: Definition: incidence of asymptomatic distal DVT Time points measured: 10 to 14 days (treatment), 4 ± 2 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 10 to 14 days (treatment), 4 ± 2 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Liver enzymes elevation: Definition: incidence of ALT > 3 times ULN Time points measured: 10 to 14 days (treatment), 4 ± 2 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) RESULTS Symptomatic VTE: Summary data (10 to 14 days): see Analysis 9.3 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 40 missing participants (6 unexplained) ‐Subgroups reported: none Summary data (4 ± 2 weeks follow‐up): not reported Fatal VTE: Summary data (10 to 14 days): betrixaban 15 mg twice daily ‐ 40 mg twice daily ‐ enoxaparin 30 mg twice daily: 0/70 ‐ 0/65 ‐ 0/40 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 40 missing participants (6 unexplained) ‐Subgroups reported: none Summary data (4 ± 2 weeks follow‐up): not reported Asymptomatic distal DVT: Summary data (10 to 14 days): see Analysis 9.10 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 40 missing participants (6 unexplained) ‐Subgroups reported: none Summary data (4 ± 2 weeks follow‐up): not reported Fatal bleeding: Summary data (10 to 14 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 1 missing participant ‐Subgroups reported: none Summary data (4 ± 2 weeks follow‐up): not reported Liver enzymes elevation: Summary data (10 to 14 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 1 missing participant ‐Subgroups reported: none Summary data (4 ± 2 weeks follow‐up): not reported
Notes	Source of funding: Portola Pharmaceuticals Inc. Conflicts of interest: authors (8) had financial relationships with several pharmaceutical companies, including Portola Pharmaceuticals Inc. Published protocol: registered in ClinicalTrials.gov as NCT00375609 Observations: none Missing data requested from authors: specific groups blinded, description of the method of randomisation and concealment of allocation, definition of serious AEs, treatment period summary data for serious AEs, follow‐up summary data for all‐cause mortality, fatal VTE, symptomatic VTE, asymptomatic distal DVT, fatal bleeding, serious AEs and liver enzymes elevation, treatment period number of participants for all‐cause mortality, treatment period missing participants for fatal VTE, symptomatic VTE and asymptomatic distal DVT, protocol of the study Missing data obtained from authors: none For the efficacy set, 18.6% of the randomised participants were excluded (21.5% in the betrixaban groups and 7.0% in the enoxaparin group). Reasons for exclusion were reported except for 5 (2.3%) participants. Reason for missing outcome data is likely to be related to adverse event outcome, with imbalance in reasons for missing data across intervention groups. For the safety set, 0.5% of the randomised participants were excluded (0.6% in the betrixaban groups and 0% in the enoxaparin group). Reasons for exclusion were reported. Imbalance in reasons for missing data across intervention groups.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	High risk	Study intervention was open‐label for the betrixaban versus enoxaparin comparisons
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded central adjudication committee. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Reasons for missing outcome data related to true outcome. Missing outcome data unbalanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Reasons for missing outcome data related to true outcome. Missing outcome data balanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

FOXTROT 2020.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Blinding: open‐label Exclusions: 50 participants from the enoxaparin and apixaban groups were excluded after randomisation. Reasons were reported Method of concealment of allocation: not mentioned Method of randomisation: participants were randomised on study day 1 (the day before surgery) using a single‐stream and block method, in which 2 lists of random numbers were generated using a standard SAS randomisation program. The first list was used to randomise participants in a 5:1:1ratio to osocimab, enoxaparin, and apixaban with a block size of 14; the second list was used to randomise participants equally between the doses of osocimab with a block size of 15. Primary efficacy analysis set: per‐protocol analysis Primary safety analysis set: per‐protocol analysis Study duration: October 2017 until January 2019 Withdrawals: 7 participants withdrew after randomisation
Participants	Baseline characteristics Apixaban group Age; mean ± SD: 64.9 ± 8.4 Gender; F/M: 65/18 Enoxaparin group Age; mean ± SD: 67 ± 8.8 Gender; F/M: 55/21 Inclusion criteria: people 18 years or older undergoing elective, primary, unilateral, total knee arthroplasty, who were willing to adhere to the study procedures Exclusion criteria: "Active bleeding or a high risk of bleeding; A history of brain, spinal, or ophthalmologic surgery within the 3 months before randomisation. Sustained uncontrolled hypertension. Body weight of more than 135 kg. Hemoglobin level of less than 10 g/dL for men and 11 g/dL for women. Platelet count of less than 150 × 109/L. Activated partial thromboplastin time above the upper limit of normal. Calculated creatinine clearance below 60 mL/min. Clinically significant liver disease. A history of prior deep vein thrombosis (DVT). Epidural analgesia after surgery for postoperative dosing, and spinal anaesthesia and epidural analgesia for preoperative dosing" Pretreatment: slight differences regarding age and gender Diagnosis: scheduled to elective, primary, unilateral TKA Location of participants: multicentre, 13 countries (54 centres) Number of participants randomised: 813 (210 for the enoxaparin and apixaban groups)
Interventions	Apixaban group Intervention: 2.5 mg (oral) twice daily Starting time: 12 to 24 hours postoperatively Duration: 10 to 13 days Concomitant interventions: none Excluded interventions: none Enoxaparin group Intervention: 40 mg (subcutaneous) once daily Starting time: started either in the evening before surgery or 6 to 8 hours postoperatively (at the investigator’s discretion) Duration: 10 to 13 days Concomitant interventions: none Excluded interventions: none Oxocimab group (factor XI inhibitor): not included in the analysis
Outcomes	All cause mortality Major VTE (bilateral venography) Symptomatic VTE Major bleeding Fatal VTE Asymptomatic distal DVT Fatal bleeding Minor adverse events
Notes	Source of funding: Bayer HealthCare All authors (8) had financial relationships with the sponsor Quote: "An independent committee, whose members were blinded to treatment assignment, adjudicated all venograms for the presence of thrombosis and all clinically suspected episodes of VTE or bleeding. A separate, independent data and safety monitoring committee reviewed trial outcomes at regular intervals." Independent/blinding statistical analyst: no. Quote: "Role of the Funder/Sponsor: Working in concert with the academic members of the steering committee, the sponsor contributed to the design and conduct of the study; collection, analysis, and interpretation of the data, and preparation, review, and approval of the manuscript. The data derived from the study are the property of the sponsor." Missing data requested from authors: none Missing data obtained from authors: none For the primary analysis set, 24.3% of the randomised participants were excluded (27.6% in the enoxaparin group and 20.9% in the apixaban group). Reasons for exclusion were reported. Reasons for exclusion unbalanced in numbers and reasons across intervention groups. Study did not report the symptomatic VTE events in all the randomised participants.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computarised randomisation was performed. Quote: "standard SAS randomization program."
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	High risk	Open‐label for the comparison between enoxaparin and apixaban
Blinding of outcome assessment (detection bias)	Low risk	Blinded committee assessed outcomes
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome. Missing outcome data unbalanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	Low risk	Reported primary efficacy and safety outcomes were included in the published NCT protocol (NCT03276143)
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

Fuji 2014‐THA.

Study characteristics
Methods	Aim: efficacy and safety Design: parallel‐group (phase II/III) Blinding: open‐label (for the darexaban versus enoxaparin comparisons); groups blinded: none Study duration: May 2009 to March 2010 Duration of intervention: 10 to 14 days Duration of follow‐up: 3 to 5 weeks Method of randomisation: no description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) underwent planned surgery, and 3) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: darexaban 15 mg twice daily: 11 (reasons not detailed); darexaban 30 mg twice daily: 17 (reasons not detailed); enoxaparin 20 mg twice daily: 7 (reasons not detailed); placebo: 11 (reasons not detailed) Exclusions: darexaban 15 mg twice daily: 33 (reasons not detailed); darexaban 30 mg twice daily: 40 (reasons not detailed); enoxaparin 20 mg twice daily: 22 (reasons not detailed); placebo: 34 (reasons not detailed); 12 discontinued treatment (allocated groups not reported); not specified if withdrawals are included in exclusions
Participants	Diagnosis: scheduled to THA, elective Location of participants: 4 countries (Japan, Taiwan, Korea, Thailand) Number of participants randomised: 622 Age (years; mean ± SD): (primary efficacy analysis population) darexaban 15 mg twice daily: 62.1 ± 10.5; darexaban 30 mg twice daily: 61.0 ± 10.8; enoxaparin 20 mg twice daily: 61.6 ± 11.0; placebo: 62.6 ± 10.3 Gender (male %): (primary efficacy analysis population) darexaban 15 mg twice daily: 23 (16.9%); darexaban 30 mg twice daily: 29 (21.6%); enoxaparin 20 mg twice daily: 20 (24.4%); placebo: 28 (21.7%) Baseline imbalances: slight differences regarding gender, type of anaesthesia and mechanical prophylaxis Inclusion criteria: Adult males and females (aged 20 years or older) scheduled for elective THA Exclusion criteria: History of DVT or PE Expected confinement to bed for at least 7 days prior to surgery Hemorrhagic or coagulation disorders, or idiopathic‐/heparin‐induced thrombocytopenia Concomitant use of antiplatelet or anticoagulant therapy, or planned treatment with these agents during the period from 1 week prior to surgery until the end of venography History of major trauma, major surgery, or eye, spinal cord, or brain surgery within 90 days prior to surgery Other scheduled lower extremity surgery during the study period Clinically significant bleeding within 90 days prior to the screening visit Use of an intrathecal or epidural catheter that could not be removed 2 hours prior to the start of treatment Diagnosed uncontrolled, moderate or severe hypertension (systolic and/or diastolic blood pressure 160 and 100 mmHg, respectively) Gastrointestinal ulcer within 180 days prior to the screening visit Myocardial infarction or stroke within 180 days prior to surgery Weight < 40 kg at screening Diagnosed bacterial endocarditis Diagnosed hepatic impairment, and/or renal impairment (defined as serum creatinine 1.5 times the ULN and/or creatinine clearance < 30 mL/min) Diagnosed retinopathy Malignancy currently under treatment
Interventions	Number of intervention groups: 4 Concomitant interventions: none as part of the study; intermittent pneumatic compression and elastic stocking were allowed Excluded interventions: none Darexaban 15 mg twice daily group: Intervention: darexaban (oral) 15 mg twice daily Starting time: after surgery: 12 to 24 hours Duration: 10 to 14 days Darexaban 30 mg twice daily group: Intervention: darexaban (oral) 30 mg twice daily Starting time: after surgery: 12 to 24 hours Duration: 10 to 14 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 20 mg twice daily Starting time: after surgery: 24‐36h Duration: 10 to 14 days Placebo group: Intervention: placebo (oral) twice daily Starting time: after surgery: 12 to 24 hour Duration: 10 to 14 days
Outcomes	Number of outcomes: According to protocol: no protocol is available OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Symptomatic VTE: Definition: incidence of symptomatic VTE Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Major bleeding: Definition: incidence of major bleeding (definition not reported) Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Serious AEs: non‐hepatic: Definition: number of patients with at least one serious AE (definition not provided), excluding those liver‐related Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Asymptomatic distal DVT: Definition: incidence of asymptomatic distal DVT Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Liver enzymes elevation: Definition: incidence of ALT > 3 times ULN Time points measured: 10 to 14 days (treatment) Time points reported: 10 to 14 days (treatment) Minor AEs: Definition: number of patients with at least one mild AE (definition not provided) Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) RESULTS All‐cause mortality: Summary data (10 to 14 days): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): darexaban 15 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 20 mg twice daily ‐ placebo (oral) twice daily: 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Major VTE: Summary data (10 to 14 days): darexaban 15 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 20 mg twice daily ‐ placebo (oral) twice daily: 1/144 ‐ 1/144 ‐ 0/NR ‐ 5/137 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, not calculable missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): not reported Symptomatic VTE: Summary data (10 to 14 days): darexaban 15 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 20 mg twice daily ‐ placebo (oral) twice daily: 0/NR ‐ 1/174 ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, not calculable missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): not reported Major bleeding: Summary data (10 to 14 days): see Analysis 1.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 13 missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): not reported Serious AEs: non‐hepatic Summary data (10 to 14 days): see Analysis 9.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 13 missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): not reported Fatal VTE: Summary data (10 to 14 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 141 missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): darexaban 15 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 20 mg twice daily ‐ placebo (oral) twice daily: 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, not calculable missing participants ‐Subgroups reported: none Asymptomatic distal DVT: Summary data (10 to 14 days): see Analysis 9.10 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 145 missing participants (4 unexplained) ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): not reported Fatal bleeding: Summary data (10 to 14 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 13 missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): darexaban 15 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 20 mg twice daily ‐ placebo (oral) twice daily: 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, not calculable missing participants ‐Subgroups reported: none Liver enzymes elevation: Summary data (10 to 14 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 22 missing participants (9 unexplained) ‐Subgroups reported: none Minor AEs: Summary data (10 to 14 days): see Analysis 9.13 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 13 missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): not reported
Notes	Source of funding: Astellas Conflicts of interest: authors (3) had financial relationships with several pharmaceutical companies, including Astellas Published protocol: registered in clinicaltrials.gov as NCT00913120 Observations: none Missing data requested from authors: specific groups blinded, description of the method of randomisation and concealment of allocation, details about withdrawals and exclusions, definition of major bleeding, serious AEs and minor AEs, follow‐up summary data for major VTE, asymptomatic distal DVT, major bleeding, serious AEs, minor AEs, treatment period number of participants for all‐cause mortality, major VTE and symptomatic VTE, follow‐up number of participants for all‐cause mortality, fatal VTE, fatal bleeding, protocol of the study Missing data obtained from authors: information regarding blinding and withdrawals; data on major bleeding (number of events) and all‐cause mortality (number of participants) For the efficacy set, > 22.7% of the randomised participants were excluded (> 21.3% in the darexaban groups and > 21.2% in the enoxaparin group). Besides, for the efficacy outcomes included in this review, data from several (not defined) sets were included. Reasons for exclusion were not reported. CONSORT flow diagram incomplete. For the safety set, 2.1% of the randomised participants were excluded (data were insufficient to calculate the proportions for each group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	High risk	Study intervention was open‐label for the darexaban versus enoxaparin comparisons
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of outcome assessment was not mentioned. Blinding of the data analysts is not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data.
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

Fuji 2014‐TKA.

Study characteristics
Methods	Aim: efficacy and safety Design: parallel‐group (phase II/III) Blinding: open‐label (for the darexaban versus enoxaparin comparisons); groups blinded: none Study duration: June 2009 to January 2010 Duration of intervention: 10 to 14 days Duration of follow‐up: 3 to 5 weeks Method of randomisation: no description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) underwent planned surgery, and 3) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: darexaban 15 mg twice daily: 4 (reasons not detailed); darexaban 30 mg twice daily: 5 (reasons not detailed); enoxaparin 20 mg twice daily: 12 (reasons not detailed); placebo: 11 (reasons not detailed) Exclusions: darexaban 15 mg twice daily: 14 (reasons not detailed); darexaban 30 mg twice daily: 17 (reasons not detailed); enoxaparin 20 mg twice daily: 25 (reasons not detailed); placebo: 25 (reasons not detailed); 17 discontinued treatment (allocated groups not reported); not specified if withdrawals are included in exclusions
Participants	Diagnosis: scheduled to TKA, elective Location of participants: 3 countries (Japan, Taiwan, Thailand) Number of participants randomised: 386 Age (years; mean ± SD): (primary efficacy analysis population) darexaban 15 mg twice daily: 72.5 ± 7.0; darexaban 30 mg twice daily: 71.2 ± 7.9; enoxaparin 20 mg twice daily: 72.3 ± 8.0; placebo: 73.9 ± 7.1 Gender (male %): (primary efficacy analysis population) darexaban 15 mg twice daily: 9 (11.1%); darexaban 30 mg twice daily: 9 (12.7%); enoxaparin 20 mg twice daily: 9 (13.6%); placebo: 7 (9.7%) Baseline imbalances: slight differences regarding underlying disease Inclusion criteria: Male and female adults (aged 20 years or older) scheduled for elective TKA Exclusion criteria: History of DVT or PE Expected confinement to bed for at least 7 days prior to surgery Hemorrhagic or coagulation disorders, or idiopathic‐/heparin‐induced thrombocytopenia Concomitant use of antiplatelet or anticoagulant therapy, or planned treatment with these agents during the period from 1 week prior to surgery until the end of venography History of major trauma, major surgery, or eye, spinal cord, or brain surgery within 90 days prior to surgery Other scheduled lower extremity surgery during the study period Clinically significant bleeding within 90 days prior to the screening visit Use of an intrathecal or epidural catheter that could not be removed 2 hours prior to the start of treatment Diagnosed uncontrolled, moderate or severe hypertension (systolic and/or diastolic blood pressure 160 and 100 mmHg, respectively) Gastrointestinal ulcer within 180 days prior to the screening visit Myocardial infarction or stroke within 180 days prior to surgery Weight < 40 kg at screening Diagnosed bacterial endocarditis Diagnosed hepatic impairment, and/or renal impairment (defined as serum creatinine 1.5 times the upper limit of normal [ULN] and/or creatinine clearance < 30 mL/min) Diagnosed retinopathy Malignancy currently under treatment
Interventions	Number of intervention groups: 4 Concomitant interventions: none as part of the study; intermittent pneumatic compression and elastic stocking were allowed Excluded interventions: none Darexaban 15 mg twice daily group: Intervention: darexaban (oral) 15 mg twice daily Starting time: after surgery: Duration: 10 to 14 days Darexaban 30 mg twice daily group: Intervention: darexaban (oral) 30 mg twice daily Starting time: after surgery: Duration: 10 to 14 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 20 mg twice daily Starting time: after surgery: 24‐36h Duration: 10 to 14 days Placebo group: Intervention: placebo (oral) twice daily Starting time: after surgery: Duration: 10 to 14 days
Outcomes	Number of outcomes: According to protocol: no protocol is available OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Symptomatic VTE: Definition: incidence of symptomatic VTE Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Major bleeding: Definition: incidence of major bleeding (definition not reported) Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Serious AEs: non‐hepatic: Definition: number of patients with at least one serious AE (definition not provided), excluding those liver‐related Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Asymptomatic distal DVT: Definition: incidence of asymptomatic distal DVT Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Liver enzymes elevation: Definition: incidence of ALT >3 times ULN Time points measured: 10 to 14 days (treatment) Time points reported: 10 to 14 days (treatment) Minor AEs: Definition: number of patients with at least one mild AE (definition not provided) Time points measured: 10 to 14 days (treatment), 3 to 5 weeks (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) RESULTS All‐cause mortality: Summary data (10 to 14 days): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): darexaban 15 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 20 mg twice daily ‐ placebo (oral) twice daily: 0/NR ‐ 1/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Major VTE: Summary data (10 to 14 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 85 missing participants (reasons not detailed) ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): not reported Symptomatic VTE: Summary data (10 to 14 days): darexaban 15 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 20 mg twice daily ‐ placebo (oral) twice daily: 1/91 ‐ 0/NR ‐ 2/90 ‐ 0/NR ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, not calculable missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): not reported Major bleeding: Summary data (10 to 14 days): see Analysis 1.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 20 missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): not reported Serious AEs: non‐hepatic Summary data (10 to 14 days): see Analysis 9.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 20 missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): not reported Fatal VTE: Summary data (10 to 14 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 96 missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): darexaban 15 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 20 mg twice daily ‐ placebo (oral) twice daily: 0/NR ‐ 1/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, not calculable missing participants ‐Subgroups reported: none Asymptomatic distal DVT: Summary data (10 to 14 days): see Analysis 9.10 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 106 missing participants (10 unexplained) ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): not reported Fatal bleeding: Summary data (10 to 14 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 20 missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): darexaban 15 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 20 mg twice daily ‐ placebo (oral) twice daily: 0/NR ‐ 1/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, not calculable missing participants ‐Subgroups reported: none Liver enzymes elevation: Summary data (10 to 14 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 21 missing participants (1 unexplained) ‐Subgroups reported: none Minor AEs: Summary data (10 to 14 days): see Analysis 9.13 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 20 missing participants ‐Subgroups reported: none Summary data (3 to 5 weeks' follow‐up): not reported
Notes	Source of funding: Astellas Conflicts of interest: authors (3) had financial relationships with several pharmaceutical companies, including Astellas Published protocol: registered in ClinicalTrials.gov as NCT00917254 Observations: none Missing data requested from authors: specific groups blinded, description of the method of randomisation and concealment of allocation, details about withdrawals and exclusions, definition of major bleeding, serious AEs and minor AEs, follow‐up summary data for major VTE, asymptomatic distal DVT, major bleeding, serious AEs, minor AEs, treatment period number of participants for all‐cause mortality, major VTE and symptomatic VTE, follow‐up number of participants for all‐cause mortality, fatal VTE, fatal bleeding, protocol of the study Missing data obtained from authors: information regarding blinding and withdrawals; data on all‐cause mortality (number of participants) For the efficacy set, > 20.4% of the randomised participants were excluded (> 16.9% in the darexaban groups and > 27.5% in the enoxaparin group). Besides, for the efficacy outcomes included in this review, data from several (not defined) sets were included. Reasons for exclusion were not reported. CONSORT flow diagram incomplete. For the safety set, 5.2% of the randomised participants were excluded (data were insufficient to calculate the proportions for each group). Reasons for exclusion were not reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	High risk	Study intervention was open‐label for the darexaban versus enoxaparin comparisons
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of outcome assessment was not mentioned. Blinding of the data analysts is not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

Hoseinzadeh 2022.

Study characteristics
Methods	Aim: compare the effects of injectable enoxaparin with oral rivaroxaban in DVT prophylaxis in patients with femoral peritrochanteric fracture Design: parallel‐group Blinding: single‐blind enoxaparin vs rivaroxaban Study duration: 28 days Duration of intervention: 28 days Duration of follow‐up: 4 weeks Method of randomisation: randomisation table Method of concealment of allocation: using https://www.random.org and Gaussian function, the participants were randomly assigned into two groups of enoxaparin and rivaroxaban. Primary efficacy analysis set: thrombotic complications. D‐dimer. Ultrasound scan Primary safety analysis set: not mentioned Withdrawals: none Exclusions: renal failure, sensitivity to one of the prescribed drugs in the intervention, endocarditis, liver disease, uncontrolled blood pressure, blood diseases, history of thromboembolism in the past three months treated with anticoagulants, receiving corticosteroids five days before the surgery, hip replacement surgery during previous six months, and pregnancy.
Participants	Diagnosis: femoral peritrochanteric fracture Location of participants: Shohada Hospital of Tabriz, Iran from January 2019 to December 2019 Number of participants randomised: 88 Age (years): adults Rivaroxaban mean age of 68.79 ± 4.19 years Enoxaparin mean age of 67.45 ± 4.49 years Gender (male %): rivaroxaban group: 44 participants (males 65.90%) enoxaparin group: 44 participants (males 61.36%) Baseline imbalances: not found Inclusion criteria: patients with surgery indication Exclusion criteria: "1. renal failure, 2. sensitivity to one of the prescribed drugs, 3. endocarditis, 4. liver disease, 5. uncontrolled BP, 6. blood diseases, 7. history of thromboembolism in the past three months treated with anticoagulants, 8. corticosteroids five days before the surgery, 9. hip replacement surgery during the last six months, 10. pregnancy."
Interventions	Interventions: parallel assignment Number of intervention groups: 02 Concomitant interventions: procedures to treat pain. Except epidural catheter implantation was performed during and after the surgery. The surgical procedure was performed in the same way for all patients. Excluded interventions: the anaesthesiologist was informed of the study protocol to avoid use of the epidural catheter implantation method to treat pain. Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: immediately after surgery Duration: 28 days (Three days after receiving medication in the hospital, the investigators checked daily by telephone to see if medication had been received) Enoxaparin group: Intervention: enoxaparin 40 mg (subcutaneous) once daily Starting time: immediately after surgery Duration: 28 days (Three days after receiving medication in the hospital, the investigators checked daily by telephone to see if medication had been received)
Outcomes	Number of outcomes: 03 According to protocol: no protocol is available Reported: not stated Primary outcomes included: DVT Secondary outcomes included: symptoms and haematologic parameters Reported outcomes relevant to this review: none. (Symptomatic DVT with no events; outcome not addressed in this review) Primary outcome: DVT (not specified if proximal or distal; nor was PE evaluated) Definition: "Diagnostic Vascular specialist with US [ultrasound] scan (The patients were followed daily in hospitalization by the resident; when he detected suspicion of DVT, he called a vascular specialist, who, with the US, verified or ruled out the diagnosis. The resident suspected, based on symptoms, of 5 cases on Rivaroxaban and 3 on enoxaparin; all were ruled out.)" Time points measured: daily for 3 days Time points reported: none
Notes	Source of funding: granted by Tabriz University of Medical Sciences. Conflicts of interest: the author(s) declared no potential conflicts of interest Published protocol: not found Observations: no DVT and bleeding events reported. No missing participants reported Missing data requested from authors: none Missing data obtained from authors: none No exclusion or withdrawal was reported for any study group No CONSORT flow diagram
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation table. Quote:"the patients took a piece of paper from a bag containing 100 pieces of black and white papers. All patients who took a white paper were included in the study."
Allocation concealment (selection bias)	Low risk	Using https://www.random.org and Gaussian function, the participants were randomly assigned into two groups of enoxaparin and rivaroxaban.
Blinding of participants and personnel (performance bias)	High risk	A single‐blind comparison of enoxaparin versus rivaroxaban was stated; however, the study appears to be open‐label. There was no double‐dummy method for blinding participants, and the personnel contacted the participants by telephone, reminding them to take the oral or subcutaneous medication during the call.
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of outcome assessment was not mentioned. Blinding of the data analysts is not mentioned
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	Low risk	The study appears to be free of other sources of bias.

Hu 2015.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Blinding: open‐label Exclusions: none Method of concealment of allocation: not reported Method of randomisation: not reported Primary efficacy analysis set: all treated participants Primary safety analysis set: all treated participants Study duration: July 2011 to July 2014 Withdrawals: none
Participants	Baseline characteristics Rivaroxaban group Age; mean (range): 60.53 ± 3.74 Gender; F/M: 19/26 Enoxaparin group Age; mean (range): 61.52 ± 3.46 Gender; F/M: 21/24 Overall Age; mean (range): not reported Gender; F/M: not reported Inclusion criteria: "Patients who underwent total knee arthroplasty and did not take any anticoagulant. Not having any condition related to the heart, liver or lungs. Not having knee osteoarthritis" Excluded criteria: "History of vascular surgery. History of deep vein thrombosis. History of clotting disorder" Pretreatment: none Diagnosis: scheduled to total knee arthroplasty Location of participants: single centre, China Number of participants randomised: 90
Interventions	Rivaroxaban group Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery Duration: 2 weeks Concomitant interventions: none Excluded interventions: none LMWH group Intervention: enoxaparin (subcutaneous) 5000 IU once daily Starting time: 12 hours before surgery Duration: 2 weeks Concomitant interventions: none Excluded interventions: none
Outcomes	Asymptomatic distal DVT. Detected by colour Doppler ultrasound (two weeks after surgery) Volume of blood loss Outcome type: continuous outcome
Notes	Source of funding: not stated Conflicts of interest: The authors stated that the article and the content do not imply relevant conflicts of interest. Missing data requested from authors: none Missing data obtained from authors: none No exclusions or withdrawals were reported for any study group. No CONSORT flow diagram. It is not clear if there were exclusions or withdrawals.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation with random numbers table
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	High risk	The study was open‐label for the comparison rivaroxaban versus LMWH
Blinding of outcome assessment (detection bias)	Unclear risk	No blinding of any group was reported
Incomplete outcome data (attrition bias) Benefit outcomes	Unclear risk	Not enough information to detect if there were missing outcome data.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Not enough information to detect if there were missing outcome data.
Selective reporting (reporting bias)	Unclear risk	No published protocol was found
Other bias	Low risk	The study appears to be free of other sources of bias.

Hui 2013.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: open‐label; groups blinded: none Study duration: March 2011 to September 2012 Duration of intervention: 2 to 5 weeks Duration of follow‐up: 19 to 22 weeks Method of randomisation: no description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: all participants Primary safety analysis set: all participants Withdrawals: no participant withdrew Exclusions: no participant was excluded
Participants	Diagnosis: scheduled to HR, primary, unilateral Location of participants: 1 country (China) Number of participants randomised: 106 Age (years; mean ± SD): (all randomised population) rivaroxaban group: 64.40 ± 6.26; LMWH group: 64.70 ± 6.57 Gender (male %): (all randomised population) rivaroxaban group: 31 (58.5%); LMWH group: 29 (54.7%) Baseline imbalances: none Inclusion criteria: Adults older than 50 years of age Admitted for primary of unilateral HR Consent to accept two different types of anticoagulant therapy after HR Agreement of the patients and their families with the process and purpose of the experiment, and signed informed consent Exclusion criteria: Active bleeding or bleeding tendency Obvious liver disease (such as acute hepatitis, chronic active hepatitis, cirrhosis) Suffering from some of the obstacles to bilateral Doppler ultrasound examination (such as single‐leg amputation) Concomitant use of human immunodeficiency virus protease inhibitors Presence of serious infection Use of other drugs that may affect the results of the study
Interventions	Number of intervention groups: 2 Concomitant interventions: none Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 6 hours Duration: 5 weeks LMWH group: Intervention: LMWH (subcutaneous) 4100 IU once daily Starting time: after surgery: 6 hours Duration: 2 weeks
Outcomes	Number of outcomes, according to protocol: no protocol is available Asymptomatic distal DVT: Definition: incidence of asymptomatic distal DVT Time points measured: treatment period: 5 weeks for rivaroxaban and 2 weeks for LMWH; follow‐up period: 6 months after surgery Time points reported: 5 weeks for rivaroxaban and 2 weeks for LMWH; follow‐up period: 6 months after surgery Volume of blood loss: Definition: total volume of the drainage of the surgical wound until extubation Method/unit of measurement: millilitre (mL) Time points measured: not reported Time points reported: not reported RESULTS Asymptomatic distal DVT: Summary data (5 weeks for rivaroxaban and 2 weeks for LMWH): rivaroxaban 0/53; LMWH 0/53 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (follow‐up period: 6 months after surgery): see Analysis 1.10 ‐Analysis set, type of analysis and missing participants: not specified, ITT analysis, no missing participants ‐Subgroups reported: none Volume of blood loss: Summary data (time point not reported) (mean ± SD): see Analysis 1.14 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none
Notes	Source of funding: no external funding Conflicts of interest: none Published protocol: none Observations: none Missing data requested from authors: specific groups blinded, description of method of randomisation and concealment of allocation, details about withdrawals, definition of serious and minor AEs, time points of measurement of volume of blood loss, protocol of the study Missing data obtained from authors: none For the efficacy and safety set, no participant was excluded. No CONSORT flow diagram. All participants in both groups were followed for 1 month after surgery.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	High risk	Type of blinding design was not stated. Given the use of interventions with different durations (5 weeks for rivaroxaban versus 2 weeks for active control) and route of administration, we assumed the study was open‐label.
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of outcome assessment was not mentioned. Blinding of the data analysts is not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	Low risk	No exclusion or withdrawal was reported for any study group
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	No exclusion or withdrawal was reported for any study group
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	High risk	There is at least one important risk of bias. The study had a potential source of bias related to the specific study design used (inappropriate intervention as duration of prophylaxis differed markedly for each intervention group and the outcome analysis was at different time points; 5 weeks versus 2 weeks). The intervention could be favoured in the design of the study, since follow‐up was longer than 2 weeks and the analysis of the outcomes were at different times.

Jiang 2019.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Blinding: open‐label Exclusions: none Method of concealment of allocation: not mentioned Method of randomisation: not described Primary efficacy analysis set: all treated participants Primary safety analysis set: all treated participants Study duration: January 2015 to November 2017 Withdrawals: none
Participants	Baseline characteristics Apixaban group Age; mean (range): 68.7 ± 5.7 Gender; F/M: 48/62 LMWH group Age; mean (range): 70.2 ± 6.1 Gender; F/M: 58/52 Overall Age; mean (range): Gender; F/M: 106/114 Inclusion criteria: "older adults (> 60 years old). Patients with total knee arthroplasty indications. Patients without surgical contraindications. Patients with no history of DVT before surgery" Exclusion criteria: "Patients with malignant tumors. Patients with liver, kidney, heart, lung, or other important organ dysfunction. Patients with blood coagulation dysfunction. Patients whose preoperative ultrasonography of blood vessels revealed DVT. Patients who were allergic to the drugs being studied in this research. Patients with a history of mental illness. Patients who refused to provide informed consent. Patients who did not meet the experimental ethical standards" Pretreatment: no major differences Diagnosis: scheduled to TKA Location of participants: single centre, China Number of participants randomised: 220
Interventions	Apixaban group Intervention: apixaban 2.5 mg twice daily Starting time: 12 hours after surgery Duration: 5 weeks Concomitant interventions: not specified Excluded interventions: not specified LMWH group Intervention: 4000 IU once daily Starting time: 12 hours after surgery Duration: 5 weeks Concomitant interventions: not specified Excluded interventions: not specified
Outcomes	All‐cause mortality Fatal VTE (colour Doppler ultrasound) Fatal bleeding Volume of blood loss (DVT at 3 months after surgery, by colour eco‐Doppler ultrasound every 2 weeks. They found 6 in the apixaban group and 22 in the LMWH group; however, they did not mention the number of participants evaluated and measured at the time point in which DVT events happened. Events not included in the review meta‐analysis )
Notes	Source of funding: supported by the Chinese Medical Doctor Association ‐ Special Fund for Thrombosis Prevention after Major Orthopedics Surgery (No. 2015COS0810) and Nanjing General Hospital Subject (No. 2017003). Conflicts of interest: not stated. Missing data requested from authors: none Missing data obtained from authors: none No exclusions or withdrawals were reported for the efficacy and safety analysis sets. No CONSORT flow diagram. It is not clear if there were exclusions or withdrawals.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method is not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	High risk	The study was open‐label for the comparison apixaban versus LMWH
Blinding of outcome assessment (detection bias)	Unclear risk	No blinding of any group was reported
Incomplete outcome data (attrition bias) Benefit outcomes	Unclear risk	Not enough information to detect if missing outcome data existed.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Not enough information to detect if missing outcome data existed.
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	High risk	There is at least one important risk of bias. The study did not report that conflicts of interest were registered. Also, there may be a risk of bias for an inappropriate co‐intervention, but there is either insufficient information to assess whether an important risk of bias exists, or insufficient rationale or evidence that an identified problem will introduce bias.

Kanan 2008.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: double‐blind, double‐dummy; groups blinded: participants Study duration: September 2006 to April 2007 Duration of intervention: 32 to 36 days Duration of follow‐up: 29 days (for laboratory analyses only) Method of randomisation: no description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: participants who adhered to treatment Primary safety analysis set: participants who adhered to treatment Withdrawals: no participant withdrew Exclusions: 2 participants were excluded due to low adherence to treatment, allocated group was not reported
Participants	Diagnosis: scheduled to THA Location of participants: 1 country (Brazil) Number of participants randomised: 67 Age (years; mean ± SD): rivaroxaban group: 54.5 ± 13.9; enoxaparin group: 61.3 ± 12.8 Gender (male %): rivaroxaban group: 18 (55%); enoxaparin group:18 (56%) Baseline imbalances: slight differences regarding age, weight and underlying diagnosis Inclusion criteria: Adults 18 years or older Hip prosthesis with elective surgery indication Signed informed consent Exclusion criteria: Scheduled to bilateral HR surgery Active bleeding or high risk of bleeding Contraindication to any of the study drugs or a condition which prevents anticoagulant treatment Condition which prevents carrying out bilateral venography Pregnant or breastfeeding women Women of reproductive age without adequate contraception Alcohol or drug dependence Concomitant use of HIV protease inhibitors Treatment with other experimental drug within 30 days before study entry Use of IPC during study treatment period Simultaneous participation in another clinical trial Treatment with other type of anticoagulant which cannot be stopped or other prohibited medication Relevant liver disease
Interventions	Number of intervention groups: 2 Concomitant interventions: none Excluded interventions: IPC Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: not reported Duration: 32 to 36 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: 6 to 8 hours Duration: 32 to 36 days
Outcomes	Number of outcomes, according to protocol: no protocol is available OUTCOMES No outcome relevant to this review was reported RESULTS No outcome relevant to this review was reported
Notes	Source of funding: Bayer HealthCare (declared for researchers and the institution, but not for the study) Conflicts of interest: authors stated that the external financial support received did not imply conflicts of interest Published protocol: none Observations: none Missing data requested from authors: specific groups blinded, description of the method of randomisation and concealment of allocation, details about withdrawals and exclusions, gender distribution of participants, source of funding, protocol of the study Missing data obtained from authors: none For the analysis set, 3.0% of the randomised participants were excluded (data were insufficient to calculate the proportions for each group). Reasons for exclusion were not reported in enough detail (allocated group was not reported). No CONSORT flow diagram
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method is not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion for the comparison with enoxaparin. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of outcome assessment was not mentioned. Blinding of the data analysts is not mentioned
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	High risk	There is at least two important risks of bias. The authors have potential conflict of interest and the study did not report a management plan that explains the procedures or additional steps to be taken to minimize the risk of bias to ensure research integrity, such as divestures or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

Karampinas 2019.

Study characteristics
Methods	Design: randomised controlled trial Group: parallel group Blinding: open‐label Exclusions: none Method of concealment of allocation: use of sealed envelopes Method of randomisation: not described Primary efficacy analysis set: all treated participants Primary safety analysis set: all treated participants Study duration: January 2014 to January 2018 Withdrawals: none
Participants	Inclusion criteria: primary varus gonarthrosis Exclusion criteria: "Previous surgery at the knee. Thrombocytopenia. Anaemia (Hb < 10 g/dL). Warfarin therapy. Coagulopathy. Previous VTE. Significant comorbidities such as ischaemic heart disease, cerebrovascular accident, liver cirrhosis and renal disease" Group differences: differences for age and BMI were observed, no statistical analysis was performed Diagnosis: scheduled to TKA Location of participants: single centre (Greece) Number of participants randomised: 158 (104 included in the analysis) Rivaroxaban group: Age; mean (range): 74 (69 to 82) Gender; F/M: 32/14 LMWH group: Age; mean (range): 72 (35 to 80) Gender; F/M: 35/23
Interventions	Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: postoperatively Duration: 25 days Concomitant interventions: preoperative intravenous TXA, postoperative intra‐articular TXA Excluded interventions: none LMWH group: Intervention: LMWH (subcutaneous) once daily Starting time: postoperatively Duration: 25 days Concomitant interventions: preoperative intravenous TXA, postoperative intra‐articular TXA Excluded interventions: none There was another group with rivaroxaban but without concomitant intervention (not included in the review analysis).
Outcomes	Symptomatic VTE (imaging investigation not performed) Major bleeding Volume of blood loss
Notes	Source of funding: not stated All authors declare that they have no conflicts of interest. Missing data requested from authors: none Missing data obtained from authors: none For the analysis set, the number of participants included in the analysis was not reported. It is not clear if there were exclusions or withdrawals. No CONSORT flow diagram.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method is not described
Allocation concealment (selection bias)	Low risk	Allocated group was designated with a sealed envelope
Blinding of participants and personnel (performance bias)	High risk	The study was open‐label for the comparison of rivaroxaban versus LMWH
Blinding of outcome assessment (detection bias)	Unclear risk	No blinding of any group was reported
Incomplete outcome data (attrition bias) Benefit outcomes	Unclear risk	Insufficient information to detect if there were missing outcome data
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Insufficient information to detect if there were missing outcome data
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	Unclear risk	There may be a risk of bias for an inappropriate co‐intervention, but there is either insufficient information to assess whether an important risk of bias exists, or insufficient rationale or evidence that an identified problem will introduce bias.

Khalafallah 2018.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: open‐label Study duration: January 2010 to December 2012 Duration of intervention: 14 days for TKR, 35 days for THR Duration of follow‐up: 6 months after surgery Method of randomisation: "by using a specific computer programme in the Pharmacy Department with 10‐20 blocks of patients that allow random allocation of subjects" Method of concealment of allocation: treatment codes located at central sites. Personnel associated with data collection and collation were masked to treatment allocation. Primary efficacy analysis set: not reported Primary safety analysis set: postoperative bleeding and blood transfusion. Safety data were collected and reported prospectively to the Ethics Committee in Tasmania. Withdrawals: 10 participants cancelled surgery (rivaroxaban 6; enoxaparin 4) Rivaroxaban group: 13 lost to follow‐up. Reasons not provided Enoxaparin group: 5 lost to follow‐up. Reasons not provided Exclusions: no participant was excluded
Participants	Diagnosis: scheduled to elective TKA: 405 (rivaroxaban group 202 / enoxaparin group 203) or elective THA: 335 (rivaroxaban group 167 / enoxaparin group 168) Location of participants: 1 country (Australia) Number of participants randomised: 750 (375 for rivaroxaban group and 375 for enoxaparin group). 740 participants analysed (369 rivaroxaban group / 371 enoxaparin group) Age (years; mean ± SD): rivaroxaban group: 67.3 ± 10.4; enoxaparin group: 67.2 ± 9.87 Gender (male %): rivaroxaban group 369 participants: 45.8%; enoxaparin group 371 participants: 51.2% Baseline imbalances: significant difference regarding gender Inclusion criteria Arthroplasty for knee or hip disease Over 18 years of age Normal baseline platelet count, prothrombin and partial thromboplastin times Written informed consent, prior to the procedure according to the ethics code of conduct At high (post‐orthopaedic surgery) risk of VTE and the treatment plan will include anticoagulation Women of childbearing potential must be using adequate measures of contraception (as determined by the Investigator) to avoid pregnancy and should be highly unlikely to conceive during the study period Women of childbearing potential must have a negative pregnancy test at screen Exclusion criteria Surgery for acute fracture (< 4 weeks), septic joint, or extraction arthroplasty Personal history of thromboembolic disease or documented hypercoagulation disease Increased risk of haemorrhage; for example, gastrointestinal bleeding, as from active gastric ulcer, or bleeding diathesis; or persistent intestinal or urinary tract bleed Haemorrhagic stroke; brain, spinal, or ophthalmologic surgery in previous 12 months Requires chronic anticoagulation with warfarin Requires chronic platelet function suppressive therapy for coronary or peripheral artery stents Prior adverse reaction to any of the study drugs Pregnant and/or lactating women and women of child‐bearing potential not using acceptable means of contraception Participation in any other clinical trials involving investigational or marketed products within 30 days prior to entry in the study
Interventions	Number of intervention groups: 2 Concomitant interventions: IPC started in the first 24 hours after surgery Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: not reported Duration: 14 days for TKR, 35 days for THR Follow‐up period: 6 months Enoxaparin group: Intervention: LMWH (subcutaneous) 40 mg daily (originally twice daily was reported) Starting time: after surgery: after discharge Duration: 14 days for TKR, 35 days for THR Follow‐up period: 6 months
Outcomes	Number of outcomes according to protocol: 5 OUTCOMES (reported 5) Included in the review: 3 Symptomatic or asymptomatic VTE: (total VTE) Definition: incidence of symptomatic or asymptomatic venous thromboembolism Time points measured: 14 days for TKR, 35 days for THR (treatment), 6 months after surgery (follow‐up) detected by Doppler ultrasound at day 21 or if clinical symptoms of VTE were reported by monthy design questionnaires over a period of 6 months post surgery. If sympoms were reported VTE was confirmed by Doppler ultrasound or CT pulmonary angiogram. Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) a bleeding index ≥ 2.0 (bleeding index = units of red blood cells transfused + prebleed Hb − postbleed Hb); 2) bleeding at a critical site; 3) bleeding requiring an additional operation or intervention; 4) fatal bleeding Time points measured: 14 days for TKR, 35 days for THR Volume of blood loss: mL of blood drain in day one RESULTS Total VTE: Summary data (14 days for TKR, 35 days for THR, 6 months): ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 13/8 missing participants ‐Subgroups reported: none Major bleeding: no reported events. Safety set ‐ subgroups reported: none Volume of blood loss: safety set ‐ subgroups reported: none
Notes	Source of funding: partially self‐funded, external sources of funding: Launceston General Hospital, St Leukes Hospital, Launceston Conflicts of interest: athors declared not conflicts of interest. Published protocol: available in the Australian New Zealand Clinical Trials Register. ACTRN12609000762257 Observations: initially named PREVENT trial. Data obtained from LANCET Haematology manuscript draft (not peer reviewed). Targeted sample size not achieved Number of events per type of surgery not detailed ITT analyses were stated; however, the missing outcome data were not included in the analyses CONSORT flow diagram incomplete VTE events were not specified as symptomatic or asymptomatic, nor as proximal DVT, distal DVT, or PE. We included them as total VTE events. Missing data requested from authors: none Missing data obtained from authors: none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "computer programme in the Pharmacy Department with 10‐20 blocks of patients that allow random allocation of subjects"
Allocation concealment (selection bias)	Low risk	Treatment codes located at central sites. Personnel associated with data collection and collation were masked to treatment allocation.
Blinding of participants and personnel (performance bias)	High risk	Open‐label for rivaroxaban and enoxaparin
Blinding of outcome assessment (detection bias)	Low risk	Personnel associated with data collection and collation were masked to treatment allocation.
Incomplete outcome data (attrition bias) Benefit outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data balanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	Low risk	Reported primary and secondary outcomes were included in the published protocol in the Australian New Zealand Clinical Trials Register. ACTRN12609000762257
Other bias	Unclear risk	There may be a risk of bias for an inappropriate co‐intervention, but there is either insufficient information to assess whether an important risk of bias exists, or insufficient rationale or evidence that an identified problem will introduce bias.

Kim 2016.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: double‐blind, double‐dummy; groups blinded: participants, data collector Study duration: August 2011 to February 2013 Duration of intervention: 7 to 12 days Duration of follow‐up: 6 weeks Method of randomisation: computer‐generated schedule Method of concealment of allocation: not mentioned Primary efficacy analysis set: all randomised participants who completed the study Primary safety analysis set: same as efficacy set Withdrawals: rivaroxaban group (29): 16 withdrew consent, 6 because of adverse events, 7 for other reasons. Enoxaparin group (28): 15 withdrew consent, 6 because of adverse events, 7 for other reasons. Placebo group (15): 10 withdrew consent, 5 for other reasons Exclusions: no participant was excluded
Participants	Diagnosis: scheduled to elective, primary, unilateral THA Location of participants: single centre (South Korea) Number of participants randomised: 958 (758 for rivaroxaban and enoxaparin groups) Age (years; mean ± SD): rivaroxaban group: 44.4 ± 8.6 (younger than 60), 68.6 ± 6.4 (older than 60); enoxaparin group: 43.9 ± 9.4 (younger than 60), 69.4 ± 6.7 (older than 60); placebo group: 43.4 ± 9.9 (younger than 60) Gender (male %): rivaroxaban group: 53.8% (younger than 60), 38.6% (older than 60); enoxaparin group: 62% (younger than 60), 35.9% (older than 60); placebo group: 58.9% (younger than 60) Baseline imbalances: significant difference regarding body mass index Inclusion criteria: Male or female without childbearing potential Aged 20 years Scheduled for elective primary THA Exclusion criteria: Recent history of active bleeding or VTE Known genetic disorder associated with bleeding tendency or any condition related with an increased risk of bleeding, including persistent blood pressure of 160 mmHg systolic and/or 100 mmHg diastolic at baseline Myocardial infarction or cerebrovascular accident within three months of the scheduled surgery Major surgery in the prior three months Renal insufficiency with a creatinine clearance < 60 mL/minute (min) Hepatic failure combined with coagulopathy Thrombocytopenia (platelets < 100,000/mm³) Planned indwelling epidural catheter for > 6 hours (h) after the end of surgery Requiring concomitant anti‐thrombotic medication at any time during the period from one week before the surgery until the end of the thromboprophylaxis period Scheduled for simultaneous bilateral THA Anatomically deformed hip requiring osteotomy
Interventions	Number of intervention groups: 3 Concomitant interventions: IPC Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 12 hours Duration: 7 to 12 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: after surgery: 12 hours Duration: 7 to 12 days Placebo group: Intervention: tablet containing sugar (oral) or normal saline 1 mL (subcutaneous) once daily Starting time: after surgery: 12 hours Duration: 7 to 12 days
Outcomes	Number of outcomes according to protocol: no protocol is available OUTCOMES All‐cause mortality: Definition: incidence of deaths from all causes Time points measured: 9 to 14 days (treatment), 6 weeks (additionally, follow‐up) Time points reported: 9 to 14 days (treatment), 6 weeks (additionally, follow‐up) Major bleeding: Definition: incidence of digestive, intra‐articular, haemoptysis, intra‐ocular, pericardial, or retroperitoneal bleeding Time points measured: 9 to 14 days (treatment), 6 weeks (additionally, follow‐up) Time points reported: 9 to 14 days (treatment), 6 weeks (additionally, follow‐up) Fatal VTE: Definition: incidence of VTE‐related deaths Time points measured: 9 to 14 days (treatment), 6 weeks (additionally, follow‐up) Time points reported: 9 to 14 days (treatment), 6 weeks (additionally, follow‐up) Fatal bleeding: Definition: assumed as 0 given that 0 all‐cause deaths were reported Time points measured: 9 to 14 days (treatment), 6 weeks (additionally, follow‐up) Time points reported: 9 to 14 days (treatment), 6 weeks (additionally, follow‐up) Liver enzymes elevation: Definition: incidence of concurrent elevations of total bilirubin > 3 times of the baseline and ALT or AST > 5 times of the baseline Time points measured: 6 weeks (additionally, follow‐up) Time points reported: 6 weeks (additionally, follow‐up) RESULTS All‐cause mortality: Summary data (9 to 14 days, 6 weeks): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 72 missing participants ‐Subgroups reported: age groups (< 60 years; ≥ 60 years) Major bleeding: Summary data (9 to 14 days, 6 weeks): see Analysis 1.4 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 72 missing participants ‐Subgroups reported: age groups (< 60 years; ≥ 60 years) Fatal VTE: Summary data (9 to 14 days, 6 weeks): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 72 missing participants ‐Subgroups reported: age groups (< 60 years; ≥ 60 years) Fatal bleeding: Summary data (9 to 14 days, 6 weeks): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 72 missing participants ‐Subgroups reported: age groups (< 60 years; ≥ 60 years) Liver enzymes elevation: Summary data (9 to 14 days): not reported Summary data (6 weeks): see Analysis 1.12 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 72 missing participants ‐Subgroups reported: age groups (< 60 years; ≥ 60 years)
Notes	Source of funding: not reported Conflicts of interest: none declared Published protocol: none Observations: none Missing data requested from authors: description of the method of concealment of allocation, summary data for liver enzymes elevation during treatment period (9 to 14 days), source of funding, protocol of the study Missing data obtained from authors: none For the analysis set, 7.5% of the randomised participants were excluded (7.7% in the rivaroxaban group and 7.4% in the enoxaparin group). Reasons for exclusion were reported. Intolerable adverse events in 12 participants balanced amongst the study groups (no details about the type of adverse events).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random sequence was computer‐generated
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded investigator. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	Unclear risk	Missing data reported. Reasons for missing outcome data related to true outcome. Insufficient information about reasons for and numbers of missing outcome data. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing data reported. Reasons for missing outcome data related to true outcome. Missing outcome data balanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk is not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	Unclear risk	There may be a risk of bias for an inappropriate co‐intervention, but there is either insufficient information to assess whether an important risk of bias exists, or insufficient rationale or evidence that an identified problem will introduce bias.

Kunal 2021.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Blinding: not described Exclusions: none Method of concealment of allocation: not described Method of randomisation: not reported Primary efficacy analysis set: all treated participants Primary safety analysis set: all treated participants Study duration: from 1 October 2018 to 31 August 2019 Withdrawals: none
Participants	Baseline characteristics Apixaban group Age; mean (range): 58.8 Gender; F/M: 19/29 LMWH group Age; mean (range): 61.6 Gender; F/M: 24/24 Overall Age; mean (range): Not reported Gender; F/M: Not reported Inclusion criteria: undergoing either elective unilateral or simultaneous bilateral total knee arthroplasty and total hip arthroplasty and willing to undergo Doppler studies. Informed written consent obtained Exclusion criteria: "Hereditary (first degree) or acquired bleeding or coagulation disorder. Known or suspected history of heparin‐induced thrombocytopenia. Need for ongoing treatment with a parenteral or oral anticoagulant (e.g., subjects with mechanical valves, warfarin‐eligible atrial fibrillation). Known coagulopathy. Active bleeding. Active hepatobiliary disease. Any condition, in the opinion of the investigator, for which surgery or administration of an anticoagulant was contraindicated. Hypersensitivity to unfractionated heparin, low molecular weight heparin, porcine products, or iodinated contrast medium. Pre‐operative deep vein thrombosis detected in Doppler ultrasound and patients with previous known/documented history of thrombogenic events such as deep‐vein thrombosis (DVT) and pulmonary embolism." Pretreatment: no statistically significant differences were observed Diagnosis: scheduled to THA or TKA Location of participants: single centre (India) Number of participants randomised: 96
Interventions	Apixaban group Intervention: apixaban (oral) 2.5 mg twice daily Starting time: 12 to 24 hours postoperatively Duration: 2 weeks (TKA), 5 weeks (THA) Concomitant interventions: intermittent pneumatic compression devices during the period of hospital stay Excluded interventions: none LMWH group Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: 12 to 24 hours postoperatively Duration: 2 weeks (TKA), 5 weeks (THA) Concomitant interventions: intermittent pneumatic compression devices during the period of hospital stay Excluded interventions: none
Outcomes	All‐cause mortality Major VTE (DVT by Doppler ultrasound; PE by computed tomography pulmonary angiography) Symptomatic VTE Major bleeding Fatal VTE Asymptomatic distal DVT Clinically relevant non‐major bleeding (outcome not included in this review)
Notes	Funding: not stated Disclosure of potential conflicts of interest: none Missing data requested from authors: none Missing data obtained from authors: none No participant was excluded after randomisation (if there was any randomisation). No CONSORT flow diagram
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding of participants and personnel (performance bias)	High risk	Not described. Open‐label design can be assumed
Blinding of outcome assessment (detection bias)	Unclear risk	Not described. Open‐label design can be assumed
Incomplete outcome data (attrition bias) Benefit outcomes	Low risk	No exclusion or withdrawal was reported for any study group
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	No exclusion or withdrawal was reported for any study group
Selective reporting (reporting bias)	Unclear risk	No published protocol was found
Other bias	Unclear risk	There may be a risk of bias for an inappropriate co‐intervention, but there is either insufficient information to assess whether an important risk of bias exists, or insufficient rationale or evidence that an identified problem will introduce bias.

Lassen 2003.

Study characteristics
Methods	Aim: dose‐response, efficacy and safety Design: parallel‐group Blinding: double‐blind, not described; groups blinded: outcome adjudicators Study duration: not reported Duration of intervention: 10 ± 2 days Duration of follow‐up: 32 ± 2 days (until day 42) Method of randomisation: no description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: not defined Primary safety analysis set: not defined Withdrawals: not reported Exclusions: not reported
Participants	Diagnosis: scheduled to TKR, elective, primary Location of participants: not reported Number of participants randomised: not reported Age (years): not reported Gender (male %): not reported Baseline imbalances: not assessable Inclusion criteria: not reported Exclusion criteria: not reported
Interventions	Number of intervention groups: 5 Concomitant interventions: none Excluded interventions: none Razaxaban 25 mg twice daily group: Intervention: razaxaban (oral) 25 mg twice daily Starting time: after surgery: 8 hours Duration: 10 ± 2 days Razaxaban 50 mg twice daily group: Intervention: razaxaban (oral) 50 mg twice daily Starting time: after surgery: 8h Duration: 10 ± 2 days Razaxaban 75 mg twice daily group: Intervention: razaxaban (oral) 75 mg twice daily Starting time: after surgery: 8h Duration: 10 ± 2 days Razaxaban 100 mg twice daily group: Intervention: razaxaban (oral) 100 mg twice daily Starting time: after surgery: 8h Duration: 10 ± 2 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 30 mg twice daily Starting time: after surgery: 12‐24h Duration: 10 ± 2 days
Outcomes	Number of outcomes according to protocol: no protocol is available OUTCOMES Relevant for this review. No reported data RESULTS Not reported
Notes	Source of funding: not reported Conflicts of interest: not reported Published protocol: none Observations: during follow‐up period, participants received additional non‐study drug prophylaxis at the discretion of their physician Missing data requested from authors: specific groups blinded, date of start and end of study, description of method of randomisation and concealment of allocation, definition of primary efficacy and safety sets, details about withdrawals and exclusions, location of study, number of randomised patients, data on age and gender of participants, inclusion and exclusion criteria, definition of major bleeding, treatment period summary data for major bleeding, follow‐up summary data for major bleeding, source of funding, statement of conflicts of interest, protocol of the study Missing data obtained from authors: none This study is reported in abstract format only For the efficacy set, > 33.2% of the randomised participants were excluded (data were insufficient to calculate the proportions for each group). Reasons for exclusion were not reported. No CONSORT flow diagram
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Low risk	Study was stated to be double‐blinded, but no description was provided
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded central adjudication committee. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data.
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	High risk	There are at least two important risks of bias. The study did not report if conflicts of interest were registered. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

NCT01205932.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group (phase III) Blinding: double‐blind, double‐dummy; groups blinded: participants Study duration: September 2010 to August 2011 Duration of intervention: 6 to 7 ± 2 days or 34 to 35 ± 4 days Duration of follow‐up: 30 days Method of randomisation: no description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: not defined Primary safety analysis set: not defined Withdrawals: not reported Exclusions (88): allocated groups and reasons not reported
Participants	Diagnosis: scheduled to THR, elective Location of participants: 1 country (Japan): 42 centres Number of participants randomised: 402 (303 for all rivaroxaban groups and 99 for enoxaparin) Exclusions/lost to follow‐up, others: 68 for rivaroxaban groups and 20 for enoxaparin group. Age (years; mean): (safety analysis population) 63.8 Gender (male %): (safety analysis population) 65 (16.4%) Baseline imbalances: none Inclusion criteria: Male and females aged 20 years or above Undergoing elective THR (the first replacement of the applicable hip joint) Written informed consent to participation after receiving detailed verbal and written information on any study‐specific procedures in advance Exclusion criteria: Planned, staged major orthopaedic surgery within 3 months prior to elective THR or during this study History of clinically significant active bleeding (e.g. intracranial bleeding, GI bleeding), or high bleeding risk: within 3 months prior to elective THR for GI bleeding Hepatic disease which is associated with coagulopathy leading to a clinically relevant bleeding risk Severe impaired renal function (CLCR (creatinine clearance) calculated by Cockcroft‐Gault formula: < 30 mL/min) Conditions prohibiting bilateral venography (e.g. amputation of 1 leg, allergy to contrast media) Ongoing anticoagulant therapy (e.g. warfarin, heparins and Factor Xa inhibitors other than study medication) that cannot be stopped (in the opinion of the investigator/sub investigator) People in whom epidural catheters are expected to be left in for longer than 18 hours postoperatively Planned IPC during treatment period
Interventions	Number of intervention groups: 4 Concomitant interventions: none Excluded interventions: none Rivaroxaban 5 mg once daily group: Intervention: rivaroxaban (oral) 5 mg once daily Starting time: not reported Duration: 34 to 35 ± 4 days Rivaroxaban 7.5 mg once daily group: Intervention: rivaroxaban (oral) 7.5 mg once daily Starting time: not reported Duration: 34 to 35 ± 4 days Rivaroxaban 10 mg once daily group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: not reported Duration: 34 to 35 ± 4 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 20 mg twice daily Starting time: not reported Duration: 6 to 7 ± 2 days
Outcomes	Number of outcomes according to protocol: 11 Reported: 11. Reported additionally: AEs, serious AEs, liver enzymes elevation (overall treatment period). Not reported: total DVT, distal DVT (first treatment period), minor bleeding (first and overall treatment periods). No safety outcomes with enough information to include data for analysis. OUTCOMES All‐cause mortality: Definition: incidence of death from any cause (diagnostic instrument not stated) Time points measured: 9 ± 2 days (first treatment period), 36 ± 4 days (overall treatment period), 30 days (additionally, follow‐up) Time points reported: 9 ± 2 days (first treatment period) Major VTE Definition: incidence of fatal and nonfatal PE or proximal DVT Time points measured: 9 ± 2 days (first treatment period), 36 ± 4 days (overall treatment period) Time points reported: 9 ± 2 days (first treatment period) Symptomatic VTE Definition: incidence of symptomatic VTE Time points measured: 9 ± 2 days (first treatment period), 36 ± 4 days (overall treatment period), 30 days (additionally, follow‐up) Time points reported: 9 ± 2 days (first treatment period), 36 ± 4 days (overall treatment period), 30 days (additionally, follow‐up) Fatal VTE Definition: incidence of death due to VTE Time points measured: 9 ± 2 days (first treatment period), 36 ± 4 days (overall treatment period), 30 days (additionally, follow‐up) Time points reported: 9 ± 2 days (first treatment period) RESULTS All‐cause mortality: Summary data (9 ± 2 days): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 88 missing participants ‐Subgroups reported: none Summary data (36 ± 4 days): not reported Summary data (30‐day follow‐up): not reported Major VTE: Summary data (9 ± 2 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 88 missing participants ‐Subgroups reported: none Summary data (36 ± 4 days): not reported Symptomatic VTE: Summary data (9 ± 2 days): rivaroxaban 5 mg once daily ‐ 7.5 mg once daily ‐ 10 mg once daily ‐ enoxaparin 20 mg twice daily: 0/77 ‐ 0/77 ‐ 0/81 ‐ 0/79 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 88 missing participants ‐Subgroups reported: none Summary data (36 ± 4 days): see Analysis 1.3 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 88 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): rivaroxaban 5 mg once daily ‐ 7.5 mg once daily ‐ 10 mg once daily ‐ enoxaparin 20 mg twice daily: 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, not calculable missing participants ‐Subgroups reported: none Fatal VTE: Summary data (9 ± 2 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 88 missing participants ‐Subgroups reported: none Summary data (36 ± 4 days): not reported Summary data (30‐day follow‐up): not reported
Notes	Source of funding: Bayer Yakuhin, Ltd. Conflicts of interest: not reported Published protocol: registered in ClinicalTrials.gov as NCT01205932 Observations: none Missing data requested from authors: authors could not be contacted Missing data obtained from authors: none For the efficacy set, 21.9% of the randomised participants were excluded (data were insufficient to calculate the proportions by study group). Reasons for exclusion were not reported in enough detail. For the safety set, 1.5% of the randomised participants were excluded (data were insufficient to calculate the proportions by study group). Reasons for exclusion were not reported in enough detail. No CONSORT flow diagram
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method is not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of investigator was stated, but no details are provided. Blinding of outcome assessment was not mentioned. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Reasons for missing outcome data related to true outcome. Proportion of missing outcomes compared with the observed event risk was enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Reasons for missing outcome data related to true outcome. Insufficient information regarding reasons for missing outcome data.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 3 outcomes not included in the protocol were reported, 3 outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There is at least one important risk of bias. The study did not report if conflicts of interest were registered.

NCT01206972.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group (phase III) Blinding: double‐blind, double‐dummy; groups blinded: participants Study duration: October 2010 to June 2011 Duration of intervention: 11 to 12 ± 2 days or 10 to 11 ± 2 days Duration of follow‐up: 30 days Method of randomisation: no description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: not defined Primary safety analysis set: not defined Withdrawals: not reported Exclusions (89): allocated groups and reasons not reported
Participants	Diagnosis: scheduled to TKR, elective Location of participants: 1 country (Japan): 33 centres Number of participants randomised: 302 (226 for all rivaroxaban groups and 76 for enoxaparin) Exclusions/lost to follow‐up, others: 68 for rivaroxaban groups and 22 for enoxaparin group. Age (years; mean): (safety analysis population) 73 Gender (male %): (safety efficacy analysis population) 53 (17.8%) Baseline imbalances: none Inclusion criteria: Male and females aged 20 years or above Undergoing elective TKR (the first replacement of the applicable knee joint) Written informed consent to participation after receiving detailed verbal and written information on any study‐specific procedures in advance Exclusion criteria: Planned, staged major orthopaedic surgery within 3 months prior to elective TKR or during this study History of clinically significant active bleeding (e.g. intracranial bleeding, GI bleeding), or high bleeding risk: within 3 months prior to elective THR for GI bleeding Participants with hepatic disease which is associated with coagulopathy leading to a clinically relevant bleeding risk Severe impaired renal function (CLCR (creatinine clearance) calculated by Cockcroft‐Gault formula: < 30 mL/min) Conditions prohibiting bilateral venography (e.g. amputation of 1 leg, allergy to contrast media) Ongoing anticoagulant therapy (e.g. warfarin, heparins and Factor Xa inhibitors other than study medication) that cannot be stopped (in the opinion of the investigator/sub investigator) People in whom epidural catheters are expected to be left in for longer than 18 hours post‐operatively Planned IPC during treatment period
Interventions	Number of intervention groups: 4 Concomitant interventions: none Excluded interventions: none Rivaroxaban 5 mg once daily group: Intervention: rivaroxaban (oral) 5 mg once daily Starting time: not reported Duration: 11 to 12 ± 2 days Rivaroxaban 7.5 mg once daily group: Intervention: rivaroxaban (oral) 7.5 mg once daily Starting time: not reported Duration: 11 to 12 ± 2 days Rivaroxaban 10 mg once daily group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: not reported Duration: 11 to 12 ± 2 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 20 mg twice daily Starting time: not reported Duration: 10 to 11 ± 2 days
Outcomes	Number of outcomes according to protocol: 10 Reported: 12. Reported additionally: all‐cause death, AEs, serious AEs, drug‐related AEs, liver enzymes elevation (treatment period). No safety outcomes with enough information to included data for analysis. Not reported: minor bleeding, proximal DVT, distal DVT (treatment period). OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 13 ± 4 days (treatment), 30 days (additionally, follow‐up) Time points reported: 13 ± 4 days (treatment period) Major VTE Definition: incidence of fatal and nonfatal PE or proximal DVT (diagnostic instrument not stated) Time points measured: 13 ± 4 days (treatment), 30 days (additionally, follow‐up) Time points reported: 13 ± 4 days (treatment period) Symptomatic VTE Definition: incidence of symptomatic VTE Time points measured: 13 ± 4 days (treatment), 30 days (additionally, follow‐up) Time points reported: 13 ± 4 days (treatment period), 30 days (additionally, follow‐up) Fatal VTE Definition: incidence of death due to VTE Time points measured: 13 ± 4 days (treatment), 30 days (additionally, follow‐up) Time points reported: 13 ± 4 days (treatment period) RESULTS All‐cause mortality: Summary data (13 ± 4 days): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 89 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): not reported Major VTE: Summary data (13 ± 4 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 89 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): not reported Symptomatic VTE: Summary data (13 ± 4 days): see Analysis 9.3 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 89 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): rivaroxaban 5 mg once daily ‐ 7.5 mg once daily ‐ 10 mg once daily ‐ enoxaparin 20 mg twice daily: 0/NR ‐ 1/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, not calculable missing participants ‐Subgroups reported: none Fatal VTE: Summary data (13 ± 4 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 89 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): not reported
Notes	Source of funding: Bayer Yakuhin, Ltd. Conflicts of interest: not reported Published protocol: registered in ClinicalTrials.gov as NCT01206972 Observations: none Missing data requested from authors: authors could not be contacted Missing data obtained from authors: none For the efficacy set, 29.5% of the randomised participants were excluded (data were insufficient to calculate the proportions by study group). Reasons for exclusion were not reported in enough detail. For the safety set, 1.3% of the randomised participants were excluded (data were insufficient to calculate the proportions by study group). Reasons for exclusion were not reported in enough detail. No CONSORT flow diagram
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method was not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of investigator was stated, but no details are provided. Blinding of outcome assessment was not mentioned. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Reasons for missing outcome data related to true outcome. Proportion of missing outcomes compared with the observed event risk is large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Reasons for missing outcome data related to true outcome. Insufficient information regarding reasons for missing outcome.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: five outcomes not included in the protocol were reported, three outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There is at least one important risk of bias. The study did not report if conflicts of interest were registered.

ODIXa‐HIP 2007.

Study characteristics
Methods	Aim: dose‐response study, efficacy and safety Design: parallel‐group/dose‐escalation (phase IIa) Blinding: open‐label; groups blinded: outcome adjudicators Study duration: December 2002 to December 2003 Duration of intervention: 5 to 9 days Duration of follow‐up: 30 to 60 days Method of randomisation: in a 3:1 ratio (rivaroxaban:enoxaparin), rivaroxaban dose groups were established sequentially. No additional description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) had data on safety outcome, 3) underwent planned surgery, 4) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who: 1) received at least one dose of study medication, 2) had data on safety outcome Withdrawals Rivaroxaban 2.5 mg twice daily: 6 (2 due to adverse event, 1 due to death, 3 had consent withdrawn/protocol violation); 5 mg twice daily: 8 (3 due to adverse event, 5 had consent withdrawn/protocol violation); 10 mg twice daily: 0; 30 mg once daily: 9 (3 due to adverse event, 1 due to death, 5 had consent withdrawn/protocol violation); 20 mg twice daily: 8 (5 due to adverse event, 3 had consent withdrawn/protocol violation); 30 mg twice daily: 20 (14 due to adverse event, 6 had consent withdrawn/protocol violation); Enoxaparin: 7 (4 due to adverse event, 3 had consent withdrawn/protocol violation) Exclusions: rivaroxaban 2.5 mg twice daily: 8 (not detailed); 5 mg twice daily: 13 (not detailed); 10 mg twice daily: 13 (not detailed); 30 mg once daily: 9 (not detailed); 20 mg twice daily: 12 (not detailed); 30 mg twice daily: 14 (not detailed); Enoxaparin: 48 (not detailed)
Participants	Diagnosis: scheduled to THR, elective, primary Location of participants: 10 countries (Europe and Israel) Number of participants randomised: 641 Age (years; mean and range): (safety analysis population) rivaroxaban 2.5 mg twice daily / 5 mg twice daily / 10 mg twice daily / 30 mg once daily / 20 mg twice daily / 30 mg twice daily 64 (31‐86) / 67 (42‐84) / 65 (39‐89) / 66 (41‐84) / 66 (32‐84) / 64 (30‐87); enoxaparin 64 (30‐92) Gender (male %): (safety analysis population) rivaroxaban 2.5 mg twice daily / 5 mg twice daily / 10 mg twice daily / 30 mg once daily / 20 mg twice daily / 30 mg twice daily 29 (38%) / 29 (36%) / 24 (35%) / 38 (43%) / 32 (42%) / 34 (46%); enoxaparin 74 (46%) Baseline imbalances: slight differences regarding gender, surgery details and type of anaesthesia among the groups Inclusion criteria: Males aged ≥ 18 years and post‐menopausal females scheduled for elective, primary, total hip replacement surgery Exclusion criteria: DVT, PE, myocardial infarction, transient ischaemic attack or stroke within 6 months of study entry Severe hypertension Severely impaired hepatic or renal function Body weight < 45 kg Drug or alcohol abuse Therapy with any other drugs that might affect the study outcome, including anticoagulants, aspirin or other antiplatelet agents, factor Xa inhibitors other than the study medication, or any other drug affecting coagulation (except nonsteroidal anti‐inflammatory drugs with a half‐life of < 17 hours)
Interventions	Number of intervention groups: 7 Concomitant interventions: none as part of the study; graduated compression stockings were allowed Excluded interventions: IPC Rivaroxaban 2.5 mg twice daily group: Intervention: rivaroxaban (oral) 2.5 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 5 mg twice daily group: Intervention: rivaroxaban (oral) 5 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 10 mg twice daily group: Intervention: rivaroxaban (oral) 10 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 30 mg once daily group: Intervention: rivaroxaban (oral) 30 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 20 mg twice daily group: Intervention: rivaroxaban (oral) 20 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 30 mg twice daily group: Intervention: rivaroxaban (oral) 30 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: evening Duration: 5 to 9 days
Outcomes	Number of outcomes according to protocol: 8 Reported: 14. Reported additionally: all‐cause death, major VTE (proximal DVT and non‐fatal PE and fatal VTE), major bleeding, CRNM bleeding, minor bleeding, serious AEs, liver enzymes elevation (treatment period), volume of blood transfusions, volume of blood loss (treatment period, assumed) Not reported: distal DVT (treatment period), DVT and non‐fatal PE and fatal VTE (follow‐up period), symptomatic VTE (treatment and follow‐up periods). OUTCOMES Major VTE (bilateral venography) Symptomatic VTE Major bleeding Fatal VTE Fatal bleeding Liver enzymes elevation Volume of blood loss RESULTS All cause mortality: Summary data (7 to 11 days): rivaroxaban 2.5 mg twice daily ‐ 5 mg twice daily ‐ 10 mg twice daily ‐ 30 mg once daily ‐ 20 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 40 mg twice daily: 1/63 ‐ 0/63 ‐ 0/55 ‐ 1/73 ‐ 0/59 ‐ 0/46 ‐ 0/107 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 175 missing participants (117 not detailed) ‐Subgroups reported: none Summary data (30 to 60‐day follow‐up): not reported Major VTE: Summary data (7 to 11 days): rivaroxaban 2.5 mg twice daily ‐ 5 mg twice daily ‐ 10 mg twice daily ‐ 30 mg once daily ‐ 20 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 40 mg twice daily: 7/63 ‐ 5/63 ‐ 2/55 ‐ 1/73 ‐ 0/59 ‐ 2/46 ‐ 5/107 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 175 missing participants (117 not detailed) ‐Subgroups reported: none Summary data (30 to 60‐day follow‐up): not reported Major bleeding: Summary data (7 to 11 days): rivaroxaban 2.5 mg twice daily ‐ 5 mg twice daily ‐ 10 mg twice daily ‐ 30 mg once daily ‐ 20 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 40 mg twice daily: 0/76 ‐ 2/80 ‐ 2/68 ‐ 4/88 ‐ 5/77 ‐ 8/74 ‐ 0/162 ‐Analysis set, type of analysis and missing participants: safety set, modified intention‐to‐treat (mITT) analysis, 16 missing participants ‐Subgroups reported: none Summary data (30 to 60‐day follow‐up): not reported Fatal VTE: Summary data (7 to 11 days): rivaroxaban 2.5 mg twice daily ‐ 5 mg twice daily ‐ 10 mg twice daily ‐ 30 mg once daily ‐ 20 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 40 mg twice daily: 1/63 ‐ 0/63 ‐ 0/55 ‐ 0/73 ‐ 0/59 ‐ 0/46 ‐ 0/107 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 175 missing participants (117 not detailed) ‐Subgroups reported: none Summary data (30 to 60‐day follow‐up): not reported Fatal bleeding: Summary data (5 to 9 days): rivaroxaban 2.5 mg twice daily ‐ 5 mg twice daily ‐ 10 mg twice daily ‐ 30 mg once daily ‐ 20 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 40 mg twice daily: 0/76 ‐ 0/80 ‐ 0/68 ‐ 0/88 ‐ 0/77 ‐ 0/74 ‐ 0/162 ‐Analysis set, type of analysis and missing participants: safety set, modified intention‐to‐treat (mITT) analysis, 16 missing participants ‐Subgroups reported: none Summary data (30 to 60‐day follow‐up): not reported Liver enzymes elevation: Summary data (5 to 9 days): rivaroxaban 2.5 mg twice daily ‐ 5 mg twice daily ‐ 10 mg twice daily ‐ 30 mg once daily ‐ 20 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 40 mg twice daily: 4/76 ‐ 3/78 ‐ 1/66 ‐ 6/86 ‐ 1/75 ‐ 3/72 ‐ 7/152 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 36 missing participants (20 unexplained) ‐Subgroups reported: none Summary data (30 to 60‐day follow‐up): not reported Volume of blood loss: Summary data (time point not reported) (mean ± SD): rivaroxaban 2.5 mg twice daily ‐ 5 mg twice daily ‐ 10 mg twice daily ‐ 30 mg once daily ‐ 20 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 40 mg twice daily: 652±394 ‐ 579±427 ‐ 632±483 ‐ 597±360 ‐ 621±400 ‐ 738±533 ‐ 603±330 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 44 missing participants (28 unexplained) ‐Subgroups reported: none
Notes	Source of funding: not stated Conflicts of interest: not stated Published protocol: registered in ClinicalTrials.gov as NCT00839826 Observations: during follow‐up period, patients received additional prophylaxis at the discretion of the investigator Missing data requested from authors: specific groups blinded, description of method of randomisation and concealment of allocation, definition of serious AEs, time points for measurement of volume of blood loss, treatment period summary data for serious AEs, follow‐up summary data for all‐cause mortality, fatal VTE, major VTE, fatal bleeding, major bleeding, liver enzymes elevation, treatment period missing participants for all‐cause mortality, fatal VTE, major VTE, liver enzymes elevation, volume of blood loss Missing data obtained from authors: none For the efficacy set (per‐protocol analysis), 27.3% of the randomised participants were excluded (25.1% in the rivaroxaban groups and 34.0% in the enoxaparin group). Reasons for exclusion were not reported in enough detail for 117 (18.3%) participants. For the safety set, 2.5% of the randomised participants were excluded (3.3% in the rivaroxaban group and 0% in the enoxaparin group). Reasons for exclusion were reported. However, for liver enzymes elevation and volume of blood loss, data from other sets (not defined, reasons not explained) were reported. CONSORT flow diagram incomplete.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Randomisation was stated, but the method was not described. Of note, under the dose‐escalation design, the probability of participants being allocated to each rivaroxaban group was different.
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	High risk	Study intervention was open‐label for the rivaroxaban versus enoxaparin comparisons
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by a blinded outcome adjudication committee (except for liver enzymes elevation and volume of blood loss). Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data unbalanced in reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: nine outcomes not included in the protocol were reported, three outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There is at least two important risks of bias. The study did not report if conflicts of interest were registered. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

ODIXa‐HIP2 2006.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: double‐blind, double‐dummy; groups blinded: participants, outcome adjudicators Study duration: January 2004 to August 2005 Duration of intervention: 5 to 9 da Duration of follow‐up: 30‐60d Method of randomisation: no description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set (per‐protocol population): participants who: 1) received at least one dose of study medication, 2) had data on safety outcome, 3) underwent planned surgery, 4) had a primary efficacy outcome that could be evaluated, 5) did not show any major protocol violations Primary safety analysis set: participants who: 1) received at least one dose of study medication, 2) had data on safety outcomes Withdrawals: not reported Exclusions: rivaroxaban 2.5 mg twice daily: 31 (3 did not receive study drug, 25 had inadequate efficacy assessment, 3 had protocol violations); 5 mg twice daily: 29 (2 did not receive study drug, 25 had inadequate efficacy assessment, 2 had protocol violations); 10 mg twice daily: 37 (5 did not receive study drug, 30 had inadequate efficacy assessment, 2 had protocol violations); 20 mg twice daily: 38 (3 did not receive study drug, 33 had inadequate efficacy assessment, 2 had protocol violations); 30 mg twice daily: 8 (8 had inadequate efficacy assessment); enoxaparin: 29 (3 did not receive study drug, 23 had inadequate efficacy assessment, 3 had protocol violations)
Participants	Diagnosis: scheduled to THR, elective, primary Location of participants: 13 countries (Europe and Israel) Number of participants randomised: 722 Age (years; mean and range): (safety analysis population) rivaroxaban 2.5 mg twice daily / 5 mg twice daily / 10 mg twice daily / 20 mg twice daily / 30 mg twice daily 66 (26‐86) / 64 (31‐84) / 65 (43‐93) / 65 (35‐85) / 67 (51‐87); enoxaparin 65 (27‐82) Gender (male %): (safety analysis population) rivaroxaban 2.5 mg twice daily / 5 mg twice daily / 10 mg twice daily / 20 mg twice daily / 30 mg twice daily 47 (36%) / 63 (46%) / 53 (40%) / 51 (38%) / 16 (43%); enoxaparin 54 (41%) Baseline imbalances: slight differences regarding gender and type of anaesthesia Inclusion criteria: Males (aged ≥ 18 years) and postmenopausal female patients undergoing elective primary THR Exclusion criteria: Any bleeding disorder or taking drugs that may affect the study outcome, such as anticoagulants, platelet‐aggregation inhibitors, or any other drug influencing coagulation (except nonsteroidal anti‐inflammatory drugs with a half‐life of < 17 hours) DVT, PE, myocardial infarction, transient ischaemic attack, or ischaemic stroke during the 6 months before the study Severe hypertension; severe liver or renal impairment; medical conditions that may interfere with the study Body weight < 45 kg Alcohol or drug abuse
Interventions	Number of intervention groups: 6 Concomitant interventions: none as part of the study; elastic compression stockings were allowed Excluded interventions: none Rivaroxaban 2.5 mg twice daily group: Intervention: rivaroxaban (oral) 2.5 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 5 mg twice daily group: Intervention: rivaroxaban (oral) 5 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 10 mg twice daily group: Intervention: rivaroxaban (oral) 10 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 20 mg twice daily group: Intervention: rivaroxaban (oral) 20 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 30 mg twice daily group: Intervention: rivaroxaban (oral) 30 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: evening Duration: 5 to 9 days
Outcomes	Number of outcomes according to protocol: 8 Reported: 13. Reported additionally: major VTE, fatal VTE, major bleeding, fatal bleeding, CRNM bleeding, minor bleeding, non‐hepatic serious AEs (treatment period), all‐cause mortality (treatment and follow‐up periods), volume of blood transfusions, volume of blood loss (treatment period, assumed) Not reported: distal DVT (treatment period), symptomatic VTE, symptomatic PE, symptomatic DVT, DVT and PE and fatal VTE (follow‐up period) OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 5 to 9 days (treatment), 30 to 60 days (additionally, follow‐up) Time points reported: 5 to 9 days (treatment), 30 to 60 days (additionally, follow‐up) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT (bilateral venography) Time points measured: 5 to 9 days (treatment), 30 to 60 days (additionally, follow‐up) Time points reported: 5 to 9 days (treatment) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 5 to 9 days (treatment), 30 to 60 days (additionally, follow‐up) Time points reported: 5 to 9 days (treatment) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥2 g/dL within 24h; 2) transfusion of ≥ 2 U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding; 6) bleeding warranting treatment cessation Time points measured: 7 to 11 days (treatment, 2 days after the last study drug intake), 30 to 60 days (additionally, follow‐up) Time points reported: 7 to 11 days (treatment, 2 days after the last study drug intake) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 5 to 9 days (treatment), 30 to 60 days (additionally, follow‐up) Time points reported: 5 to 9 days (treatment) Fatal bleeding: Volume of blood loss: Definition: volume of post‐operative drainage Method/unit of measurement: mL Time points measured: not reported Time points reported: not reported RESULTS All‐cause mortality: Summary data (5 to 9 days): rivaroxaban 2.5 mg twice daily ‐ 5 mg twice daily ‐ 10 mg twice daily ‐ 20 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 40 mg once daily: 0/104 ‐ 0/109 ‐ 0/101 ‐ 0/99 ‐ 0/29 ‐ 0/106 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 174 missing participants (2 unexplained) ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): rivaroxaban 2.5 mg twice daily ‐ 5 mg twice daily ‐ 10 mg twice daily ‐ 20 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 40 mg once daily: 0/NR ‐ 1/NR ‐ 1/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, not calculable missing participants ‐Subgroups reported: none Major VTE: Summary data (5 to 9 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 174 missing participants (2 unexplained) ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): not reported Major bleeding: Summary data (7 to 11 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 18 missing participants ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): not reported Fatal VTE: Summary data (5 to 9 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 174 missing participants (2 unexplained) ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): not reported Fatal bleeding: Summary data (5 to 9 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 18 missing participants ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): not reported Volume of blood loss: Summary data (time point not reported) (mean ± SD): see Analysis 1.14 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 18 missing participants ‐Subgroups reported: none
Notes	Source of funding: Bayer HealthCare Conflicts of interest: authors had financial relationships with several pharmaceutical companies, including Bayer Healthcare Published protocol: registered in clinicaltrials.gov as NCT00398905 Observations: during follow‐up period, participants received additional prophylaxis at the discretion of the investigator Missing data requested from authors: specific groups blinded, description of method of randomisation and concealment of allocation, details about withdrawals, definition of serious AEs, time point for measurement of volume of blood loss, follow‐up summary data for fatal VTE, major VTE, fatal bleeding and major bleeding, treatment period number of participants for serious AEs, follow‐up number of participants for all‐cause mortality, treatment period missing participants for all‐cause mortality, fatal VTE, major VTE Missing data obtained from authors: none For the efficacy set (per‐protocol analysis), 24.1% of the randomised participants were excluded (24.6% in the rivaroxaban groups and 22.1% in the enoxaparin group). Reasons for exclusion were reported in except for 2 (0.3%) participants. For the safety set, 2.5% of the randomised participants were excluded (2.4% in the rivaroxaban group and 2.9% in the enoxaparin group). Reasons for exclusion were reported. CONSORT flow diagram incomplete.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method was not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by a blinded outcome adjudication committee (except for volume of blood loss). Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data balanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 10 outcomes not included in the protocol were reported, 4 outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

ODIXa‐HIP‐OD 2006.

Study characteristics
Methods	Aim: efficacy and safety Design: parallel‐group (phase II) Blinding: double‐blind, double‐dummy; groups blinded: participants, outcome adjudicators Study duration: November 2004 to July 2005 Duration of intervention: 5 to 9 days Duration of follow‐up: 30 to 60 days Method of randomisation: in 1:1:1:1:1 and 2:1:1:1:1:1 ratios (before and after the inclusion of the rivaroxaban 5 mg once daily group), no additional description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set (per‐protocol population): participants who: 1) received at least one dose of study medication, 2) had data on safety outcome, 3) underwent planned surgery, 4) had a primary efficacy outcome that could be evaluated, 5) did not show any major protocol violations Primary safety analysis set: participants who: 1) received at least one dose of study medication, 2) had data on safety outcome Withdrawals: not reported Exclusions: rivaroxaban 5 mg once daily: 39 (3 did not receive study drug, 30 had inadequate efficacy assessment, 6 had protocol violations); 10 mg once daily: 34 (5 did not receive study drug, 28 had inadequate efficacy assessment, 1 had protocol violations); 20 mg once daily: 36 (1 did not receive study drug, 1 adverse event, 26 had inadequate efficacy assessment, 7 had protocol violations, 1 had no safety data); 30 mg once daily: 41 (2 did not receive study drug, 31 had inadequate efficacy assessment, 7 had protocol violations, 1 had no safety data); 40 mg once daily: 52 (3 did not receive study drug, 37 had inadequate efficacy assessment, 7 had protocol violations, 5 had no safety data); enoxaparin: 53 (3 did not receive study drug, 43 had inadequate efficacy assessment, 7 had protocol violations)
Participants	Diagnosis: scheduled to THR, elective, primary Location of participants: 11 countries (Europe and Israel) Number of participants randomised: 873 Age (years; mean and range): (safety analysis population) rivaroxaban 5 mg once daily / 10 mg once daily / 20 mg once daily / 30 mg once daily / 40 mg once daily 64.8 (28‐84) / 64.0 (27‐87) / 65.0 (27‐93) / 65.4 (31‐86) / 64.7 (27‐83); enoxaparin 65.6 (30‐89) Gender (male %): (safety analysis population) rivaroxaban 5 mg once daily / 10 mg once daily / 20 mg once daily / 30 mg once daily / 40 mg once daily 56 (44%) / 53 (37%) / 57 (41%) / 69 (49%) / 56 (41%); enoxaparin 57 (36%) Baseline imbalances: slight differences regarding gender Inclusion criteria: Men aged 18 years and postmenopausal women scheduled for elective, primary total hip replacement surgery Exclusion criteria: DVT, PE, myocardial infarction, transient ischaemic attack, or ischaemic stroke during the 6 months before the study Intracerebral, intraocular, or gastrointestinal bleeding in the previous 6 months Taking drugs that might have affected the study outcome, such as other anticoagulants, platelet‐aggregation inhibitors, or any other drug influencing coagulation (except nonsteroidal antiinflammatory drugs with a half‐life < 17 hours) Severe hypertension, severe liver or renal impairment, medical conditions that may interfere with the study, or body weight < 45 kg Drug or alcohol abuse
Interventions	Number of intervention groups: 6 Concomitant interventions: none Excluded interventions: IPC Rivaroxaban 5 mg once daily group: Intervention: rivaroxaban (oral) 5 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 10 mg once daily group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 20 mg once daily group: Intervention: rivaroxaban (oral) 20 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 30 mg once daily group: Intervention: rivaroxaban (oral) 30 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 40 mg once daily group: Intervention: rivaroxaban (oral) 40 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: evening Duration: 5 to 9 days
Outcomes	Number of outcomes according to protocol: 10 Reported: 13. Reported additionally: major bleeding, CRNM bleeding, minor bleeding, liver enzymes elevation (treatment period), volume of blood transfusions, volume of blood loss (treatment period, assumed), major VTE (proximal DVT and non‐fatal PE and fatal VTE) (follow‐up period), all‐cause death (treatment and follow‐up period) Not reported: symptomatic DVT, symptomatic PE (follow‐up period), symptomatic VTE (treatment and follow‐up period) OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 5 to 9 days (treatment), 30 to 60 days (additionally, follow‐up) Time points reported: 5 to 9 days (treatment), 30 to 60 days (additionally, follow‐up) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT (bilateral venography) Time points measured: 5 to 9 days (treatment), 30 to 60 days (additionally, follow‐up) Time points reported: 5 to 9 days (treatment), 30 to 60 days (additionally, follow‐up) Symptomatic VTE: Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2g/dL within 24h; 2) transfusion of ≥ 2 U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding; 6) bleeding warranting treatment cessation Time points measured: 7 to 11 days (treatment, 2 days after the last study drug intake), 30 to 60 days (additionally, follow‐up) Time points reported: 7 to 11 days (treatment, 2 days after the last study drug intake) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 5 to 9 days (treatment), 30 to 60 days (additionally, follow‐up) Time points reported: 5 to 9 days (treatment) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 5 to 9 days (treatment), 30 to 60 days (additionally, follow‐up) Time points reported: 5 to 9 days (treatment) Liver enzymes elevation: Definition: incidence of ALT > 3 times ULN Time points measured: 5 to 11 days (treatment, 2 days after the last study drug intake) Time points reported: 5 to 11 days (treatment, 2 days after the last study drug intake) Volume of blood loss: Definition: volume of postoperative drainage Method/unit of measurement: mL Time points measured: not reported Time points reported: not reported RESULTS All‐cause mortality: Summary data (5 to 9 days): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 255 missing participants ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): rivaroxaban 5 mg once daily ‐ 10 mg once daily ‐ 20 mg once daily ‐ 30 mg once daily ‐ 40 mg once daily ‐ enoxaparin 40 mg once daily: 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, not calculable missing participants ‐Subgroups reported: none Major VTE: Summary data (5 to 9 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 255 missing participants ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): rivaroxaban 5 mg once daily ‐ 10 mg once daily ‐ 20 mg once daily ‐ 30 mg once daily ‐ 40 mg once daily ‐ enoxaparin 40 mg once daily: 0/NR ‐ 0/NR ‐ 1/NR ‐ 0/NR ‐ 2/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, not calculable missing participants ‐Subgroups reported: none Symptomatic VTE: Major bleeding: Summary data (7 to 11 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 28 missing participants Summary data (30 to 60 days' follow‐up): not reported Fatal VTE: Summary data (5 to 9 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 255 missing participants ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): not reported Fatal bleeding: Summary data (5 to 9 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 28 missing participants ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): not reported Liver enzymes elevation: Summary data (7 to 11 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 92 missing participants (64 unexplained) Volume of blood loss: Summary data (time point not reported) (mean ± SD): see Analysis 9.14 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 28 missing participants ‐Subgroups reported: none
Notes	Source of funding: Bayer HealthCare Conflicts of interest: authors had financial relationships with several pharmaceutical companies, including Bayer Healthcare Published protocol: registered in the EU clinical trials register as EUCTR2004‐001341‐14 and in clinicaltrials.gov as NCT00396786 Observations: during follow‐up period, participants received additional prophylaxis at the discretion of the investigator Missing data requested from authors: specific groups blinded, description of method of randomisation and concealment of allocation, details about withdrawals, time points for measurement of volume of blood loss, follow‐up summary data for fatal VTE, fatal bleeding, major bleeding, follow‐up number of participants for all‐cause mortality and major VTE, treatment period missing participants for liver enzymes elevation Missing data obtained from authors: none For the efficacy set (per‐protocol analysis), 29.2% of the randomised participants were excluded (28.3% in the rivaroxaban groups and 33.1% in the enoxaparin group). Reasons for exclusion were reported. For the safety set, 3.2% of the randomised participants were excluded (3.5% in the rivaroxaban group and 1.9% in the enoxaparin group). Reasons for exclusion were reported in enough detail. However, for liver enzymes elevation, data from other set (not defined, reasons not explained) were reported, for which 10.5% of participants were excluded.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method was not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by a blinded outcome adjudication committee (not stated for liver enzymes elevation and volume of blood loss). Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data balanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov and the EU clinical trials register. Reported outcomes differed from those included in the protocol: 8 outcomes not included in the protocol were reported, 3 outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

ODIXa‐KNEE 2005.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: double‐blind, double‐dummy; groups blinded: participants, outcome adjudicators Study duration: February 2004 to November 2004 Duration of intervention: 5 to 9 days Duration of follow‐up: 30 days (60 days for some participants, reasons not reported) Method of randomisation: computer‐generated randomisation list Method of concealment of allocation: use of an interactive voice response system for allocation of participants Primary efficacy analysis set (per‐protocol population): participants who: 1) received at least one dose of study medication, 2) had a primary efficacy outcome that could be evaluated, 3) did not show any major protocol violations Primary safety analysis set: participants who: 1) received at least one dose of study medication Withdrawals: not reported Exclusions: rivaroxaban groups (not detailed according to dose group): 186 (7 did not receive study drug, 97 had no venography, 82 had uninterpretable venography); enoxaparin group: 31 (1 did not receive study drug, 13 had no venography, 17 had uninterpretable venography)
Participants	Diagnosis: scheduled to TKR, elective Location of participants: 2 countries (Canada and USA) Number of participants randomised: 621 Age (years; mean and range): (safety analysis population) rivaroxaban 2.5 mg twice daily / 5 mg twice daily / 10 mg twice daily / 20 mg twice daily / 30 mg twice daily 66 (48‐84) / 66 (45‐86) / 67 (49‐84) / 68 (45‐92) / 66 (39‐86); enoxaparin 66 (47‐83) Gender (male %): (safety analysis population) rivaroxaban 2.5 mg twice daily / 5 mg twice daily / 10 mg twice daily / 20 mg twice daily / 30 mg twice daily 41 (41%) / 37 (36%) / 37 (36%) / 31 (32%) / 43 (41%); enoxaparin 47 (45%) Baseline imbalances: none Inclusion criteria: Males aged 18 years or above and postmenopausal females scheduled for elective TKR Signed an informed consent Exclusion criteria: Any bleeding disorder Current use of drugs that may affect study outcome such as anticoagulants, platelet aggregation inhibitors (e.g. aspirin, clopidogrel, dipyridamole), or any other drug influencing coagulation (except non‐steroidal anti‐inflammatory drugs with half‐life < 17 h) DVT, PE, myocardial infarction, transient ischaemic attack, or ischaemic stroke within the last 6 months Severe hypertension; severe liver or renal impairment Body‐weight < 45 kg Alcohol or drug abuse
Interventions	Number of intervention groups: 6 Concomitant interventions: none as part of the study; elastic compression stockings were allowed Excluded interventions: none Rivaroxaban 2.5 mg twice daily group: Intervention: rivaroxaban (oral) 2.5 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 5 mg twice daily group: Intervention: rivaroxaban (oral) 5 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 10 mg twice daily group: Intervention: rivaroxaban (oral) 10 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 20 mg twice daily group: Intervention: rivaroxaban (oral) 20 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Rivaroxaban 30 mg twice daily group: Intervention: rivaroxaban (oral) 30 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 5 to 9 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 30 mg twice daily Starting time: after surgery: morning Duration: 5 to 9 days
Outcomes	Number of outcomes according to protocol: 8 Reported: 17. Reported additionally: major VTE, major bleeding, fatal bleeding, CRNM bleeding, minor bleeding, non‐hepatic serious AEs, volume of blood transfusions, volume of blood loss (treatment period), all‐cause death, fatal VTE (treatment and follow‐up periods) Not reported: DVT and PE and fatal VTE (treatment period) OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 5 to 9 days (treatment), 30 days (additionally, follow‐up) Time points reported: 5 to 9 days (treatment), 30 days (additionally, follow‐up) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT (bilateral venography) Time points measured: 5 to 9 days (treatment), 30 days (additionally, follow‐up) Time points reported: 5 to 9 days (treatment) Symptomatic VTE: Definition: incidence of symptomatic VTE Time points measured: 5 to 9 days (treatment), 30 days (additionally, follow‐up) Time points reported: 5 to 9 days (treatment) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥2 g/dL within 24 hours; 2) transfusion of ≥ 2U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding; 6) bleeding warranting treatment cessation Time points measured: 5 to 11 days (treatment, 2 days after the last study drug intake), 30 days (additionally, follow‐up) Time points reported: 5 to 11 days (treatment, 2 days after the last study drug intake) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 5 to 9 days (treatment), 30 days (additionally, follow‐up) Time points reported: 5 to 9 days (treatment), 30 days (additionally, follow‐up) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 5 to 9 days (treatment), 30 days (additionally, follow‐up) Time points reported: 5 to 9 days (treatment) Volume of blood loss: Definition: volume of post‐operative drainage Method/unit of measurement: mL Time points measured: 5 to 9 days (treatment) Time points reported: 5 to 9 days (treatment) RESULTS All‐cause mortality: Summary data (5 to 9 days): rivaroxaban 2.5 mg twice daily ‐ 5 mg twice daily ‐ 10 mg twice daily ‐ 20 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 30 mg twice daily: 0/63 ‐ 0/57 ‐ 0/60 ‐ 0/57 ‐ 0/59 ‐ 0/70 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 255 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): rivaroxaban 2.5 mg twice daily ‐ 5 mg twice daily ‐ 10 mg twice daily ‐ 20 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 30 mg once daily: 2/NR ‐ 0/NR ‐ 1/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, not calculable missing participants ‐Subgroups reported: none Major VTE: Summary data (5 to 9 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 255 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): not reported Symptomatic VTE: Summary data (5 to 9 days): see Analysis 1.3 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 255 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): not reported Major bleeding: Summary data (7‐11d): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 8 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): not reported Fatal VTE: Summary data (5 to 9 days): rivaroxaban 2.5 mg twice daily ‐ 5 mg twice daily ‐ 10 mg twice daily ‐ 20 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 30 mg twice daily: 0/63 ‐ 0/57 ‐ 0/60 ‐ 0/57 ‐ 0/59 ‐ 0/70 ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, 255 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): rivaroxaban 2.5 mg twice daily ‐ 5 mg twice daily ‐ 10 mg twice daily ‐ 20 mg twice daily ‐ 30 mg twice daily ‐ enoxaparin 30 mg once daily: 1/NR ‐ 0/NR ‐ 1/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: per‐protocol set, per‐protocol analysis, not calculable missing participants ‐Subgroups reported: none Fatal bleeding: Summary data (5 to 9 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 8 missing participants ‐Subgroups reported: none Summary data (30‐day follow‐up): not reported Volume of blood loss: Summary data (5 to 9 days) (mean±SD): see Analysis 1.14 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 8 missing participants ‐Subgroups reported: none
Notes	Source of funding: Bayer HealthCare AG Conflicts of interest: authors had financial relationships with several pharmaceutical companies, including Bayer Healthcare Published protocol: registered in ClinicalTrials.gov as NCT00402467 Observations: during follow‐up period, participants received additional prophylaxis at the discretion of the treating physician Missing data requested from authors: specific groups blinded, details about follow‐up period, details about withdrawals and exclusions, definition of serious AEs, treatment period summary data for serious AEs, follow‐up summary data for symptomatic VTE, fatal bleeding and major bleeding, follow‐up number of participants for all‐cause mortality and fatal VTE Missing data obtained from authors: none For the efficacy set (per‐protocol analysis), 41.1% of the randomised participants were excluded (42.6% in the rivaroxaban groups and 33.3% in the enoxaparin group). Reasons for exclusion were reported. For the safety set, 1.3% of the randomised participants were excluded (1.4% in the rivaroxaban groups and 1.0% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random sequence was computer‐generated.
Allocation concealment (selection bias)	Low risk	Participants were allocated using an interactive voice response system.
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers.
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by a blinded outcome adjudication committee. Blinding of the data analysts was not mentioned.
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data balanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 9 outcomes not included in the protocol were reported, 1 outcome included in the protocol was not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

ONYX‐1 2007.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group (phase IIa) Blinding: open‐label; groups blinded: outcome adjudicators Study duration: not reported Duration of intervention: 7 to 10 days Duration of follow‐up: approximately 4 weeks Method of randomisation: computer‐generated random sequence; in a 4:1 ratio (darexaban doses:enoxaparin), darexaban dose groups were established sequentially; stratified by country Method of concealment of allocation: it was stated that random sequence was "generated centrally", but no additional description was provided regarding allocation concealment Primary efficacy analysis set: not defined Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: darexaban 3 mg once daily: 4 (2 withdrew before treatment, 2 discontinued treatment); 10 mg once daily: 2 (2 discontinued treatment); 30 mg once daily: 0; 60 mg once daily: 2 (2 withdrew before treatment); enoxaparin: 0 Exclusions: darexaban 3 mg once daily: 4 (excluding withdrawals, reasons not reported); 10 mg once daily: 2 (excluding withdrawals, reasons not reported); 30 mg once daily: 5 (excluding withdrawals, reasons not reported); 60 mg once daily: 7 (excluding withdrawals, reasons not reported); enoxaparin: 5 (excluding withdrawals, reasons not reported)
Participants	Diagnosis: scheduled to HR, elective, primary Location of participants: 4 countries (Denmark, Sweden, Poland, Czech Republic) Number of participants randomised: 178 Age (years; mean and range): (population not specified, not detailed according to study group) 63.3 (35‐87) Gender (male %): (population not specified, not detailed according to study group) 80 (45%) Baseline imbalances: not assessable Inclusion criteria: Scheduled for elective primary THR surgery Aged 18 years or over Provided written informed consent Exclusion criteria: Any factor considered to be associated with an increased risk of VTE (documented history of DVT or PE, bedridden for longer than 7 days prior to the operation, malignancy currently under cytostatic or cytotoxic treatment, or being the reason for HR) or bleeding (known hemorrhagic disorder, thrombocytopenia, history of recent clinically important bleeding or major trauma, major surgery, or eye, spinal cord, or brain surgery, acute bacterial endocarditis, severe hypertension, retinopathy, planned indwelling intrathecal or epidural catheter for more than 6 hours after the end of surgery) Concomitant use of anticoagulants/antiplatelet agents Hypersensitivity to iodinated contrast media or enoxaparin Myocardial infarction or stroke within the last 6 months Body weight less than 50 kg, and ALT and AST above upper limit of normal, and/or creatinine greater or equal to 120 μmol/L
Interventions	Number of intervention groups: 5 Concomitant interventions: none Excluded interventions: none Darexaban 3 mg once daily group: Intervention: darexaban (oral) 3 mg once daily Starting time: after surgery: 6 to 10 hours Duration: 7 to 10 days Darexaban 10 mg once daily group: Intervention: darexaban (oral) 10 mg once daily Starting time: after surgery: 6 to 10 hours Duration: 7 to 10 days Darexaban 30 mg once daily group: Intervention: darexaban (oral) 30 mg once daily Starting time: after surgery: 6 to 10 hours Duration: 7 to 10 days Darexaban 60 mg once daily group: Intervention: darexaban (oral) 60 mg once daily Starting time: after surgery: 6 to 10 hours Duration: 7 to 10 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: 12 hours Duration: 7 to 10 days
Outcomes	Number of outcomes according to protocol: no protocol available OUTCOMES Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT (bilateral venography) Time points measured: 7 to 10 days (treatment), 4 weeks (additionally, follow‐up) Time points reported: 7 to 10 days (treatment) Symptomatic VTE: Definition: incidence of symptomatic VTE Time points measured: 7 to 10 days (treatment), 4 weeks (additionally, follow‐up) Time points reported: 7 to 10 days (treatment), 4 weeks (additionally, follow‐up) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) a bleeding index ≥ 2.0 (bleeding index = units of red blood cells transfused and prebleed Hb ‐ postbleed Hb); 2) bleeding at a critical site (intracranial, intraocular, pericardial, or retroperitoneal); 3) bleeding into the operated joint, requiring an additional operation or intervention; 4) fatal bleeding Time points measured: 7 to 10 days (treatment), 4 weeks (additionally, follow‐up) Time points reported: 7 to 10 days (treatment) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 7 to 10 days (treatment), 4 weeks (additionally, follow‐up) Time points reported: 7 to 10 days (treatment) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 7 to 10 days (treatment), 4 weeks (additionally, follow‐up) Time points reported: 7 to 10 days (treatment) RESULTS Major VTE: Summary data (7 to 10 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: not specified, not specified, 31 missing participants (23 unexplained) ‐Subgroups reported: none Summary data (4‐week follow‐up): not reported Symptomatic VTE: Summary data (7 to 10 days): darexaban 3 mg once daily ‐ 10 mg once daily ‐ 30 mg once daily ‐ 60 mg once daily ‐ enoxaparin 40 mg once daily: 0/27 ‐ 0/31 ‐ 0/31 ‐ 0/27 ‐ 0/31 ‐Analysis set, type of analysis and missing participants: not specified, not specified, 31 missing participants (23 unexplained) ‐Subgroups reported: none Summary data (4‐week follow‐up): darexaban 3 mg once daily ‐ 10 mg once daily ‐ 30 mg once daily ‐ 60 mg once daily ‐ enoxaparin 40 mg once daily: 0/NR ‐ 0/NR ‐ 1/NR ‐ 1/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Major bleeding: Summary data (7 to 10 days): see Analysis 1.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 3 missing participants ‐Subgroups reported: none Summary data (4‐week follow‐up): not reported Fatal VTE: Summary data (7 to 10 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: not specified, not specified, 31 missing participants (23 unexplained) ‐Subgroups reported: none Summary data (4‐week follow‐up): not reported Fatal bleeding: Summary data (7 to 10 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 3 missing participants ‐Subgroups reported: none Summary data (4‐week follow‐up): not reported
Notes	Source of funding: Astellas Pharma Inc. Conflicts of interest: all authors had financial relationships with several pharmaceutical companies, including Astellas Pharma Inc. Published protocol: none Observations: none Missing data requested from authors: specific groups blinded, date of start and end of study, description of method of randomisation and concealment of allocation, reasons for exclusion of participants, data on age and gender of participants, treatment period summary data for serious AEs, follow‐up summary data for all‐cause mortality, fatal VTE, major VTE, fatal bleeding, major bleeding, serious AEs, treatment period number of participants for all‐cause mortality, follow‐up number of participants for symptomatic VTE, treatment period missing participants for fatal VTE, major VTE and symptomatic VTE, protocol of the study Missing data obtained from authors: none For the efficacy set, 17.4% of the randomised participants were excluded (18.3% in the darexaban groups and 13.9% in the enoxaparin group). Reasons for exclusion were not reported in enough detail for 23 (12.9%) participants. For the safety set, 1.3% of the randomised participants were excluded (1.4% in the darexaban groups and 1.0% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random sequence was centrally computer‐generated
Allocation concealment (selection bias)	Unclear risk	Although the use of a centrally generated random sequence was stated for allocation of participants, this description does not imply that the allocation itself was carried out centrally.
Blinding of participants and personnel (performance bias)	High risk	Study intervention was open‐label for the darexaban versus enoxaparin comparisons.
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded central adjudication committee. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Reasons for missing outcome data related to true outcome. Missing outcome data unbalanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data balanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

ONYX‐2 2010.

Study characteristics
Methods	Aim: dose‐response study, efficacy and safety Design: parallel‐group Blinding: open‐label (for the darexaban versus enoxaparin comparisons); groups blinded: outcome adjudicators Study duration: June 2006 to April 2007 Duration of intervention: 5 weeks Duration of follow‐up: 4 to 5 weeks (4 additional weeks for liver laboratory analyses) Method of randomisation: computer‐generated random sequence; stratified by country Method of concealment of allocation: it was stated that the random sequence was generated "centrally", but no additional description was provided regarding allocation concealment Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) underwent surgery, 3) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who: 1) received at least one dose of study medication, 2) underwent surgery Withdrawals: not reported Exclusions: darexaban 5 mg once daily: 52 (11 did not undergo planned surgery, 41 not detailed); 10 mg once daily: 51 (7 did not undergo planned surgery, 44 not detailed); 30 mg once daily: 56 (13 did not undergo planned surgery, 43 not detailed); 60 mg once daily: 51 (8 did not undergo planned surgery, 43 not detailed); 120 mg once daily: 58 (12 did not undergo planned surgery, 46 not detailed); enoxaparin: 41 (3 did not undergo planned surgery, 38 not detailed)
Participants	Diagnosis: scheduled to THA, elective, primary Location of participants: 17 countries (Europe) Number of participants randomised: 1017 Age (years; mean and range): (safety analysis population) darexaban 5 mg once daily / 10 mg once daily / 30 mg once daily / 60 mg once daily / 120 mg once daily 60.2 (34‐82) / 60.3 (29‐84) / 57.9 (23‐85) / 61.3 (24‐84) / 60.2 (26‐83); enoxaparin 58.1 (22‐85) Gender (male %): (safety analysis population) darexaban 5 mg once daily / 10 mg once daily / 30 mg once daily / 60 mg once daily / 120 mg once daily 68 (43.7%) / 73 (45.3%) / 77 (49.4%) / 73 (44.8%) / 77 (49.4%); enoxaparin 80 (48.2%) Baseline imbalances: slight differences regarding surgical technique and type of anaesthesia Inclusion criteria: People (18 years or older) scheduled for elective primary total hip arthroplasty Exclusion criteria: Factors likely to be associated with an increased risk of VTE (bedridden for longer than 7 days before surgery, malignancies) and a documented history of VTE (major exclusion criteria) Factors associated with an increased risk of bleeding (known hemorrhagic or thrombotic disorders, history of recent clinically important bleeding or major surgery) Any contraindication for use of enoxaparin besides liver enzymes above the ULN Surgery planned for a bilateral hip arthroplasty or surgery of the contralateral hip within 10 weeks after enrolment Pregnant, breastfeeding women and women of childbearing potential who were not taking adequate contraceptive precautions
Interventions	Number of intervention groups: 6 Concomitant interventions: none Excluded interventions: none Darexaban 5 mg once daily group: Intervention: darexaban (oral) 5 mg once daily Starting time: after surgery: 6 to 10 hours Duration: 5 weeks Darexaban 10 mg once daily group: Intervention: darexaban (oral) 10 mg once daily Starting time: after surgery: 6 to 10 hours Duration: 5 weeks Darexaban 30 mg once daily group: Intervention: darexaban (oral) 30 mg once daily Starting time: after surgery: 6 to 10 hours Duration: 5 weeks Darexaban 60 mg once daily group: Intervention: darexaban (oral) 60 mg once daily Starting time: after surgery: 6 to 10 hours Duration: 5 weeks Darexaban 120 mg once daily group: Intervention: darexaban (oral) 120 mg once daily Starting time: after surgery: 6 to 10 hours Duration: 5 weeks Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: 12 ± 2 hours Duration: 5 weeks
Outcomes	Number of outcomes: According to protocol: 8 Reported: 16. Reported additionally: fatal VTE, major VTE, PE, proximal DVT, distal DVT, symptomatic DVT, total bleeding, fatal bleeding (primary treatment period) Not reported: none OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 6 to 9 days (primary treatment period), 6 weeks (up to 7 days post‐treatment), 4 to 5 weeks (additionally, follow‐up) Time points reported: 6 to 9 days (primary treatment period) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT (bilateral venography) Time points measured: 6 to 9 days (primary treatment period), 6 weeks (up to 7 days post‐treatment), 4 to 5 weeks (additionally, follow‐up) Time points reported: 6 to 9 days (primary treatment period) Symptomatic VTE: Definition: incidence of symptomatic VTE Time points measured: 6 to 9 days (primary treatment period), 6 weeks (up to 7 days post‐treatment), 4 to 5 weeks (additionally, follow‐up) Time points reported: 6 to 9 days (primary treatment period) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) a bleeding index ≥ 2.0 (bleeding index = units of red blood cells transfused and prebleed Hb ‐ postbleed Hb); 2) bleeding at a critical site (intracranial, intraocular, pericardial, or retroperitoneal); 3) bleeding into the operated joint, requiring an additional operation or intervention; 4) fatal bleeding Time points measured: 6 to 9 days (primary treatment period), 6 weeks (up to 7 days post‐treatment), 4 to 5 weeks (additionally, follow‐up) Time points reported: 6 to 9 days (primary treatment period) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 6 to 9 days (primary treatment period), 6 weeks (up to 7 days post‐treatment), 4 to 5 weeks (additionally, follow‐up) Time points reported: 6 to 9 days (primary treatment period) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 6 to 9 days (primary treatment period), 6 weeks (up to 7 days post‐treatment), 4 to 5 weeks (additionally, follow‐up) Time points reported: 6 to 9 days (primary treatment period) RESULTS All‐cause mortality: Summary data (6 to 9 days): see Analysis 9.1 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 309 missing participants (255 not detailed) ‐Subgroups reported: none Summary data (6 weeks): not reported Summary data (4 to 5 weeks' follow‐up): not reported Major VTE: Summary data (6 to 9 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 309 missing participants (255 not detailed) ‐Subgroups reported: none Summary data (6 weeks): not reported Summary data (4 to 5 weeks' follow‐up): not reported Symptomatic VTE: Summary data (6 to 9 days): see Analysis 9.3 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 309 missing participants (255 not detailed) ‐Subgroups reported: none Summary data (6 weeks): not reported Summary data (4 to 5 weeks' follow‐up): not reported Major bleeding: Summary data (6 to 9 days): darexaban 5 mg once daily ‐ 10 mg once daily ‐ 30 mg once daily ‐ 60 mg once daily ‐ 120 mg once daily ‐ enoxaparin 40 mg once daily: 0/158 ‐ 0/161 ‐ 0/156 ‐ 1/163 ‐ 0/156 ‐ 1/166 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 57 missing participants ‐Subgroups reported: none Summary data (6 weeks): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 57 missing participants ‐Subgroups reported: none Summary data (4 to 5 weeks' follow‐up): not reported Fatal VTE: Summary data (6 to 9 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 309 missing participants (255 not detailed) ‐Subgroups reported: none Summary data (6 weeks): not reported Summary data (4 to 5 weeks' follow‐up): not reported Fatal bleeding: Summary data (6 to 9 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 57 missing participants ‐Subgroups reported: none Summary data (6 weeks): not reported Summary data (4 to 5 weeks' follow‐up): not reported
Notes	Source of funding: Astellas Pharma Inc. Conflicts of interest: all authors had financial relationships with several pharmaceutical companies, including Astellas Pharma Inc. Published protocol: registered in clinicaltrials.gov as NCT00353678, and in ICTRP as EUCTR2005‐002457‐41‐AT Observations: treatment lasted 5 weeks, but first outcome measure time point was at 6 to 9 days Missing data requested from authors: specific groups blinded, description of method of concealment of allocation, details about withdrawals and exclusions, follow‐up summary data for all‐cause mortality, fatal VTE, major VTE, symptomatic VTE, fatal bleeding, major bleeding, treatment period missing participants for all‐cause mortality, fatal VTE, major VTE and symptomatic VTE Missing data obtained from authors: none For the efficacy set, 30.4% of the randomised participants were excluded (31.6% in the darexaban groups and 24.4% in the enoxaparin group). Reasons for exclusion were not reported in enough detail for 255 (25.1%) participants. For the safety set, 5.6% of the randomised participants were excluded (6.5% in the darexaban groups and 1.2% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random sequence was centrally computer‐generated
Allocation concealment (selection bias)	Unclear risk	Although the use of a centrally generated random sequence was stated for allocation of participants, this description does not imply that the allocation itself was carried out centrally
Blinding of participants and personnel (performance bias)	High risk	Study intervention was open‐label for the darexaban versus enoxaparin comparisons
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent central adjudication committee, but it was not explicitly stated as blinded to treatment allocation. Data analysis was performed by the sponsor, and blinding was not stated
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in numbers across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 8 outcomes not included in the protocol were reported, 0 outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

ONYX‐3 2014.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group (phase IIb) Blinding: double‐blind, double‐dummy, with the use of double‐blind labelled supplies provided by a central site; groups blinded: participants, clinicians/providers, outcome adjudicators Study duration: April 2009 to August 2010 Duration of intervention: 35 days Duration of follow‐up: 30 days Method of randomisation: computer‐generated random sequence; in a 1:1:1:1:1 ratio Method of concealment of allocation: not mentioned Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) underwent planned surgery, 3) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who: 1) received at least one dose of study medication, 2) underwent planned surgery Withdrawals: not reported Exclusions (not detailed according to intervention group): 546 (55 did not receive study drug or planned surgery, 15 excluded from safety set, 386 had no evaluable venography, 90 had incomplete venography)
Participants	Diagnosis: scheduled to hip arthroplasty, elective, primary Location of participants: 28 countries (Australia, Austria, Brazil, Canada, Colombia, Czech Republic, Denmark, Estonia, Finland, Germany, Hungary, India, Israel, Italy, Latvia, Lithuania, The Netherlands, Norway, Poland, Romania, Russian Federation, Slovakia, South Africa, Spain, Sweden, UK, Ukraine, USA) Number of participants randomised: 1992 Age (years; mean ± SD): (safety analysis population) darexaban 15 mg twice daily / 30 mg once daily / 30 mg twice daily / 60 mg once daily 60.4 ± 10.95 / 59.5 ± 11.73 / 59.8 ± 12.13 / 59.5 ± 11.82; enoxaparin 61.1 ± 11.79 Gender (male %): (safety analysis population) darexaban 15 mg twice daily / 30 mg once daily / 30 mg twice daily / 60 mg once daily 188 (50.3%) / 191 (49.9%) / 204 (52.7%) / 182 (47.3%); enoxaparin 166 (42.2%) Baseline imbalances: slightly higher proportion of female and osteoarthritis patients in enoxaparin group Inclusion criteria: Males and females aged > 18 years who were scheduled for elective primary hip arthroplasty Exclusion criteria: Active bleeding or any condition associated with an increased risk of bleeding, including known haemorrhagic disorder or thrombocytopenia (platelets < 100,000/mm3) at screening History of myocardial infarction or stroke within three months before the planned surgery Persistent blood pressure of 160 mmHg systolic and/or 100 mmHg diastolic at baseline, with or without medication History of major trauma, major surgery, or eye, spinal cord or brain surgery within three months of the planned surgery Scheduled major surgery or other invasive procedures with the potential for uncontrolled bleeding during the study Planned indwelling intrathecal or epidural catheter for > 6 hours after the end of surgery Creatinine clearance < 60 mL/minute (as calculated using the Cockroft−Gault formula) Hepatic impairment (according to alanine aminotransferase [ALT], aspartate aminotransferase [AST] and bilirubin measurements) Any concurrent illness that could interfere with the conduct and outcomes of the study
Interventions	Number of intervention groups: 5 Concomitant interventions: none Excluded interventions: none Darexaban 15 mg twice daily group: Intervention: darexaban (oral) 15 mg twice daily Starting time: after surgery: 6 to 10 hours Duration: 35 days Darexaban 30 mg once daily group: Intervention: darexaban (oral) 30 mg once daily Starting time: after surgery: 6 to 10 hours Duration: 35 days Darexaban 30 mg twice daily group: Intervention: darexaban (oral) 30 mg twice daily Starting time: after surgery: 6 to 10 hours Duration: 35 days Darexaban 60 mg once daily group: Intervention: darexaban (oral) 60 mg once daily Starting time: after surgery: 6 to 10 hours Duration: 35 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: 12 ± 2 hours Duration: 35 days
Outcomes	Number of outcomes: According to protocol: 3 Reported: 3. Reported additionally: fatal VTE, major VTE (primary treatment period), major bleeding (primary treatment, overall treatment, follow‐up and entire study periods) Not reported: composite of symptomatic VTE and all‐cause death (primary treatment period), VTE (primary treatment, overall treatment, follow‐up and entire study periods), bleeding (overall treatment, follow‐up and entire study periods) OUTCOMES Major VTE: Definition: incidence of fatal and nonfatal PE or VTE (bilateral venography) Time points measured: 8 to 12 days (primary treatment period), 35 days (treatment), 30 days (additionally, follow‐up) Time points reported: 8 to 12 days (primary treatment period) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2 g/dL within 24h; 2) transfusion of ≥ 2 U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding; 6) bleeding warranting treatment cessation Time points measured: 8 to 12 days (primary treatment period), 35 days (treatment), 30 days (additionally, follow‐up) Time points reported: 8 to 12 days (primary treatment period), 35 days (treatment), 30 days (additionally, follow‐up) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 8 to 12 days (primary treatment period), 35 days (treatment), 30 days (additionally, follow‐up) Time points reported: 8 to 12 days (primary treatment period) RESULTS Major VTE: Summary data (8 to 12 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 456 missing participants (90 participants excluded from the primary efficacy set were included in these data, reasons were not reported) ‐Subgroups reported: none Summary data (35 days): not reported Summary data (30‐day follow‐up): not reported Major bleeding: Summary data (8 to 12 days): darexaban 15 mg twice daily ‐ 30 mg once daily ‐ 30 mg twice daily ‐ 60 m once daily ‐ enoxaparin 40 mg once daily: 4/374 ‐ 4/383 ‐ 9/387 ‐ 7/385 ‐ 8/393 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 70 missing participants ‐Subgroups reported: none Summary data (35 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 70 missing participants ‐Subgroups reported: none Summary data (30 days follow‐up): see Analysis 1.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 70 missing participants ‐Subgroups reported: none Fatal VTE: Summary data (8 to 12 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 546 missing participants ‐Subgroups reported: none Summary data (35 days): not reported Summary data (30‐day follow‐up): not reported
Notes	Source of funding: Astellas Conflicts of interest: authors (7) had financial relationships several pharmaceutical companies, including with Astellas Published protocol: registered in clinicaltrials.gov as NCT00902928, and in ICTRP as EUCTR2008‐004416‐13‐NL Observations: none Missing data requested from authors: specific groups blinded, description of method of concealment of allocation, details about withdrawals and exclusions, follow‐up summary data for fatal and major VTE, treatment period missing participants for major VTE Missing data obtained from authors: none For the efficacy set, 27.4% of the randomised participants were excluded (28.7% in the darexaban groups and 22.5% in the enoxaparin group). Reasons for exclusion were reported. However, for major VTE, data from other set (not defined, reasons not explained) were reported. For the safety set, 3.5% of the randomised participants were excluded (3.7% in the darexaban groups and 3.0% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random sequence was centrally computer‐generated
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Low risk	Participants and clinicians/providers were blinded with the use of double‐blind labelled supplies provided by a central site
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent central adjudication committee, but it was not explicitly stated as blinded to treatment allocation. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in numbers across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 3 outcomes not included in the protocol were reported, 3 outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

Özler 2015.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: open‐label; groups blinded: none Study duration: March 2011 to July 2013 Duration of intervention: 10 days for TKR, 30 days for THR Duration of follow‐up: 32 days for TKR, 12 days for THR (6 weeks after surgery) Method of randomisation: no description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: all randomised participants Primary safety analysis set: all randomised participants Withdrawals: no participants withdrew Exclusions: no participant was excluded
Participants	Diagnosis: scheduled to TKA or THA Location of participants: 1 country (Turkey) Number of participants randomised: 180 Age (years; mean and range): (all randomised population) rivaroxaban: 65 (45 to 80); enoxaparin: 67 (40 to 87); dabigatran: 68 (49 to 82) Gender (male %): (all randomised population) rivaroxaban: 17 (28.3%); enoxaparin: 22 (36.7%); dabigatran: 23 (38.3%) Baseline imbalances: slight differences regarding gender and type of arthroplasty Inclusion criteria: Having undergone TKA or THA Body weight > 50 kg Age ≥ 18 years Exclusion criteria: Inherited or acquired clinically significant bleeding disorder Major surgery, uncontrolled hypertension, or myocardial infarction within the last 3 months History of hemorrhagic stroke GI or urogenital bleeding within the last 6 months Severe liver disease Severe renal insufficiency (creatinine clearance < 30 mL/min) Active malignant disease Platelet count < 100x109/L
Interventions	Number of intervention groups: 3 Concomitant interventions: enoxaparin 40 mg twice daily during hospital stay, initiated 12 hours after surgery Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: after discharge Duration: until completion of 10/30 days of DVT prophylaxis Enoxaparin group: Intervention: LMWH (oral) 30 mg twice daily Starting time: after surgery: after discharge Duration: until completion of 10/30 days of DVT prophylaxis Dabigatran group: Intervention: aspirin (subcutaneous) 100 mg once daily Starting time: after surgery: after discharge Duration: until completion of 10/30 days of DVT prophylaxis
Outcomes	Number of outcomes according to protocol: no protocol is available OUTCOMES Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT (diagnostic instrument not specified, probably Doppler ultrasound) Time points measured: 10 or 30 days (treatment), period to complete 6 weeks after surgery (additionally, follow‐up) Time points reported: 10 or 30 days (treatment), period to complete 6 weeks after surgery (additionally, follow‐up) Symptomatic VTE: Definition: incidence of symptomatic VTE Time points measured: 10 or 30 days (treatment), period to complete 6 weeks after surgery (additionally, follow‐up) Time points reported: 10 or 30 days (treatment), period to complete 6 weeks after surgery (additionally, follow‐up) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2g/dL within 24 hours; 2) transfusion of ≥ 4 U (approximately 800 mL) of blood; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding Time points measured: 10 or 30 days (treatment), period to complete 6 weeks after surgery (additionally, follow‐up) Time points reported: 10 or 30 days (treatment) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 10 or 30 days (treatment), period to complete 6 weeks after surgery (additionally, follow‐up) Time points reported: 10 or 30 days (treatment), period to complete 6 weeks after surgery (additionally, follow‐up) Asymptomatic distal DVT: Definition: incidence of asymptomatic distal DVT Time points measured: 10 or 30 days (treatment), period to complete 6 weeks after surgery (additionally, follow‐up) Time points reported: 10 or 30 days (treatment), period to complete 6 weeks after surgery (additionally, follow‐up) RESULTS Major VTE: Summary data (10 or 30 days): rivaroxaban ‐ enoxaparin ‐ dabigatran: 0/60 ‐ 0/60 ‐ 0/60 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (period to complete 6 weeks after surgery): see Analysis 1.2 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Symptomatic VTE: Summary data (10 or 30 days): rivaroxaban ‐ enoxaparin ‐ dabigatran: 0/60 ‐ 0/60 ‐ 0/60 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (period to complete 6 weeks after surgery): see Analysis 1.3 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Major bleeding: Summary data (10 or 30 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (period to complete 6 weeks after surgery): not reported Fatal VTE: Summary data (10 or 30 days): rivaroxaban ‐ enoxaparin ‐ dabigatran: 0/60 ‐ 0/60 ‐ 0/60 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (period to complete 6 weeks after surgery): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITTanalysis, no missing participants ‐Subgroups reported: none Asymptomatic distal DVT: Summary data (10 or 30 days): rivaroxaban ‐ enoxaparin ‐ dabigatran: 0/60 ‐ 0/60 ‐ 0/60 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (period to complete 6 weeks after surgery): see Analysis 1.10 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none
Notes	Source of funding: not stated Conflicts of interest: none Published protocol: none Observations: although stated to be randomised, the description of the method of participant selection gives the impression that distribution was sequential rather than random. Missing data requested from authors: specific groups blinded, description of method of randomisation and concealment of allocation, details about withdrawals, follow‐up summary data for major bleeding, source of funding, protocol of the study Missing data obtained from authors: none For the efficacy and analysis set, all randomised participants were included. No CONSORT flow diagram.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method was not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	High risk	Study intervention was open‐label for the rivaroxaban versus enoxaparin comparison
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of outcome assessment was not mentioned. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	Low risk	No exclusion or withdrawal was reported for any study group.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	No exclusion or withdrawal was reported for any study group.
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	High risk	There is at least one important risk of bias. The study did not report that conflicts of interest were registered. Also, there may be a risk of bias for an inappropriate co‐intervention, but there is either insufficient information to assess whether an important risk of bias exists, or insufficient rationale or evidence that an identified problem would introduce bias.

Rahman 2020.

Study characteristics
Methods	Design: randomised controlled trial Group: parallel group Blinding: open‐label Exclusions: no participant was excluded (after randomisation) Method of concealment of allocation: none (the surgeon was informed of allocation by telephone) Method of randomisation: 1:1 permuted block randomisation Primary efficacy analysis set: all treated participants Primary safety analysis set: all treated participants Study duration: January 2019 to January 2020 Withdrawals: none
Participants	Inclusion criteria: aged < 70 years undergoing primary THA for the management of primary or secondary hip osteoarthritis Age: 41.53; range (19 to 63 years) Rivaroxaban group: 42.95 ± 10.6 years. Male: 36 (45.0%) LMWH group: 40.1 ± 14.7 years. Male: 44 (55.0%) Exclusion criteria: history of VTE; receiving anticoagulation for any reason before the surgery; decompensated hepatic or renal disease Group differences: none Diagnosis: scheduled to THA, primary Location of participants: single centre (Egypt) Number of participants randomised: 160
Interventions	Rivaroxaban group: Intervention: 10 mg once daily Starting time: 12 hours after surgery Duration: 14 days Concomitant interventions: none Excluded interventions: none LMWH group: Intervention: 40 IU once daily Starting time: 12 hours after surgery Duration: 14 days Concomitant interventions: none Excluded interventions: none
Outcomes	All‐cause mortality Major bleeding Serious adverse events: hepatic Fatal VTE (computed tomography venography) Fatal bleeding
Notes	The authors stated they did not receive any financial support for the research, authorship, and/or publication of the article. The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of the article. Missing data requested from authors: none Missing data obtained from authors: none None of the participants were lost to follow‐up. No CONSORT flow diagram
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A randomisation technique was used. An allocation ratio of 1:1 by using a permuted block randomisation technique. The randomisation process was performed by a research assistant who kept all the data in a secure computer. The intervention group assignment of the participants was performed by a research assistant, who informed the surgeons of the group assignments by phone after the end of the procedure.
Allocation concealment (selection bias)	High risk	Allocation concealment was not performed
Blinding of participants and personnel (performance bias)	High risk	The study was open‐label for both groups
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of outcome assessment was not mentioned. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	Low risk	No exclusion or withdrawal was reported for any study group
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	No exclusion or withdrawal was reported for any study group
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	Low risk	The study appears to be free of other sources of bias.

Raskob 2010.

Study characteristics
Methods	Aim: dose‐response study, efficacy and safety Design: parallel‐group Blinding: double‐blind, double‐dummy; groups blinded: participants, outcome adjudicators Study duration: not reported Duration of intervention: 7 to 10 days Duration of follow‐up: 30 to 60 days Method of randomisation: in a 1:1:1:1:1 ratio, no additional description was provided Method of concealment of allocation: use of an interactive voice recognition system for allocation of participants Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who: 1) received at least one dose of study medication, 2) had data on safety outcome Withdrawals: not reported Exclusions: edoxaban 15 mg once daily: 23 (1 did not receive study drug, 8 had no venography, 14 had uninterpretable venography); 30 mg once daily: 20 (1 did not receive study drug, 6 had no venography, 13 had uninterpretable venography); 60 mg once daily: 30 (2 did not receive study drug, 12 had no venography, 16 had uninterpretable venography); 90 mg once daily: 26 (14 had no venography, 12 had uninterpretable venography); dalteparin: 31 (3 did not receive study drug, 10 had no venography, 18 had uninterpretable venography)
Participants	Diagnosis: scheduled to THR, elective, primary, unilateral Location of participants: 7 countries (USA, Canada, Russia, Ukraine, Denmark, Latvia, UK) Number of participants randomised: 903 Age (years; mean ± SD): (all randomised population) edoxaban 15 mg once daily / 30 mg once daily / 60 mg once daily / 90 mg once daily 57.6±12.4 / 57.3±12.5 / 57.3±12.3 / 58.3±11.6; dalteparin 58.5±12.3 Gender (male %): (all randomised population) edoxaban 15 mg once daily / 30 mg once daily / 60 mg once daily / 90 mg once daily 107 (61.1%) / 113 (58.5%) / 111 (64.9%) / 119 (63.6%); dalteparin 93 (52.5%) Baseline imbalances: slight differences regarding gender among the groups Inclusion criteria: 18 years of age or older undergoing elective primary unilateral THR Exclusion criteria: Known or suspected bleeding or coagulation disorder Haemorrhagic stroke or conditions predisposing to intracranial haemorrhage such as previous intraocular haemorrhage, non‐haemorrhagic stroke within the past three months, or intracranial malignancy GI bleeding or a documented peptic ulcer within the past 3 months Abnormal preoperative baseline prothrombin time or activated partial thromboplastin time Uncontrolled hypertension (systolic > 180 mmHg and/or diastolic > 110 mm Hg) Anaemia (haemoglobin < 9 g/dL) or thrombocytopenia (platelets < 100 x 109/L) Receiving a VKA or having received therapeutic doses of an anticoagulant or fibrinolytic agent within 10 days prior to the study Requiring ongoing treatment with aspirin (> 100 mg/day), clopidogrel, ticlopidine, dipyridamole, or non‐steroidal anti‐inflammatory drugs Hypersensitivity to dalteparin, heparin, or pork products Contraindication to venography such as a serum creatinine > 132 μM, a history of allergy to iodine or radiopaque dye, or inadequate venous access for the procedure Having elevations of ALT or AST ≥ 1.5 times the upper limit of the reference range, bilirubin above the upper limit of the reference range, or gamma glutamyltransferase (GGT) ≥ 2.0 times the upper limit of the reference range Positive hepatitis A, B, or C serology Known drug or alcohol dependence within the past 12 months Estimated survival less than 12 months or having received an investigational drug in the previous month or having been previously enroled in a study of edoxaban Judged by the investigator to be unable to complete the study requirements or comply with the protocol Unable or unwilling to provide written informed consent Women who were pregnant or lactating
Interventions	Number of intervention groups: 5 Concomitant interventions: none Excluded interventions: none Edoxaban 15 mg once daily group: Intervention: edoxaban (oral) 15 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 7 to 10 days Edoxaban 30 mg once daily group: Intervention: edoxaban (oral) 30 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 7 to 10 days Edoxaban 60 mg once daily group: Intervention: edoxaban (oral) 60 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 7 to 10 days Edoxaban 90 mg once daily group: Intervention: edoxaban (oral) 90 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 7 to 10 days Dalteparin group: Intervention: dalteparin (subcutaneous) 5000 IU once daily Starting time: after surgery: 6 to 8 hours Duration: 7 to 10 days
Outcomes	Number of outcomes: According to protocol: 4 Reported: 5. Reported additionally: all‐cause mortality, major VTE, major bleeding, serious adverse events (non‐hepatic), liver enzymes elevation (treatment period) Not reported: VTE, change in PT (prothrombin time), change in aPTT (activated partial thromboplastin time), major and CRNM bleeding (treatment period) OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 7 to 10 days (treatment), 30 to 60 days (additionally, follow‐up) Time points reported: 7 to 10 days (treatment) Major VTE: Definition: incidence of fatal and nonfatal PE or VTE (bilateral venography) Time points measured: 7 to 10 days (treatment), 30 to 60 days (additionally, follow‐up) Time points reported: 7 to 10 days (treatment) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2g/dL within 24 hours; 2) transfusion of ≥ 2U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding; 6) bleeding warranting treatment cessation Time points measured: 10 days (treatment), 30 to 60 days (additionally, follow‐up) Time points reported: 10 days (treatment) Serious AEs: non‐hepatic: Definition: number of patients with at least one serious AE (definition not reported), excluding those liver‐related Time points measured: 10 days (treatment), 30 to 60 days (additionally, follow‐up) Time points reported: 10 days (treatment) Liver enzymes elevation: Definition: incidence of ALT/AST > 3 times ULN on same date Time points measured: 10 days (treatment), 30 to 60 days (additionally, follow‐up) Time points reported: 10 days (treatment) RESULTS All‐cause mortality: Summary data (7 to 10 days): see Analysis 9.1 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 7 missing participants ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): not reported Major VTE: Summary data (7 to 10 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 129 missing participants ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): not reported Major bleeding: Summary data (10 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 7 missing participants ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): not reported Serious AEs: non‐hepatic Summary data (10 days): see Analysis 9.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 7 missing participants ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): not reported Liver enzymes elevation: Summary data (10 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 7 missing participants ‐Subgroups reported: none Summary data (30 to 60 days' follow‐up): not reported
Notes	Source of funding: Daiichi Sankyo Pharma Development Conflicts of interest: authors had financial relationships with several pharmaceutical companies, including Daiichi Sankyo Published protocol: registered in clinicaltrials.gov as NCT00398216, and in ICTRP as EUCTR2006‐000758‐29‐DK Observations: none Missing data requested from authors: specific groups blinded, date of start and end of study, description of method of randomisation, details about withdrawals, definition of serious AEs, follow‐up summary data for all‐cause mortality, major VTE, major bleeding, serious AEs and liver enzymes elevation Missing data obtained from authors: none For the efficacy set, 14.4% of the randomised participants were excluded (13.6% in the darexaban groups and 17.7% in the enoxaparin group). Reasons for exclusion were reported. However, for all‐cause mortality, data from safety set were reported. For the safety set, 0.8% of the randomised participants were excluded (0.5% in the darexaban groups and 1.7% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method was not described
Allocation concealment (selection bias)	Low risk	Participants were allocated with an interactive voice recognition system
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded central adjudication committee. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data balanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 5 outcomes not included in the protocol were reported, 4 outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

RECORD1 2008.

Study characteristics
Methods	Aim: non‐inferiority & superiority, efficacy and safety Design: parallel‐group (phase III) Blinding: double‐blind, double‐dummy; groups blinded: participants, outcome adjudicators Study duration: February 2006 to March 2007 Duration of intervention: 31 to 39 days Duration of follow‐up: 30 to 35 days Method of randomisation: computer‐generated random sequence, with permuted blocks; stratified by study site Method of concealment of allocation: use of a central telephone system for allocation of participants Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) underwent planned surgery, 3) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: not reported Exclusions: rivaroxaban group (671): 57 (2.5%) did not receive study drug, 17 (0.8%) did not undergo planned surgery, 1 (< 0.1%) received wrong study drug, 588 (25.9%) had inadequate venography, 8 (0.4%) had source data not verified; enoxaparin group (717): 51 (2.2%) did not receive study drug, 21 (0.9%) did not undergo planned surgery, 2 (< 0.1%) received wrong study drug, 635 (27.9%) had inadequate venography, 8 (0.4%) had source data not verified
Participants	Diagnosis: scheduled to THA, elective Location of participants: 27 countries (Argentina, Australia, Austria, Belgium, Brazil, Canada, Chile, Colombia, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Israel, Italy, Lithuania, the Netherlands, Norway, Poland, Slovakia, South Africa, Spain, Sweden, Turkey, USA) Number of participants randomised: 4541 Age (years; mean and range): (safety analysis population) rivaroxaban 63.1 (18 to 91); enoxaparin 63.3 (18 to 93) Gender (male %): (safety analysis population) rivaroxaban 989 (44.8%); enoxaparin 982 (44.2%) Baseline imbalances: none Inclusion criteria: Men and women of at least 18 years of age who were scheduled to undergo elective THA Exclusion criteria: Scheduled to undergo staged, bilateral hip arthroplasty Pregnant or breastfeeding Active bleeding or a high risk of bleeding Contraindication for prophylaxis with enoxaparin or a condition that might require an adjusted dose of enoxaparin Conditions preventing bilateral venography Substantial liver disease Severe renal impairment (creatinine clearance, < 30 mL per minute) Concomitant use of protease inhibitors for the treatment of human immunodeficiency virus infection Planned IPC Requirement for anticoagulant therapy that could not be stopped
Interventions	Number of intervention groups: 2 Concomitant interventions: none Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 35 ± 4 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: 12 hours Duration: 35 ± 4 days
Outcomes	Number of outcomes: According to protocol: 7 Reported: 16. Reported additionally: all‐cause mortality (treatment and follow‐up periods), fatal VTE, PE, fatal bleeding, CRNM bleeding, serious adverse events (non‐hepatic), any adverse event, liver enzymes elevation, volume of blood loss, volume of blood transfusion, post‐operative wound infection (treatment period) Not reported: VTE and all‐cause mortality, DVT, major VTE and major bleeding, major VTE and all‐cause mortality OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 30 to 42 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 30 to 42 days (treatment), 30 to 35 days (additionally, follow‐up) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT (bilateral venography) Time points measured: 30 to 42 days (treatment) Time points reported: 30 to 42 days (treatment) Symptomatic VTE: Definition: incidence of symptomatic VTE Time points measured: 30 to 42 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 30 to 42 days (treatment), 30 to 35 days (additionally, follow‐up) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2g/dL within 24 hours; 2) transfusion of ≥ 2 U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding Time points measured: 33 to 41 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 33 to 41 days (treatment, 2 days after the last study drug intake) Serious AEs: non‐hepatic: Definition: number of participants with at least one serious AE (definition not reported), excluding those liver‐related Time points measured: 33 to 41 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 33 to 41 days (treatment, 2 days after the last study drug intake) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 30 to 42 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 30 to 42 days (treatment) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 33 to 41 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 33 to 41 days (treatment, 2 days after the last study drug intake) Liver enzymes elevation: Definition: incidence of ALT >3 times ULN Time points measured: 33 to 41 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 33 to 41 days (treatment, 2 days after the last study drug intake) RESULTS All‐cause mortality: Summary data (30 to 42 days): see Analysis 9.1 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 1388 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 1388 missing participants ‐Subgroups reported: none Major VTE: Summary data (30 to 42 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: not specified, not specified, 1177 missing participants ‐Subgroups reported: none Symptomatic VTE: Summary data (30 to 42 days): see Analysis 9.3 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 142 missing participants (34 did not undergo planned surgery) ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): see Analysis 1.3 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 142 missing participants (34 did not undergo planned surgery) ‐Subgroups reported: none Major bleeding: Summary data (33 to 41 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 108 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Serious AEs: non‐hepatic Summary data (33 to 41 days): see Analysis 9.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 108 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Fatal VTE: Summary data (30 to 42 days): rivaroxaban 2/NR; enoxaparin 1/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Fatal bleeding: Summary data (33 to 41 days): see Analysis 9.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 108 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Liver enzymes elevation: Summary data (33 to 41 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 284 missing participants (176 unexplained) ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported
Notes	Source of funding: Bayer HealthCare, Johnson & Johnson Conflicts of interest: all authors had financial relationships with several pharmaceutical companies, including Bayer HealthCare Published protocol: registered in clinicaltrials.gov as NCT00329628, in anzctr.org.au as ACTRN12606000068561, and in ICTRP as EUCTR2005‐004351‐35‐LT Observations: during follow‐up period, participants received additional prophylaxis at the discretion of the investigator Missing data requested from authors: specific groups blinded, details about withdrawals, definition of serious AEs, time points for measurement of volume of blood loss, treatment period summary data for volume of blood loss, follow‐up summary data for fatal bleeding, major bleeding, serious AEs and liver enzymes elevation, treatment period number of participants for fatal VTE, treatment period missing participants for liver enzymes elevation Missing data obtained from authors: none For the efficacy set, 30.6% of the randomised participants were excluded (29.6% in the rivaroxaban group and 31.5% in the enoxaparin group). Reasons for exclusion were reported. Of note, for symptomatic VTE, data from other set (3.1% of randomised participants excluded, reasons explained) were reported. For the safety set, 2.4% of the randomised participants were excluded (2.5% in the rivaroxaban group and 2.2% in the enoxaparin group). Reasons for exclusion were reported. However, for liver enzymes elevation and volume of blood loss, data from other sets (6.3% and 18.9% of randomised participants excluded, reasons not fully explained) were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random sequence was computer‐generated, with permuted blocks
Allocation concealment (selection bias)	Low risk	Participants were allocated using a central telephone system
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded central adjudication committee. Data analysis was performed by the sponsor, and blinding was not stated
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Reasons for missing outcome data related to true outcome. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 11 outcomes not included in the protocol were reported, four outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

RECORD2 2008.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group (phase III) Blinding: double‐blind, double‐dummy; groups blinded: participants, outcome adjudicators Study duration: February 2006 to April 2007 Duration of intervention: 31 to 39 days Duration of follow‐up: 30 to 35 days Method of randomisation: computer‐generated random sequence, with permuted blocks; stratified by study site Method of concealment of allocation: use of a central telephone system for allocation of participants Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) underwent planned surgery, 3) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: not reported Exclusions: rivaroxaban group (388): 24 (1.9%) did not receive study drug, 16 (1.3%) did not undergo the planned surgery, 348 (27.8%) had uninterpretable venography (155 (12.4%) not done, 57 (4.6%) unilateral, 127 (10.1%) indeterminate/unevaluable, 9 (0.7%) not in time window); enoxaparin group (388): 28 (2.2%) did not receive study drug, 22 (1.8%) did not undergo the planned surgery, 338 (26.9%) had uninterpretable venography (159 (12.6%) not done, 57 (4.5%) unilateral, 111 (8.8%) indeterminate/unevaluable, 11 (0.9%) not in time window)
Participants	Diagnosis: scheduled to THA, elective Location of participants: 21 countries (Australia, Brazil, Canada, China, Colombia, Denmark, Estonia, India, Indonesia, Latvia, Lithuania, Mexico, New Zealand, Norway, Peru, Portugal, South Africa, South Korea, Sweden, UK, USA) Number of participants randomised: 2509 Age (years; mean and range): (safety analysis population) rivaroxaban 61.4 (18‐93); enoxaparin 61.6 (19‐93) Gender (male %): (safety analysis population) rivaroxaban 561 (45.7%); enoxaparin 578 (47%) Baseline imbalances: none Inclusion criteria: Aged 18 years or over, and scheduled to undergo elective THA Exclusion criteria: Scheduled to undergo staged bilateral hip arthroplasty Active bleeding or a high risk of bleeding Any condition contraindicating the use of enoxaparin or that might require enoxaparin dose adjustment, including severe renal impairment Significant liver disease Pregnancy or breastfeeding Concomitant use of HIV protease inhibitors, use of fibrinolytic therapy or planned IPC during the study period Conditions preventing bilateral venography, or the requirement for an anticoagulant that could not be discontinued
Interventions	Number of intervention groups: 2 Concomitant interventions: none Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 35 ± 4 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: 12 hours Duration: 10 to 14 days
Outcomes	Number of outcomes: According to protocol: 7 Reported: 16. Reported additionally: all‐cause mortality (treatment and follow‐up periods), fatal VTE, PE, fatal bleeding, CRNM bleeding, any bleeding, serious adverse events (non‐hepatic), any adverse event, liver enzymes elevation, volume of blood loss, volume of blood transfusions, post‐operative wound infection (treatment period) Not reported: DVT, major VTE and major bleeding, major VTE and all‐cause death (treatment period) OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 30 to 42 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 30 to 42 days (treatment), 30 to 35 days (additionally, follow‐up) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT (bilateral venography) Time points measured: 30 to 42 days (treatment) Time points reported: 30 to 42 days (treatment) Symptomatic VTE: Definition: incidence of symptomatic VTE Time points measured: 30 to 42 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 30 to 42 days (treatment), 30 to 35 days (additionally, follow‐up) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2g/dL within 24 hours; 2) transfusion of ≥ 2 U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding Time points measured: 33 to 41 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 33 to 41 days (treatment, 2 days after the last study drug intake) Serious AEs: non‐hepatic: Definition: number of participants with at least one serious AE (definition not reported), excluding those liver‐related Time points measured: 33 to 41 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 33 to 41 days (treatment, 2 days after the last study drug intake) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 30 to 42 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 30 to 42 days (treatment), 30 to 35 days (additionally, follow‐up) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 33 to 41 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 33 to 41 days (treatment, 2 days after the last study drug intake) Liver enzymes elevation: Definition: incidence of ALT > 3 times ULN Time points measured: 31 to 39 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 31 to 39 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) RESULTS All‐cause mortality: Summary data (30 to 42 days): see Analysis 9.1 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 776 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 52 missing participants ‐Subgroups reported: none Major VTE: Summary data (30 to 42 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: not specified, not specified, 586 missing participants ‐Subgroups reported: none Symptomatic VTE: Summary data (30 to 42 days): see Analysis 9.3 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 90 missing participants (39 did not undergo planned surgery) ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): see Analysis 1.3 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 90 missing participants (38 did not undergo planned surgery) ‐Subgroups reported: none Major bleeding: Summary data (33 to 41 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 52 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Serious AEs: non‐hepatic Summary data (33 to 41 days): see Analysis 9.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 52 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Fatal VTE: Summary data (30 to 42 days): rivaroxaban 0/NR; enoxaparin 1/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): rivaroxaban 0/NR; enoxaparin 1/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Fatal bleeding: Summary data (33 to 41 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 52 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Liver enzymes elevation: Summary data (31 to 39 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 178 missing participants (126 unexplained) ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): see Analysis 1.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 311 missing participants (259 unexplained) ‐Subgroups reported: none
Notes	Source of funding: Bayer HealthCare AG and Johnson & Johnson Pharmaceutical Research and Development LLC Conflicts of interest: all authors had financial relationships with several pharmaceutical companies, including Bayer HealthCare Published protocol: registered in clinicaltrials.gov as NCT00332020, and in ICTRP as EUCTR2005‐004691‐20‐PT Observations: during follow‐up period, participants received additional prophylaxis at the discretion of the investigator Missing data requested from authors: specific groups blinded, details about withdrawals, definition of serious AEs, time points for measurement of volume of blood loss, treatment period summary data for volume of blood loss, follow‐up summary data for fatal bleeding, major bleeding, serious AEs, treatment period and follow‐up number of participants for fatal VTE, treatment period and follow‐up missing participants for liver enzymes elevation Missing data obtained from authors: none For the efficacy set, 30.9% of the randomised participants were excluded (31.0% in the rivaroxaban group and 30.9% in the enoxaparin group). Reasons for exclusion were reported. Of note, for symptomatic VTE, data from other set (3.6% of randomised participants excluded, reasons explained) were reported. For the safety set, 2.1% of the randomised participants were excluded (1.9% in the rivaroxaban group and 2.2% in the enoxaparin group). Reasons for exclusion were reported. However, for liver enzymes elevation and volume of blood loss, data from other sets (7.0% and 37.0% of randomised participants excluded, reasons not fully explained) were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random sequence was computer‐generated, with permuted blocks
Allocation concealment (selection bias)	Low risk	Participants were allocated with a central telephone system
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded central adjudication committee. Data analysis was performed by the sponsor, and blinding was not stated
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in numbers across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 12 outcomes not included in the protocol were reported, 3 outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least three important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had two potential sources of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects; and inappropriate intervention as the duration of prophylaxis markedly differed for each intervention group and the outcome analyses were at different time points; 5 weeks versus 2 weeks). The intervention could be favoured in the design of the study, since follow‐up was longer than 2 weeks and the analyses of the outcomes were at different times.

RECORD3 2008.

Study characteristics
Methods	Aim: non‐inferiority & superiority, efficacy and safety Design: parallel‐group (phase III) Blinding: double‐blind, double‐dummy; groups blinded: participants, outcome adjudicators Study duration: February 2006 to January 2007 Duration of intervention: 10 to 14 days Duration of follow‐up: 30 to 35 days Method of randomisation: no description was provided Method of concealment of allocation: use of a central telephone system for allocation of participants Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) underwent planned surgery, 3) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: not reported Exclusions: rivaroxaban group (430): 34 (2.7%) did not receive study drug, 19 (1.5%) did not undergo the planned surgery, 376 (30%) had uninterpretable venography; enoxaparin group (399): 38 (3%) did not receive study drug, 22 (1.7%) did not undergo the planned surgery, 339 (26.5%) had uninterpretable venography
Participants	Diagnosis: scheduled to TKA Location of participants: 19 countries (Austria, Belgium, Canada, China, Colombia, Czech Republic, Denmark, France, Germany, Israel, Italy, Mexico, the Netherlands, Norway, Peru, Poland, South Africa, Spain, Sweden) Number of participants randomised: 2531 Age (years; mean and range): (safety analysis population) rivaroxaban 67.6 (28 to 91); enoxaparin 67.6 (30 to 90) Gender (male %): (safety analysis population) rivaroxaban 363 (29.8%); enoxaparin 418 (33.7%) Baseline imbalances: the proportion of female participants was different between the groups (P = 0.03) Inclusion criteria: 18 years of age or older and scheduled for TKA Exclusion criteria: Active bleeding or a high risk of bleeding that contraindicated the use of LMWH Any contraindication to the use of enoxaparin or any condition necessitating adjustment of its dose Conditions preventing bilateral venography Clinically significant liver disease Concomitant use of protease inhibitors of the human immunodeficiency virus or fibrinolytic agents Planned IPC Requirement of ongoing anticoagulant therapy Pregnancy or breast‐feeding
Interventions	Number of intervention groups: 2 Concomitant interventions: none Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 10 to 14 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: 12 hours Duration: 10 to 14 days
Outcomes	Number of outcomes: According to protocol: 7 Reported: 16. Reported additionally: all‐cause mortality (treatment and follow‐up periods), fatal VTE, PE, fatal bleeding, CRNM bleeding, any bleeding, serious adverse events (non‐hepatic), any adverse event, liver enzymes elevation, volume of blood loss, volume of blood transfusions, post‐operative wound infection (treatment period) Not reported: DVT, major VTE and major bleeding, major VTE and all‐cause death (treatment period) OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 13 to 17 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 13 to 17 days (treatment), 30 to 35 days (additionally, follow‐up) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT (bilateral venography) Time points measured: 13 to 17 days (treatment) Time points reported: 13 to 17 days (treatment) Symptomatic VTE: Definition: incidence of symptomatic VTE Time points measured: 13 to 17 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 13 to 17 days (treatment), 30 to 35 days (additionally, follow‐up) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2 g/dL within 24 hours; 2) transfusion of ≥ 2 U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake) Serious AEs: non‐hepatic: Definition: number of participants with at least one serious AE (definition not reported), excluding those liver‐related Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake) Liver enzymes elevation: Definition: incidence of ALT > 3 times ULN Time points measured: 10 to 14 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) RESULTS All‐cause mortality: Summary data (13 to 17 days): see Analysis 9.1 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 829 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 113 missing participants (41 did not undergo planned surgery) ‐Subgroups reported: none Major VTE: Summary data (13 to 17 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: not specified, not specified, 698 missing participants ‐Subgroups reported: none Symptomatic VTE: Summary data (13 to 17 days): see Analysis 9.3 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 113 missing participants (41 did not undergo planned surgery) ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): see Analysis 1.3 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 113 missing participants (41 did not undergo planned surgery) ‐Subgroups reported: none Major bleeding: Summary data (12 to 16 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 72 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Serious AEs: non‐hepatic Summary data (12 to 16 days): see Analysis 9.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 72 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Fatal bleeding: Summary data (12 to 16 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 72 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Liver enzymes elevation: Summary data (10 to 14 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 225 missing participants (153 unexplained) ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported
Notes	Source of funding: Bayer HealthCare AG and Johnson & Johnson Pharmaceutical Research and Development LLC Conflicts of interest: all authors had financial relationships with several pharmaceutical companies, including Bayer HealthCare Published protocol: registered in clinicaltrials.gov as NCT00361894, and in ICTRP as EUCTR2005‐004620‐40‐IT Observations: during follow‐up period, participants received additional prophylaxis at the discretion of the investigator Missing data requested from authors: specific groups blinded, description of method of randomisation, details about withdrawals, definition of serious AEs, time points for measurement of volume of blood loss, treatment period summary data for volume of blood loss, follow‐up summary data for fatal bleeding, major bleeding, serious AEs and liver enzymes elevation, treatment period and follow‐up number of participants for fatal VTE, treatment period missing participants for liver enzymes elevation Missing data obtained from authors: none For the efficacy set, 32.8% of the randomised participants were excluded (34.3% in the rivaroxaban group and 31.2% in the enoxaparin group). Reasons for exclusion were reported. Of note, for symptomatic VTE, data from other set (4.5% of randomised participants excluded, reasons explained) were reported. For the safety set, 2.8% of the randomised participants were excluded (2.7% in the rivaroxaban group and 3.0% in the enoxaparin group). Reasons for exclusion were reported. However, for liver enzymes elevation and volume of blood loss, data from other sets (8.9% and 17.3% of randomised participants excluded, reasons not fully explained) were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were randomly assigned to study drug through a central telephone system
Allocation concealment (selection bias)	Low risk	Participants were allocated with a central telephone system
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded central adjudication committee. Data analysis was performed by the sponsor, and blinding was not stated
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 12 outcomes not included in the protocol were reported, 3 outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

RECORD4 2009.

Study characteristics
Methods	Aim: non‐inferiority & superiority, efficacy and safety Design: parallel‐group (phase III) Blinding: double‐blind, double‐dummy; groups blinded: participants, outcome adjudicators Study duration: June 2006 to October 2007 Duration of intervention: 10 to 15 days Duration of follow‐up: 30 to 35 days Method of randomisation: with permuted blocks (size of 4), stratified by study site, no additional description was provided Method of concealment of allocation: use of a central telephone system for allocation of participants Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) underwent planned surgery, 3) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: not reported Exclusions: rivaroxaban group (619): 58 (3.7%) did not receive study drug, 2 (0.1%) did not undergo the planned surgery, 559 (35.3%) had uninterpretable venography (189 (33.8%) not done, 116 (20.8%) unilateral, 244 (43.6%) indeterminate/unevaluable, 10 (1.8%) not in time window); enoxaparin group (605): 56 (3.6%) did not receive study drug, 3 (0.2%) did not undergo the planned surgery, 546 (34.9%) had uninterpretable venography (184 (33.7%) not done, 105 (19.2%) unilateral, 253 (46.3%) indeterminate/unevaluable, 4 (0.7%) not in time window)
Participants	Diagnosis: scheduled to TKA Location of participants: 12 countries (Bulgaria, Canada, Denmark, India, Israel, Lithuania, Mexico, Pakistan, Poland, Sri Lanka, Sweden, USA) Number of participants randomised: 3148 Age (years; mean ± SD): (safety analysis population) rivaroxaban 64.4 ± 9.7; enoxaparin 64.7 ± 9.7 Gender (male %): (safety analysis population) rivaroxaban 519 (34%); enoxaparin 541 (35.9%) Baseline imbalances: none Inclusion criteria: Aged 18 years or older and scheduled for TKA Exclusion criteria: Active bleeding or a high risk of bleeding Any disorder contraindicating the use of enoxaparin or that might necessitate enoxaparin dose adjustment Disorders preventing bilateral venography Clinically significant liver disease Severe renal impairment (creatinine clearance < 30 mL per min) Concomitant use of drugs that strongly inhibit cytochrome P450, such as protease inhibitors or ketoconazole Pregnancy or breastfeeding Planned IPC Requirement for ongoing anticoagulant therapy
Interventions	Number of intervention groups: 2 Concomitant interventions: none Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 11 to 15 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 30 mg twice daily Starting time: after surgery: 12 to 24 hours Duration: 10 to 14 days
Outcomes	Number of outcomes: According to protocol: 8 Reported: 12. Reported additionally: all‐cause mortality (treatment and follow‐up periods), asymptomatic distal DVT, CRNM bleeding, any bleeding, serious adverse events, post‐operative wound infection, liver enzymes elevation, volume of blood loss, volume of blood transfusions (treatment period) Not reported: major VTE and all‐cause death, any VTE, any DVT, major VTE and major bleeding (treatment period) OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 11 to 17 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 11 to 17 days (treatment), 30 to 35 days (additionally, follow‐up) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT (bilateral venography) Time points measured: 11 to 17 days (treatment) Time points reported: 11 to 17 days (treatment) Symptomatic VTE: Definition: incidence of symptomatic VTE Time points measured: 11 to 17 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 11 to 17 days (treatment) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2 g/dL within 24 hours; 2) transfusion of ≥ 2 U of packed RBC; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding Time points measured: 12 to 17 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 12 to 17 days (treatment, 2 days after the last study drug intake) Serious AEs: non‐hepatic: Definition: number of participants with at least one serious AE (definition not reported), excluding those liver‐related Time points measured: 12 to 17 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 12 to 17 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 11 to 17 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 11 to 17 days (treatment), 30 to 35 days (additionally, follow‐up) Asymptomatic distal DVT: Definition: incidence of asymptomatic distal DVT Time points measured: 11 to 15 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 11 to 15 days (treatment) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 12 to 17 days (treatment, 2 days after the last study drug intake), 30 to 35 days (additionally, follow‐up) Time points reported: 12 to 17 days (treatment, 2 days after the last study drug intake) Liver enzymes elevation: Definition: incidence of ALT > 3 times ULN Time points measured: 10 to 15 days (treatment), 30 to 35 days (additionally, follow‐up) Time points reported: 10 to 15 days (treatment) Volume of blood loss: Definition: volume in postoperative drain Method/unit of measurement: mL Time points measured: not reported Time points reported: not reported RESULTS All‐cause mortality: Summary data (11 to 17 days): see Analysis 9.1 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 114 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 114 missing participants ‐Subgroups reported: none Major VTE: Summary data (11 to 17 days): see Analysis 9.2 ‐Analysis set, type of analysis and missing participants: not specified, not specified, 698 missing participants ‐Subgroups reported: none Symptomatic VTE: Summary data (11 to 17 days): see Analysis 9.3 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 114 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Major bleeding: Summary data (11 to 17 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 114 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Serious AEs: non‐hepatic Summary data (11 to 17 days): see Analysis 9.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 114 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): see Analysis 1.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 114 missing participants ‐Subgroups reported: none Fatal VTE: Summary data (11 to 17 days): rivaroxaban 1/NR; enoxaparin 0/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): rivaroxaban 0/NR; enoxaparin 0/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Asymptomatic distal DVT: Summary data (11 to 15 days): see Analysis 9.10 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 1224 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Fatal bleeding: Summary data (11 to 17 days): see Analysis 9.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 114 missing participants ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Liver enzymes elevation: Summary data (11 to 17 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 226 missing participants (112 unexplained) ‐Subgroups reported: none Summary data (30 to 35 days' follow‐up): not reported Volume of blood loss: Summary data (time point not reported) (mean ± SD): see Analysis 1.14 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 1123 missing participants (those for whom data were not available) ‐Subgroups reported: none
Notes	Source of funding: Bayer Schering Pharma AG and Johnson & Johnson Pharmaceutical Research and Development Conflicts of interest: all authors had financial relationships with several pharmaceutical companies, including Bayer Schering Pharma AG Published protocol: registered in clinicaltrials.gov as NCT00362232, and in ICTRP as EUCTR2006‐002402‐60‐SE Observations: during follow‐up period, participants received additional prophylaxis at the discretion of the investigator; safety events were considered only after first study drug intake Missing data requested from authors: specific groups blinded, description of method of randomisation, details about withdrawals, definition of serious AEs, time points for measurement of volume of blood loss, treatment period summary data for volume of blood loss, follow‐up summary data for symptomatic VTE, asymptomatic distal DVT, fatal bleeding, major bleeding, serious AEs and liver enzymes elevation, treatment period and follow‐up number of participants for fatal VTE, treatment period missing participants for liver enzymes elevation Missing data obtained from authors: none Data from safety set were reported for the efficacy outcomes (except for asymptomatic distal DVT, for which 38.9% of the randomised participants were excluded). For this set, 3.6% of the randomised participants were excluded (3.6% in the rivaroxaban group and 3.6% in the enoxaparin group). Reasons for exclusion were reported. For the safety set, 3.6% of the randomised participants were excluded (3.6% in the rivaroxaban group and 3.6% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were randomly assigned to study drug through a central telephone system, stratified by centre with permuted blocks of four participants.
Allocation concealment (selection bias)	Low risk	Participants were allocated via a central telephone system
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded central adjudication committee. Data analysis was performed by the sponsor, and blinding was not stated
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data balanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 9 outcomes not included in the protocol were reported, 4 outcomes included in the protocol were not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least three important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects). The study has been claimed to have been fraudulent.

STARS E‐3 2014.

Study characteristics
Methods	Aim: non‐inferiority & superiority, efficacy and safety Design: parallel‐group Blinding: double‐blind, double‐dummy; groups blinded: participants, clinicians/providers, outcome adjudicators Study duration: February 2009 to October 2009 Duration of intervention: 10 to 14 days Duration of follow‐up: 25 to 35 days Method of randomisation: computer‐generated random sequence (with the SAS function RANUNI), with permuted blocks (size of 4); stratified by study site Method of concealment of allocation: not mentioned Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) underwent planned surgery, 3) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: edoxaban group: 2 did not visit the study site for the follow‐up examination, enoxaparin group: 0; not specified if considered among reported post‐randomisation exclusions Exclusions: edoxaban group (61): 6 did not receive study drug, 36 had no venography, 19 had uninterpretable venography; enoxaparin group (61): 7 did not receive study drug, 37 had no venography, 17 had uninterpretable venography
Participants	Diagnosis: scheduled to TKA, unilateral, excluding revision arthroplasty Location of participants: 2 countries (Japan, Taiwan) Number of participants randomised: 716 Age (years; mean and range): (primary efficacy analysis population) rivaroxaban 72.6 (36 to 84); enoxaparin 72.1 (30 to 84) Gender (male %): (primary efficacy analysis population) rivaroxaban 54 (18.1%); enoxaparin 66 (22.4%) Baseline imbalances: none Inclusion criteria: Men and women aged 20 to 84 years scheduled to undergo unilateral TKA, excluding revision arthroplasty Exclusion criteria: Increased risk of bleeding High risk for thromboembolism Body weight < 40 kg Severe renal impairment (creatinine clearance < 30 mL/min) Hepatic dysfunction Pregnant or lactating women
Interventions	Number of intervention groups: 2 Concomitant interventions: none as part of the study; IPC and elastic stockings were allowed Excluded interventions: none Edoxaban group: Intervention: edoxaban (oral) 30 mg once daily Starting time: after surgery: 6 to 24 hours Duration: 11 to 14 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 20 mg twice daily Starting time: after surgery: 24 to 36 hours Duration: 11 to 14 days
Outcomes	Number of outcomes: According to protocol: 7 Reported: 9. Reported additionally: serious AE, liver enzymes elevation (treatment period) Not reported: none OUTCOMES Symptomatic VTE: Definition: incidence of symptomatic VTE (unilateral venography) Time points measured: 10 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2g/dL within 24 hours; 2) transfusion of ≥ 4 U (approximately 800 mL) of blood; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding Time points measured: 10 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Serious AEs: non‐hepatic: Definition: number of participants with at least one serious AE, excluding those liver‐related. Serious AE were those which: 1) resulted in death, 2) were life‐threatening, 3) required hospitalisation at the hospital or clinic or prolonged existing hospitalisation, 4) resulted in disability/incapacity, 5) were likely to result in disability/incapacity, 6) congenital anomaly/birth defect Time points measured: 10 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) Asymptomatic distal DVT: Definition: incidence of asymptomatic distal DVT (unilateral venography) Time points measured: 10 to 14 days (treatment) Time points reported: 10 to 14 days (treatment) Liver enzymes elevation: Definition: incidence of ALT ≥ 3 times ULN Time points measured: 10 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 10 to 14 days (treatment) RESULTS All‐cause mortality: Summary data (10 to 14 days): edoxaban 0/NR; enoxaparin 0/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Summary data (30‐ to 35‐day follow‐up): not reported Symptomatic VTE: Summary data (10 to 14 days): see Analysis 9.3 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 122 missing participants ‐Subgroups reported: none Summary data (30‐ to 35‐day follow‐up): not reported Major bleeding: Summary data (10 to 14 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 13 missing participants ‐Subgroups reported: none Summary data (30‐ to 35‐day follow‐up): not reported Serious AEs: non‐hepatic Summary data (10 to 14 days): see Analysis 9.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 13 missing participants ‐Subgroups reported: none Summary data (30‐ to 35‐day follow‐up): not reported Asymptomatic distal DVT: Summary data (10 to 14 days): see Analysis 9.10 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 122 missing participants ‐Subgroups reported: none Liver enzymes elevation: Summary data (10 to 14 days): edoxaban 0.6%; enoxaparin 5.7% (number of participants per group was not reported) ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, not calculable missing participants ‐Subgroups reported: none Summary data (30‐ to 35‐day follow‐up): not reported
Notes	Source of funding: Daiichi Sankyo Co., Ltd Conflicts of interest: authors (9) had financial relationships with several pharmaceutical companies, including Daiichi Sankyo Co., Ltd Published protocol: registered in clinicaltrials.gov as NCT01181102, and in ICTRP as JPRN‐JapicCTI‐090727 Observations: none Missing data requested from authors: specific groups blinded, description of method of concealment of allocation, details about withdrawals, definition of serious AEs, follow‐up summary data for all‐cause mortality, fatal VTE, symptomatic VTE, fatal bleeding, major bleeding, serious AEs and liver enzymes elevation, treatment period number of participants for all‐cause mortality, fatal VTE, fatal bleeding and liver enzymes elevation Missing data obtained from authors: information regarding blinding, withdrawals, outcome (serious adverse event) definition, and data on some outcomes For the efficacy set, 17.2% of the randomised participants were excluded (16.9% in the edoxaban group and 17.1% in the enoxaparin group). Reasons for exclusion were reported. For the safety set, 1.8% of the randomised participants were excluded (1.7% in the edoxaban group and 2.0% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random sequence was computer‐generated (with the SAS function RANUNI), with permuted blocks
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded events adjudication committee (not stated for liver enzymes elevation). Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	Low risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data balanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was not large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 2 outcomes not included in the protocol were reported, no outcome included in the protocol was not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

STARS J‐2 2014.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group (phase II) Blinding: open‐label (for the darexaban versus enoxaparin comparisons); groups blinded: outcome adjudicators Study duration: June 2008 to December 2008 Duration of intervention: 11 to 14 days Duration of follow‐up: mean of 29.14 days (0 to 104) Method of randomisation: use of the biased coin‐based method; stratified by study site Method of concealment of allocation: use of a registration centre contacted by fax for allocation of participants Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) had a primary efficacy outcome that could be evaluated, 3) did not show significant protocol violations Primary safety analysis set: participants who: 1) received at least one dose of study medication, 2) had data on safety outcome Withdrawals: edoxaban 15 mg once daily: 6 (reasons not detailed); 30 mg once daily: 3 (reasons not detailed); enoxaparin 20 mg twice daily: 9 (reasons not detailed) Exclusions: edoxaban 15 mg once daily: 11 (had uninterpretable venography); 30 mg once daily: 14 (1 did not receive study drug, 13 had uninterpretable venography); enoxaparin 20 mg twice daily: 15 (2 did not receive study drug, 13 had uninterpretable venography)
Participants	Diagnosis: scheduled to THA, unilateral Location of participants: 2 countries (Japan, Taiwan) Number of participants randomised: 264 Age (years; mean ± SD): (primary efficacy analysis population) edoxaban 15 mg once daily: 61.3±10.3; 30 mg once daily: 60.6±9.6; enoxaparin 20 mg twice daily: 58.9±10.7 Gender (male %): (primary efficacy analysis population) edoxaban 15 mg once daily: 15 (19.2%); 30 mg once daily: 3 (4.2%); enoxaparin 20 mg twice daily: 15 (20.3%) Baseline imbalances: significant difference regarding gender Inclusion criteria: Aged 20 to < 85 years old, undergoing unilateral THA Exclusion criteria: Revision THA or bilateral THA History of intracranial bleeding, comorbid intraocular bleeding, or intracranial malignant tumour Comorbid GI bleeding or peptic ulcer that occurred within 90 days before the day of informed consent History of symptomatic DVT or PE Weight < 40 kg Use of antithrombotic therapy Severe renal disorder (creatinine clearance < 30 mL/min) Evidence of hepatic dysfunction AST or ALT ≥ 2 x ULN of the study site, total bilirubin ≥ 1.5 x ULN of the study site)
Interventions	Number of intervention groups: 3 Concomitant interventions: none as part of the study; physiotherapy was allowed Excluded interventions: none Edoxaban 15 mg once daily group: Intervention: edoxaban (oral) 15 mg once daily Starting time: after surgery: 6 to 24 hours Duration: 11 to 14 days Edoxaban 30 mg once daily group: Intervention: edoxaban (oral) 30 mg once daily Starting time: after surgery: 6 to 24 hours Duration: 11 to 14 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 20 mg twice daily Starting time: after surgery: 24 to 36 hours Duration: 11 to 14 days
Outcomes	Number of outcomes: According to protocol: 7 Reported: 9. Reported additionally: serious AE, liver enzymes elevation (treatment period) Not reported: none OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 11 to 14 days (treatment) Time points reported: 11 to 14 days (treatment) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT (bilateral venography) Time points measured: 11 to 14 days (treatment) Time points reported: 11 to 14 days (treatment) Symptomatic VTE: Definition: incidence of symptomatic VTE (bilateral venography) Time points measured: 11 to 14 days (treatment) Time points reported: 11 to 14 days (treatment) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2 g/dL within 24 hours; 2) transfusion of ≥ 4 U (approximately 800mL) of blood; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) life‐threatening bleeding Time points measured: 11 to 14 days (treatment) Time points reported: 11 to 14 days (treatment) Serious AEs: non‐hepatic: Definition: number of patients with at least one serious AE, excluding those liver‐related. Serious AE were those which: 1) resulted in death, 2) were life‐threatening, 3) required hospitalisation at the hospital or clinic or prolonged existing hospitalisation, 4) resulted in disability/incapacity, 5) were likely to result in disability/incapacity, 6) congenital anomaly/birth defect Time points measured: 11 to 14 days (treatment) Time points reported: 11 to 14 days (treatment) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 11 to 14 days (treatment) Time points reported: 11 to 14 days (treatment) Asymptomatic distal DVT: Definition: incidence of asymptomatic distal DVT (bilateral venography) Time points measured: 11 to 14 days (treatment) Time points reported: 11 to 14 days (treatment) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 11 to 14 days (treatment) Time points reported: 11 to 14 days (treatment) Liver enzymes elevation: Definition: incidence of ALT >3 times ULN Time points measured: 11 to 14 days (treatment) Time points reported: 11 to 14 days (treatment) RESULTS All‐cause mortality: Summary data (11 to 14 days): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 40 missing participants ‐Subgroups reported: none Major VTE: Summary data (11 to 14 days): see Analysis 1.2 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 40 missing participants ‐Subgroups reported: none Symptomatic VTE: Summary data (11 to 14 days): see Analysis 1.3 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 40 missing participants ‐Subgroups reported: none Major bleeding: Summary data (11 to 14 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 3 missing participants ‐Subgroups reported: none Serious AEs: non‐hepatic Summary data (11 to 14 days): see Analysis 9.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 3 missing participants ‐Subgroups reported: none Fatal VTE: Summary data (11 to 14 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 40 missing participants ‐Subgroups reported: none Asymptomatic distal DVT: Summary data (11 to 14 days): see Analysis 9.10 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 40 missing participants ‐Subgroups reported: none Fatal bleeding: Summary data (11 to 14 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 3 missing participants ‐Subgroups reported: none Liver enzymes elevation: Summary data (11 to 14 days): edoxaban 15 mg once daily ‐ 30 mg once daily ‐ enoxaparin 20 mg twice daily: 3/NR ‐ 2/NR ‐ 9/NR ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, not calculable missing participants ‐Subgroups reported: none
Notes	Source of funding: Daiichi Sankyo Co., Ltd Conflicts of interest: authors (5) had financial relationships with several pharmaceutical companies, including Daiichi Sankyo Published protocol: registered in clinicaltrials.gov as NCT01203098, and in ICTRP as JPRN‐JapicCTI‐090827 Observations: none Missing data requested from authors: specific groups blinded, details about follow‐up period, details about withdrawals, definition of serious AEs, treatment period number of participants for liver enzymes elevation Missing data obtained from authors: information on outcome (serious adverse events) definition, and data on duration of follow‐up For the efficacy set, 15.2% of the randomised participants were excluded (14.3% in the edoxaban groups and 16.9% in the enoxaparin group). Reasons for exclusion were reported. For the safety set, 1.1% of the randomised participants were excluded (0% in the edoxaban groups and 2.2% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random sequence was obtained with a biased coin‐based method
Allocation concealment (selection bias)	Unclear risk	Participants were allocated by a patient registration centre, contacted by fax
Blinding of participants and personnel (performance bias)	High risk	Study intervention was open‐label for the edoxaban versus enoxaparin comparisons
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded events adjudication committee (not stated for all‐cause mortality and liver enzymes elevation). Data analysis was performed by the sponsor, and blinding was not stated
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data balanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: two outcomes not included in the protocol were reported, no outcome included in the protocol was not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

STARS J‐4 2014.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: open‐label; groups blinded: outcome adjudicators Study duration: October 2008 to August 2009 Duration of intervention: 11 to 14 days Duration of follow‐up: 25 to 35 days Method of randomisation: in a 2:1 ratio (edoxaban:enoxaparin), no additional description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) had a primary efficacy outcome that could be evaluated, 3) did not show significant protocol violations Primary safety analysis set: participants who: 1) received at least one dose of study medication, 2) had data on safety outcome, 3) did not show significant protocol violations Withdrawals: not reported Exclusions: edoxaban group (19): 3 did not receive study drug, 16 had uninterpretable venography; enoxaparin group (4): 1 did not receive study drug, 3 had uninterpretable venography
Participants	Diagnosis: scheduled to hip fracture (trochanteric or subtrochanteric) surgery Location of participants: 1 country (Japan) Number of participants randomised: 92 Age (years; mean ± SD): (safety analysis population) edoxaban 76.5 ± 11; enoxaparin 75.6 ± 12 Gender (male %): (safety analysis population) edoxaban 11 (18.6%); enoxaparin 7 (24.1%) Baseline imbalances: differences in most of the baseline characteristics measured Inclusion criteria: Men and women of at least 20 years of age who provided written informed consent and were scheduled to undergo surgery within 10 days for inner or outer femoral neck (trochanteric or subtrochanteric) fracture Exclusion criteria: Increased risk for bleeding (e.g. a history of intracranial bleeding or recent GI bleeding) or increased risk for VTE (i.e. had a prior VTE, history of fracture or prosthetic replacement of lower limbs within 6 months, or recent occurrence of myocardial infarction, cerebral infarction, or transient ischaemic attack) Body weight of < 40 kg Current use of antithrombotic therapy for another indication Severe renal impairment (creatinine clearance < 30 mL/min) or evidence of hepatic impairment Conditions preventing bilateral venography Pregnancy or lactation Contraindications to enoxaparin
Interventions	Number of intervention groups: 2 Concomitant interventions: none as part of the study; IPC and elastic stockings were allowed Excluded interventions: none Edoxaban group: Intervention: edoxaban (oral) 30 mg once daily Starting time: after surgery: 6 to 24 hours Duration: 11 to 14 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 20 mg twice daily Starting time: after surgery: 24 to 36 hours Duration: 11 to 14 days
Outcomes	Number of outcomes: According to protocol: 2 Reported: 18. Reported additionally: all‐cause death, fatal VTE, major VTE, PE, symptomatic VTE, symptomatic DVT, asymptomatic proximal DVT, asymptomatic distal DVT, any DVT, major bleeding, CRNM bleeding, minor bleeding, any bleeding, any AE, serious AE, liver enzymes elevation (treatment period) Not reported: none OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 11 to 14 days (treatment) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT (bilateral venography) Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 11 to 14 days (treatment) Symptomatic VTE: Definition: incidence of symptomatic VTE (bilateral venography) Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 11 to 14 days (treatment) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2g/dL within 24 hours; 2) transfusion of ≥ 4 U (approximately 800 mL) of blood; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 11 to 14 days (treatment) Serious AEs: non‐hepatic: Definition: number of participants with at least one serious AE, excluding those liver‐related. Serious AE were those which: 1) resulted in death, 2) were life‐threatening, 3) required hospitalisation at the hospital or clinic or prolonged existing hospitalisation, 4) resulted in disability/incapacity, 5) were likely to result in disability/incapacity, 6) congenital anomaly/birth defect Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 36 to 49 days (overall study duration) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 11 to 14 days (treatment) Asymptomatic distal DVT: Definition: incidence of asymptomatic distal DVT Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 11 to 14 days (treatment) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 11 to 14 days (treatment) Liver enzymes elevation: Definition: incidence of ALT increased Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 36 to 49 days (overall study duration) RESULTS All‐cause mortality: Summary data (11 to 14 days): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 19 missing participants ‐Subgroups reported: none Summary data (25 to 35 days' follow‐up): not reported Major VTE: Summary data (11 to 14 days): see Analysis 1.2 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 19 missing participants ‐Subgroups reported: none Summary data (25 to 35 days' follow‐up): not reported Symptomatic VTE: Summary data (11 to 14 days): see Analysis 1.3 ‐Analysis set, type of analysis and missing participants: efficacy set, modified intention‐to‐treat (mITT) analysis, 19 missing participants ‐Subgroups reported: none Summary data (25 to 35 days' follow‐up): not reported Major bleeding: Summary data (11 to 14 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 4 missing participants ‐Subgroups reported: none Summary data (25 to 35 days' follow‐up): not reported Serious AEs: non‐hepatic Summary data (11 to 14 days): not reported Summary data (25 to 35 days' follow‐up): not reported Summary data (36 to 49 days, overall study duration): see Analysis 1.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 4 missing participants ‐Subgroups reported: none Fatal VTE: Summary data (11 to 14 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 19 missing participants ‐Subgroups reported: none Summary data (25 to 35 days' follow‐up): not reported Asymptomatic distal DVT: Summary data (11 to 14 days): see Analysis 9.10 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 19 missing participants ‐Subgroups reported: none Summary data (25 to 35 days' follow‐up): not reported Fatal bleeding: Summary data (11 to 14 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 4 missing participants ‐Subgroups reported: none Summary data (25 to 35 days' follow‐up): not reported Liver enzymes elevation: Summary data (11 to 14 days): not reported Summary data (25 to 35 days' follow‐up): not reported Summary data (36 to 49 days, overall study duration): see Analysis 1.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 4 missing participants ‐Subgroups reported: none
Notes	Source of funding: Daiichi Sankyo Co., Ltd Conflicts of interest: all authors had financial relationships with several pharmaceutical companies, including Daiichi Sankyo Published protocol: registered in clinicaltrials.gov as NCT01181141, and in ICTRP as JPRN‐JapicCTI‐090798 Observations: none Missing data requested from authors: specific groups blinded, description of method of randomisation and concealment of allocation, details about withdrawals, definition of serious AEs, treatment period summary data for serious AEs and liver enzymes elevation, follow‐up summary data all‐cause mortality, fatal VTE, major VTE, symptomatic VTE, asymptomatic distal DVT, fatal bleeding, major bleeding, serious AEs and liver enzymes elevation Missing data obtained from authors: information regarding blinding, outcome (serious adverse events) definition, and data on some outcomes For the efficacy set, 20.7% of the randomised participants were excluded (25.8% in the edoxaban group and 10.0% in the enoxaparin group). Reasons for exclusion were reported. For the safety set, 4.3% of the randomised participants were excluded (4.8% in the edoxaban group and 3.3% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method was not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	High risk	Study intervention was open‐label for the edoxaban versus enoxaparin comparison
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded events adjudication committee (not stated for all‐cause mortality and liver enzymes elevation). Data analysis was performed by the sponsor, and blinding was not stated
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data unbalanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data balanced in numbers and reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 16 outcomes not included in the protocol were reported, no outcome included in the protocol was not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

STARS J‐V 2015.

Study characteristics
Methods	Aim: non‐inferiority, efficacy and safety Design: parallel‐group (phase III) Blinding: double‐blind, double‐dummy; groups blinded: participants, clinicians/providers, outcome adjudicators Study duration: May 2009 to January 2010 Duration of intervention: 11 to 14 days Duration of follow‐up: 25 to 35 days Method of randomisation: in a 1:1 ratio, no additional description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who received at least one dose of study medication Withdrawals: edoxaban group (23): 9 due to AEs, 6 voluntarily withdrew, 3 became ineligible, 2 because of investigator's decision, 3 unexplained; enoxaparin group (32): 10 due to AEs, 8 voluntarily withdrew, 3 because of investigator's decision, 11 unexplained Exclusions: edoxaban group (52): 4 did not receive study drug, 48 had uninterpretable venography; enoxaparin group (55): 2 did not receive study drug, 53 had uninterpretable venography
Participants	Diagnosis: scheduled to THA, elective, primary, unilateral Location of participants: 1 country (Japan) Number of participants randomised: 610 Age (years; mean ± SD): (primary efficacy analysis population) edoxaban 62.8 ± 9.6; enoxaparin 62.8 ± 9.7 Gender (male %): (primary efficacy analysis population) edoxaban 35 (13.7%); enoxaparin 36 (14.5%) Baseline imbalances: none Inclusion criteria: people 20 to 85 years of age undergoing unilateral THA Exclusion criteria: Risk for bleeding (e.g. history of intracranial bleeding, comorbid GI bleeding or peptic ulcer within the previous 3 months, or prothrombin time (PT) prolongation above the ULN) Risk for thromboembolism (e.g. history of symptomatic DVT or PE, coagulation disease at high risk for thromboembolism, or history of lower limb fracture within the previous 6 months) Previous TKA Expected joint replacement of the other lower limb at the same time of anaesthesia Weight < 40 kg Severe renal impairment (creatinine clearance (CLCR) of < 30 mL/min) Evidence of hepatic dysfunction (serum AST or serum ALT levels ≥ 2 times the ULN or total bilirubin ≥ 1.5 times the ULN) Previous treatment with edoxaban Current antithrombotic therapy for another complication
Interventions	Number of intervention groups: 2 Concomitant interventions: none as part of the study; IPC and elastic stockings were allowed Excluded interventions: none Edoxaban group: Intervention: edoxaban (oral) 30 mg once daily Starting time: after surgery: 6 to 24 hours Duration: 11 to 14 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 20 mg twice daily Starting time: after surgery: 24 to 36 hours Duration: 11 to 14 days
Outcomes	Number of outcomes: According to protocol: 7 Reported: 13. Reported additionally: major VTE, minor bleeding, any bleeding, any AE, serious AE, liver enzymes elevation Not reported: none OUTCOMES Symptomatic VTE: Definition: incidence of symptomatic VTE (bilateral venography) Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 11 to 14 days (treatment) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2 g/dL within 24 hours; 2) transfusion of ≥ 4 U (approximately 800 mL) of blood; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 11 to 14 days (treatment) Serious AEs: non‐hepatic: Definition: number of patients with at least one serious AE, excluding those liver‐related. Serious AE were those which: 1) resulted in death, 2) were life‐threatening, 3) required hospitalisation at the hospital or clinic or prolonged existing hospitalisation, 4) resulted in disability/incapacity, 5) were likely to result in disability/incapacity, 6) congenital anomaly/birth defect Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 11 to 14 days (treatment) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 11 to 14 days (treatment) Asymptomatic distal DVT: Definition: incidence of asymptomatic distal DVT Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 11 to 14 days (treatment) Liver enzymes elevation: Definition: incidence of ALT/AST > 3 times ULN on same date Time points measured: 11 to 14 days (treatment), 25 to 35 days (additionally, follow‐up) Time points reported: 11 to 14 days (treatment) RESULTS Symptomatic VTE: Summary data (11 to 14 days): see Analysis 1.3 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 107 missing participants ‐Subgroups reported: none Summary data (25 to 35 days' follow‐up): not reported Major bleeding: Summary data (11 to 14 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 6 missing participants ‐Subgroups reported: none Summary data (25 to 35 days' follow‐up): not reported Serious AEs: non‐hepatic Summary data (11 to 14 days): see Analysis 9.8 ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Summary data (25 to 35 days' follow‐up): not reported Fatal VTE: Summary data (11 to 14 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 107 missing participants ‐Subgroups reported: none Summary data (25 to 35 days' follow‐up): not reported Asymptomatic distal DVT: Summary data (11 to 14 days): see Analysis 9.10 ‐Analysis set, type of analysis and missing participants: efficacy set, modified intention‐to‐treat (mITT) analysis, 107 missing participants ‐Subgroups reported: none Summary data (25 to 35 days' follow‐up): not reported Liver enzymes elevation: Summary data (11 to 14 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 6 missing participants ‐Subgroups reported: none Summary data (25 to 35 days' follow‐up): not reported
Notes	Source of funding: Daiichi Sankyo Co., Ltd Conflicts of interest: all authors had financial relationships with several pharmaceutical companies, including Daiichi Sankyo Published protocol: registered in clinicaltrials.gov as NCT01181167, and in ICTRP as JPRN‐JapicCTI‐090771 Observations: none Missing data requested from authors: specific groups blinded, description of method of randomisation and concealment of allocation, details about withdrawals, definition of serious AEs, follow‐up summary data for fatal VTE, symptomatic VTE, asymptomatic distal DVT, major bleeding, serious AEs and liver enzymes elevation, treatment period number of participants for serious AEs Missing data obtained from authors: information on blinding, outcome (serious adverse events) definition, and data on some outcomes For the efficacy set, 17.5% of the randomised participants were excluded (16.9% in the edoxaban group and 18.2% in the enoxaparin group). Reasons for exclusion were reported. For the safety set, 1.0% of the randomised participants were excluded (1.3% in the edoxaban group and 0.7% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method was not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Low risk	Participants were blinded with the use of placebo in a double‐dummy fashion. No description was provided for the method of blinding of clinicians/providers
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded events adjudication committee (not stated for liver enzymes elevation). Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Missing outcome data balanced in numbers and reasons across intervention groups.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: six outcomes not included in the protocol were reported, no outcome included in the protocol was not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

Tang 2017.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: double‐blind, not described; groups blinded: participants Study duration: September 2011 to September 2016 Duration of intervention: 28 days Duration of follow‐up: 30 days Method of randomisation: segmental random balance method to generate a random number table. Random numbers were computer‐generated Method of concealment of allocation: use of sealed envelopes Primary efficacy analysis set: all randomised participants who completed the study Primary safety analysis set: same as efficacy set Withdrawals (not detailed according to intervention group): 11 lost to follow‐up Exclusions: not reported
Participants	Diagnosis: scheduled to hip fracture surgery Location of participants: single centre (China) Number of participants randomised: 298 Age (years; mean ± SD): rivaroxaban group 72 ± 14; enoxaparin group 68 ± 17; sequential therapy group 69 ± 12 Gender (male %): rivaroxaban group 35.4%; enoxaparin group 42.1%; sequential therapy group 37.5% Baseline imbalances: none Inclusion criteria: Factures caused by fall‐induced damage Admitted to the hospital within 24 hours following injury Diagnosed by X‐ray and/or computed tomography Received internal fixation Exclusion criteria: Lower extremity DVT that was confirmed by pre‐operative imaging History of thromboembolic disease with anticoagulant therapy Haemorrhagic diseases and/or major bleeding history (such as intracranial haemorrhage or GI bleeding that required blood transfusion) Severe liver disease Coagulation disorders and/or contraindications to anticoagulation Contraindication to rivaroxaban and/or LMWH
Interventions	Number of intervention groups: 3 Concomitant interventions: skin traction Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 6 hours Duration: 28 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 4000IU once daily Starting time: after surgery: 12 hours Duration: 28 days Sequential therapy group: Intervention: enoxaparin (subcutaneous) 4000 IU once daily for 1 week followed by rivaroxaban (oral) 10 mg once daily for 3 weeks Starting time: after surgery: 12 hours Duration: 28 days
Outcomes	Number of outcomes: According to protocol: 5 Reported: 15. Reported additionally: proximal DVT, distal DVT, symptomatic PE, total PE, fatal PE, Hb reduction > 20 g/L, other bleeding, volume of blood loss (postoperative drainage), time to stitch removal, adverse incisions Not reported: none OUTCOMES Fatal VTE: Definition: incidence of death due to VTE Time points measured: 30 days Time points reported: 30 days Volume of blood loss: Definition: volume of postoperative drainage Method/unit of measurement: mL Time points measured: not reported Time points reported: not reported RESULTS Fatal VTE: Summary data (30 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 11 missing participants ‐Subgroups reported: none Volume of blood loss: Summary data (time point not reported) (mean±SD): see Analysis 1.14 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 11 missing participants ‐Subgroups reported: none
Notes	Source of funding: self‐collected funds Conflicts of interest: none Published protocol: registered in the Chinese Clinical Trial Registry as ChiCTR‐INR‐17010495 Observations: none Missing data requested from authors: specific groups blinded, details about withdrawals (according to allocated group), time point measured for volume of blood loss Missing data obtained from authors: none For the analysis set, 3.7% of the randomised participants were excluded (not detailed according to intervention group). Exclusions were due to loss to follow‐up. Study did not report the numbers of participants initially randomised to each group. No missing outcome data provided for each group. CONSORT flow diagram incomplete
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random numbers were computer‐generated with segmental random balance method
Allocation concealment (selection bias)	Low risk	Allocated group was designated with a sealed envelope
Blinding of participants and personnel (performance bias)	Low risk	Study was stated to be double‐blinded (participants and researchers), but no description was provided
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded events adjudication committee (not stated for liver enzymes elevation). Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data.
Selective reporting (reporting bias)	High risk	The protocol was published in chictr.org.cn. Reported outcomes differed from those included in the protocol: 10 outcomes not included in the protocol were reported. Selective reporting involved 1 outcome included in the review
Other bias	Low risk	The study appears to be free of other sources of bias.

Wasko 2015.

Study characteristics
Methods	Aim: safety Design: parallel‐group Blinding: double‐blind, not described; groups blinded: investigators, outcome adjudicators Study duration: September 2013 to July 2014 Duration of intervention: 3 days Duration of follow‐up: there was no follow‐up beyond treatment period Method of randomisation: no description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: not defined Primary safety analysis set: not defined Withdrawals: not reported Exclusions: not reported
Participants	Diagnosis: scheduled to THR, primary, unilateral Location of participants: not reported Number of participants randomised: 60 Age (years): not reported Gender (male %): not reported Baseline imbalances: not assessable Inclusion criteria: primary, end‐stage hip osteoarthritis requiring THR Exclusion criteria: Inflammatory arthropathies Liver disorders Neoplastic conditions Clotting disorders
Interventions	Number of intervention groups: 3 Concomitant interventions: none as part of the study; physiotherapy was allowed Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: not reported Duration: 28 days (outcome measure was performed 3 days after surgery) Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: not reported Duration: 28 days (outcome measure was performed 3 days after surgery) Dabigatran group: Intervention: dabigatran (oral) 110 mg twice daily Starting time: after surgery: not reported Duration: 27 days (outcome measure was performed 3 days after surgery)
Outcomes	Number of outcomes: According to protocol: 3 Reported: 2. Reported additionally: none Not reported: drop in the haemoglobin value (3rd postoperative day) OUTCOMES Volume of blood loss: Definition: volume of blood loss calculated according to the Nadler formula Method/unit of measurement: mL Time points measured: 3 days Time points reported: 3 days RESULTS Volume of blood loss: Summary data (3 days) (mean ± SD): see Analysis 1.14 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none
Notes	Source of funding: Centre of Postgraduate Medical Education Conflicts of interest: not reported Published protocol: registered in ClinicalTrials.gov as NCT02085824 Observations: none Missing data requested from authors: specific groups blinded, description of method of randomisation and concealment of allocation, definition of primary efficacy and safety sets, details about withdrawals and exclusions, location of study, age and gener distribution of participants, statement of conflicts of interest Missing data obtained from authors: none This study is reported in abstract format only. No efficacy outcome was reported. For the safety set, no participant was excluded. No CONSORT flow diagram.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method was not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Low risk	Study was stated to be double‐blinded, but no description was provided
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of outcome assessment was not mentioned. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	Unclear risk	Insufficient information to detect if missing outcome data existed.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	No exclusion or withdrawal was reported for any study group
Selective reporting (reporting bias)	Low risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: no outcomes not included in the protocol were reported, 1 outcome included in the protocol was not reported. Selective reporting did not involve any outcome included in the review
Other bias	High risk	There was at least one important risk of bias. The study did not report if conflicts of interest were registered.

Weitz 2010.

Study characteristics
Methods	Aim: dose‐response study, efficacy and safety Design: parallel‐group (phase II) Blinding: open‐label (for the letaxaban versus enoxaparin comparisons); groups blinded: outcome adjudicators Study duration: October 2007 to October 2008 Duration of intervention: 10 to 14 days Duration of follow‐up: not reported Method of randomisation: no description was provided Method of concealment of allocation: use of an interactive voice recognition system for allocation of participants Primary efficacy analysis set: participants who: 1) received at least one dose of study medication, 2) had a primary efficacy outcome that could be evaluated Primary safety analysis set: participants who: 1) received at least one dose of study medication Withdrawals: letaxaban 10 mg twice daily: 9 (3 AE, 0 protocol deviation, 2 voluntary withdrawal, 4 other); 20 mg twice daily: 18 (7 AE, 1 protocol deviation, 4 voluntary withdrawal, 6 other); 40 mg once daily: 13 (6 AE, 6 voluntary withdrawal, 1 other); 40 mg twice daily: 21 (12 AE, 1 protocol deviation, 1 voluntary withdrawal, 7 other); 80 mg once daily: 7 (5 AE, 2 voluntary withdrawal); 80 mg twice daily: 7 (6 AE, 1 protocol deviation); enoxaparin: 14 (9 AE, 1 protocol deviation, 4 other) Exclusions: letaxaban 10 mg twice daily: 28 (1 did not receive study drug, 27 had no evaluable adjudication result); 20 mg twice daily: 43 (1 did not receive study drug, 42 had no evaluable adjudication result); 40 mg once daily: 48 (1 did not receive study drug, 47 had no evaluable adjudication result); 40 mg twice daily: 51 (1 did not receive study drug, 50 had no evaluable adjudication result); 80 mg once daily: 51 (3 did not receive study drug, 48 had no evaluable adjudication result); 80 mg twice daily: 43 (1 did not receive study drug, 42 had no evaluable adjudication result); enoxaparin: 56 (2 did not receive study drug, 54 had no evaluable adjudication result)
Participants	Diagnosis: scheduled to TKR, elective, unilateral, primary Location of participants: 2 countries (USA and Canada) Number of participants randomised: 1048 Age (years; mean and range): (all randomised population) letaxaban 10 mg twice daily / 20 mg twice daily / 40 mg once daily / 40 mg twice daily / 80 mg once daily / 80 mg twice daily 64.5 (43‐88) / 65.5 (42‐83) / 65.6 (44‐87) / 65.5 (44‐85) / 64.8 (38‐86) / 65.1 (41‐90); enoxaparin 63.2 (41‐89) Gender (male %): (all randomised population) letaxaban 10 mg twice daily / 20 mg twice daily / 40 mg once daily / 40 mg twice daily / 80 mg once daily / 80 mg twice daily 44 (41.9%) / 45 (34.9%) / 61 (37.4%) / 53 (32.5%) / 54 (33.1%) / 62 (38.3%); enoxaparin 63.2 70 (42.9%) Baseline imbalances: slight differences regarding gender among the groups Inclusion criteria: Aged 18 years or older and scheduled to undergo elective, unilateral, primary TKR surgery Signed informed consent Exclusion criteria: Known bleeding diathesis Known inherited thrombophilic disorder Requirement for ongoing treatment with an anticoagulant, aspirin (> 162 mg/day) or other antiplatelet agents (clopidogrel, ticlopidine or dipyridamole), fibrinolytic agents or other agents that affect coagulation (except for non‐steroidal anti‐inflammatory drugs with a half‐life of less than 17 hour) History of intra‐cerebral, intraocular or GI bleeding, or active gastric or duodenal ulceration within the past 6 months History of major surgery within the past 3 months History of VTE, myocardial infarction, stroke or transient ischaemic attack within the past 6 months History of hypersensitivity or allergies to other factor Xa inhibitors, enoxaparin, or iodine or contrast dye used for venography Multiple (more than 3) or traumatic epidural or spinal punctures immediately prior to randomisation Requirement for an indwelling epidural catheter for post‐operative analgesia Severe hypertension (systolic blood pressure > 180 mmHg and/or diastolic blood pressure > 110 mmHg at screening) Moderate to severe renal dysfunction (based on a calculated creatinine clearance < 45 mL/minute) at screening ALT greater than 2 times the ULN, active liver disease, or jaundice at screening Anaemia (haemoglobin < 10.0 g/dL) or thrombocytopenia (platelet count < 100x109/L) at screening History of cancer that had not been in remission for at least 5 years (excluding basal or stage I squamous cell carcinoma of the skin) Women who were pregnant or lactating
Interventions	Number of intervention groups: 7 Concomitant interventions: none as part of the study; elastic compression stockings were allowed Excluded interventions: none Letaxaban 10 mg twice daily group: Intervention: letaxaban (oral) 10 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 10 to 14 days Letaxaban 20 mg twice daily group: Intervention: letaxaban (oral) 20 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 10 to 14 days Letaxaban 40 mg once daily group: Intervention: letaxaban (oral) 40 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 10 to 14 days Letaxaban 40 mg twice daily group: Intervention: letaxaban (oral) 40 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 10 to 14 days Letaxaban 80 mg once daily group: Intervention: letaxaban (oral) 80 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 10 to 14 days Letaxaban 80 mg twice daily group: Intervention: letaxaban (oral) 80 mg twice daily Starting time: after surgery: 6 to 8 hours Duration: 10 to 14 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 30 mg twice daily Starting time: after surgery: 12 to 24 hours Duration: 6 to 14 days
Outcomes	Number of outcomes: According to protocol: 7 Reported: 15. Reported additionally: all‐cause death, major bleeding, major and CRNM bleeding, volume of blood transfusions, volume of blood loss, serious AE, any AE, liver enzymes elevation (treatment period) Not reported: none OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake) Symptomatic VTE: Definition: incidence of symptomatic VTE (bilateral venography) Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2g/dL within 24 hours; 2) transfusion of ≥ 4 U (approximately 800 mL) of blood; 3) bleeding at a critical site (intracranial, intraspinal, intraocular, pericardial, or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or intervention; 5) fatal bleeding Time points measured: 10 to 14 days (treatment) Time points reported: 10 to 14 days (treatment) Serious AEs: non‐hepatic: Definition: number of patients with at least one severe AE (definition not reported), excluding those liver‐related Time points measured: 10 to 14 days (treatment) Time points reported: 10 to 14 days (treatment) Fatal VTE: Definition: incidence of death due to VTE Time points measured: 12 to 16 days (treatment, 2 days after the last study drug intake), 30 days (additionally, follow‐up) Time points reported: 12 to 16 days (treatment, 2 days after the last study drug intake), 30 days (additionally, follow‐up) Fatal bleeding: Definition: incidence of death due to bleeding Time points measured: 10 to 14 days (treatment) Time points reported: 10 to 14 days (treatment) Liver enzymes elevation: Definition: incidence of ALT > 3 times ULN Time points measured: 10 to 14 days (treatment) Time points reported: 10 to 14 days (treatment) Volume of blood loss: Definition: volume of post‐operative drainage Method/unit of measurement: mL Time points measured: not reported Time points reported: not reported RESULTS All‐cause mortality: Summary data (12 to 16 days): see Analysis 9.1 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 318 missing participants (2 participants excluded from the primary efficacy set were included in these data, reasons were not reported) ‐Subgroups reported: none Symptomatic VTE: Summary data (12 to 16 days): see Analysis 9.3 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 318 missing participants (2 participants excluded from the primary efficacy set were included in these data, reasons were not reported) ‐Subgroups reported: none Major bleeding: Summary data (10 to 14 days): see Analysis 9.4 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 10 missing participants ‐Subgroups reported: none Serious AEs: non‐hepatic: Summary data (10 to 14 days): see Analysis 9.8 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 10 missing participants ‐Subgroups reported: none Fatal VTE: Summary data (12 to 16 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, mITT analysis, 318 missing participants (2 participants excluded from the primary efficacy set were included in these data, reasons were not reported) ‐Subgroups reported: none Summary data (30‐day follow‐up): letaxaban 10 mg twice daily ‐ 20 mg twice daily ‐ 40 mg once daily ‐ 40 mg twice daily ‐ 80 mg once daily ‐ 80 mg twice daily ‐ enoxaparin 30 mg twice daily: 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐ 0/NR ‐Analysis set, type of analysis and missing participants: not specified, not specified, not calculable missing participants ‐Subgroups reported: none Fatal bleeding: Summary data (10 to 14 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 10 missing participants ‐Subgroups reported: none Liver enzymes elevation: Summary data (10 to 14 days): see Analysis 9.12 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 10 missing participants ‐Subgroups reported: none Volume of blood loss: Summary data (time point not reported) (mean±SD): see Analysis 1.14 ‐Analysis set, type of analysis and missing participants: safety set, mITT analysis, 10 missing participants ‐Subgroups reported: none
Notes	Source of funding: Takeda Global Research and Development, Inc Conflicts of interest: all authors had financial relationships with several pharmaceutical companies, including Takeda Published protocol: registered in ClinicalTrials.gov as NCT00641732 Observations: none Missing data requested from authors: specific groups blinded, details about follow‐up period, description of method of randomisation, details about withdrawals, definition of serious AEs, time point of measurement of volume of blood loss, follow‐up number of participants for fatal VTE, treatment period missing participants for all‐cause mortality, fatal VTE, symptomatic VTE, protocol of the study Missing data obtained from authors: none For the efficacy set, 30.3% of the randomised participants were excluded (29.8% in the letaxaban group and 33.1% in the enoxaparin group). Reasons for exclusion were reported. For the safety set, 1.0% of the randomised participants were excluded (0.9% in the letaxaban group and 1.2% in the enoxaparin group). Reasons for exclusion were reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was stated, but the method was not described
Allocation concealment (selection bias)	Low risk	Participants were allocated by the use of an interactive voice recognition system
Blinding of participants and personnel (performance bias)	High risk	Study intervention was open‐label for the letaxaban versus enoxaparin comparison
Blinding of outcome assessment (detection bias)	Unclear risk	Outcome assessment was performed by an independent, blinded adjudication committee (not stated for liver enzymes elevation and volume of blood loss). Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Reasons for missing outcome data related to true outcome. Missing outcome data unbalanced in reasons across intervention groups. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	High risk	Missing outcome data reported. Reasons for missing outcome data related to true outcome. Missing outcome data unbalanced in reasons across intervention groups.
Selective reporting (reporting bias)	High risk	The protocol was published in ClinicalTrials.gov. Reported outcomes differed from those included in the protocol: 8 outcomes not included in the protocol were reported, no outcome included in the protocol was not reported. Selective reporting involved several outcomes included in the review
Other bias	High risk	There are at least two important risks of bias. The authors have potential conflicts of interest and the study did not report a management plan explaining procedures or additional steps taken to minimise the risk of bias to ensure research integrity, such as divestiture or independent data review. Also, the study had a potential source of bias related to the specific study design used (insensitive instrument to measure outcomes which can lead to underestimation of beneficial effects).

Wing 2020.

Study characteristics
Methods	Study design: randomised controlled trial Group: parallel group Blinding: not described Exclusions: none Method of concealment of allocation: not described Method of randomisation: not described Primary efficacy analysis set: all treated participants Primary safety analysis set: all treated participants Study duration: from January 2017 to December 2018 Withdrawals: none
Participants	Inclusion criteria: scheduled to single‐sided hip joint replacement Exclusion criteria: not reported Group differences: no differences in terms of sex, age, or body mass index (BMI) Diagnosis: scheduled to hip replacement (THR) Location of participants: single centre (China) Number of participants randomised: 50 Rivaroxaban group Age; mean (range): 65.40 ± 5.26 Gender; F/M: 12/13 LMWH group Age; mean (range): 63.70 ± 5.57 Gender; F/M: 12/13 Overall Age; mean (range): not reported Gender; F/M: Not reported
Interventions	Rivaroxaban group Intervention: rivaroxaban (oral) 10 mg once daily Starting time: 6 hours postoperatively Duration: 5 weeks Concomitant interventions: hip and knee exercise 3 days after hip replacement Excluded interventions: none LMWH group Intervention: LMWH (subcutaneous) 4100 IU once daily Starting time: 6 hours postoperatively Duration: 2 weeks Concomitant interventions: hip and knee exercise 3 days after hip replacement Excluded interventions: none
Outcomes	Symptomatic VTE (colour Doppler ultrasound) Volume of blood loss
Notes	Missing data requested from authors: none Missing data obtained from authors: none No participant was excluded after randomisation. No CONSORT flow diagram.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	High risk	Blinding method not described. Open‐label design can be assumed
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of outcome assessment was not mentioned. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	Low risk	No exclusion or withdrawal was reported for any study group
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	No exclusion or withdrawal was reported for any study group
Selective reporting (reporting bias)	Unclear risk	No published protocol was found
Other bias	High risk	There is at least one important risk of bias. The study had a potential source of bias related to the specific study design used (inappropriate intervention as the duration of prophylaxis differed markedly for each intervention group, and the outcome analysis was at different time points; 5 weeks versus 2 weeks). The intervention could be favoured in the design of the study, since follow‐up was longer than 2 weeks and the analyses of the outcomes were at different times.

Xie 2017.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: open‐label Study duration: August 2015 to March 2016 Duration of intervention: 15 days Duration of follow‐up: 3 months Method of randomisation: computer‐generated randomisation Method of concealment of allocation: use of sequentially numbered sealed opaque envelopes Primary efficacy analysis set: all randomised participants who completed the study Primary safety analysis set: same as efficacy set Withdrawals: rivaroxaban group: 0; enoxaparin group: 2 lost to follow‐up Exclusions: not reported
Participants	Diagnosis: scheduled to TKR, unilateral, primary Location of participants: single centre (China) Number of participants randomised: 196 Age (years ± SD): rivaroxaban group 65.2 ± 5.5, enoxaparin group: 66.8 ± 7.4 Gender (male %): rivaroxaban group 22.9%, enoxaparin group: 12.2% Baseline imbalances: none Inclusion criteria: Aged 18 years and older Scheduled for primary unilateral TKR for osteoarthritis Exclusion criteria: Cardiovascular problems (history of myocardial infraction, angina, and atrial fibrillation) Cerebrovascular conditions (history of previous stroke) Thromboembolic disorders (history of DVT or PE) Clotting disorders Discontinuation of oral NSAID less than 1 week before Active peptic ulcer or organic damage with bleeding tendency Anticoagulants intake Known allergy to TXA, LMWH, or rivaroxaban
Interventions	Number of intervention groups: 2 Concomitant interventions: intravenous and intra‐articular TXA. Intravenous dose of 15 mg/kg TXA administrated 5 to 15 minutes before tourniquet deflation, and 1 g TXA injected into the articular cavity through the drainage tube, which remained clamped for 2 hours and was removed the next morning postoperatively Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 6 to 21 hours Duration: 15 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 4000 IU once daily Starting time: after surgery: 6 to 14 hours Duration: 15 days
Outcomes	Number of outcomes according to protocol: no protocol is available OUTCOMES All‐cause mortality: Definition: incidence of death from any cause Time points measured: 15 days (treatment period), 3 months (additional follow‐up) Time points reported: 15 days (treatment period) Major VTE: Definition: incidence of fatal and nonfatal PE or proximal DVT (colour Doppler ultrasound) Time points measured: 15 days (treatment period) Time points reported: 15 days (treatment period) Symptomatic VTE: Definition: VTE event detected by routine ultrasonic examination postoperatively which was not associated with suggestive symptoms Time points measured: 15 days (treatment period) Time points reported: 15 days (treatment period) Major bleeding: Definition: incidence of clinically overt bleeding with (1 or more): 1) haemoglobin decreased by ≥ 2 g/dL compared with pre‐randomisation level; 2) transfusion of ≥ 2 U of packed RBC; 3) bleeding at a critical site (intracerebral, intraocular, intraspinal, pericardial or retroperitoneal); 4) bleeding into the operated joint, requiring an additional operation or any unusual medical intervention or procedure for relief (e.g., draining or puncture of a haematoma at the surgical site, transfer to an ICU or emergency room); 5) fatal bleeding; 6) warrant of treatment cessation Time points measured: 15 days (treatment period) Time points reported: 15 days (treatment period) Fatal VTE: Definition: incidence of death from VTE Time points measured: 15 days (treatment period) Time points reported: 15 days (treatment period) Asymptomatic distal DVT: Definition: incidence of asymptomatic distal DVT Time points measured: 15 days (treatment period) Time points reported: 15 days (treatment period) Fatal bleeding: Definition: incidence of death from bleeding Time points measured: 15 days (treatment period) Time points reported: 15 days (treatment period) RESULTS All‐cause mortality: Summary data (15 days): see Analysis 1.1 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 2 missing participants ‐Subgroups reported: none Major VTE: Summary data (15 days): see Analysis 1.2 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 2 missing participants ‐Subgroups reported: none Symptomatic VTE: Summary data (15 days): see Analysis 1.3 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 2 missing participants ‐Subgroups reported: none Major bleeding: Summary data (15 days): see Analysis 1.4 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 2 missing participants ‐Subgroups reported: none Fatal VTE: Summary data (15 days): see Analysis 1.9 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 2 missing participants ‐Subgroups reported: none Asymptomatic distal DVT: Summary data (15 days): see Analysis 1.10 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 2 missing participants ‐Subgroups reported: none Fatal bleeding: Summary data (15 days): see Analysis 1.11 ‐Analysis set, type of analysis and missing participants: efficacy set, ITT analysis, 2 missing participants ‐Subgroups reported: none
Notes	Source of funding: National Health and Family Planning Commission of the People’s Republic of China (CN) Program (201302007) Conflicts of interest: none Published protocol: none Observations: none Missing data requested from authors: method of blinding of data collectors and analysts, follow‐up summary data for all‐cause mortality, protocol of the study Missing data obtained from authors: none For the analysis set, 1% of the randomised participants were excluded (2% in the rivaroxaban group and 0% in the enoxaparin group). Reasons for exclusion were reported. No CONSORT flow diagram.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random sequence was computer‐generated
Allocation concealment (selection bias)	Low risk	Allocation was designated with sequentially numbered sealed opaque envelopes
Blinding of participants and personnel (performance bias)	High risk	Study was open‐label for the rivaroxaban versus enoxaparin comparison
Blinding of outcome assessment (detection bias)	Unclear risk	Data collectors are stated to have been blinded to treatment allocation, but the method was not described. Data analysts are stated to have been blinded to treatment allocation, but the method was not described
Incomplete outcome data (attrition bias) Benefit outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Incomplete outcome data (attrition bias) Harm outcomes	High risk	Missing outcome data reported. Insufficient information regarding reasons for missing outcome data. Proportion of missing outcomes compared with the observed event risk was large enough to have had a clinically relevant impact on the intervention effect estimate.
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	Unclear risk	There may be a risk of bias for an inappropriate co‐intervention, but there is either insufficient information to assess whether an important risk of bias exists, or insufficient rationale or evidence that an identified problem will introduce bias.

Yun‐Fei 2018.

Study characteristics
Methods	Design: randomised controlled trial Group: parallel group Blinding: open‐label Exclusions: none Method of concealment of allocation: not described Method of randomisation: random numbers table Primary efficacy analysis set: all treated participants Primary safety analysis set: all treated participants Study duration: from January 2015 to January 2017 Withdrawals: none
Participants	Inclusion criteria: Scheduled to primary total hip replacement surgery Age between 60 and 80 years Weight between 35 and 75 kg Absence of deep vein thrombosis by lower limb Doppler ultrasonography Rivaroxaban group: age: 69.28+‐10.42 years; gender: F/M: 48/50 LMWH group: age: 69.33+‐11.84 years; gender: F/M: 49/49 Exclusion criteria: Abnormal clotting function of other blood disorders Severe disease of the circulatory system Liver or kidney failure Malignancy Venous valve insufficiency by lower limb Doppler ultrasound Any postoperative complication which warrants stopping anticoagulant medication Group differences: no differences were observed Diagnosis: scheduled to THR Location of participants: single centre (China) Number of participants randomised: 196
Interventions	Rivaroxaban group Intervention: rivaroxaban (oral) 10 mg once daily Starting time: 6 hours postoperatively Duration: 2 weeks Concomitant interventions: none Excluded interventions: none LMWH group Intervention: LMWH (subcutaneous) 40 mg once daily Starting time: 6 hours postoperatively Duration: 5 weeks Concomitant interventions: none Excluded interventions: none
Outcomes	Symptomatic VTE
Notes	Missing data requested from authors: none Missing data obtained from authors: none No exclusions nor withdrawals were reported. No CONSORT flow diagram.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was performed with random numbers table
Allocation concealment (selection bias)	Unclear risk	The method of concealment of allocation was not described
Blinding of participants and personnel (performance bias)	High risk	The study was open‐label for the comparison of rivaroxaban and LMWH
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of outcome adjudicators was not reported
Incomplete outcome data (attrition bias) Benefit outcomes	Low risk	No exclusions nor withdrawals were reported.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	No exclusions nor withdrawals were reported.
Selective reporting (reporting bias)	Unclear risk	No published protocol was found
Other bias	Unclear risk	There may be a risk of bias for an inappropriate intervention, but there is either insufficient information to assess whether an important risk of bias exists, or insufficient rationale or evidence that an identified problem will introduce bias. The control may have been favoured in the design of the study, since follow‐up was longer than 2 weeks and the time analysis of the outcomes was not stated.

Zhou 2023.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: double‐blind, double‐dummy; groups blinded: participants, trial participants, outcome assessors, and data collectors were blinded to allocation. Study duration: 7 days Duration of intervention: 7 days Duration of follow‐up: not mentioned Method of randomisation: computer‐generated random sequence by Randomization.com prepared by a statistician who was not involved in the clinical trial. "Method of concealment of allocation: the randomisation assignments were placed within sequentially numbered opaque sealed envelopes in the custody of a certified study pharmacist. At the end of the procedure, the envelopes were opened, and the corresponding medications were handled by an investigator who was blinded to the patient’s care." Primary efficacy analysis set: thrombotic complications Primary safety analysis set: haemoglobin, total blood loss, transfusion rate, postoperative drainage, bleeding complications. Withdrawals: none Exclusions: none
Participants	Diagnosis: scheduled to primary unilateral TKA Location of participants: 1 country (China) from October 2020 to December 2021. Number of participants randomised: 180 Age (years): Group 1 mean age of 64.8±7.2 years Group 2 mean age of 64.1±6.7years Group 3 mean age of 66.4±7.6years Gender (male %): Rivaroxaban group: 60 participants (male 45%) Dalteparin group: 60 participants (male 48%) ASA group: 60 participants (male 43%) Baseline imbalances: not found Inclusion criteria: "patients (1) aged 55–80 years; (2) diagnosed with knee osteoarthritis according to the Kellgren‐Lawrence grading system (greater than or equal to grade III); and (3) willing to undergo primary unilateral TKA." Exclusion criteria: "(1) secondary osteoarthritis (2) systemic or local infection; (3) blood system diseases; (4) previous or current use of antithrombotic drugs; (5) previous history of thrombosis or thrombosis discovered on color Doppler ultrasound of both lower extremities; (6) high‐risk cardiovascular disease (CVD) with thromboses, including cerebral infarction, myocardial infarction, atrial fibrillation, heart failure and post‐stenting; (7) use of NSAIDs; and (8) history of epilepsy or severe liver and kidney insufficiency."
Interventions	Number of intervention groups: 3 Concomitant interventions: "60 mg/kg TXA intravenously 5min prior to prosthesis placement. The combined lumbar and rigid anesthesia was chosen, and a medial parapatellar approach was selected with a median skin incision made anterior to the knee under the control of a balloon tourniquet. TKA was subsequently performed according to the routine procedure, one drainage tube was placed, then the incision was closed, and pressure bandaging was completed. Prophylactic antibiotics were administered intravenously 30min before excision, and antibiotics were applied continuously for 7days postoperatively. After recovery from anaesthesia, each patient was instructed to lie fat and keep the afected [sic] limb straight for ankle pump exercises. Compression bandaging of the afected [sic] limb and ice packs were recommended 24h after surgery. Multimodal analgesia using analgesic pumps combined with NSAIDs was recommended. On the second day after surgery, patients were encouraged to perform active and passive knee fexion [sic] and extension exercises under the help of the surgeon and walk with the help of a walker. Allogeneic blood transfusions are given if the Hb is <70g/L in asymptomatic patients or 70–100g/L in symptomatic patients." Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 12 hours Duration: 30 days Dalteparin group: Intervention: Dalteparin sodium (2500 IU, (subcutaneous) once daily Starting time: after surgery: 12 hours Duration: 30 days ASA group: Intervention: ASA 100 mg once daily Starting time: after surgery: 12 hours Duration: 30 days
Outcomes	Number of outcomes: According to protocol: no protocol is available Reported: 7 Primary outcomes included: post‐treatment drainage volume thrombotic complication rate Secondary outcomes included: haematological parameters blood transfusion rate TBL (total blood loss) intraoperative blood loss bleeding complication rate Reported outcomes relevant to this review: Blood transfusion rate: Definition: "blood transfusion rate was calculated by dividing the number of patients receiving blood transfusions in each group by the total number of patients." Time points measured: 1d (treatment) Time points reported: 1d (treatment) Intraoperative blood loss: Definition: "Intraoperative blood loss was determined by the net increase in gauze used intraoperatively plus the amount of fuid [sic] in the drainage bottle at the end of the procedure minus the amount of fuid [sic] used for fushing." Time points measured: 1d (treatment) Time points reported: 1d (treatment) Bleeding complications: Definition: Bleeding complications included bleeding from the incision, subcutaneous petechiae, and bleeding from other parts of the body. Time points measured: not stated Time points reported: not stated Post‐treatment drainage volume: Definition: Post‐treatment drainage volume was defined as the total amount of fluid drained from the drainage tube 24h post‐treatment (including the 24th h). Time points measured: 1d (treatment) Time points reported: 1d (treatment) Thrombotic complication rate: Definition: thrombotic complication rate was determined by dividing the number of patients with thrombosis in each group by the total number of patients. Patients underwent Doppler ultrasonography prior to TKA to assess the presence or absence of DVT. Moreover, patients received lower extremity arterio‐venous ultrasound on days 3, 7, 30, 60 and 90 postoperatively, and DVT was defined if the results were positive. If PE‐associated symptoms were found, the diagnosis of PE was confirmed by the positive results of CT pulmonary angiography (CTPA). Time points measured: 3d‐90d (treatment) Time points reported: not defined. RESULTS Rivaroxaban group Postoperative distal DVT: 4 /60 patients Postoperative proximal DVT: 0 /60 patients Pulmonary embolism: 0/60 patients Major bleeding: 1/60 (bleeding of digestive tract) Minor bleeding: 15/60 (subcutaneous ecchymosis 10 or Incision bleeding 5) Intraoperative blood loss: (269.7±SD88.0) Drainage volume after treatment: (243.4±SD72.5) Dalteparin group: Postoperative distal DVT: 4 /60 patients Postoperative proximal DVT: 1 /60 patients Pulmonary embolism: 0/60 patients Major bleeding: 0/60 (bleeding of digestive tract) Minor bleeding: 6/60 (subcutaneous ecchymosis 4 or Incision bleeding 2) Intraoperative blood loss: (274.1±SD87.3) Drainage volume after treatment: (295.4±SD72.5) Aspirin group: Postoperative distal DVT: 7 /60 patients Postoperative proximal DVT: 0 /60 patients Pulmonary embolism: 0/60 patients Major bleeding: 1/60 (bleeding of digestive tract) Minor bleeding: 4/60 (subcutaneous ecchymosis 1 or Incision bleeding 3) Intraoperative blood loss: (256.4±SD83.5) Drainage volume after treatment: (205.2±SD69.0)
Notes	Source of funding: authors state that this study did not receive any external funding. Conflicts of interest: authors declared no potential conflicts of interest Published protocol: not found Observations: first time registered at the Chinese Clinical Trials Registry (ChiCTR2200060169) after study initiated. The stated randomisation method differs between the Chinese trial registration form (random number table method by Excel) and the published clinical trial. Missing data requested from authors: none Missing data obtained from authors: none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated random sequence by Randomization.com prepared by a statistician who was not involved in this clinical trial.
Allocation concealment (selection bias)	Low risk	Quote: "The randomization assignments were placed within sequentially numbered opaque sealed envelopes in the custody of a certified study pharmacist. At the end of the procedure, the envelopes were opened, and the corresponding medications were handled by an investigator who was blinded to the patient’s care."
Blinding of participants and personnel (performance bias)	Low risk	Double‐blind, double‐dummy; groups blinded: trial participants, outcome assessors, and data collectors were blinded to allocation.
Blinding of outcome assessment (detection bias)	Low risk	Outcome assessors, and data collectors were blinded to allocation. Statistician was not involved in this clinical trial.
Incomplete outcome data (attrition bias) Benefit outcomes	Low risk	No exclusions nor withdrawals were reported.
Incomplete outcome data (attrition bias) Harm outcomes	Low risk	No exclusions nor withdrawals were reported.
Selective reporting (reporting bias)	Unclear risk	No published protocol was found
Other bias	Unclear risk	There may be a risk of bias for an inappropriate co‐intervention, but there is either insufficient information to assess whether an important risk of bias exists, or insufficient rationale or evidence that an identified problem will introduce bias.

Zou 2014.

Study characteristics
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: not mentioned; groups blinded: none Study duration: July 2011 to July 2013 Duration of intervention: 14 days Duration of follow‐up: 2 weeks Method of randomisation: computer‐generated random number table (in Microsoft Office Excel 2010) Method of concealment of allocation: not mentioned Primary efficacy analysis set: not defined Primary safety analysis set: not defined Withdrawals: not reported Exclusions: not reported
Participants	Diagnosis: scheduled to TKA, unilateral, primary Location of participants: 1 country (China) Number of participants randomised: 324 Age (years; mean and range): (all randomised population) rivaroxaban: 63.5 (50 to 82); LMWH: 65.7 (54 to 80); aspirin: 62.7 (47 to 79) Gender (male %): (all randomised population) rivaroxaban: 32 (31.4%); LMWH: 20 (17.9%); aspirin: 28 (25.5%) Baseline imbalances: slight differences in gender among the groups Inclusion criteria: Diagnosed with knee osteoarthritis Initially underwent unilateral TKA DTV‐negative according to the preoperative colour Doppler ultrasonography on the deep veins of both lower extremities Gave informed consent for the therapeutic schedule Exclusion criteria: History of hemorrhagic disease or a bleeding tendency during the preoperative coagulation test Medical history of VTE Infused with over 2000 mL of fluids 24 hours after surgery Underwent knee arthroplasty Used a combination of other drugs that might impact the findings
Interventions	Number of intervention groups: 3 Concomitant interventions: none Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg twice daily Starting time: after surgery: 12 hours Duration: 14 days LMWH group: Intervention: LMWH (oral) 40 mg once daily Starting time: after surgery: 12 hours Duration: 14 days Aspirin group: Intervention: aspirin (subcutaneous) 100 mg once daily Starting time: after surgery: 12 hours Duration: 14 days
Outcomes	Number of outcomes according to protocol: no protocol is available Symptomatic VTE: Definition: incidence of symptomatic VTE Time points measured: 14 days (treatment), 2 weeks (additionally, follow‐up) Time points reported: 14 days (treatment), 2 weeks (additionally, follow‐up) RESULTS Symptomatic VTE: Summary data (14 days): see Analysis 9.3 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none Summary data (2 weeks' follow‐up): see Analysis 1.3 ‐Analysis set, type of analysis and missing participants: all randomised participants, ITT analysis, no missing participants ‐Subgroups reported: none
Notes	Source of funding: not stated Conflicts of interest: none Published protocol: none Observations: none Missing data requested from authors: type pf blinding, specific groups blinded, description of method of concealment of allocation, definition of primary efficacy and safety, details about withdrawals and exclusion, source of funding, protocol of the study Missing data obtained from authors: none For the efficacy set, the number of participants included in the analyses was not reported. For the safety set, the number of participants included in the analyses was not reported. Study did not report a CONSORT flow diagram, nor provide the number of participants initially randomised to each group. No missing outcome data provided for each group.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random sequence was computer‐generated
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias)	Unclear risk	Blinding of the study was not stated. No blinding of any group was reported
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of outcome assessment was not mentioned. Blinding of the data analysts was not mentioned
Incomplete outcome data (attrition bias) Benefit outcomes	Unclear risk	Insufficient information to detect if missing outcome data existed.
Incomplete outcome data (attrition bias) Harm outcomes	Unclear risk	Insufficient information to detect if missing outcome data existed.
Selective reporting (reporting bias)	Unclear risk	Published protocol was not found
Other bias	Low risk	The study appears to be free of other sources of bias.

AE: adverse event
ALT: alanine transaminase
AST: aspartate transaminase
CRNM: clinically relevant non‐major
DBP: diastolic blood pressure
DVT: deep venous thrombosis
GI: gastrointestinal
h: hours
Hb: haemoglobin
HR: hip replacement
ICU: intensive care unit
INR: international normalised ratio
IPC: intermittent pneumatic compression
ITT: intention‐to‐treat
LMWH: low molecular weight heparin
mITT: modified intention‐to‐treat
NR: not reported
NSAID: non‐steroidal anti‐inflammatory drug
PE: pulmonary embolism
RBC: red blood cell
SBP: systolic blood pressure
THA: total hip arthroplasty
THR: total hip replacement
TKR: total knee replacement
TXA: tranexamic acid
UFH: unfractionated heparin
ULN: upper limit of normal
VTE: venous thromboembolism

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Çiçek 2021	Ineligible study design: retrospective
CTRI/2011/07/001881	Ineligible study design
Deng 2020	Ineligible study design
DRIVE 2008	Ineligible intervention: SR123781A (anti‐factor Xa and IIa)
Eriksson 2000	Ineligible intervention
Fuji 2010	Ineligible comparator: placebo
Jiang 2018	Ineligible population: people with venous thromboembolism
Kwong 2007	Ineligible study design: economic study of RECORD3 2008
Kwong 2008	Ineligible study design: economic study of RECORD4 2009
Velik‐Salchner 2011	Ineligible study design: prospective observational study
Verhamme 2013	Ineligible comparator: factor VIII antibody
Zhou 2019	Ineligible intervention: enoxaparin included in both treatment arms

Characteristics of studies awaiting classification [ordered by study ID]

Hu 2022.

Methods	Aim: investigate the influence of enoxaparin and rivaroxaban on postoperative anticoagulation and bone metabolism in elderly patients undergoing joint replacement Design: not stated. Control group and observation group Blinding: not stated Study duration: not stated Duration of intervention: January 2020 to January 2021 Duration of follow‐up: not mentioned Method of randomisation: randomised. Not described Method of concealment of allocation: not stated Primary efficacy analysis set: not stated Primary safety analysis set: not stated Withdrawals: not stated Exclusions: not stated
Participants	Diagnosis: joint replacement Location of participants: China Number of participants randomised: 60 Age (years): elderly patients. Not stated Gender (male %): not stated Baseline imbalances: not stated Inclusion criteria: not stated Exclusion criteria: not stated Interventions: control group and observation group Number of intervention groups: 02 Concomitant interventions: not stated Excluded interventions: not stated
Interventions	Enoxaparin group Intervention: enoxaparin after surgery: no mention of doses Starting time: not stated. Duration: not stated. Rivaroxaban group Intervention: rivaroxaban after surgery: no mention of doses Starting time: not stated Duration: not stated
Outcomes	Number of outcomes: 01 According to protocol: none. No protocol is available Reported: none Primary outcomes included: not stated Secondary outcomes included: not stated Reported outcomes relevant to this review: no information available Outcomes: incidence of thrombosis Definition: not stated Time points measured: not stated Time points reported: not stated No results available
Notes	Source of funding: not mentioned Conflicts of interest: not mentioned Published protocol: not stated Observations: abstract with no data Missing data requested from authors: none Missing data obtained from authors: none

Wang 2021.

Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: double‐blind, double‐dummy; groups blinded: not detailed Study duration: 5 weeks Duration of intervention: 5 weeks Duration of follow‐up: not mentioned Method of randomisation: not mentioned Method of concealment of allocation: not mentioned Primary efficacy analysis set: not mentioned Primary safety analysis set: not mentioned Withdrawals: not reported Exclusions: not reported
Participants	Diagnosis: scheduled for primary unilateral hip arthroplasty Location of participants: 1 country (China) Number of participants randomised: 50 Age (years): not mentioned Gender (male %): not mentioned Baseline imbalances: not mentioned Inclusion criteria: not mentioned Exclusion criteria: not mentioned
Interventions	Number of intervention groups: 2 Concomitant interventions: not mentioned Excluded interventions: not mentioned Apixaban group: Intervention: apixaban (oral) 5 mg once daily Starting time: after surgery: 6 hours Duration: 5 weeks LMWH group: Intervention: low molecular weight heparin (subcutaneous) 4100 units once daily Starting time: after surgery: 6 hours Duration: 2 weeks
Outcomes	Number of outcomes: According to protocol: no protocol is available Reported: not mentioned Reported outcomes relevant to this review: not mentioned Time points measured: not mentioned Time points reported: not mentioned RESULTS Not mentioned. Authors stated: "There were no significant differences between two groups (p > 0.05)."
Notes	Source of funding: not mentioned Published protocol: not found Missing data requested from authors: none Missing data obtained from authors: none Unable to obtain a copy of study paper; only the abstract was available from a web page with no validated security certificates.

Wei 2008.

Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: double‐blind, double‐dummy; groups blinded: not detailed Study duration: not reported Duration of intervention: 10 to 39 days Duration of follow‐up: not mentioned Method of randomisation: no description was provided Method of concealment of allocation: not mentioned Primary efficacy analysis set: modified ITT population, not further specified Primary safety analysis set: not reported Withdrawals: not reported Exclusions: not reported
Participants	Diagnosis: scheduled to THA Location of participants: 1 country (China) Number of participants randomised: 325 Age (years): not reported Gender (male %): not reported Baseline imbalances: not reported Inclusion criteria: not reported Exclusion criteria: not reported
Interventions	Number of intervention groups: 2 Concomitant interventions: none Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 6 to 8 hours Duration: 35 ± 4 days Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: before surgery: evening Duration: 12 ± 2 days
Outcomes	Number of outcomes: According to protocol: no protocol is available Reported: not reported OUTCOMES No outcome relevant to this review was reported RESULTS No outcome relevant to this review was reported
Notes	Source of funding: not stated Conflicts of interest: none Published protocol: none Observations: none Abstract only

ITT: intention‐to‐treat; LMWH: low molecular weight heparin; THA: total hip arthroplasty; TKR: total knee replacement

Characteristics of ongoing studies [ordered by study ID]

ChiCTR1800016829.

Study name	Administering aspirin and rivaroxaban to prevent deep venous thrombosis after total hip arthroplasty: a prospective randomised control trial
Methods	Aim: to evaluate the effect of different anti‐coagulation drugs after THR Design: parallel‐group Blinding: not reported Study duration: June 2018 to November 2018 Duration of intervention: not reported Duration of follow‐up: not reported Method of randomisation: computer‐generated random sequence Method of concealment of allocation: not reported Primary efficacy analysis set: not reported Primary safety analysis set: not reported
Participants	Diagnosis: scheduled to THR Location of participants: 1 country (China) Number of participants randomised: 240 Inclusion criteria Unilateral total hip arthroplasty for osteoarthritis Willing to participate in study and sign informed consent form Exclusion criteria History of haemorrhagic disease or a bleeding tendency during the preoperative coagulation test Medical history of VTE or lower limb atherosclerosis, or used a combination of other drugs that might impact the findings
Interventions	Number of intervention groups: 3 Concomitant interventions: not reported Excluded interventions: not reported Aspirin group: Intervention: aspirin, dose not reported Starting time: not reported Duration: not reported LMWH group: Intervention: LMWH, dose not reported Starting time: not reported Duration: not reported Rivaroxaban group: Intervention: rivaroxaban, dose not reported Starting time: not reported Duration: not reported
Outcomes	Number of outcomes according to protocol: 4 OUTCOMES PT (prothrombin time) APTT (activated partial thromboplastin time) INR factor X antibody: Definition: not reported Time points measured: not reported DVT: Definition: not reported Time points measured: not reported Complication: Definition: not reported Time points measured: not reported Function: Definition: not reported Time points measured: not reported
Starting date	30 June 2018
Contact information	Author: Luo Zeyu Address: 37 Guoxuexiang, Chengdu, Sichuan, China Email: dr_zeyu@163.com
Notes	Source of funding: self‐funded Conflicts of interest: not reported Published protocol: available in www.chictr.org.cn Observations: ethics approval from 1990, recruiting ended in November 2018 but no published results yet We discovered a different code, ChiCTR1800016830, for the same study. The only difference we noticed is that the scheduled surgery under this code is TKR instead of THR.

IRCT20161121031003N4.

Study name	A clinical trial to compare the effect of rivaroxaban with enoxaparin on the prophylaxis of venous thromboembolism in patients undergoing knee or hip arthroplasty surgery
Methods	Aim: not specified Design: parallel group Blinding: single blinded Study duration: ongoing (not reported) Duration of intervention: 14 days for enoxaparin; 12 days for rivaroxaban Duration of follow‐up: 24 weeks Method of randomisation: sequentially numbered sealed opaque envelopes will be used to conceal the sequencing Method of concealment of allocation: not reported Primary efficacy analysis: not reported Primary safety analysis: not reported
Participants	Diagnosis: TKA or THA Location: 1 country (Iran) Number of participants randomised: target size 150 Inclusion criteria: > 18 years of age; candidate for hip or knee arthroplasty surgery Exclusion criteria: Severe side effects of a drug Moderate to severe hepatic disorder Cardiovascular diseases Age minimum: 18 years Age maximum: no limit Gender: both
Interventions	Number of intervention groups: 2 Concomitant interventions: ASA (acetylsalicylic acid/aspirin) use for 14 days after exposure to study drug Excluded interventions: none Enoxaparin group Intervention: enoxaparin 40 mg once daily SQ Starting time: not reported Duration: 14 days Rivaroxaban group Intervention: rivaroxaban 10 mg oral daily Starting time: not reported Duration: 12 days
Outcomes	Number of outcomes according to protocol: 6 Deep vein thrombosis Incidence of DVT Time points: 6, 12, and 24 weeks after the intervention. Method of measurement: Doppler ultrasound. Skin complications Time points: 6, 12, and 24 weeks after the intervention. Method of measurement: questionnaire. Severe bleeding Time points: 6, 12, and 24 weeks after the intervention. Method of measurement: questionnaire. The side effects of wounds Time points: 6, 12, and 24 weeks after the intervention. Method of measurement: questionnaire. Gastrointestinal complications Time points: 6, 12, and 24 weeks after the intervention. Method of measurement: questionnaire. Wound infection Time points: 6, 12, and 24 weeks after the intervention. Method of measurement: questionnaire.
Starting date	18 March 2020
Contact information	Mostafa Salehpour Shafa Yahyaian Hospital, Mojahedin Eslam Street 4816633131, Tehran, Iran (Islamic Republic of) Telephone: +98 21 3354 2001 Email: gharanizadeh.k@iums.ac.ir Affiliation: Iran University of Medical Sciences
Notes	https://en.irct.ir/trial/38758

IRCT20181128041784N1.

Study name	Comparison of the effectiveness of 4 drugs (aspirin 80, aspirin 325, enoxaparin and rivaroxaban) in preventing coagulation disorders after knee replacement surgery
Methods	Aim: superiority Design: parallel group Blinding: not blinded Study duration: ongoing (not reported) Duration of intervention: 3 weeks for enoxaparin; 3 weeks for rivaroxaban Duration of follow‐up: 6 weeks Method of randomisation: block randomisation method Method of concealment of allocation: not reported Primary efficacy analysis: not reported Primary safety analysis: not reported
Participants	Diagnosis: scheduled for TKR Location of participants: 1 country (Iran) Number of participants randomised: target size 124 Inclusion criteria "Patients with a definite diagnosis of arthritis, rheumatism or destruction of the joint surface, patients who are candidates for knee joint replacement surgery due to pain and severe movement limitation" Exclusion criteria "History of skin, respiratory or allergy to aspirin, history of taking anticoagulants and bleeding diseases, history of malignancy, history of liver and kidney diseases" No age limit
Interventions	Number of intervention groups: 4 Concomitant interventions: not reported Excluded interventions: none Enoxaparin group Intervention: enoxaparin 40 mg daily subcutaneously Starting time: not reported Duration: 3 weeks Rivaroxaban group Intervention: rivaroxaban 10 mg daily oral Starting time: not reported Duration: 3 weeks Aspirin 80 group Intervention: aspirin 80 mg twice a day oral Starting time: not reported Duration: 3 weeks Aspirin 325 group Intervention: aspirin 325 mg daily oral Starting time: not reported Duration: 3 weeks
Outcomes	Number of outcomes according to protocol: 3 Deep vein thrombosis Measurement periods at 1, 2 and 6 weeks By venous Doppler ultrasound Pulmonary thromboembolism Measurement periods at 1, 2 and 6 weeks By computed tomography angiography Bleeding Not specified
Starting date	23 December 2018
Contact information	Hamedan University of Medical Sciences Gholamreza Ghorbani Amjad Resalat Square, Ayatollah Motahari Boulevard, Ba'ath Hospital Hamedan, Iran Phone +98 81 3264 0020 Fax +98 81 3265 1515
Notes	Study protocol in Arabic; translation to English by Google web translator

IRCT20190325043107N16.

Study name	Comparison the Injectual Enoxaparin with Oral Rivaroxaban in DVT Prophylaxis of Patients with Pre‐Trochanteric Hip Fracture
Methods	Aim: prevention, efficacy Design: control and parallel group Blinding: open study Study duration: 2020‐01‐05 to 2021‐01‐04 Duration of intervention: not mentioned Duration of follow‐up: not mentioned Method of randomisation: "All patients will be randomly assigned to one of the two groups of enoxaparin or rivaroxaban with the help of an online randomization list" Method of concealment of allocation: not mentioned Primary efficacy analysis set: not reported Primary safety analysis set: not reported
Participants	Diagnosis: acute embolism and thrombosis of deep veins of lower extremity Location of participants: not mentioned Number of participants randomised: 80 Inclusion criteria: presence of peri‐trochanteric femoral fractures Exclusion criteria: heparin sensitivity; endocarditis; liver disease; haematologic diseases
Interventions	Number of intervention groups: 2 Concomitant interventions: not mentioned Excluded interventions: none Intervention group: "in the intervention group, after surgery, they will receive 40 mg of enoxaparin subcutaneously for 30 days (one daily). The enoxaparin ampoule will be purchased from Iran Daroo Company ready and given by a nurse every day. Blood will be taken from the patient before the intervention and on the last day of the intervention and serum D‐Dimer levels will be measured." Control group: "patients will be enrolled after random assignment. In the intervention group, after surgery, 10 mg rivaroxaban will be given orally for 30 days (one daily). Rivaroxaban tablets will be purchased from Iran Daroo Company and will be administered daily by a nurse. Blood will be taken from the patient before the intervention and on the last day of the intervention and serum D‐Dimer levels will be measured."
Outcomes	Number of outcomes according to protocol: not mentioned
Starting date	5/01/2020
Contact information	Principal investigator Mohammad Bazavar Tabriz University of Medical Sciences, Azadi Ave
Notes	Source of funding: not mentioned Conflicts of interest: the authors declare no conflicts of interest. Published protocol: available in IRCT Iranian Registry of Clinical Trials Observations: not mentioned

JPRN‐UMIN000026819.

Study name	A randomized controlled trial of low‐dose Enoxaparin versus low‐dose Edoxavan for prevention of venous thromboembolism after total hip arthroprasty in elderly or underweight or moderate renal insufficient patients.
Methods	Aim: ageing population, with low body weight or renal dysfunction. Safety Design: parallel group Blinding: not reported Study duration: November 2016 to November 2019 Duration of intervention: 10 days after surgery Duration of follow‐up: not reported Method of randomisation: not reported Method of concealment of allocation: not reported Primary efficacy analysis set: not reported Primary safety analysis set: not reported
Participants	Diagnosis: osteoarthritis of the hip, osteonecrosis of the femoral head, femoral neck fracture, rheumatoid arthritis, rapidly destructive coxarthropathy, septic arthritis of the hip Location of participants: 1 country (Japan) Number of participants randomised: 80 Number of participants enroled: 76 Inclusion criteria Arthroplasty for hip disease Over 20 years of age Elderly; over 75 years old and/or Low weight; 40 kg < BW < 50 kg and/or Moderate renal function disorder; 30 mL/min < creatinine clearance rate < 50 mL/min Exclusion criteria Creatinine clearance rate﹤30 mL/min Common use of anticoagulant, antiplatelet and the P glucoprotein repressor Perioperative amount of bleeding more than 600 mL Allergic to iodine contrast agent Under 20 years old Inappropriate as a study participant subject
Interventions	Number of intervention groups: 2 Concomitant interventions: not mentioned Excluded interventions: not mentioned Edoxaban group: Intervention: edoxaban (oral) 15 mg once daily Starting time: not reported Duration: 10 days Enoxaparin group: Intervention: enoxaparin 2000 international units per day Starting time: after surgery Duration: 10 days
Outcomes	Not reported Number of outcomes: not reported
Starting date	21 November 2016
Contact information	Tomonori, Tetsunaga Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Sciences and Okayama University Hospital, Ethics Committee, 2‐5‐1, Shikata‐cho, Kita‐ku, Okayama City, Okayama,700‐8558, Japan. 086‐235‐650. mae6605@adm.okayama‐u.ac.jp
Notes	Funding source: none Conflicts of interest: not reported Published protocol: available in the Japanese Clinical Trials Registry Observations: last follow‐up: 30 November 2019

JPRN‐UMIN000033422.

Study name	Elucidation of individual difference factors in the pharmacokinetics and clinical effects of anticoagulants edoxaban and enoxasaparin [sic]
Methods	Aim: efficacy (presence of thrombus by computed tomography imaging) Design: randomised Blinding: open Study duration: 4 June 2015 to 7 August 2018 Duration of intervention: not reported Duration of follow‐up: not reported Method of randomisation: not reported Method of concealment of allocation: not mentioned Primary efficacy analysis set: not reported Primary safety analysis set: not reported
Participants	Diagnosis: TKA Location of participants: 1 country (Japan) Number of participants randomised: 100 Inclusion criteria Undergoing knee replacement Over 20 years of age Exclusion criteria History of DVT and PTE Disease with abnormal coagulopathy No written informed consent Contraindicated for edoxaban or enoxaparin Taking warfarin Other people judged inappropriate by the attending physician
Interventions	Number of intervention groups: 2 Concomitant interventions: not reported Excluded interventions: not reported Edoxaban group: Intervention: not reported Starting time: not reported Duration: not reported Enoxaparin group: Intervention: not reported Starting time: not reported Duration: not reported
Outcomes	Not reported Number of outcomes: not reported
Starting date	04 June 2015
Contact information	Principal investigator: Kaneko, Hironori Kitasato University, Kitasato Institute Hospital. 5‐9‐1 shirokane minato‐ku tokyo. 03‐3444‐616. hiroka@insti.kitasato‐u.ac.jp
Notes	Funding source: Kitasato University Conflicts of interest: not reported Published protocol: available in the Japanese Clinical Trials Registry. Unpublished Observations: Last follow‐up date 07 August 2018. Registered date 17 July 2018. Last modified 19 March 2021

NCT00408239.

Study name	PEARL‐1 trial (Factor Xa inhibitor for prevention of venous thromboembolism in patients undergoing elective total knee replacement)
Methods	Aim: dose‐escalation study, efficacy and safety Design: parallel‐group Blinding: open‐label Study duration: December 2006 to March 2009 Duration of intervention: not reported Duration of follow‐up: 2 weeks Method of randomisation: not described Method of concealment of allocation: not mentioned Primary efficacy analysis set: not reported Primary safety analysis set: not reported
Participants	Diagnosis: scheduled to TKR Location of participants: 8 countries (Indonesia, Japan, Republic of Korea, Malaysia, Philippines, Singapore, Taiwan, Thailand) Number of participants randomised: 367 Inclusion criteria Scheduled for elective primary total knee replacement surgery Legal minimum age requirement (country‐specific) Written informed consent has been obtained Exclusion criteria Documented history or considered to be at increased risk of VTE Considered to be at increased risk of bleeding: Known haemorrhagic disorder and/or coagulation disorder Thrombocytopaenia Clinically important bleeding occurred within 3 months prior to the screening visit Acute bacterial endocarditis Severe hypertension Retinopathy Concomitant use of anticoagulants / antiplatelet agents (including homeopathic drugs) and/or anticipated postoperative need for other reasons than prevention of DVT during the study period.
Interventions	Number of intervention groups: 3 Concomitant interventions: none Excluded interventions: none YM150 dose 1 group: Intervention: YM150 (oral), dose not reported Starting time: not reported Duration: not reported YM150 dose 2 group: Intervention: YM150 (oral), dose not reported Starting time: not reported Duration: not reported Enoxaparin group: Intervention: enoxaparin (subcutaneous), dose not reported Starting time: not reported Duration: not reported
Outcomes	Number of outcomes according to protocol: 4 All‐cause mortality: Definition: incidence of death from any cause Time points measured: 2 weeks Major bleeding: Definition: not reported Time points measured: 2 weeks
Starting date	December 2006
Contact information	Principal investigator: not reported
Notes	Source of funding: Astellas Pharma Inc Conflicts of interest: not reported Published protocol: available in ClinicalTrials.gov Observations: actual study completion date reported in March 2009, but no published results were found

NCT02085824.

Study name	NCT02085824
Methods	Aim: compare rivaroxaban and enoxaparin in terms of preventing DV and PE, and also in this study, we compared the complications due to these drugs in patients undergoing elective arthroplasty. Design: parallel assignment Blinding: double (Investigator; outcomes assessor) Study duration: recruiting. "Actual Study Start Date July 2012. Estimated Study Completion Date: April 2014" Duration of intervention: not mentioned Duration of follow‐up: not mentioned Method of randomisation: not mentioned Method of concealment of allocation: not mentioned Primary efficacy analysis set: not mentioned Primary safety analysis set: not mentioned Withdrawals: not mentioned Exclusions: not mentioned
Participants	Diagnosis: total hip replacement (THR) Location of participants: Poland Number of participants randomised: estimated 60 Age (years): 18 years and older Gender (male %): not stated Baseline imbalances: not mentioned Inclusion criteria: primary, end‐stage hip osteoarthritis requiring total hip arthroplasty Exclusion criteria: inflammatory arthropathies, liver disorders, neoplastic conditions, clotting disorders
Interventions	Number of intervention groups: 03 Concomitant interventions: not mentioned Excluded interventions: not mentioned Enoxaparin group Intervention: enoxaparin 40 mg subcutaneously daily, once pre‐operatively Starting time: after surgery: once pre‐operatively and then daily Duration: 28 days Rivaroxaban group Intervention: rivaroxaban 10 mg given orally daily Starting time: after surgery Duration: 28 days Dabigatran group Intervention: dabigatran 110 mg given orally daily Starting time: postoperatively and then twice daily Duration: 27 days after the surgery Outcomes Number of outcomes: 03 According to protocol: protocol no available Reported: not stated Primary outcomes included: not stated Secondary outcomes included: not stated
Outcomes	Reported outcomes relevant to this review: 01 "OUTCOME 1: total blood loss by the 3rd postoperative day Definition: measured for each individual. Measured according to the formula described by Nadler, Hidalgo and Bloch (Prediction of blood volume in normal human adults. Surgery 1962;51:224‐32)" Time points measured: 3rd day postoperatively Time points reported: not reported "OUTCOME 2: drop in the haemoglobin value between 3rd day postop and preoperative Definition: difference between haemoglobin on the 3rd postoperative day and day before operation" Time points measured: 3rd day postoperatively Time points reported: not reported "OUTCOME 3: wound healing disturbances according to the definition of Centers for Disease Control and Prevention Definition: wound healing disturbances ‐ diagnosis to be made according to the definition of Centers for Disease Control and Prevention (Mangram, Horan, Pearson, Silver and Jarvis. Guideline for prevention of surgical site infection, 1999. Hospital Infection Control Practices Advisory Committee. Infect Control Hosp Epidemiol. 1999 Apr;20(4):250‐78)." Time points measured: 3 months postoperatively Time points reported: not reported
Starting date	July 2012
Contact information	Marcin K Wasko, MD, PhD +48227794031 ext 384 m.wasko@cmkp.edu.pl
Notes	Source of funding: Centre of Postgraduate Medical Education Conflicts of interest: not mention Published protocol: registered in ClinicalTrials.gov as NCT02085824 Observations: actual study completion date reported as April 2014, but no published results were found Missing data requested from authors: none Missing data obtained from authors: none

NCT02379663.

Study name	Prophylaxis of deep vein thrombosis following total hip arthroplasty
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: triple (participants, care provider, outcomes adjudicator) Study duration: January 2012 to January 2013 Duration of intervention: not reported Duration of follow‐up: 2 weeks Method of randomisation: not described Method of concealment of allocation: not mentioned Primary efficacy analysis set: not reported Primary safety analysis set: not reported
Participants	Diagnosis: scheduled to THA Location of participants: not reported Number of participants randomised: 639 Inclusion criteria Male or female without childbearing potential aged ≥ 20 years who were scheduled for elective primary THA Exclusion criteria Recent history of active bleeding or VTE Known genetic disorder associated with bleeding tendency or any condition related with an increased risk of bleeding Persistent blood pressure of ≥ 160 mmHg systolic and/or ≥ 100 mmHg diastolic at baseline Myocardial infarction or cerebrovascular accident within three months of the scheduled surgery Major surgery in the prior 3 months Renal insufficiency with a creatinine clearance < 60 mL/min, hepatic failure combined with coagulopathy, or thrombocytopaenia (platelets < 100,000/mm³), or planned indwelling epidural catheter for > 6 hours after the end of surgery
Interventions	Number of intervention groups: 3 Concomitant interventions: none Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: not reported Duration: not reported Enoxaparin group: Intervention: enoxaparin (subcutaneous) 40 mg once daily Starting time: not reported Duration: not reported Placebo group: Intervention: normal saline (subcutaneous) 1cc once daily Starting time: not reported Duration: not reported
Outcomes	Number of outcomes according to protocol: 1 OUTCOMES No outcome relevant to this review was reported
Starting date	January 2012
Contact information	Principal investigator: not reported
Notes	Source of funding: Samsung Medical Center Conflicts of interest: not reported Published protocol: available in ClinicalTrials.gov Observations: actual study completion date reported as January 2013, but no published results were found

NCT03088358.

Study name	Safety and efficacy of TeaRx Xa factor direct inhibitor versus enoxaparin as a venous thromboembolic events (VTE) prevention following total knee replacement
Methods	Aim: efficacy and safety Design: parallel assignment Masking: double (participants, investigator) Blinding: not reported Study duration: August 2013 to September 2015 Duration of intervention: 12 days Duration of follow‐up: not mentioned Method of randomisation: not mentioned Method of concealment of allocation: not mentioned Primary efficacy analysis set: not reported Primary safety analysis set: not reported
Participants	Diagnosis: TKR Location of participants: Russian Number of participants randomised: 200 Inclusion criteria Male or female, age ≥ 18 years; Planned total knee replacement surgery; Signed informed consent form; Willing to comply with the protocol; Willing to use adequate contraception during the trial. Exclusion criteria Surgery for acute fracture 4 weeks before screening; history of septic inflammation in the joint; prosthesis revision or one leg missing History of venous thrombosis of any location or PE History of heparin induced thrombocytopenia or other thrombocytopathy; hemorrhagic diathesis History of evident coagulopathy or in a relative Congenital thrombophilia Bleeding within 6 months of screening; increased risk of bleeding Body mass index less than 18.5 or more than 40 kg/m2 Systolic blood pressure > 180 mmHg and/or diastolic blood pressure > 110 mmHg registered twice within 15 to 30 minutes Hb ≤ 10.5 g/dL in females or ≤ 11.5 g/dL in males Platelets < 100 000/mm3 Clinical significant abnormalities of activated partial thromboplastin time (aPTT) and/or international normalised ratio (INR) Glomerular filtration rate (GFR) < 30 mL/min/1.73 m2 ALT or AST ≥ 2 x ULN or total bilirubin ≥ 1.5 x ULN
Interventions	Number of intervention groups: 4 Concomitant interventions: not mentioned Excluded interventions: not mentioned Experimental: TeaRx 50 mg TeaRx: 50 mg per day orally (active 25 mg + placebo 50 mg every 12 hours) for 12 days (± 2 days) Experimental: TeaRx 100 mg TeaRx: 100 mg per day orally (placebo 25 mg + active 50 mg every 12 hours) for 12 days (± 2 days) Experimental: TeaRx 150 mg TeaRx: 150 mg per day orally (active 25 mg + active 50 mg every 12 hours) for 12 days (± 2 days) Active comparator: enoxaparin Enoxaparin 40 mg subcutaneously per day (24‐hour interval) for 12 days (± 2 days)
Outcomes	Number of outcomes according to protocol: 2 "Primary Outcome Measures: 1. Incidence of deep venous thrombosis (DVT) (efficacy of the selected TeaRx dose) [ Time Frame: 6 weeks following total knee replacement ] 2. Incidence of nonfatal pulmonary embolism (PE) (efficacy of the selected TeaRx dose) [ Time Frame: 6 weeks following total knee replacement ] 3. Incidence of symptomatic venous thromboembolic events (DVT, PE) (efficacy of the selected TeaRx dose) [ Time Frame: 6 weeks following total knee replacement ] 4. venous thromboembolic events (VTE) caused mortality (efficacy of the selected TeaRx dose) [ Time Frame: 6 weeks following total knee replacement ] 5. Non‐VTE caused mortality (efficacy of the selected TeaRx dose) [ Time Frame: 6 weeks following total knee replacement ] 6. AEs and SAEs from subject complaints, physical examination, vital signs, laboratory results (efficacy of the selected TeaRx dose) [ Time Frame: 6 weeks following total knee replacement Secondary Outcome Measures : 1. Incidence of bleeding (safety of selected TeaRx dose) [ Time Frame: 6 weeks following total knee replacement ] 2. Major and clinically relevant non‐major bleeding" Time points measured: not mentioned
Starting date	1/08/2013
Contact information	Principal investigator: Prof. Mikhail N. Zamyatin Address: not mentioned
Notes	Source of funding: not mentioned Conflicts of interest: not reported Published protocol: not mentioned Observations: Phase 2 study with no publications available. Last update posted at ClinicalTrials.gov on 23 March 2017

NCT05189002.

Study name	A multicenter randomized double‐blind placebo‐controlled prospective study to evaluate the safety and efficacy of the direct factor Xa inhibitor dimolegin (DD217) in prevention of venous thromboembolic complications during knee replacement
Methods	Aim: efficacy and safety. "A formal conclusion about superiority will be made if the lower limit of the specified confidence intervals exceeds the value of 0.0. A formal conclusion on non‐inferiority will be made if the lower limit of the specified confidence intervals exceeds the value of ‐0.05 (‐5.0 %)." Design: parallel‐group: "...in groups of 80 patients will be investigated." Blinding: triple (participant, care provider, investigator) Study duration: active. Not recruiting. "Actual Study Start Date May 31, 2019. Estimated Study Completion Date: December 31, 2022" Duration of intervention: 14 days Duration of follow‐up: 6 weeks Method of randomisation: randomised. Not described Method of concealment of allocation: "Dose Dimolegin ‐ DD217 in each group was blinded for a patient and investigator (double‐blind method) using six tablets for each administration, some of them contained Dimolegin ‐ DD217, and others were masked as Placebo of the study product." Primary efficacy analysis set: not stated Primary safety analysis set: not stated Withdrawals: not stated Exclusions: not stated
Participants	Diagnosis: knee replacement Location of participants: Russian Federation (15 locations) Number of participants randomised: "It is planned to randomize 240 patients (160 patients in two different groups of Dimolegin ‐ DD217 therapy and 80 patients in the reference group of Fragmin (INN: dalteparin). The maximum number of patients who can be included in the study at the first stage is 320." Age (years): 18 years and older Gender (male %): not stated Baseline imbalances: not stated Inclusion criteria: "Men and women aged 18 years and older who need elective primary total knee replacement (cement or cementless arthroplasty). Signed informed consent. Ability to comply with all protocol requirements. Patients' consent to use adequate methods of contraception throughout the study" Exclusion criteria: "Surgery for an acute fracture 4 weeks before screening; septic inflammation of the joint, revision of the prosthesis or the absence of one leg Venous thrombosis of any localization or a confirmed PE episode at the present time or in the medical history Heparin‐induced thrombocytopenia or other thrombocytopathies currently or in the history, hemorrhagic diathesis Obvious coagulopathy ongoing or in the history of the patient or a blood relative Congenital thrombophilia according to the medical history (deficiency of antithrombin III, protein C, protein S, Leiden mutation of coagulation factor V, increased level of coagulation factor VIII, mutation of prothrombin G20210A) Active bleeding (intracranial, intraocular, nasal, digestive or other localization) at present or within 6 months prior to screening, high risk of bleeding Collection of at least one volume unit of donated blood (> 500 mL) or blood transfusion during the previous 12 weeks Surgery on the brain or spinal cord, spine, ophthalmic or major surgery or injury in the last 90 days Gastrointestinal tract disorders that can disrupt the absorption of the study drug (Crohn's disease, ulcerative colitis, irritable bowel syndrome) Acute gastric or duodenal ulcer, erosive gastritis with increased risk of bleeding Significant cardiovascular diseases ongoing or within 6 months prior to screening, including: chronic heart failure of class III or IV (according to the classification of the New York Heart Association), myocardial infarction, unstable angina, surgery on the heart and coronary vessels (including percutaneous coronary intervention with or without coronary artery stenting), significant diseases of the heart valves, hemodynamically significant cardiac arrhythmias, transient ischemic attack, ischemic or hemorrhagic stroke, uncontrolled hypertension Active liver disease (viral hepatitis B or C, cirrhosis of the liver) and biliary tract disease, with the exception of non‐alcoholic steatohepatitis with normal levels of hepatic transaminases (ALT and AST) Nephrotic syndrome, significant kidney diseases with the events of nephrotic syndrome (decreased filtration renal function with decreased estimated glomerular filtration rate (eGFR) < 60 according to the MDRD formula (MDRD) Malignant neoplasms during the last 5 years (with the exception of basal cell carcinoma for which radical treatment was carried out). Positive test for HIV, syphilis, hepatitis B or C markers (HBsAg and Anti‐HCV) Significant drug or alcohol abuse according to the Investigator in the history or currently The development of trophic disorders of the lower extremities that do not respond to medical treatment Any condition, in which, according to the Investigator, surgical intervention or anticoagulants are contraindicated Body mass index (BMI) less than 18.5 or more than 40 kg/m2. Body weight above 130 kg Systolic BP > 180 mmHg and/or diastolic BP >110 mmHg, reported with two consecutive measurements for 15‐30 minutes Hemoglobin < 105 g/L in women or < 115 g/L in men Abnormal laboratory parameters of the coagulation system (platelets, activated partial thromboplastin time (APTT), international normalized ratio (INR) and D‐dimer), which, according to the Investigator and Medical Expert of the study, cause suspicion of blood clotting or problems in the hemostasis system in the patient eGFR < 60 mL/min/1.73 m2 (by MDRD formula) ALT or AST > 2 x upper limit of normal (ULN) or total bilirubin > 1.5 x ULN Hypersensitivity or contraindications to Dimolegin ‐ DD217, dalteparin sodium, unfractionated heparin or warfarin; pig tissue preparations, radiopaque preparations (for multislice spiral computed tomography (MSCT) with intravenous enhancing) The need for constant use of parenteral or oral anticoagulants (for example, patients with artificial heart valves, patients with atrial fibrillation who are indicated for warfarin therapy) Systemic therapy with azole group drugs (ketoconazole, fluconazole, etc.), as well as other CYP3A4 inhibitors 7 days before and during screening. Previous and concomitant therapy: taking antiplatelet drugs, therapy with vitamin K antagonists, therapy with unfractionated heparin, low molecular weight heparin (LMWH), direct oral anticoagulants, the use of NSAIDs should be stopped at least 7 days before the start of the study therapy, systemic therapy with strong CYP3A4 and P‐glycoprotein inhibitors, systemic therapy with strong inducers of CYP3A4 and P‐glycoprotein Women who are pregnant or breastfeeding Women planning pregnancy during a clinical trial (including women who received a positive pregnancy test result during screening or before taking the study drug) Women of childbearing potential (including non‐sterilized surgically and in the postmenopausal period less than 2 years) who do not want or cannot use adequate methods of contraception throughout the study. Adequate methods of contraception include the use of a condom or diaphragm (barrier method) with spermicide Participation in another clinical trial currently or within 30 days prior to screening, use of any investigational drug for 30 days or 5 half‐lives (which is longer) prior to screening Affiliation to the investigational site: close relatives of the Investigator, dependent persons (for example, an employee or a person studying at the investigational site) Inability to read or write; unwillingness to understand and follow the study protocol procedures; non‐compliance with the regimen of treatment or procedures which, according to Investigator, may affect the study results or patient's safety and prevent the patient from further participating in the study; any other concomitant medical or serious mental conditions, which make the patient ineligible for the clinical study, restrict validity of the consent or may affect the patient's ability to participate in the study"
Interventions	Interventions: parallel assignment Number of intervention groups: 3 Concomitant interventions: not stated Excluded interventions: not stated 1a‐Dimolegin 40 mg group Intervention: product once a day in doses 40 mg orally and saline solution subcutaneously. "Dose was blinded for a patient and investigator (double‐blind method) using six tablets for each administration, some of them contained Dimolegin ‐ DD217, and others were masked as Placebo of the study product." Starting time: not stated Duration: not stated 1b‐Dimolegin 60 mg group Intervention: product once a day in doses 60 mg orally and saline solution subcutaneously. "Dose was blinded for a patient and investigator (double‐blind method) using six tablets for each administration, some of them contained Dimolegin ‐ DD217, and others were masked as Placebo of the study product." Starting time: not stated Duration: not stated 2‐Fragmin group Intervention: "Fragmin in accordance with the instruction for medical use (5000 IU s/c once a day with 24‐hour interval) and six Placebo tablets masked as study product Dimolegin ‐ DD217 once a day" Starting time: not stated Duration: not stated
Outcomes	Number of outcomes: 4 According to protocol: not stated Reported: not stated Primary outcomes included: total venous thromboembolism (VTE) index (14 days) Secondary outcomes included: 2. total venous thromboembolism (VTE) index (Week 6) "Other Outcome Measures: 1. The frequency of cumulative major and clinically significant minor bleeding on Visits V1‐ V13 [ Time Frame: 14 Days ] 2. The frequency of cumulative major and clinically significant minor bleeding on Visits V1‐ V13 (Days D2‐D14±1) ‐ primary safety endpoint. 3. The frequency of cumulative major and clinically significant minor bleeding before Visit V15 [ Time Frame: 6 Weeks ] 4. The frequency of cumulative major and clinically significant minor bleeding before Visit V15 (W6±3 days) ‐ secondary safety endpoint." Reported outcomes relevant to this review: Primary outcomes: total venous thromboembolism (VTE) index Definition: "The frequency of composite endpoint, which includes: 1.1. Asymptomatic deep vein thrombosis (DVT) detected by bilateral phlebography (preferably) or USDS at Visit V13 (Day D14±1); Objectively confirmed by phlebography, USDS, computed tomography or other method symptomatic or asymptomatic DVT before Visit V13 (Day D14±1) inclusive; 1.2. Non‐fatal PE before Visit V13 (Day D14±1) inclusive; 1.3. Fatal PE before Visit V13 (Day D14±1) inclusive; 1.4 Unexplained death, in which pulmonary embolism (PE) cannot be excluded before the Visit V13 (Day D14 ± 1) inclusive." Time points measured: 14 Days Time points reported: none Secondary outcomes: total venous thromboembolism (VTE) index (Week 6) Definition: "The frequency of composite endpoint, including: 2.1 Symptomatic VTE, 2.2 non‐fatal PE, 2.3 fatal PE, 2.4 unexplained death, in which PE cannot be excluded before the Visit V15 (W6 ±3 days) inclusive; Objectively confirmed by phlebography, USDS, computed tomography or other method symptomatic or asymptomatic DVT before Visit V15 (W6±3 days) inclusive." Time points measured: 6 weeks Time points reported: none
Starting date	May 2019. Estimated study completion: 31 December 2022
Contact information	Study Chair: Dmitry A Napalkov, Professor Department of Faculty Therapy No. 1 of the Sechenov University
Notes	Source of funding: PharmaDiall Ltd. Conflicts of interest: not stated Published protocol: registered in ClinicalTrials.gov as NCT05189002 Observations: Active. Not recruiting. Actual study completion date estimated as 31 December 2022, but no published results were found Missing data requested from authors: none Missing data obtained from authors: none

Pellegrini 2022.

Study name	PEPPER trial (Comparative effectiveness of pulmonary embolism prevention after hip and knee replacement)
Methods	Aim: superiority, efficacy and safety Design: parallel‐group Blinding: open‐label Study duration: December 2016 to August 2020 (estimated) Duration of intervention: 30 days Duration of follow‐up: 6 months Method of randomisation: not described Method of concealment of allocation: not mentioned Primary efficacy analysis set: not reported Primary safety analysis set: not reported
Participants	Diagnosis: scheduled to TKR or THR Location of participants: 2 countries (USA and Canada) Number of participants randomised: 25,000 (estimated) Inclusion criteria 21 years of age or older Undergoing elective primary, revision or second stage re‐implantation total hip/knee replacement or uni‐compartmental knee replacement or hip resurfacing arthroplasty Has necessary mental capacity to participate and is able to comply with study protocol requirements Eligible for randomisation to at least two of the three study regimens Is not pregnant on the day of surgery Has signed the consent form Is willing to be randomised and participate in the study Exclusion criteria Undergoing bilateral hip or knee replacement Has been previously enroled Is pregnant or breastfeeding Is on chronic anticoagulation other than antiplatelet medications Concurrently enroled in another active interventional clinical trial testing a drug or intervention known or believed to interact with aspirin, warfarin, or rivaroxaban Has documented gastrointestinal, cerebral, or other haemorrhage within 3 months Has a known diagnosis of defective haemostasis and past history of clinical bleeding requiring transfusion and treatment Has had an operative procedure involving the eye, ear, or central nervous system within 1 month Has uncontrolled hypertension with systolic blood pressure > 220 mmHg or diastolic blood pressure > 120 mmHg Body weight of less than 41 kg at baseline visit Member of a vulnerable patient population
Interventions	Number of intervention groups: 3 Concomitant interventions: IPC Excluded interventions: none Rivaroxaban group: Intervention: rivaroxaban (oral) 10 mg once daily Starting time: after surgery: 24 hours Duration: not reported Warfarin group: Intervention: warfarin (oral) 2.5 mg to 7.5 mg according to body weight, once daily to an INR target of 2.0 Starting time: before surgery: day of surgery Duration: 30 days Aspirin group: Intervention: aspirin (oral) 162 mg as first dose, 81 mg once daily thereafter Starting time: before surgery: day of surgery Duration: 30 days
Outcomes	Number of outcomes according to protocol: 6 All‐cause mortality: Definition: incidence of death from any cause Time points measured: 6 months Major bleeding: Definition: not reported Time points measured: 6 months
Starting date	December 2016
Contact information	Principal investigator: Carol A Lambourne Medical University of South Carolina (US)
Notes	Source of funding: Medical University of South Carolina Conflicts of interest: not reported Published protocol: available in ClinicalTrials.gov Observations: none

ALT: alanine transaminase; AST: aspartate transaminase; DVT: deep vein thrombosis; Hb: haemoglobin; INR: international normalised ratio; IPC: intermittent pneumatic compression; LMWH: low molecular weight heparin; PE: pulmonary embolism; RBC: red blood cell; THA: total hip arthroplasty; THR: total hip replacement; TKR: total knee replacement; ULN: upper limit of the normal range; VTE: venous thromboembolism

Differences between protocol and review

We considered total joint 'replacement' and total joint 'arthroplasty' as equivalent procedures, given the interchangeable use of these terms in the literature (Fawaz 2020).

We assessed the methodological quality of the included studies with the original Cochrane risk of bias approach (Higgins 2011b). We present detailed methodological explanations regarding each domain in the risk of bias analysis in the review. The planned assessment with the 2002 tool developed by Handoll and colleagues was considered unnecessary (Handoll 2002). We created summary of findings tables, in accordance with current Cochrane guidelines.

We analysed all the included studies instead of excluding those we considered to have a high risk of bias for randomisation sequence generation and concealment of allocation, as planned in the protocol. To evaluate the impact of including studies with a high risk of bias in these domains, we performed a sensitivity analysis (see Analysis 10: Direct factor Xa inhibitors versus LMWH: sensitivity analysis excluding studies with a high risk of bias in random sequence generation or allocation concealment).

The proposed three‐group classification of duration of follow‐up was replaced by a sensitivity analysis restricted to results assessed during the treatment period (as opposed to follow‐up time points).

We have presented more detailed methodological explanations regarding:

• the risk of attrition bias (which led us to present two tables in the appendices section with the analysis of three scenarios for missing participants);

• other possible sources of bias (which led us to perform three sensitivity analyses);

• heterogeneity;

• meta‐analysis model effect;

• funnel plot asymmetry;

• subgroup analyses;

• criteria we used for downgrading evidence certainty using the GRADE approach.

We also present a table with the parameters of the four effect size categories for each outcome.

As a result of the subgroup analyses and exploration of heterogeneity, we discovered differences in major bleeding events according to type of drug, which led us to present this outcome separately for rivaroxaban and the other direct factor Xa inhibitors.

For subgroup analyses, we included the type of drug and frequency of administration as additional variables. We re‐designated our planned sensitivity analysis according to the type of surgery as a subgroup analysis.

We included RCTs assessing the effects of direct factor Xa inhibitors, irrespective of the daily dose studied. The planned restriction to "standard doses" was included as a subgroup analysis according to all approved doses or one of three daily dose groups and, when necessary, an individual dose analysis.

We removed the sensitivity analysis titled "Re‐analysis of the data using other statistical approaches (using random‐effects meta‐analysis instead of fixed and vice versa)" because it does not contribute to the interpretation of results.

We also removed two outcomes (total VTE and total bleeding) from the review, as they did not provide relevant additional information.

We summarised treatment effects using risk ratios (RR) instead of odds ratios (OR), given their easier interpretation and the prospective design of the included studies.

Contributions of authors

CAS: drafted protocol, independently searched for trials and evaluated the titles and abstracts, independently evaluated the methodological quality of trials and extracted data, and drafted review, abstract and plain language summary.
GNM: independently evaluated the methodological quality of trials and extracted data, and drafted review.
GM: independently evaluated the methodological quality of trials and extracted data, and drafted review.
RB: drafted protocol, independently evaluated the methodological quality of trials and extracted data, and drafted review.
JEB: independently searched for trials and evaluated the titles and abstracts, independently evaluated the methodological quality of trials and extracted data, and drafted review and abstract.

Sources of support

Internal sources

• Source of support, Peru

  none

External sources

• Chief Scientist Office, Scottish Government Health Directorates, The Scottish Government, UK, UK

  The Cochrane Vascular editorial base is supported by the Chief Scientist Office.

Declarations of interest

CAS: none known
GNM: none known
GM: none known
RB: none known
JEB: none known

New