Interrupted versus uninterrupted anticoagulation therapy for catheter ablation in adults with arrhythmias

Abstract

Background

The management of anticoagulation therapy around the time of catheter ablation (CA) procedure for adults with arrhythmia is critical and yet is variable in clinical practice. The ideal approach for safe and effective perioperative management should balance the risk of bleeding during uninterrupted anticoagulation while minimising the risk of thromboembolic events with interrupted therapy.

Objectives

To compare the efficacy and harms of interrupted versus uninterrupted anticoagulation therapy for catheter ablation (CA) in adults with arrhythmias.

Search methods

We searched CENTRAL, MEDLINE, Embase, and SCI‐Expanded on the Web of Science for randomised controlled trials on 5 January 2021. We also searched three registers on 29 May 2021 to identify ongoing or unpublished trials. We performed backward and forward searches on reference lists of included trials and other systematic reviews and contacted experts in the field. We applied no restrictions on language or publication status.

Selection criteria

We included randomised controlled trials comparing uninterrupted anticoagulation with any modality of interruption with or without heparin bridging for CA in adults aged 18 years or older with arrhythmia.

Data collection and analysis

Two review authors conducted independent screening, data extraction, and assessment of risk of bias. A third review author resolved disagreements. We extracted data on study population, interruption strategy, ablation procedure, thromboembolic events (stroke or systemic embolism), major and minor bleeding, asymptomatic thromboembolic events, cardiovascular and all‐cause mortality, quality of life (QoL), length of hospital stay, cost, and source of funding. We used GRADE to assess the certainty of the evidence. 

Main results

We identified 12 studies (4714 participants) that compared uninterrupted periprocedural anticoagulation with interrupted anticoagulation. Studies performed an interruption strategy by either a complete interruption (one study) or by a minimal interruption (11 studies), of which a single‐dose skipped strategy was used (nine studies) or two‐dose skipped strategy (two studies), with or without heparin bridging.

Studies included participants with a mean age of 65 years or greater, with only two studies conducted in relatively younger individuals (mean age less than 60 years). Paroxysmal atrial fibrillation (AF) was the primary type of AF in all studies, and seven studies included other types of AF (persistent and long‐standing persistent). Most participants had CHADS2 or CHADS2‐VASc demonstrating a low–moderate risk of stroke, with almost all participants having normal or mildly reduced renal function. Ablation source using radiofrequency energy was the most common (seven studies).

Ten studies (2835 participants) were conducted in East Asian countries (Japan, China, and South Korea), while the remaining two studies were conducted in the USA. Eight studies were conducted in a single centre. Postablation follow‐up was variable among studies at less than 30 days (three studies), 30 days (six studies), and more than 30 days postablation (three studies).

Overall, the meta‐analysis showed high uncertainty of the effect between the interrupted strategy compared to uninterrupted strategy on the primary outcomes of thromboembolic events (risk ratio (RR) 1.76, 95% confidence interval (CI) 0.33 to 9.46; I² = 59%; 6 studies, 3468 participants; very low‐certainty evidence). However, subgroup analysis showed that uninterrupted vitamin K antagonist (VKA) is associated with a lower risk of thromboembolic events without increasing the risk of bleeding. There is also uncertainty on the outcome of major bleeding events (RR 1.10, 95% CI 0.59 to 2.05; I² = 6%; 10 studies, 4584 participants; low‐certainty evidence). The uncertainty was also evident for the secondary outcomes of minor bleeding (RR 1.01, 95% CI 0.46 to 2.22; I² = 87%; 9 studies, 3843 participants; very low‐certainty evidence), all‐cause mortality (RR 0.34, 95% CI 0.01 to 8.21; 442 participants; low‐certainty evidence) and asymptomatic thromboembolic events (RR 1.45, 95% CI 0.85 to 2.47; I² = 56%; 6 studies, 1268 participants; very low‐certainty evidence). There was a lower risk of the composite endpoint of thromboembolic events (stroke, systemic embolism, major bleeding, and all‐cause mortality) in the interrupted compared to uninterrupted arm (RR 0.23, 95% CI 0.07 to 0.81; 1 study, 442 participants; low‐certainty evidence).

In general, the low event rates, different comparator anticoagulants, and use of different ablation procedures may be the cause of imprecision and heterogeneity observed.

Authors' conclusions

This meta‐analysis showed that the evidence is uncertain to inform the decision to either interrupt or continue anticoagulation therapy around CA procedure in adults with arrhythmia on outcomes of thromboembolic events, major and minor bleeding, all‐cause mortality, asymptomatic thromboembolic events, and a composite endpoint of thromboembolic events (stroke, systemic embolism, major bleeding, and all‐cause mortality). 

Most studies in the review adopted a minimal interruption strategy which has the advantage of reducing the risk of bleeding while maintaining a lower level of anticoagulation to prevent periprocedural thromboembolism, hence low event rates on the primary outcomes of thromboembolism and bleeding. The one study that adopted a complete interruption of VKA showed that uninterrupted VKA reduces the risk of thromboembolism without increasing the risk of bleeding. Hence, future trials with larger samples, tailored to a more generalisable population and using homogeneous periprocedural anticoagulant therapy and ablation source are required to address the safety and efficacy of the optimal management of anticoagulant therapy prior to ablation. 

Plain language summary

The continuation or discontinuation of oral anticoagulants for people with arrhythmia and undergoing catheter ablation

Review question

Should doctors continue or stop blood thinner medicine before catheter ablation procedures in people with arrhythmia?

Background

Irregular heartbeats (also known as arrhythmia) come in different types; the most common type of which is uncoordinated irregular beats (also known as atrial fibrillation). People with irregular heartbeats are at risk of stroke and hence they require long‐term use of blood thinner medicines to prevent blood clots. Many of these patients may need to undergo a procedure called 'ablation' where doctors use surgery or a catheter (a thin flexible tube) to restore the normal rhythm of the heart. Before this procedure, the doctor may decide to continue the blood thinner or stop it temporarily. Such a decision should consider two things. First, the risk of a blood clot developing at the time of the ablation if the blood thinner is stopped. Second, the risk of bleeding during the procedure if the blood thinner is continued. There is no clear evidence that one decision is better than the other in terms of safety. Due to this uncertainty, we undertook this review to investigate the benefit and harms of continuing versus stopping the blood thinner at the time of the ablation procedure.

Study characteristics

We searched scientific databases and identified 12 clinical trials involving 4714 participants aged 18 years and older with atrial fibrillation. The participants were randomly (like a flip of a coin) put into one of two or more groups. The groups were: continuing the blood thinner at the time of the procedure or stopping the blood thinner either for several days or by one or two doses before the procedure with or without giving additional heparin. 

The trials reported on the risks of stroke, major bleeding, and minor bleeding after a follow‐up period of about 30 days after the procedure. The included trials examined different ablation procedures as well as different blood thinners, including Coumadin and the new oral blood thinners such as apixaban, rivaroxaban, edoxaban, and dabigatran. Ten trials were conducted in East Asian populations (Japan, China, and South Korea). Drug companies funded two trials. The evidence is current to 5 January 2021.

Key results

Due to the uncertainty of the evidence, after combining the results from the 12 clinical trials in this review, we are unsure about advising one strategy over another with the current evidence. The decision to continue, completely stop or withhold one or two doses of the blood thinner before the procedure should be individualised based on the patient's risks of stroke and bleeding until more well‐designed clinical trials are available to inform a conclusive decision.

Quality of the evidence

The evidence should be interpreted with caution. The review results are limited by variations in most trial comparisons, low numbers of events, trials being conducted in only one medical setting, and mostly described the intervention in East Asian populations. Further high‐quality trials are needed to find the ideal strategy for handling blood thinners at the time of the ablation procedure without increasing harm.

Summary of findings

Summary of findings 1. Interrupted versus uninterrupted anticoagulation therapy for catheter ablation in adults with arrhythmias.

Interrupted versus uninterrupted anticoagulation therapy for catheter ablation in adults with arrhythmias
Patient or population: adults with arrhythmias undergoing catheter ablation Setting: hospital Intervention: interrupted anticoagulation Comparison: uninterrupted anticoagulation
Outcomes and follow‐up	Anticipated absolute effects* (95% CI)		Relative effect(95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with uninterrupted anticoagulation	Risk with interrupted
Thromboembolic events Follow‐up: 48 hours to 1 year	3 per 1000	6 per 1000 (1 to 33)	RR 1.76 (0.33 to 9.46)	3468 (6 studies)	㊉◯◯◯ Very lowa	In addition, 6 studies reported no thromboembolic events occurred in either the interrupted or uninterrupted arms.
Major bleeding Follow‐up: 1 month to 1 year	10 per 1000	12 per 1000 (6 to 21)	RR 1.10 (0.59 to 2.05)	4584 (10 studies)	㊉㊉◯◯ Lowb	In addition, 1 study reported no major bleeding events occurred in either arm (Xing 2017). Tabish 2010 did not report on this outcome. Follow‐up was unclear for Di Biase 2014 and  Yoshimura 2017 for this outcome.
Composite endpoint of thromboembolic events (stroke or systemic embolism), major bleeding events, and all‐cause mortality Follow‐up: 3 months	59 per 1000	13 per 1000 (4 to 47)	RR 0.23 (0.07 to 0.81)	442 (1 study)	㊉㊉◯◯ Lowc	—
Minor bleeding Follow‐up: 1 month to 6 months	56 per 1000	57 per 1000 (26 to 125)	RR 1.01 (0.46 to 2.22)	3843 (9 studies)	㊉◯◯◯ Very lowa	Nogami 2019 did not report minor bleeding as a separate outcome but within a composite endpoint. Follow‐up was unclear for Di Biase 2014.
All‐cause mortality Follow‐up: 3 months	0/220 in the interrupted group vs 1/222 in the uninterrupted group died from any cause		RR 0.34 (0.01 to 8.21)	442 (1 study)	㊉㊉◯◯ Lowd	There was only 1 death in the uninterrupted arm in Nogami 2019, and this was used to calculate the RR. In addition, Reynolds 2018 reported that there were 0 mortality events in either arm.
Asymptomatic thromboembolic events Follow‐up: 1 hour to 1 month	148 per 1000	215 per 1000 (126 to 366)	RR 1.45 (0.85 to 2.47)	1268 (6 studies)	㊉◯◯◯ Very lowe	Follow‐up was unclear for Di Biase 2014 and Xing 2017.
*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; RR: risk ratio.
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.

^aDowngraded three levels. Once for risk of bias in included studies, specifically in the domains of randomisation and allocation concealment; once level for inconsistency owing to a moderate level of statistical heterogeneity, and once for imprecision due to wide confidence intervals in the effect estimate that could be explained by the low event rates in included trials.
^bDowngraded two levels. Once for risk of bias in included studies, specifically in the domains of randomisation and allocation concealment; and once for imprecision due to wide confidence intervals in the effect estimate that could be explained by the low event rates in included trials.
^cDowngraded two levels. Once for risk of bias in the included study, specifically in the domains of blinding; and once for imprecision as only one study with a low event rate reported this outcome.
^dDowngraded two levels. Once for risk of bias in the included study, specifically in the domains of blinding; and once for imprecision caused by low event rate and wide confidence interval.
^eDowngraded three levels. Once for risk of bias in included studies, specifically in the domains of randomisation and allocation concealment; once for inconsistency owing to a moderate level of statistical heterogeneity; and once for imprecision due to wide confidence intervals in the effect estimate that could be explained by the variability in assessing this outcome in the included study.

Background

Description of the condition

Arrhythmia is a problem with the rate or rhythm of the heartbeat, causing the heart to beat in a slow, rapid, or irregular fashion (Rakel 2017). Some of the cardiac arrhythmias can be serious and life‐threatening, while others may not be symptomatic or significant (Rakel 2017). Supraventricular arrhythmias are generated above the ventricles in the atrial or atrioventricular nodal tissues; they include atrial fibrillation (AF), atrial flutter, paroxysmal supraventricular tachycardia, and automatic atrial tachycardias (Rakel 2017). In contrast, ventricular arrhythmias are mostly due to secondary cardiac causes (such as structural heart diseases) and they are associated with high mortality if not treated immediately (Al‐Khatib 2018). Ventricular arrhythmias encompass premature ventricular complexes, ventricular tachycardia, and ventricular fibrillation (Ludhwani 2019). Treatment modalities for arrhythmias vary according to the type, presence of symptoms, risk stratification, and the impact of abnormal automaticity on patient outcomes (e.g. stroke, sudden death).

In clinical practice, AF is the most common arrhythmia and its incidence increases with age (Zimetbaum 2017). People with AF are five times more likely to experience a stroke than the rest of the population (Pisters 2010; Wolf 1991); the risk is higher in the presence (and severity) of underlying heart disease, as well as the comorbid existence of diabetes, hypertension, chronic kidney disease, and progressive ageing. The Global Anticoagulant Registry in the FIELD‐Atrial Fibrillation (GARFIELD‐AF) showed that in newly diagnosed people with AF at two‐year follow‐up the rate of stroke was 1.25 (95% confidence interval (CI) 1.13 to 1.38), bleeding was 0.70 (95% CI 0.62 to 0.81), and all‐cause mortality was 3.83 (95% CI 3.62 to 4.05) per 100 person‐years (Bassand 2016). Stroke is also costly due to increased healthcare resource utilisation for the management of AF and its complications (Blomstrom 2011; Murray 2012).

The management of AF focuses on three clinical approaches. First is the use of oral anticoagulants (OACs) for stroke prevention (January 2019; Kirchhof 2016; Lip 2018). Most guidelines recommend chronic anticoagulation for people identified to be at high risk of stroke using the CHA2DS2‐VASc score (January 2019; Kirchhof 2016; Lip 2018). Until recently, vitamin K antagonists (VKAs) were the main OAC recommended. However, novel direct oral anticoagulants (DOACs) – dabigatran, rivaroxaban, apixaban, and edoxaban – are now the preferred agents for stroke prevention in eligible people with AF (Steffel 2018). The second approach in the management of AF is the use of rate‐control drugs to regulate the ventricular rate. The third approach can include rhythm‐control interventions (cardioversion, medications, or ablation) for restoration and maintenance of normal sinus rhythm in selected patients (Andrade 2018; January 2019; Jones 2014; Lip 2018). Evidence is accumulating in support of catheter ablation (CA) as first‐line therapy for AF, especially in people with symptoms. Guidelines give a strong recommendation (class I, level A) for offering ablation therapy to people who failed treatment with antiarrhythmic drugs, have no structural heart disease, and are still symptomatic (Calkins 2018a; January 2019).

Atrial flutter shares a significant clinical resemblance to AF and it commonly coexists in the same patient. CA is the most effective and first‐line therapy in atrial flutter (Brugada 2019; Calkins 2018b; Page 2016). Antithrombotic therapy is also recommended based on stroke risk assessment, similar to AF (Kirchhof 2016; Lip 2018).

In people with ventricular arrhythmias, treatment approaches mainly focus on treating the underlying cardiac pathology and preventing sudden cardiac arrest and cardiac death (Shivkumar 2019). CA is also one of the main strategies of management in people with sustained ventricular tachycardia (Cronin 2019; Shivkumar 2019). Although anticoagulants are not the main therapy for people with ventricular tachycardia, they could be used for people with a non‐urgent indication for ablation, especially if a mobile thrombus is present (Cronin 2019). In such cases, treatment with OACs for a limited period of time after extensive endocardial ventricular tachycardia ablation may be considered (Cronin 2019). Furthermore, device therapy such as implantable defibrillators, pacemakers and re‐synchronisation therapy defibrillator had become routine therapy for primary and secondary prevention in people with ventricular arrhythmia (January 2019; Lau 2017). CA procedures are carried out percutaneously through vascular access (venous or arterial, or both) and they carry the risk of thromboembolism due to interruption of anticoagulant therapy and the introduction of foreign materials to the circulation (Widimsky 2016).

Description of the intervention

During CA procedures, the continuation of OAC therapy presents a clinical decision that should take into account the procedure as well as the patient's risks of bleeding and thromboembolic events (Clark 2018; Doherty 2017). Therefore, periprocedural management of OACs is paramount to improve clinical outcomes and reduce risks of bleeding and thrombosis in people with AF undergoing CA (Abed 2016; Calkins 2018a). Two major strategies are frequently used in current clinical practice (Balouch 2017; De Heide 2018; Di Biase 2019; Efremidis 2015; Ha 2018; Nagao 2018; Nogami 2019; Xing 2017; Xing 2018). In the first strategy, OAC therapy is interrupted or minimally interrupted (defined as withholding up to two doses of the anticoagulant) before the start of the procedure, with or without the use of bridging therapy with intravenous heparin or subcutaneous low molecular weight heparin (LMWH) (Balouch 2017; De Heide 2018; Di Biase 2019; Ha 2018). Following the procedure, an OAC is resumed once haemostasis is achieved (Calkins 2018b; Doherty 2017). In the second strategy, the procedure is performed without interruption or reduction of OAC therapy (Kaiser 2013; Tapanainen 2013).

There are major differences between VKAs and DOACs that should be considered in preprocedural planning (Barnes 2018). The half‐life of clotting factors determines the antithrombotic effect of VKAs. More specifically, the reduction in the level of factor II/prothrombin, which has the longest half‐life of about 96 hours, is required for the complete antithrombotic effect of VKAs. Standard interruption for VKAs is usually five days to reverse the anticoagulant effect of warfarin and achieve normal haemostasis (Barnes 2018). The American College of Cardiology (ACC) 2017 consensus statement recommends using individualised international normalised ratios (INRs) to plan preprocedure warfarin interruption (Doherty 2017). Bridging therapy with LMWH or heparin has previously been used and recommended when interrupting VKAs during the preoperative period; however, recent studies have argued against this approach (Garwood 2011).

Unlike VKAs, DOACs have different pharmacokinetic properties – including having a short time to peak effect, somewhat short half‐lives (around 12 hours) – and are renally eliminated (Steffel 2018). Since the extent of renal clearance varies between DOACs, calculating the patient's creatinine clearance is important when developing a preprocedural plan (Barnes 2018). Generally, bridging therapy with either heparin or LMWH is not recommended when interrupting DOACs before CA procedures. The timing of interrupting DOACs differs depending on the agent. For dabigatran, it is currently recommended to discontinue therapy one to two days prior to CA procedures when creatinine clearance is more than 50 mL/minute, and three to five days prior to CA procedures when creatinine clearance is less than 50 mL/minute (Barnes 2018). For rivaroxaban, apixaban, and edoxaban, if the interruption is planned, the treatment should be stopped for at least 24 hours prior to CA and once there is adequate haemostasis, the treatment can be resumed (Raval 2017).

How the intervention might work

Physicians face a dilemma when deciding to withhold or continue OACs for people with AF having CA procedures (Deharo 2016; Nascimento 2014). Interruption of OACs, although recommended by various practice guidelines, has been associated with negative outcomes (which include an increased risk of thromboembolism and device‐pocket haematoma), especially when bridging with heparin is used (Garwood 2017; Hirao 2018). These complications can lead to a prolonged cessation of OACs, an increase in hospital stay, the need for surgery, and an increased risk of infection (Essebag 2016; Kojima 2018; Proietti 2015). Therefore, uninterrupted OAC therapy has been recommended with evidence suggesting this to be a safe and feasible option (Ha 2018; Kaiser 2013; Kwak 2010; Zhao 2018). However, the risk of bleeding remains a concern when continuing OACs in people with AF going through CA procedures.

Why it is important to do this review

Anticoagulation management around procedures such as CA is of critical importance and must be dealt with carefully to minimise the risk of thromboembolic and bleeding complications. Recommendations from international guidelines concerning whether to interrupt or continue OAC therapy for people undergoing surgical procedures diverge. The 2017 Heart Rhythm Society/European Heart Rhythm Association/European Cardiac Arrhythmia Society/Asia Pacific Heart Rhythm Society/Latin American Society of Electrophysiology and Cardiac Stimulation (HRS/EHRA/ECAS/APHRS/SOLAECE) guideline stated a class I recommendation and a level A of evidence for performing the ablation procedure under uninterrupted anticoagulant therapy for people with AF on therapeutic doses of warfarin or dabigatran (Calkins 2018b). However, the American College of Chest Physicians (CHEST) guideline in 2018 provided a weak recommendation and low quality evidence on continuing VKAs, dabigatran, or rivaroxaban in people with AF who are planned for CA (Lip 2018). Moreover, several meta‐analyses and systematic reviews have explored the safety and efficacy of a single strategy (uninterrupted OAC therapy) around CA procedures, with most reviews including randomised controlled trials (RCTs) as well as controlled prospective studies (Bin Abdulhak 2013; Cardoso 2018; Elgendy 2017; Garg 2016; Romero 2018; Ukaigwe 2017; Vallakati 2016; Vamos 2016; Wu 2016; Zhao 2017; Zhao 2018).

In recent years, several RCTs have explored the safety and efficacy of interrupted versus uninterrupted OAC therapy around procedures such as CA (Nagao 2018; Nogami 2019; Reynolds 2018; Nakamura 2018). The COMPARE trial showed a significantly lower risk of periprocedural symptomatic thromboembolic events in the uninterrupted VKA group compared to the interrupted group  (Di Biase 2014). The  uninterrupted VKA group was also associated with a higher risk of bleeding possibly due to bridge therapy. The evidence on DOACs is still emerging; so far, studies have shown a decreased risk of thromboembolic events when DOACs were uninterrupted (Nakamura 2018; Nakamura 2019a; Nakamura 2019b), or minimally interrupted (Nagao 2018; Nogami 2019; Reynolds 2018; Yu 2019), with no significant increase in the risk of major bleeding.

Since the current uncertain status of the evidence along with the publication of several new studies merits a systematic review and meta‐analysis, we planned to undertake this review to investigate the efficacy and safety of interrupted versus uninterrupted preprocedural anticoagulant therapy.

Objectives

To systematically compare the efficacy and harms of interrupted versus uninterrupted anticoagulation therapy for catheter ablation (CA) in adults with arrhythmias.

Methods

Criteria for considering studies for this review

Types of studies

We included RCTs that compared the efficacy and harms of uninterrupted anticoagulation with any modality of interruption with or without heparin bridging for CA in adults aged 18 years or older with arrhythmia. We included studies published as full‐text, abstract only, and those that were not published. We planned to include cluster‐randomised trials; however, we did not identify any such trials. We did not consider cross‐over trials for inclusion due to the nature of the intervention of our review.

Types of participants

We included trials of people aged 18 years and older with arrhythmia – including AF, atrial flutter, or ventricular tachycardia – who were receiving long‐term treatment with OACs (VKAs or DOACs) and were undergoing CA. In the protocol, we planned to contact study authors if only a subset of eligible participants was presented in a trial and if authors were unwilling or unable to provide additional information, we planned to include studies in which at least 70% of included participants were eligible in our review (Bawazeer 2019). We planned to perform a sensitivity analysis in which such trials were removed from the primary analysis. However, we did not encounter any such trials.

Types of interventions

We included all comparisons of interrupted and uninterrupted oral anticoagulation (VKAs or DOACs), including any modalities of interruption (complete interruption with or without bridge therapy, minimal interruption of anticoagulants, or any other modality).

Types of outcome measures

We extracted definitions and measurement of clinical events (e.g. stroke, thromboembolic events, bleeding) according to the individual trials. We excluded studies that did not meet inclusion criteria. We contacted the authors for further information on study outcomes and characteristics. We did not encounter any trials that measured the outcomes but did not report it. We encountered one study that reported one of the outcomes (minor bleeding) in an unusable format within a composite point but not separately (Nogami 2019). We contacted the author for the results of minor bleeding. All studies were included in the final analysis in one or more of the outcomes. We assessed study outcomes at the longest follow‐up point.

Primary outcomes

• Thromboembolic events (stroke or systemic embolism).

• Major bleeding, defined as bleeding requiring intervention (or as reported in included trials).

Secondary outcomes

• Composite endpoint of thromboembolic events (stroke or systemic embolism), major bleeding events, and all‐cause mortality.

• Minor bleeding, defined as bleeding not requiring intervention.

• Cardiovascular mortality.

• All‐cause mortality.

• Asymptomatic thromboembolism as described by studies.

• Patient QoL as measured by any validated health‐related QoL instruments (e.g. Seattle Angina Questionnaire (Wyrwich 2014), 36‐item Short Form Health Survey (SF‐36) (Ware 1992), EuroQOL/EQ‐5D (Balestroni 2012), or participant's subjective perception of improvement (yes or no) as reported by the study authors). If we are unable to pool data on QoL due to use of different measurements, we included them in a narrative form.

• Economic costs (if data on economic cost could not be pooled, we included them in a narrative form).

• Length of hospital stay.

Search methods for identification of studies

Electronic searches

We searched the following electronic databases on 5 January 2021: 

• Cochrane Central Register of Controlled Trials (CENTRAL) (Cochrane Library 2020, Issue 1);

• Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, MEDLINE Daily and MEDLINE (Ovid, 1946 to 4 January 2021);

• Embase (Ovid, 1980 to week 53, 2021);

• Science Citation Index Expanded (SCI‐E) on the Web of Science (Clarivate Analytics, 1900 to 5 January 2021).

The search strategies are presented in Appendix 1. We applied the Cochrane sensitivity‐maximising RCT filter (Lefebvre 2011) to MEDLINE (Ovid) and adaptations of it to the other databases, except CENTRAL. We imposed no restrictions on language or publication status.

To identify ongoing or unpublished trials, we searched the following clinical trials registers on 29 May 2021: 

• World Health Organization (WHO) International Clinical Trials Registry Platform (www.who.int/ictrp/en);

• ClinicalTrials.gov (clinicaltrials.gov);

• Clinical Trials Register EU (www.clinicaltrialsregister.eu).

Searching other resources

We handsearched reference lists of included trials and reviews on the topic of management of OACs around ablation in arrhythmia. We performed backwards and forwards citation analysis on all included studies to identify other potentially relevant studies. As we did not encounter any ongoing clinical trials through searching trials registries, we did not contact researchers and pharmaceutical companies who produced anticoagulant drugs to request information on any unpublished trials. We will consider this in the update of this review. We planned to examine any relevant retraction statements and errata for included studies; however, we identified neither. 

Data collection and analysis

We performed the review and meta‐analysis according to the recommendations in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We included all eligible trials in the analysis, regardless of publication status.

Selection of studies

Three review authors (AAA, NOB, AMA) independently screened titles and abstracts of all the potential studies and coded them as 'retrieve' (eligible or potentially eligible/unclear) or 'do not retrieve'. We retrieved the full‐text publications and three review authors (AAA, NOB, AMA) independently screened these studies for inclusion, and identified and recorded reasons for exclusion of the ineligible studies. We resolved any disagreements through discussion or a fourth review author (GB or HK) arbitrated any disagreement. Figure 1 summarises studies flow and search processes. We excluded duplicates and collated multiple reports of the same study, so that each study, rather than each report, was the unit of interest in the review. We used a standardised process using COVIDENCE to extract information. We recorded the selection process in sufficient detail to complete a PRISMA flow diagram and Characteristics of excluded studies table (Liberati 2009).

1.

Data extraction and management

Three review authors (AAA, NOB, AMA)  independently extracted study characteristics twice for each trial and recorded the information using a data collection form for study characteristics and outcome data that had been piloted on one study in the review. We extracted the following study characteristics.

• Methods: study design, total duration of study, number of centres and location, study setting, and study date.

• Participants: number of participants in each treatment group, number randomised, number lost to follow‐up/withdrawn, number analysed, mean age, age range, gender, the type of arrhythmia, comorbidities, CHADS2 or CHA2DS2VASc score, bleeding risk (e.g. HAS‐BLED (Hypertension. Abnormal renal and liver function. Stroke. Bleeding) score or as reported in individual trials), inclusion criteria, and exclusion criteria.

• Interventions: generic name and dose(s) of OAC for both the intervention and comparison, concomitant medications, excluded medications, duration of anticoagulant therapy in the trial, intensity of anticoagulation dose‐adjusted using the prothrombin time ratio (PTR) or INR and adherence to anticoagulant treatment, time of interruption of the anticoagulants prior to the procedure, procedure type, device type, heparin therapy (intraprocedural or as bridge therapy), time and strategy of resumption of OAC after the procedure, and duration of follow‐up.

• Outcomes: we collected data on planned and reported outcomes in each trial.

• Funding and notable conflicts of interest of trial authors.

Three review authors (AAA, NOB, AMA) independently extracted outcome data from the included studies. We resolved disagreements by consensus or by involving a fourth review author (GAB or HAK). One review author (AAA)  transferred the data into Review Manager 5 (Review Manager 2014). The other two review authors (NOB, AMA)  double‐checked entered data by comparing the data presented in the systematic review with the data extraction form. We contacted the lead authors for three included trials for further clarification on study outcomes and definitions (Di Biase 2014; Tabish 2010; Yoh 2019).

Assessment of risk of bias in included studies

Three review authors (AAA, NOB, AMA) independently assessed the risk of bias for each study using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2017). We resolved any disagreements by discussion or by involving a third review author (GAB or HAK). We assessed the risk of bias according to the following domains.

• Random sequence generation.

• Allocation concealment.

• Blinding of participants and personnel.

• Blinding of outcome assessment.

• Incomplete outcome data.

• Selective outcome reporting.

For each of the domains, we rated each study at high, low, or unclear risk of bias. We provided a quote from the study report, together with a justification for our judgement, in the risk of bias table. We summarised the risk of bias judgements across different studies for each of the domains listed. Where information on the risk of bias related to unpublished data or correspondence with a study author, we noted this in the risk of bias table. When considering treatment effects, we took into account the risk of bias for the studies that contributed to that outcome. We did not encounter any cluster‐RCT, however, for any potential future updates, we will use recommendations from the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2017). We will consider additional bias arising from the following: randomisation taking place at the cluster level; recruitment of participants after cluster‐level treatment allocation is known, and missing cluster‐level outcome data (Higgins 2019).

Measures of treatment effect

We presented dichotomous outcomes as risk ratios (RRs) with 95% CIs. We used intention‐to‐treat (ITT) analysis to assess study outcomes. However, in cases where the method of analysis was not specified or followed per‐protocol analysis, we used the number randomised as the denominator (for dichotomous outcomes) according to the ITT analysis. We planned to extract continuous data using mean difference (MD) with standard deviation from each trial to calculate the average MD with 95% CIs. In cases where studies used different scales to measure the same continuous outcome, we planned to use the standardises mean difference (SMD) with 95% CI.  However, we did not encounter studies with continuous outcomes.

Unit of analysis issues

We did not encounter a unit of analysis issues in the included studies. We planned for cluster‐RCTs to use the recommendation from the Cochrane Handbook and ensure that the adequate sample size is calculated using the intra‐cluster correlation coefficient (Higgins 2019). However, we did not identify any cluster‐RCTs. We did include one study (Yoh 2019) that had two arms which we combined as a single comparison for the main analysis. None of the included studies had multiple time points; hence all data were considered for the follow‐up period specified in each trial. To investigate the effect of combining the multiple arms or using only the longest follow‐up period, we planned to conduct a subgroup analysis; but we did not have an adequate number of studies (at least 10 studies) to perform such analysis (Deeks 2021). 

Dealing with missing data

We contacted the lead authors for three studies for further clarification on outcomes and definitions as these trials were published as abstract only (Di Biase 2014; Tabish 2010; Yoh 2019).

Assessment of heterogeneity

We visually inspected forest plots for the direction and magnitude of effects and the degree of overlap between CIs. We also used the P value from the Chi² test and measured the heterogeneity using the I² statistic (percentage of total variation across studies due to heterogeneity).

We used the following guidelines for interpreting the I² value (Deeks 2017).

• 0% to 40%: might not be important.

• 30% to 60%: may indicate moderate heterogeneity.

• 50% to 90%: may indicate substantial heterogeneity.

• 75% to 100%: indicates considerable heterogeneity.

When we identified substantial or considerable heterogeneity (I² greater than 50%), we reported and explored possible causes by prespecified subgroup analysis.

Assessment of reporting biases

Our meta‐analysis included fewer than 10 trials in each specified outcome and a funnel plot was not reported.

Data synthesis

We undertook meta‐analyses by pooling the appropriate data using Review Manager 5 (Review Manager 2014). We used a random‐effects model to combine data as we expected some heterogeneity in the interventions and outcome definitions. We repeated the analysis using a fixed‐effect model as a sensitivity analysis.

Subgroup analysis and investigation of heterogeneity

We explored heterogeneity using subgroup analyses (when there were enough studies) according to the following parameters.

• Gender. Gender differences may influence outcomes.

• Age (less than 75 years, 75 years or more). Older people (aged 75 years or more) have a higher risk of bleeding and this outcome is pertinent to our review.

• Anticoagulation type (VKAs, DOAC). Different anticoagulants have differences in bleeding risk.

• Baseline risk of stroke using CHA2DS2‐VASc or CHADS2 (high risk greater than 2, low‐to‐moderate risk less than 2). The baseline risk of stroke (reported as CHA2DS2‐VASc or CHADS2) is very relevant to the review scope as a higher baseline risk population may have different outcomes during interruption of anticoagulant therapy prior to the procedure.

• The ablation surgical type (e.g. cryoablation, and radiofrequency hot balloon ablation or radiofrequency ablation). Ablation surgical strategies may have different stroke risks (Di Biase 2019), with radiofrequency CA being associated with endothelial denudation compared to cryoablation ablation and radiofrequency hot balloon.

• Interruption strategy (complete interruption versus minimal interruption, interruption with heparin versus no heparin). The strategy used in the interrupted arm may also influence the outcomes we were interested in (complete interruption versus minimal interruption); also, the use of heparin as bridge therapy during the interruption period may be associated with a higher risk of bleeding, which is relevant to the review objectives.

We assessed subgroup differences using interaction tests available  in Review Manager 5 (Review Manager 2014). We reported the results of subgroup analyses quoting the Chi² statistic and P value.  In addition to the outcome major bleeding (10 studies), we decided to conduct a subgroup analysis for the outcome of thromboembolism despite the fact that only 6/12 studies reported events. We thought that such subgroup analysis may provide additional meaningful insights and may explain the heterogeneity identified in this outcome, hence better use of evidence in the clinical practice. 

Sensitivity analysis

Depending on the studies obtained from the systematic search, we conducted the following sensitivity analyses for the primary outcomes.

• Excluding studies with a high risk of bias (this constitutes studies with low risk of bias for randomisation method, which is not at high risk of bias for any other domain).

• Only including studies where there were no conflicts of interest, for example regarding the funding source of trials.

• Testing the robustness of the results by repeating the analyses using different statistical models (fixed‐effect and random‐effects models).

• Restricting the analyses to published trials.

• Repeating the analysis excluding cluster‐RCTs.

• Repeating the analysis excluding studies in which not all included participants meet the review inclusion criteria.

• Repeating the analysis excluding studies with missing data that were thought to introduce serious bias. This included studies with an attrition rate of over 30%, or where differences in attrition between groups exceed 10%, or both.

We were unable to perform sensitivity analysis based on excluding studies with a high risk of bias because all such studies were judged at high risk in one other domain. We did not encounter cluster‐RCTs, studies with participants not meeting the inclusion criteria nor studies with missing data.

Summary of findings and assessment of the certainty of the evidence

We created Table 1 using the primary outcomes and the main secondary outcomes. We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness, and publication bias) to assess the certainty of the body of evidence as it related to the studies that contributed data to the meta‐analyses. We used methods and recommendations described in Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2017), using GRADEpro GDT software (GRADEpro GDT). We justified all decisions to downgrade the certainty of the evidence using footnotes, and made comments to aid the reader's understanding of the review where necessary. Three review authors (AAA, NOB, AMA) independently judged the certainty of the evidence, with disagreements resolved by discussion or involving a fourth review author (GAB). Judgements were justified, documented, and incorporated into reporting of results for each outcome. We planned to extract study data, format our comparisons in data tables and prepared a summary of findings table before writing the results and conclusions of our review. When meta‐analysis was not possible, we presented the results as a narrative summary of findings table.

The following is a list of outcomes included in the summary of findings tables.

Primary outcomes.

• Thromboembolic events.

• Major bleeding.

Secondary outcomes.

• Composite endpoint of thromboembolic events (stroke or systemic embolism), major bleeding events, and all‐cause mortality.

• Minor bleeding.

• All‐cause mortality.

• Asymptomatic thromboembolic events.

Results

Description of studies

We identified 12 RCTs (Ando 2019; Di Biase 2014; Nagao 2019; Nakamura 2019; Nogami 2019; Reynolds 2018; Tabish 2010; Xing 2017; Yamaji 2019; Yoh 2019; Yoshimura 2017; Yu 2019). Two studies were in abstract format (Tabish 2010; Yoh 2019). See Characteristics of included studies table.

Results of the search

The searches yielded 1452 records. After removing duplicates there were 1049 unique records. We excluded 977 irrelevant studies. We retrieved full‐text reports of the remaining eligible 72 records. We excluded 44 references, see Characteristics of excluded studies table. We identified multiple reports of the same studies and determined that 12 studies (from 28 records) met the inclusion criteria. See Figure 1 for a PRISMA diagram documenting our search and decision process.

Included studies

The 12 trials compared uninterrupted periprocedural anticoagulation with interrupted anticoagulation in 4714 participants; see the Characteristics of included studies for a detailed description of the included studies.

Seven studies included people with a mean age of 65 years or above, three studies enrolled participants with a mean age less than 65 years, and two studies included participants with a mean age less than 60 years. All studies included participants with paroxysmal AF, while seven studies also included other types of AF (persistent and long‐standing persistent). One study included participants with a repeat ablation.

The anticoagulant comparators between the studies varied, with three main anticoagulant comparisons. Three studies used VKAs as the periprocedural anticoagulation (Di Biase 2014; Tabish 2010; Xing 2017), eight studies used different  DOACs (Ando 2019; Nagao 2019; Nakamura 2019; Nogami 2019; Reynolds 2018; Yoh 2019; Yoshimura 2017; Yu 2019), and one study involved comparison of  VKA and dabigatran (Yamaji 2019). Seven studies described CA using radiofrequency energy, three studies used multiple catheter sources (such as radiofrequency, cryoballoon, radiofrequency hot balloon, linear ablation, and cryoablation cavotricuspid isthmus), and two studies did not report the ablation source. 

Seven studies were undertaken in Japan, two in the US, two in China, and one in South Korea. Eight studies were conducted in a single centre. Postablation follow‐up was variable among the included studies, and it encompassed a range of follow‐up periods of 48 hours (one study), seven days (one study), 10 days (one study), 30 days (six studies), three to six months (two studies), and 12 months postablation (one study). 

Interruption strategies

Studies used two main interruption modalities; either complete interruption or minimal interruption. Eleven studies described using a minimally interrupted strategy. Nine studies used a strategy of holding one or two doses of DOAC with no heparin bridging (Ando 2019; Nagao 2019; Nakamura 2019; Nogami 2019; Reynolds 2018; Yamaji 2019; Yoh 2019; Yoshimura 2017; Yu 2019). Ando 2019, Reynolds 2018, and Yoshimura 2017 interrupted apixaban on the morning of the procedure, and gave no heparin bridging. Nagao 2019 and Nakamura 2019 interrupted the twice‐daily DOACs by giving the last dose on the evening of the day before the procedure, while for once‐daily DOACs the last dose was given on the morning of the day before the procedure, with no heparin bridging. Nogami 2019 interrupted dabigatran (D) by withholding one or two doses before ablation (A)(D‐A less than 24 or D‐A 24 hours or greater) before the procedure. It is worth mentioning that D‐A is the interval between the final dose of dabigatran and the transseptal puncture. About 35% of participants (78/220) in the interrupted group were bridged with heparin therapy, of them, 15% (20/137) were in D‐A less than 24 and 70% (58/83) in the D‐A 24 hours or greater. Yoh 2019 gave the interruption on the morning of the day before the procedure for both of the twice‐daily DOACs (dabigatran and apixaban) and the once‐daily DOACs (rivaroxaban and edoxaban) with no heparin bridging. Interruption in Yu 2019 was either by single‐dose skip or 24 hours of skip (24S). It was unclear when the last dose of DOAC (rivaroxaban, dabigatran, apixaban) was administered. In addition, people with persistent AF with a CHA2DS2‐VASc score of 2 or greater assigned to the 24S group received enoxaparin 1 mg/kg in the evening the day before the procedure. Two trials used low‐intensity warfarin (INR 1.5 to 2.0) as the minimal interruption strategy with no heparin bridging in older people and compared it to standard intensity warfarin (INR 2.0 to 3.0)  (Tabish 2010; Xing 2017). Of the 12 included studies,  Di Biase 2014 was the only one that used complete interruption of the anticoagulant. Warfarin was held two to three days before the ablation, and heparin bridging given with enoxaparin 1 mg/kg twice daily until the evening before the ablation procedure; see Figure 2 for the schematic presentation of anticoagulant management around the time of the procedure.

2.

Graphical representation of strategies used for the periprocedural anticoagulation around the ablation procedure. 
A: uninterrupted activation clotting time; B: minimally interrupted using one‐dose skipped strategy; C: minimally interrupted using low‐intensity INR strategy; D: minimally interrupted using two‐dose skipped strategy; E: completely interrupted activation clotting time. DOAC: direct oral anticoagulants; H: heparin bridging; R: resumption of the drug; S: skipped dose; VKA: vitamin K antagonist; w/o: without. Included studies: 1: Ando 2019; 2: Di Biase 2014; 3: Nagao 2019; 4: Nakamura 2019; 5: Nogami 2019; 6: Reynolds 2018; 7: Tabish 2010; 8: Xing 2017; 9: Yamaji 2019; 10: Yoh 2019; 11: Yoshimura 2017; 12: Yu 2019.

Other procedure‐related protocols were fairly similar between the trials. Anticoagulants were administered for at least four weeks before ablation, transoesophageal echocardiography (TOE) was performed on the day of the procedure, and intraprocedural use of heparin and activation clotting time (ACT) target was similar between trials. Resumption of anticoagulant therapy after the ablation varied depending on the frequency of use of the anticoagulant. Studies using DOAC resumed therapy on the evening of the procedure for a twice‐daily DOAC or in the morning after the procedure for a once‐daily DOAC with or without heparin bridging  (Ando 2019; Nagao 2019; Nakamura 2019; Nogami 2019; Reynolds 2018; Yamaji 2019; Yoh 2019; Yoshimura 2017; Yu 2019). For VKA anticoagulant, one study resumed VKA on the evening of the procedure with heparin bridging (Di Biase 2014), and two studies used lower‐intensity INR throughout and after the procedure (Tabish 2010; Xing 2017).

Outcomes

All 12 studies reported on thromboembolic event outcome (stroke, transient ischaemic attack, or systemic embolism) (Ando 2019; Di Biase 2014; Nagao 2019; Nakamura 2019; Nogami 2019; Reynolds 2018; Tabish 2010; Xing 2017; Yamaji 2019; Yoh 2019; Yoshimura 2017; Yu 2019). Six studies reported no events.

Eleven studies reported major bleeding (Ando 2019; Di Biase 2014; Nagao 2019; Nakamura 2019; Nogami 2019; Reynolds 2018; Xing 2017; Yamaji 2019; Yoh 2019; Yoshimura 2017; Yu 2019). Major bleeding definitions used across studies were International Society of Thrombosis and Hemostasis (ISTH) criteria (Ando 2019; Nogami 2019; Yu 2019), Bleeding Academic Research Consortium (BARC) 3 or greater (Reynolds 2018), bleeding into a critical anatomical site (Yoh 2019), or bleeding requiring intervention in remaining studies.

Six studies reported minor bleeding (Di Biase 2014; Nagao 2019; Nakamura 2019; Xing 2017; Yoh 2019; Yu 2019). Nogami 2019 reported on minor bleeding within a composite endpoint but not separately.

Two studies reported all cause‐mortality (Nogami 2019; Reynolds 2018); Reynolds 2018 reported zero events. 

Six studies reported asymptomatic thromboembolic events (Di Biase 2014; Nagao 2019; Nakamura 2019; Xing 2017; Yoh 2019; Yoshimura 2017). Although the trials used variable terms for the events, all trials agreed on the lack of neurological symptoms and all used magnetic resonance imaging (MRI) to detect this outcome. Deneke 2015 defined the silent cerebral event as "an acute new MRI‐detected brain lesion typical to cerebral ischaemia in a patient without clinically apparent neurological deficit." In that clinical review, the author related silent cerebral events and silent cerebral lesions to cerebral ischaemic infarcts due to embolic "fingerprint" specific to the type of ablation procedure. Therefore, we believed that despite the different terms used they reflected a relatively similar outcome. Di Biase 2014 used the term silent thromboembolic lesion (STL), detected symptoms using diffusion magnetic resonance imaging (dMRI) and not resulting in clinical symptoms. Nagao 2019 used the term silent stroke and was defined as a focal hyperintense region on the diffusion‐weighted (DW) image (DWI) despite the absence of corresponding clinical symptoms detected on MRI on the day after ablation. Nakamura 2019 used the term silent cerebral ischaemic lesions (silent thromboembolism) defined as hyperintense lesions on magnetic resonance DWI, corresponding to a reduced apparent diffusion coefficient map detected on the day after ablation. The study also described detailed characteristics of DWI and the lesion size. Xing 2017 used the term asymptomatic cerebral emboli (ACE) defined as focal hyperintense areas detected by the DW sequence without any symptoms detected by MRI seven days postablation. Yoh 2019 used the term silent cerebral ischaemia defined as a lesion detected by brain MRI after the procedure, without any symptoms. Yoshimura 2017 used the term asymptomatic cerebral micro‐thromboembolism detected by MRI on the day after the procedure. Except for Di Biase 2014, all other trials performed only postablation MRI.

Funding

Three studies reported funding by a non‐industrial entity, of which the source came from a governmental grant (Yu 2019) and a grant from a private foundation or scientific society (Di Biase 2014; Yoshimura 2017), respectively. One study received no funding (Xing 2017). Two studies declared funding by the pharmaceutical industry (Nogami 2019; Reynolds 2018). Nogami 2019 declared that the funding source had no role in any aspect of the study conduct and publication, whereas Reynolds 2018 did not declare the role of the funder. The remaining six studies (Ando 2019; Nagao 2019; Nakamura 2019; Tabish 2010; Yamaji 2019; Yoh 2019) did not report the source of funding .

Excluded studies

On inspection of full‐texts, we excluded 44 studies, and details were reported in the Characteristics of excluded studies table. Of those 44 studies, 27 studies were non‐randomised comparisons of anticoagulants (Abhishek 2011; Aoyama 2019; Baltogiannis 2016; Brinkmeier 2018; Cavalli 2019; ChiCTR‐OPN‐15006584; Di 2017; Di Biase 2014a; Di Biase 2014b; Efremidis 2015; Finlay 2010; Gunawardene 2017; Konduru 2012; Kuwahara 2013; Lane 2018; Muller 2016; Page 2011; Page 2014; Saad 2011; Sagawa 2018; Steffel 2017; Stepanyan 2014; Tscholl 2017; UMIN000029693; Wakamatsu 2020; Wazni 2007; Xing 2018; Yamaji 2013). Nine trials were not the intervention of interest (jRCTs031180249; Kim 2016; Kirchhof 2018; Oh 2013; Okumura 2016; Sakamoto 2019; UMIN000028892; Yamaji 2018). Eight trials were single‐arm with interruption (Calkins 2019; EUCTR2012‐001484‐79‐DE; EUCTR2016‐003069‐25‐HU; Hohnloser 2019; NCT01729871; NCT02504177; UMIN000013341; Yoshimoto 2020).

Risk of bias in included studies

We presented the results of risk of bias assessments across all included studies (Figure 3) and the summary of risk of bias assessment of each individual study (Figure 4).

3.

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

4.

Allocation

Random sequence generation and allocation

Of the 12 included studies, only five described the random sequence generation and were judged at low risk of bias (Di Biase 2014; Nakamura 2019; Nogami 2019; Yoshimura 2017; Yu 2019). Di Biase 2014 described using block randomisation with the study centre as the blocking variable. Then used a central randomisation algorithm to generate the randomisation code. Nakamura 2019 used a computer‐generated list of random numbers. Nogami 2019 used a randomisation registration system Mebix Inc to conduct a random allocation sequence, participant enrolment, and assignment of participants to the interventions. Yoshimura 2017 stratified participants according to the type of AF and sex, and used the table of random numbers for randomisation. Yu 2019 described a central randomisation strategy using computer‐generated random permutation sequences.

Four studies did not describe methods of randomisation and were judged at unclear risk of bias (Ando 2019; Nagao 2019; Reynolds 2018;  Xing 2017). Tabish 2010 and Yoh 2019 are abstract papers and had insufficient information; both were judged at unclear risk of bias. One study was judged at high risk of bias as they described a non‐random component in the sequence generation process (Yamaji 2019).

Allocation concealment

Three studies were judged to have a low risk for allocation concealment using central randomisation to assign participants to the intervention arms (Di Biase 2014; Nogami 2019; Yu 2019). Seven studies did not describe how randomised interventions were assigned to participants  (Ando 2019; Nagao 2019; Nakamura 2019; Reynolds 2018; Xing 2017; Yamaji 2019; Yoshimura 2017), and two studies were abstract papers and had insufficient information (Tabish 2010; Yoh 2019); all were judged at unclear risk of bias. 

Blinding

Blinding of participants and personnel

There was a high risk of performance bias in eight studies since they were open‐label (Ando 2019; Di Biase 2014; Nakamura 2019; Nogami 2019; Reynolds 2018; Xing 2017; Yoshimura 2017; Yu 2019). We judged two studies at unclear risk of bias since they did not report the blinding of participants and personnel (Nagao 2019; Yamaji 2019). Tabish 2010 and Yoh 2019 are abstract papers and had insufficient information; both were judged at unclear risk of bias.

Blinding of outcome assessment

Seven studies were at low risk of bias since an independent committee of physicians adjudicated outcomes in a blinded manner (Di Biase 2014; Nagao 2019; Nakamura 2019; Nogami 2019; Reynolds 2018; Yoshimura 2017; Yu 2019). Di Biase 2014 reported that "stroke and TIA [transient ischaemic attack] diagnoses were performed by a neurologist who was blinded to the patient's group assignment as well as diagnoses of peripheral embolic events or deep venous thrombosis were performed by other physicians blinded to the group assignment." Nagao 2019 mentioned that MRI images were analysed by an independent radiologist using blinded methods. Nakamura 2019, Nogami 2019, and  Yu 2019 mentioned that a panel of experts adjudicated all endpoints in a blinded manner. Reynolds 2018 reported that an independent endpoints committee reviewed and adjudicated all potential endpoints. A safety officer of the study sponsor and the principal investigator reviewed other procedure‐attributable adverse events not meeting endpoint criteria. Yoshimura 2017 reported that blinded radiologists diagnosed cerebral MRI including DW‐ and T2W‐MRI with or without new or old thromboembolism. 

Three studies provided no information about the blinding of outcome assessors, hence were judged at unclear risk of bias (Ando 2019; Xing 2017; Yamaji 2019). Tabish 2010 and Yoh 2019 are abstract papers and had insufficient information; both were judged at unclear risk of bias.

Incomplete outcome data

We evaluated seven studies at low risk of attrition bias since they had clear participants flow charts and sufficient details about the analysis of participants where all receiving the ablation were analysed (Ando 2019; Di Biase 2014; Nakamura 2019; Nogami 2019; Reynolds 2018; Yoshimura 2017; Yu 2019). The remaining studies provided no participants flow charts to make a judgement  and are at unclear risk of attrition bias (Nagao 2019; Tabish 2010; Xing 2017; Yamaji 2019; Yoh 2019).  

Selective reporting

To judge a study at low risk of reporting bias, we required that the trial was registered, or a protocol was published, where those prespecified outcomes were reported. Four studies were at low risk of reporting bias (Di Biase 2014; Nogami 2019; Reynolds 2018; Yu 2019). In Di Biase 2014, although no prespecified plan was reported for STLs outcome, we did not consider this a major bias as the primary outcomes matched the published protocol for the outcomes of thromboembolic and major bleeding events. Nogami 2019 reported outcomes as per the published protocol. However, the length of hospital stays (one of the secondary outcomes stated in their protocol) was not mentioned in the results (we did not consider this a major bias). In Reynolds 2018 and Yu 2019, outcomes reported matched the published protocols. 

We could not judge the risk of reporting bias for the remaining eight studies because we did not identify study protocols (Ando 2019; Nagao 2019; Nakamura 2019; Tabish 2010; Xing 2017; Yamaji 2019; Yoh 2019; Yoshimura 2017).

Other potential sources of bias

We did not identify any other potential sources of bias.

Effects of interventions

See: Table 1

See Table 1.

Primary outcomes

Thromboembolic events

All 12 studies reported on the primary outcome of thromboembolic events; however, six studies reported zero events and, therefore, six studies were included in the meta‐analysis. Overall, events rates were low in both comparison groups. Pooling the data from the six studies found very low certainty evidence for the risk of thromboembolic events in the interrupted or minimally interrupted compared to uninterrupted anticoagulation (RR 1.76, 95% CI 0.33 to 9.46; I² = 59%; 6 studies, 3468 participants; Analysis 1.1). We downgraded the certainty of the evidence three levels. First, due to the risk of bias in included studies, specifically in the domains of randomisation and allocation concealment. Second, due to heterogeneity that led to inconsistency of results. Several factors could explain the moderate heterogeneity observed in the analysis such as several included studies were relatively small (i.e. fewer than 200 participants), use of different ablation sources and surgical techniques, use of different types of anticoagulation and strategies of interruption, and the difference in populations studied. Third, because of imprecision around the estimate of the effect that could have been driven by the relatively few participants and few events in most of the included studies.

1.1. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 1: Thromboembolic events

Subgroup analysis

Although we planned to perform subgroup analysis if the outcome is reported in 10 studies, we thought that the heterogeneity between the different anticoagulants used and how the interruption was carried out was worth an inspection due to the distinct nature of each type of anticoagulant. Hence, we conducted selective subgroup analysis based on the anticoagulant used and the type of interruption strategy. 

Anticoagulant type

Of the six studies that reported events, we compared the two studies that used VKA in both the interrupted and uninterrupted arms (Di Biase 2014; Xing 2017), the three studies that used DOAC in both arms (Nagao 2019; Nakamura 2019), and the one study that compared VKA in one arm and DOAC in the other arm (Nogami 2019). The test for subgroup differences was not significant (Chi² = 0.86; degrees of freedom (df) = 2 (P = 0.65); I² = 0%; Analysis 1.2).

1.2. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 2: Thromboembolic events (subgroup analysis: anticoagulant type)

Interruption strategy

We analysed two types of interruption strategies, complete interruption with heparin bridging using VKA in one study (Di Biase 2014), and the other five studies that used minimal interruption strategies (Nagao 2019; Nakamura 2019; Nogami 2019; Reynolds 2018; Xing 2017). The test for subgroup differences was significant (Chi² = 10.16; df = 1 (P = 0.001); I² = 90.2%; Analysis 1.3).

1.3. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 3: Thromboembolic events (subgroup analysis: interruption strategy)

Sensitivity analysis

Removing studies with industrial funding

Two studies reported funding from an industrial body (Nogami 2019; Reynolds 2018); excluding these studies did not cause major changes in the results (Analysis 1.4).

1.4. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 4: Thromboembolic events (sensitivity analysis: non‐industrial  funded studies only)

Analysis using the fixed‐effect model

Sensitivity analysis using a fixed‐effect model suggested that interruption of anticoagulation therapy increases the risk of thromboembolic events compared to uninterrupted strategy (RR 5.80, 95% CI 2.68 to 12.54; Analysis 1.5). However, due to moderate heterogeneity and the very low certainty of the evidence, there is uncertainty around this finding. Specifically that the study by Di Biase 2014 differed greatly in the interruption strategy as it is the only study with complete interruption of the anticoagulant VKA with heparin bridging before and after the procedure. It also contributed more weight to the analysis due to its large sample size and high event rate.

1.5. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 5: Thromboembolic events (sensitivity analysis: fixed‐effect model)

Major bleeding

Eleven studies reported major bleeding events (Ando 2019; Di Biase 2014; Nagao 2019; Nakamura 2019; Nogami 2019; Reynolds 2018;  Xing 2017; Yamaji 2019; Yoh 2019; Yoshimura 2017; Yu 2019). We pooled data from 10 studies as Xing 2017 reported zero events therefore it was not included in the meta‐analysis. There is uncertainty in the evidence for the risk of major bleeding in the interrupted compared to uninterrupted anticoagulation (RR 1.10, 95% CI 0.59 to 2.05; I² = 6%; 10 studies, 4584 participants; low‐certainty evidence; Analysis 1.6). We downgraded the certainty of the evidence two levels due to risk of bias, which was observed in the domains of randomisation and allocation concealment in most included studies, and due to imprecision.

1.6. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 6: Major bleeding

Subgroup analyses

Anticoagulation type 

Eight of 10 studies used DOAC (Ando 2019; Nagao 2019; Nakamura 2019; Reynolds 2018; Yamaji 2019; Yoh 2019; Yoshimura 2017; Yu 2019), one study used VKA (Di Biase 2014), and one study included both DOAC and VKA (Nogami 2019). The test for subgroup differences was significant (Chi² = 6.61, df = 2 (P = 0.04); I² = 69.7%; Analysis 1.5).

Baseline risk of stroke using CHADS‐VASc or CHADS2

Only the study by Nogami 2019 reported the outcome of major bleeding using the CHADS‐VASc or CHADS2. In participants with CHADS‐VASc or CHADS2 score of 2 or higher, there were three major bleeding events in the interrupted and uninterrupted arms. For those with CHA2DS2‐VASc or CHADS₂ less than 2, there were eight major bleeding events in the uninterrupted arm and no events in the interrupted arm.

Ablation source of energy, and ablation surgical type

The outcome of major bleeding events in each subgroup, and the tests for subgroup differences were not significant (Analysis 1.8; Analysis 1.9).

1.8. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 8: Major bleeding (subgroup analysis: ablation source of energy (radiofrequency ablation vs cryoablation))

1.9. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 9: Major bleeding (subgroup analysis: ablation surgical type (pulmonary vein isolation vs mixed types))

Interruption strategy 

The outcome of major bleeding events and the test of subgroup differences by interruption strategy were not significant (Analysis 1.10).

1.10. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 10: Major bleeding (subgroup analysis: interruption strategy (complete interruption vs minimal interruption))

Use of heparin bridging

The outcome of major bleeding events and the test of subgroup differences by use of heparin bridging were not significant (Analysis 1.11).

1.11. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 11: Major bleeding (subgroup analysis: interruption strategy (interruption with heparin bridging vs interruption without heparin bridging))

Sensitivity analysis

Removing studies with industrial funding

Excluding industry‐funded studies resulted in no major changes in the results of major bleeding (Analysis 1.12).

1.12. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 12: Major bleeding (sensitivity analysis: non‐industrial funded studies only)

Analysis using the fixed‐effect model

There were no changes in major bleeding events using the fixed‐effect model (RR 1.08, 95% CI 0.63 to 1.85; Analysis 1.13).

1.13. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 13: Major bleeding (sensitivity analysis: fixed‐effect model)

Removing unpublished studies

 We excluded one study that was published as an abstract (Yu 2019). Removing the study did not change the results (Analysis 1.14).

1.14. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 14: Major bleeding (sensitivity analysis: published studies only)

Secondary outcomes

Composite endpoint of thromboembolic events, major bleeding, and all‐cause mortality

One study reported on the composite endpoint of interest, therefore meta‐analysis was not performed for this outcome (Nogami 2019). Three participants in the interrupted arm developed the composite endpoint versus 13 participants in the uninterrupted arm. The analysis demonstrated a lower risk of the composite endpoint in the interrupted compared to uninterrupted arm (RR 0.23, 95% CI 0.07 to 0.81; 1 study, 442 participants; low‐certainty evidence; Analysis 1.15). We downgraded the certainty of the evidence due to imprecision probably caused by the low event rate.

1.15. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 15: Composite endpoint of thromboembolic events, major bleeding, and all‐cause mortality

Minor bleeding

Ten studies reported minor bleeding events (Ando 2019; Di Biase 2014; Nagao 2019; Nakamura 2019; Nogami 2019; Reynolds 2018; Tabish 2010; Xing 2017; Yamaji 2019; Yoh 2019); however, Nogami 2019 did not report minor bleeding as a separate outcome but within a composite endpoint. The analysis suggested that there is uncertainty around the risk of minor bleeding in the interrupted compared to uninterrupted anticoagulation (RR 1.01, 95% CI 0.46 to 2.22; I² = 87%; 9 studies, 3843 participants; very low‐certainty evidence; Analysis 1.16). We downgraded the certainty of the evidence due to the risk of bias, imprecision, and high heterogeneity of the included studies, especially as Di Biase 2014 was the only one of the nine studies that used heparin bridging in the interrupted arm. Also, the different ablation surgical techniques, different types of anticoagulation used, and the different populations studied could explain the high heterogeneity observed in the analysis. Due to the low number of studies that reported the outcome of minor bleeding, subgroup analyses were not performed.

1.16. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 16: Minor bleeding

Cardiovascular mortality

None of the included studies reported participant cardiovascular mortality.

All‐cause mortality

Two studies reported all‐cause mortality (Nogami 2019; Reynolds 2018). None of the participants in either treatment arm died in Reynolds 2018. Only one participant in the uninterrupted arm in Nogami 2019 died, and there were no deaths in the interrupted arm (RR 0.34, 95% CI 0.01 to 8.21; 1 study, 442 participants; low‐certainty evidence; Analysis 1.17). We downgraded the evidence due to imprecision caused by the low event rate and the wide CIs.  No subgroup or sensitivity analyses were performed for this outcome, as only one study reported on the all‐cause mortality endpoint.

1.17. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 17: All‐cause mortality

Asymptomatic thromboembolic events

Six studies reported asymptomatic thromboembolic events (Di Biase 2014; Nagao 2019; Nakamura 2019; Xing 2017; Yoh 2019; Yoshimura 2017). Pooling the data showed that there is uncertainty around the risk of asymptomatic thromboembolism in the interrupted compared to uninterrupted arm (RR 1.45, 95% CI 0.85 to 2.47; I² = 56%; 6 studies, 1268 participants; very low‐certainty evidence; Analysis 1.18). We downgraded the certainty of the evidence by three levels. First due to the risk of bias in included studies, specifically in the domains of randomisation and allocation concealment. Second due to inconsistency owing to a moderate level of statistical heterogeneity. Third due to imprecision due to a wide CI in the effect estimate, which could be explained by the variability in assessing this outcome in the included study. Due to the low number of studies that reported on the outcome of minor bleeding, subgroup analyses were not performed.

1.18. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 18: Asymptomatic thromboembolic events

Patient quality of life

None of the included studies reported on patient QoL.

Economic costs

None of the included studies reported on economic costs.

Length of hospital stay

None of the included studies reported on length of hospital stay.

Discussion

Summary of main results

The current systematic review and meta‐analysis included 12 studies with 4714 participants. The meta‐analysis showed that there is uncertainty in the evidence on the effect of interrupted and uninterrupted anticoagulation therapy for CA in adults with arrhythmias, for the outcomes of thromboembolic events (stroke, systemic embolism), major and minor bleeding, and asymptomatic thromboembolism. There was also uncertainty in the evidence on all‐cause mortality and composite endpoint (thromboembolic events, major bleeding, and all‐cause mortality). However, not all studies reported on all the review outcomes, therefore the number of participants varied across different analyses; furthermore, for most outcomes event rates were small. Overall, the certainty of the evidence for the major outcomes assessed by GRADE ranged from very low to low, mainly due to the risk of bias in included studies, heterogeneity that led to inconsistency of results, and due to imprecision around the estimate of effect (Table 1). 

Overall completeness and applicability of evidence

Several features in the included studies need to be carefully considered before drawing any generalisation. Across all studies, participants with a high risk of stroke or bleeding (or both) were not well‐represented (less than 10%). Ten studies were conducted in the East‐Asian population: Japanese (seven studies), Chinese (two studies), and South Korean (one study). The interethnic variabilities of drug action and the potential risk of bleeding are highly likely. In addition, manufacture recommended doses as well as local guidelines of dose adjustment of DOAC in Asian populations further limit the generalisability of the results. Paroxysmal AF was the most represented type of AF in the studies. Females made up less than a third of any population in the included studies. Representation of various comorbidities was similar between studies, but there was a low number of participants with prior stroke and almost or no representation of participants with renal impairment.

Di Biase 2014 was the only trial that used complete interruption (warfarin held two or three days before ablation and bridging with heparin), whereas the rest of the studies implemented minimal interruption by skipping one or two doses of DOAC with or without heparin bridging or used low‐intensity warfarin (INR 1.5 to 2.0) (surrogate to minimal interruption). It is also the largest study with most events, while the remaining studies reported no or few events.

The protocol of periprocedural management was fairly similar in all studies. All studies prescribed anticoagulant for at least four weeks before the procedure, and eight studies performed TOE prior to ablation. Two studies gave heparin bridging during the interruption of DOAC. This applied to participants with persistent AF (Yu 2019), and in those who had their dabigatran therapy discontinued at least 24 hours before the ablation (Nogami 2019). One study only implemented postprocedural heparin bridging until DOAC was initiated the next morning (Nakamura 2019).

The intra‐procedural use of heparin was reasonably similar between studies, but studies used mixed ablation procedures and diverse sources of ablation energy. Most studies that used mixed DOAC randomised participants based on the interruption strategy rather than the DOAC type. One study included participants who had received radiofrequency (Yoshimura 2017). One study assessed compliance with preprocedural DOAC anticoagulant by self‐reporting (Reynolds 2018). Studies that used VKAs tested the INR of participants prior to the procedure. In Xing 2017, the mean INR in the low‐intensity (INR 1.5 to 2.0), minimally interrupted arm was 1.80 (SD 0.12) on the morning of the procedure.

The heterogeneity observed in some of the study outcomes could have limited the review findings. We performed several subgroup analyses when possible to explore this heterogeneity where the stratification by interruption strategy, the type of anticoagulant, and the use of heparin bridging yielded different results for several of the study outcomes. However, these results need to be interpreted with caution, due to the observational nature of this analysis and the availability of too few studies in each subgroup arm. Di Biase 2014 was the only study that was sufficiently powered to show a difference between the two strategies, and it is currently is the only high‐quality study that showed the safety of uninterrupted VKA compared to the interrupted strategy with heparin bridging.

Eight studies were conducted in single centres; however, these centres were described as experienced with CA procedures. Most studies had underpowered study design limitations. The lack of blindness across all the trials raises the concern of ascertainment bias. However, seven studies used blinded assessment of the outcome (Di Biase 2014; Nagao 2019; Nakamura 2019; Nogami 2019; Reynolds 2018; Yoshimura 2017; Yu 2019). None of the studies had loss to follow‐up, but there was considerable variability in the period of follow‐up, which varied from one hour to 12 months postablation. 

Quality of the evidence

Overall risk of bias across trials can be interpreted as plausible bias with 8/12 included trials having a high or unclear risk of bias in several domains (open‐label design, unclear allocation concealment, and lack of published protocol in eight trials). Studies used different anticoagulant  drugs with a potential difference in their clinical effects. Patients' baseline characteristics regarding risks of stroke, bleeding risk, and study duration varied between trials, and 8/12 trials were conducted in single centres and 10/12 trials described interventions in Asian populations.

Limitations in study design or execution (risk of bias)

For all the outcomes, we downgraded the certainty of evidence by one level due to the risk of bias in included studies, specifically in the domains of randomisation, allocation concealment, and blinding.

Inconsistency of results

We downgraded the certainty of the evidence for thromboembolic events by one level for the inconsistency of results (I² = 59%). We also downgraded the certainty of the evidence for minor bleeding by one level for inconsistency, because of the very high unexplained heterogeneity (I² = 87%). Subgroup analyses were not performed for these outcomes due to the small number of studies reporting them. Although such analyses (e.g. comparing the effect of outcomes by different types of anticoagulation) might have explained the high heterogeneity observed.

Imprecision

We downgraded the certainty of evidence one level for all outcomes due to the width of the CIs that did not exclude potential benefit and harm, and due to the low event rate in general. 

Publication bias

For all outcomes, we did not downgrade the certainty of the evidence for publication bias, as we did not detect it, although the small number of included studies may have prevented this detection.

Potential biases in the review process

We included only randomised trials. Most of these studies described the limitation of sample size, were open‐label, and had unclear allocation concealment. The population studied, the mix of comparators anticoagulants, and ablation procedures were different among studies. We did not conduct a funnel plot to assess publication bias as the number was fewer than 10 trials in each outcome; however, we believe that unpublished data is a potential bias given the small sample size and the non‐significant findings of most of the included studies. 

Agreements and disagreements with other studies or reviews

There are several other systematic reviews with meta‐analyses of various comparisons of interrupted, minimally interrupted, and completely interrupted DOAC and VKA (Basu‐Ray 2020; Gorla 2018; Ha 2018; Ottóffy 2020; Santarpia 2015; Yang 2020; Zhao 2017). However, the relevance of the findings from these meta‐analyses is not directly relevant to our review because they included both RCTs and observational studies and there was no pooling of data from RCTs that matched our inclusion criteria.  Although Ottóffy 2020 described pooled analysis from RCTs and showed no difference between interrupted and uninterrupted strategies in thromboembolic events (Peto odds ratio (OR) 0.66, 95% CI 0.17 to 2.64), and lower risk of major bleeding events (Peto OR 0.36, 95% CI 0.21 to 0.62) in the uninterrupted and minimally interrupted DOAC. However, this meta‐analysis included single strategy studies, which we excluded in our review.

Mao 2020 conducted a meta‐analysis on observational studies and six RCTs (all RCTs are included in our review) evaluating the safety and efficacy of minimally interrupted and uninterrupted DOAC. Pooling data from the six RCTs showed no significant difference on outcomes of major bleeding  (OR 0.74, 95% CI 0.32 to 1.70; P = 0.47; I² = 0%) and minor bleeding (OR 1.06, 95% CI 0.70 to 1.60; P = 0.80; I² = 0%). The meta‐analysis did not include Nogami 2019 or Yamaji 2019, both are randomised and evaluated minimally interrupted versus interrupted DOAC. A second meta‐analysis by Mao 2021 included observational studies, six RCTs, and one randomised registry (all are included in our review). Data from pooled RCTs produced similar findings on major bleeding with no significant difference between the two strategies (OR 1.41, 95% CI 0.73 to 2.73; P = 0.30; I² = 0%). Although the author concluded that uninterrupted DOAC or VKA had fewer SCI compared to minimally interrupted, there was no pooled effect from RCTs only. Moreover, the meta‐analysis also did not include Yamaji 2019.

Authors' conclusions

Implications for practice.

In adults with atrial fibrillation undergoing catheter ablation, we found uncertain evidence to support one strategy over another. This uncertainty precluded us from providing a strong recommendation for the medical community. Nonetheless, reflecting on the types of anticoagulants used and how these drugs were interrupted is worth a close inspection. For completely interrupted strategy, the single largest study using vitamin K antagonists, which contributed the most events in this review, concluded that uninterrupted strategy was associated with a lower risk of preprocedural thromboembolism without increasing the risk of major bleeding, particularly in people with long‐standing persistent atrial fibrillation. For studies that utilised minimal interruption, skipping one or two doses of the direct oral anticoagulants or using low‐intensity international normalised ratios for vitamin K antagonists, the low event rate in outcomes of thromboembolism and major bleeding can be attributed to the nature of the strategy. Minimal interruption implies that the anticoagulant effect may still be persistent to prevent thromboembolism without increasing bleeding risks, which in general terms is similar to the uninterrupted approach (DeLoughery 2011). At this time, we advise clinicians and patients to select either strategy while considering each patient's characteristics before embarking on either one.

Implications for research.

Based on the 12 studies included in this review, the evidence is insufficient to inform the decision about the safety and harms of one strategy over the other (interrupted versus uninterrupted). The certainty of the evidence was very low for the thromboembolic risk and low for the major bleeding risk because of the low event rates observed in the available trials; therefore, there is a need for large well‐powered and well‐designed clinical trials to overcome this issue. Furthermore, future trials should consider including patients with comorbidities and higher risks and most importantly compare the same anticoagulant in standard doses. Moreover, trials investigating the optimal length of interruption before the procedure, especially for DOAC, 24 versus 48 hours, and the needs for heparin bridging are still needed. Furthermore, our findings cannot be extrapolated to other invasive cardiac and non‐cardiac procedures such as percutaneous coronary and non‐coronary interventions, invasive gastrointestinal and urological procedures, and others. These need to be addressed individually through well‐designed clinical trials and systematic reviews. Additionally, studying other important factors such as cost‐effectiveness and patient convenience and preferences deserve consideration by future clinical trials. 

What's new

Date	Event	Description
5 November 2021	Amended	Minor edits to abstract and plain language summary.

History

Protocol first published: Issue 12, 2019
Review first published: Issue 10, 2021

Appendices

Appendix 1. Search strategies

CENTRAL

#1 MeSH descriptor: [Anticoagulants] explode all trees

#2 Anticoagulants*

#3 (OAC or OACs or DOAC or DOACs)

#4 (vitamin k NEAR/3 antagonist*)

#5 vitamin k inhibitor*

#6 vka

#7 antivitamin k

#8 Heparin

#9 LMWH

#10 warfarin

#11 dabigatran

#12 rivaroxaban

#13 apixaban

#14 edoxaban

#15 #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #9 or #10 or #11 or #12 or #13 or #14

#16 MeSH descriptor: [Arrhythmias, Cardiac] explode all trees

#17 Arrhythmia*

#18 atrial flutter*

#19 tachycardia*

#20 tachyarrhythmia*

#21 Dysrhythmia*

#22 MeSH descriptor: [Atrial Fibrillation] this term only

#23 ((atrial or atrium or auricular) NEAR/2 fibrillat*)

#24 #16 or #17 or #18 or #19 or #20 or #21 or #22 or #23

#25 MeSH descriptor: [Catheter Ablation] this term only

#26 catheter ablat*

#27 (percutaneous NEAR/3 catheter*)

#28 #25 or #26 or #27

#29 #15 and #24 and #28

MEDLINE Ovid

1 exp Anticoagulants/

2 Anticoagulants*.tw.

3 (OAC or OACs or DOAC or DOACs).tw.

4 (vitamin k adj3 antagonist*).tw.

5 vitamin k inhibitor*.tw.

6 vka.tw.

7 antivitamin k.tw.

8 Heparin.tw.

9 LMWH.tw.

10 warfarin.tw.

11 dabigatran.tw.

12 rivaroxaban.tw.

13 apixaban.tw.

14 edoxaban.tw.

15 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14

16 exp Arrhythmias, Cardiac/

17 Arrhythmia*.tw.

18 atrial flutter*.tw.

19 tachycardia*.tw.

20 tachyarrhythmia*.tw.

21 Dysrhythmia*.tw.

22 Atrial Fibrillation/

23 ((atrial or atrium or auricular) adj2 fibrillat*).tw.

24 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23

25 Catheter Ablation/

26 catheter ablat*.tw.

27 (percutaneous adj3 catheter*).tw.

28 25 or 26 or 27

29 15 and 24 and 28

30 randomized controlled trial.pt.

31 controlled clinical trial.pt.

32 randomized.ab.

33 placebo.ab.

34 drug therapy.fs.

35 randomly.ab.

36 trial.ab.

37 groups.ab.

38 30 or 31 or 32 or 33 or 34 or 35 or 36 or 37

39 exp animals/ not humans.sh.

40 38 not 39

41 29 and 40

Embase Ovid

1 exp anticoagulant agent/

2 Anticoagulants*.tw.

3 (OAC or OACs or DOAC or DOACs).tw.

4 (vitamin k adj3 antagonist*).tw.

5 vitamin k inhibitor*.tw.

6 vka.tw.

7 antivitamin k.tw.

8 Heparin.tw.

9 LMWH.tw.

10 warfarin.tw.

11 dabigatran.tw.

12 rivaroxaban.tw.

13 apixaban.tw.

14 edoxaban.tw.

15 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14

16 exp heart arrhythmia/

17 Arrhythmia*.tw.

18 atrial flutter*.tw.

19 tachycardia*.tw.

20 tachyarrhythmia*.tw.

21 Dysrhythmia*.tw.

22 atrial fibrillation/

23 ((atrial or atrium or auricular) adj2 fibrillat*).tw.

24 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23

25 catheter ablation/

26 catheter ablat*.tw.

27 (percutaneous adj3 catheter*).tw.

28 25 or 26 or 27

29 15 and 24 and 28

30 random$.tw.

31 factorial$.tw.

32 crossover$.tw.

33 cross over$.tw.

34 cross‐over$.tw.

35 placebo$.tw.

36 (doubl$ adj blind$).tw.

37 (singl$ adj blind$).tw.

38 assign$.tw.

39 allocat$.tw.

40 volunteer$.tw.

41 crossover procedure/

42 double blind procedure/

43 randomized controlled trial/

44 single blind procedure/

45 30 or 31 or 32 or 33 or 34 or 35 or 36 or 37 or 38 or 39 or 40 or 41 or 42 or 43 or 44

46 (animal/ or nonhuman/) not human/

47 45 not 46

48 29 and 47

SCI‐EXPANDED

#27 #26 AND #25

#26 TS=(random* or blind* or allocat* or assign* or trial* or placebo* or crossover* or cross‐over*)

#25 #24 AND #21 AND #14

#24 #23 OR #22

#23 TS=(percutaneous near/3 catheter*)

#22 TS=catheter ablat*

#21 #20 OR #19 OR #18 OR #17 OR #16 OR #15

#20 TS=((atrial or atrium or auricular) near/2 fibrillat*)

#19 TS=Dysrhythmia*

#18 TS=tachyarrhythmia*

#17 TS=tachycardia*

#16 TS=atrial flutter*

#15 TS=Arrhythmia*

#14 #13 OR #12 OR #11 OR #10 OR #9 OR #8 OR #7 OR #6 OR #5 OR #4 OR #3 OR #2 OR #1

#13 TS=edoxaban

#12 TS=apixaban

#11 TS=rivaroxaban

#10 TS=dabigatran

#9 TS=warfarin

#8 TS=LMWH

#7 TS=Heparin

#6 TS=antivitamin k

#5 TS=vka

#4 TS=vitamin k inhibitor*

#3 TS=(vitamin k near/3 antagonist*)

#2 TS=(OAC or OACs or DOAC or DOACs)

#1 TS=Anticoagulants*

ClinicalTrials.gov

Condition or disease: arrhythmia

Other terms: randomised

Study type: Interventional Studies

Study Results: All Studies

Age: adult (18–64), and older adult (65+)

Intervention/treatment: catheter ablation

World Health Organization International Clinical Trials Registry Platform

In the Condition: arrhythmia

In the Intervention: catheter ablation 

Recruitment Status: ALL

Clinical Trials Register EU

Search terms: arrhythmia AND catheter ablation

Age range: adult and elderly 

Data and analyses

Comparison 1. Interrupted versus uninterrupted anticoagulation.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Thromboembolic events	6	3468	Risk Ratio (M‐H, Random, 95% CI)	1.76 [0.33, 9.46]
1.2 Thromboembolic events (subgroup analysis: anticoagulant type)	6	3468	Risk Ratio (M‐H, Random, 95% CI)	1.76 [0.33, 9.46]
1.2.1 Vitamin K antagonist vs vitamin K antagonist	2	1685	Risk Ratio (M‐H, Random, 95% CI)	3.20 [0.06, 180.32]
1.2.2 Direct oral anticoagulant vs direct oral anticoagulant	3	1341	Risk Ratio (M‐H, Random, 95% CI)	1.36 [0.26, 7.23]
1.2.3 Vitamin K antagonist vs direct oral anticoagulant	1	442	Risk Ratio (M‐H, Random, 95% CI)	0.34 [0.01, 8.21]
1.3 Thromboembolic events (subgroup analysis: interruption strategy)	6	3468	Risk Ratio (M‐H, Random, 95% CI)	1.76 [0.33, 9.46]
1.3.1 Completely interrupted	1	1584	Risk Ratio (M‐H, Random, 95% CI)	19.60 [4.75, 80.89]
1.3.2 Minimally interrupted	5	1884	Risk Ratio (M‐H, Random, 95% CI)	0.82 [0.22, 3.14]
1.4 Thromboembolic events (sensitivity analysis: non‐industrial  funded studies only)	4	2731	Risk Ratio (M‐H, Random, 95% CI)	2.74 [0.34, 22.11]
1.5 Thromboembolic events (sensitivity analysis: fixed‐effect model)	6	3468	Risk Ratio (M‐H, Fixed, 95% CI)	5.80 [2.68, 12.54]
1.6 Major bleeding	10	4584	Risk Ratio (M‐H, Random, 95% CI)	1.10 [0.59, 2.05]
1.7 Major bleeding (subgroup analysis: anticoagulation type (direct oral anticoagulant vs vitamin K antagonist))	10	4584	Risk Ratio (M‐H, Random, 95% CI)	1.10 [0.59, 2.05]
1.7.1 Direct oral anticoagulant	8	2558	Risk Ratio (M‐H, Random, 95% CI)	1.34 [0.61, 2.95]
1.7.2 Vitamin K antagonist	1	1584	Risk Ratio (M‐H, Random, 95% CI)	2.68 [0.71, 10.07]
1.7.3 Direct oral anticoagulant and vitamin K antagonist	1	442	Risk Ratio (M‐H, Random, 95% CI)	0.28 [0.08, 0.97]
1.8 Major bleeding (subgroup analysis: ablation source of energy (radiofrequency ablation vs cryoablation))	10	4584	Risk Ratio (M‐H, Random, 95% CI)	1.10 [0.59, 2.05]
1.8.1 Radiofrequency ablation	2	1784	Risk Ratio (M‐H, Random, 95% CI)	2.72 [0.80, 9.25]
1.8.2 Cryoablation	1	97	Risk Ratio (M‐H, Random, 95% CI)	0.49 [0.03, 7.62]
1.8.3 Mixed ablation sources	7	2703	Risk Ratio (M‐H, Random, 95% CI)	0.88 [0.41, 1.85]
1.9 Major bleeding (subgroup analysis: ablation surgical type (pulmonary vein isolation vs mixed types))	10	4584	Risk Ratio (M‐H, Random, 95% CI)	1.10 [0.59, 2.05]
1.9.1 Pulmonary vein isolation	4	2822	Risk Ratio (M‐H, Random, 95% CI)	1.86 [0.79, 4.36]
1.9.2 Mixed	6	1762	Risk Ratio (M‐H, Random, 95% CI)	0.69 [0.29, 1.68]
1.10 Major bleeding (subgroup analysis: interruption strategy (complete interruption vs minimal interruption))	10	4584	Risk Ratio (M‐H, Random, 95% CI)	1.10 [0.59, 2.05]
1.10.1 Complete interruption	1	1584	Risk Ratio (M‐H, Random, 95% CI)	2.68 [0.71, 10.07]
1.10.2 Minimal interruption	8	2674	Risk Ratio (M‐H, Random, 95% CI)	0.91 [0.43, 1.91]
1.10.3 Mixed	1	326	Risk Ratio (M‐H, Random, 95% CI)	0.72 [0.12, 4.26]
1.11 Major bleeding (subgroup analysis: interruption strategy (interruption with heparin bridging vs interruption without heparin bridging))	10	4584	Risk Ratio (M‐H, Random, 95% CI)	1.10 [0.59, 2.05]
1.11.1 Interruption with heparin bridging	1	1584	Risk Ratio (M‐H, Random, 95% CI)	2.68 [0.71, 10.07]
1.11.2 Interruption without heparin bridging	7	2232	Risk Ratio (M‐H, Random, 95% CI)	1.57 [0.65, 3.77]
1.11.3 Interruption with heparin bridging in subset of participants	2	768	Risk Ratio (M‐H, Random, 95% CI)	0.38 [0.14, 1.07]
1.12 Major bleeding (sensitivity analysis: non‐industrial funded studies only)	8	3847	Risk Ratio (M‐H, Random, 95% CI)	1.62 [0.78, 3.37]
1.13 Major bleeding (sensitivity analysis: fixed‐effect model)	10	4584	Risk Ratio (M‐H, Fixed, 95% CI)	1.08 [0.63, 1.85]
1.14 Major bleeding (sensitivity analysis: published studies only)	9	4479	Risk Ratio (M‐H, Random, 95% CI)	1.06 [0.55, 2.05]
1.15 Composite endpoint of thromboembolic events, major bleeding, and all‐cause mortality	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected
1.16 Minor bleeding	9	3843	Risk Ratio (M‐H, Random, 95% CI)	1.01 [0.46, 2.22]
1.17 All‐cause mortality	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected
1.18 Asymptomatic thromboembolic events	6	1268	Risk Ratio (M‐H, Random, 95% CI)	1.45 [0.85, 2.47]

1.7. Analysis.

Comparison 1: Interrupted versus uninterrupted anticoagulation, Outcome 7: Major bleeding (subgroup analysis: anticoagulation type (direct oral anticoagulant vs vitamin K antagonist))

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Ando 2019.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Total duration of the study: July 2014 to February 2017 Duration of follow‐up: 30 days after procedure Number of study centres and location: 1 in Japan
Participants	Inclusion criteria: anticoagulation with apixaban 5 mg twice daily; aged 20–85 years; any gender Exclusion criteria: received low‐dose apixaban or other OACs, and an inappropriately reduced dose of apixaban, which may lead to an underestimation of the complication risk; aged < 20 or > 85 years; presence of intracavitary thrombus; uncontrolled heart failure; with prosthetic heart valve or haemodynamically significant valvular disease; advanced liver disease; estimated glomerular filtration rate < 15 mL/minute/1.73 m2; any contraindication of the procedure Total number of participants: 97 Number of randomised participants: 97 Number lost to follow‐up/withdrawn: 0 Number of analysed participants: 97 Number of participants in each treatment group: interrupted apixaban: 65; uninterrupted apixaban: 32 Baseline characteristics Interrupted anticoagulation Age, years, mean: 66.4 (SD 10.2) Men, n (%): 49 (75.4) CHA2DS2‐VASc score, n (%): 0–1 = 19 (29.2) 2 = 17 (26.2) 3–4 = 25 (38.4) 5–9 = 4 (6.2) CHADS2 score, n (%): 0–1 = 49 (75.4) 2 = 10 (15.4) 3–6 = 6 (9.2) HAS‐BLED score, n (%): 0–2 = 60 (92.3) 3–9 = 5 (7.7) Comorbidities, n (%): diabetes mellitus: 9 (13.9) hypertension: 28 (56) stroke or TIA, or both: 6 (9.2) previous bleeding: 3 (4.6) dyslipidaemia: 17 (26.2) CHF: 5 (7.7) structural heart disease: 8 (12.25) Concomitant medications, n (%): anti‐arrhythmic drugs: 38 (58.5) beta‐blockers: 31 (47.7) ACE inhibitors/ARBs: 23 (35.4) antiplatelet drug: 7 (10.8) proton‐pump inhibitors: 20 (30.8) NSAIDs: 1 (1.5) statins: 19 (29.2) Uninterrupted anticoagulation Age, years, mean: 67.2 (SD 10.8) Men, n (%): 26 (81.3) CHA2DS2‐VASc score, n (%): 0–1 = 14 (43.8) 2 = 8 (25.0) 3–4 = 7 (21.9) 5–9 = 3 (9.3) CHADS2 score, n (%): 0–1 = 23 (71.9) 2 = 5 (15.6) 3–6 = 4 (12.5) HAS‐BLED score, n (%): 0–2 = 30 (93.7) 3–9 = 2 (6.3) Comorbidities, n (%): diabetes mellitus: 4 (12.5) hypertension: 18 (56.3) stroke or TIA, or both: 2 (6.3) previous bleeding: 1 (3.1) dyslipidaemia: 6 (18.8) CHF: 2 (6.3) structural heart disease: 3 (9.3) Concomitant medications, n (%): anti‐arrhythmic drugs: 16 (50.0) beta‐blockers: 16 (50.0) ACE inhibitors/ARBs: 8 (25.0) antiplatelet drug: 2 (6.3) proton‐pump inhibitors: 10 (31.3) NSAIDs: 0 (0.0) statins: 6 (18.8) Group differences: baseline characteristics and risk factors well‐balanced between groups
Interventions	Periprocedural anticoagulation Anticoagulant used: apixaban Dose: 5 mg twice daily Duration of anticoagulant therapy in trial: all participants treated with apixaban for ≥ 4 weeks before ablation procedure Intensity of anticoagulation or dose adjustment: none Adherence to anticoagulant treatment: not reported Time of interruption of the anticoagulants prior to procedure: apixaban interrupted on morning of procedure Heparin bridge therapy: not used Time and strategy of resumption of interrupted OAC after procedure: apixaban was administered in the evening of the procedure as usual in both groups 1 and 2 TOE: performed in all participants before ablation Ablation procedure Type of ablation: PVI Ablation energy source: cryoablation Intraprocedural anticoagulation Heparin IV 100 U/kg bolus dose immediately after inserting all the sheaths. ACT monitored every 20 minutes. Additional heparin administered to maintain ACT at 300–350 seconds. Protamine used to reverse the heparin effect at end of procedure, and then, all the sheaths were removed from participant
Outcomes	Primary outcome Prothrombotic response assessed on the basis of d‐dimer level Secondary outcomes Bleeding complications defined by ISTH Major bleeding: fatal bleeding or symptomatic bleeding (or both) in critical areas or organs, such as intracranial, intraspinal, intraocular, retroperitoneal, intra‐articular or pericardial, or intramuscular bleeding with compartment syndrome or bleeding that caused a decrease in haemoglobin level of ≥ 20 g/L (or both), or led to transfusion of ≥ 2 units of whole blood or red cells Minor bleeding: small groin haematomas and pericardial effusion that did not require intervention Thromboembolic events Symptomatic ischaemic strokes and TIAs, after ruling out intracranial haemorrhage by computed tomography, and systemic emboli
Notes	Investigators' conflicts of interest: not reported Funding: not reported Country: Japan Setting: Cardiovascular Center of Nagoya Daini Red Cross Hospital Comments: none Author's name: Yasuya Inden Institution: Department of Cardiology, Nagoya University Graduate School of Medicine Email: inden@med.nagoya‐u.ac.jp Address: Department of Cardiology, Nagoya University Graduate School of Medicine, 65 Tsurumai‑cho, Showa‑ku, Nagoya, Aichi 466‑8550, Japan
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Authors stated they used blocked randomisation but the process of selecting the blocks, such as a random number table or a computer random number generator, was not specified.
Allocation concealment (selection bias)	Unclear risk	Methods of allocation concealment not reported.
Blinding of participants and personnel (performance bias) all outcomes	High risk	Open‐label study.
Blinding of outcome assessment (detection bias) all outcomes	Unclear risk	No information provided to make judgement.
Incomplete outcome data (attrition bias) all outcomes	Low risk	Participants' flow charts were provided, there was no loss to follow‐up and outcomes were reported on all included participants.
Selective reporting (reporting bias)	Unclear risk	No published protocol. However, all expected outcomes were reported, unlikely outcome of interest were missing.

Di Biase 2014.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Total duration of the study: January 2010 to April 2014 Duration of follow‐up: 48 hours for the stroke events Number of study centres and location: 7 centres (Texas Cardiac Arrhythmia Research Foundation, University of Kansas, California Pacific Medical Center, Stanford University, Case Western Reserve University, Southlake Regional Health Centre, Catholic University, Italy)
Participants	Inclusion criteria: aged ≥ 18 years, INR 2.0–3.0 at 3–4 weeks before ablation, and CHADS2 score ≥ 1 Exclusion criteria: known bleeding disorders or inherited thrombophilic disorder, oral contraceptives or oestrogen replacement therapy, prosthetic heart valves, and contraindications to warfarin therapy Number of randomised participants: 1584 Total number of participants: 1584 Number lost to follow‐up/withdrawn: 0 Number of analysed participants: 1584 Number of participants in each treatment group: warfarin interrupted: 790; warfarin uninterrupted: 794 Baseline characteristics Interrupted anticoagulation Age, years, mean: 61 (SD 10) Men, n (%): 602 (76) Type of AF, n (%): paroxysmal: 229 (29) persistent: 174 (22) long‐standing persistent: 387 (49) Comorbidities, n (%): CAD: 182 (23) hypertension: 640 (81) CHF: 118 (15) diabetes mellitus: 302 (38) prior stroke/TIA: 55 (7) CHADS2 score, n (%): 1 = 229 (29) 2 = 268 (34) 3 = 170 (22) 4 = 94 (12) ≥ 5 = 32 (4.1) Uninterrupted anticoagulation Age, years, mean: 62 (SD 12) Men, n (%): 590 (74) Type of AF, n (%): paroxysmal: 200 (25) persistent: 189 (24) long‐standing persistent: 405 (51) Comorbidities, n (%): CAD: 206 (26) hypertension: 660 (83) CHF: 136 (17) diabetes mellitus: 318 (40) prior stroke/TIA: 64 (8) CHADS2 score, n (%): 1 = 206 (26) 2 = 284 (36) 3 = 152 (19) 4 = 101 (13) ≥ 5 = 48 (6.0) Group differences: baseline characteristics and risk factors well balanced between groups
Interventions	Periprocedural anticoagulation Anticoagulant used: warfarin Dose: as per the participant's scheduled dose Duration of anticoagulant therapy in trial: all participants received warfarin before the procedure to achieve 3–4 weeks of therapeutic INRs Intensity of anticoagulation or dose adjustment: participants had to have a therapeutic INR. If, on the day of the procedure, participants had an INR > 3.5, they were excluded. If the INR was 3–3.5, fresh‐frozen plasma was administered a few hours before the procedure. Some participants presented on the day of the procedure with a subtherapeutic INR and were not excluded Adherence to anticoagulant treatment: not reported Time of interruption of the anticoagulants prior to procedure: warfarin was discontinued 2–3 days before the ablation Heparin bridge therapy: participants in group 1 (off‐warfarin) were bridged with enoxaparin 1 mg/kg twice daily (a low molecular weight heparin) until the evening before the ablation procedure. After the procedure, enoxaparin 0.5 mg/kg twice daily was routinely started. It was stopped when the INR was > 2. Warfarin was restarted the night of the procedure Time and strategy of resumption of OAC after procedure: warfarin was restarted the night of the procedure Duration of follow‐up: 48 hours' postprocedure TOE: performed in all participants in group 1 and when the participant presented with a subtherapeutic INR on the day of the procedure in group 2 Ablation procedure Type of ablation: pulmonary vein antrum isolation Ablation energy source: RFCA Intraprocedural anticoagulation Group 1 received 15,000 international units IV and continuous infusion of heparin 1000 units/hour was started to maintain an ACT > 350 seconds Group 2 received bolus of 10,000 IU in men and 8000 IU in women. ACT was maintained at > 300 seconds. In both groups, transseptal sheaths were pulled when the ACT was < 200 seconds
Outcomes	Primary outcome Thromboembolic events (during 48 hours after ablation), defined as stroke/TIA or systemic thromboembolism Stroke was defined as the onset of a new neurological deficit that occurred anytime during or within 48 hours of the procedure. If the duration of the deficit was < 24 hours, it was defined as a TIA Secondary outcomes Bleeding complications defined as major (requiring intervention) occurrence of cardiac tamponade or haemopericardium requiring intervention, causing symptoms, or requiring transfusion; haematoma requiring intervention; massive haemoptysis; haemothorax; and retroperitoneal bleeding minor (not requiring intervention) bleeding occurrence of haematoma or any bleeding that did not require any intervention or did not cause any symptoms Pericardial effusions were analysed separately as a secondary endpoint of the study Silent thromboembolic lesion: no definition reported
Notes	Investigators' conflicts of interest: quote: "Dr Di Biase serves as a consultant for Hansen Medical, Biosense‐Webster, and St. Jude Medical. Dr Di Biase also received speaker honoraria from Biotronik and Atricure. Dr Gallinghouse is a consultant for Hansen Medical. Dr Natale has received honoraria for serving on the speakers’ bureau for St. Jude Medical, Boston Scientific, Medtronic, and Biosense‐Webster. Dr Natale is consultant for Biosense Webster and St. Jude Medical. The other authors report no conflicts" Funding: Texas Cardiac Arrhythmia Research Foundation Country: US Setting: St David's Medical Center Comments: none Author's name: Andrea Natale Institution: Texas Cardiac Arrhythmia Institute Email: dr.natale@gmail.com Address: 3000 N I‐35, Ste 720, Austin, Texas, USA, 78705
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "block randomization was performed with study center as the blocking variable. A central randomization algorithm was used to generate the randomization code."
Allocation concealment (selection bias)	Low risk	Used a central randomisation algorithm.
Blinding of participants and personnel (performance bias) all outcomes	High risk	Quote: "operators were not blinded to the anticoagulation management, which introduced a bias in the study."
Blinding of outcome assessment (detection bias) all outcomes	Low risk	Quote: "Stroke and TIA diagnoses were performed by a neurologist who was blinded to the patient's group assignment. Diagnoses of peripheral embolic events or deep venous thrombosis were performed by other physicians blinded to the group assignment."
Incomplete outcome data (attrition bias) all outcomes	Low risk	Participant flow chart was provided, no lost to follow‐up, no missing data.
Selective reporting (reporting bias)	Low risk	The primary outcome matched the protocol based on the history of changes. Although there was no prespecified plan for silent thromboembolic lesions outcome, we did not consider this a major bias given the importance of primary outcomes.

Nagao 2019.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Total duration of the study: April 2015 to January 2018 Duration of follow‐up: 1 month Number of study centres and location: single centre in Japan
Participants	Inclusion criteria: people with non‐valvular AF prescribed DOAC at the Chubu Rosai Hospital Exclusion criteria: with CrCl < 15 mL/minute Number of randomised participants: 200 Total number of participants: 200 Number lost to follow‐up/withdrawn: 0 Number of analysed participants: 200 Number of participants in each treatment group: uninterrupted use group: 100; interrupted by 1 dose group: 100 Baseline characteristics Interrupted anticoagulation Age, years, mean: 70 (SD 28) Men, n (%): 62 (62) Paroxysmal AF, n (%): 59 (59) Comorbidities, n (%): CAD: 9 (9) hypertension: 50 (50) diabetes mellitus: 29 (29) history of heart failure: 14 (14) prior stroke/TIA: 10 (10) CHADS2 score, n (%): 0 = 26 (26) 1 = 31 (31) ≥ 2 = 43 (43) CHA2DS2‐VASc score, mean: 2.6 (SD 1.5) Concomitant medications, n (%): rivaroxaban/edoxaban, n (%): 53 (53) apixaban, n (%): 47 (47) antiplatelet drugs: 8 (8) PPI/H2RA: 31 (31) beta‐blocker: 43 (43) ACE inhibitor/ARB: 26 (26) antiarrhythmic drug: 29 (29) Uninterrupted anticoagulation Age, years, mean: 70 (SD 29) Men, n (%): 64 (64) Paroxysmal AF, n (%): 57 (57) Comorbidities, n (%): CAD: 12 (12) hypertension: 55 (55) diabetes mellitus: 33 (33) history of heart failure: 15 (15) prior stroke/TIA: 7 (7) CHADS2 score, n (%): 0 = 23 (23) 1 = 27 (27) ≥ 2 = 50 (50) CHA2DS2‐VASc score, mean: 2.8 (SD 1.6) Concomitant medications, n (%): rivaroxaban/edoxaban: 49 (49) apixaban: 51 (51) antiplatelet drugs: 6 (6) PPI/H2RA: 27 (27) beta‐blocker: 51 (51) ACE inhibitor/ARB: 28 (28) antiarrhythmic drug: 33 (33) Group differences: overall, no significant difference between groups
Interventions	Periprocedural anticoagulation Anticoagulants studied: apixaban, rivaroxaban, and edoxaban Dose: apixaban twice daily (at 7 a.m. and 7 p.m.), rivaroxaban once daily (at 7 a.m.), and edoxaban once daily (at 7 a.m.) Duration of anticoagulant therapy in trial: not reported Intensity of anticoagulation or dose adjustment: doses were adjusted based on renal function: low‐dose rivaroxaban (10 mg once daily) administered to people with mild renal dysfunction (Accra 30–50 mL/minute). Apixaban dose was decided according to age, bodyweight, or renal function. For example, low apixaban dose (2.5 mg twice a day) was administered to people with any 2 of the following characteristics: advanced age (≥ 80 years), renal dysfunction (serum creatinine concentration ≥ 1.5 mg/dL), and lower bodyweight (≤ 60 kg). Moreover, a low dose of edoxaban (30 mg once a day) was used in people with 1 of the following characteristics: mild renal dysfunction (CrCl 30–50 mL/minute), lower bodyweight (≤ 60 kg), and concomitant use of P‐glycoprotein inhibitors Adherence to anticoagulant treatment: not reported Time of interruption of the anticoagulants prior to procedure: DOAC was withheld for 1 dose before the procedure Heparin bridge therapy: not used Time and strategy of resumption of interrupted OAC after the procedure: resumption of anticoagulation was the next schedule dose after procedure TOE: performed in all participants Ablation procedure Type of ablation: cavotricuspid isthmus ablation and PVI Ablation energy source: RFCA Intraprocedural anticoagulant IV bolus heparin 100 U/kg immediately after inserting all sheaths. ACT was monitored every 15 minutes, target ACT was 300 seconds. Additional heparin was administered by bolus injection every 15 minutes until the target ACT was achieved. ACT was measured every 20 minutes after achieving the target. The additional dose of heparin was decided as 60 U/kg for ACT < 200 seconds, 50 U/kg for ACT 200–250 seconds, or 40 U/kg for ACT 250–300 seconds
Outcomes	Incidence of silent stroke detected by postoperative MRI Perioperative trends in coagulation markers compared with the interrupted strategy Major bleeding complications: defined as any bleeding requiring blood transfusion, surgical intervention, and pericardial effusion with drainage Minor bleeding complications: defined as small groin or subclavian haematoma and pericardial effusion that did not require any intervention Thromboembolic complications: defined as symptomatic ischaemic strokes and TIAs were classified as thromboembolic complications after intracranial haemorrhage was ruled out by CT Periprocedural complications defined as adverse events that occurred within 1 month before or after the ablation procedure
Notes	Investigators' conflicts of interest: not reported Funding: not reported Country: Japan Setting: Chubu Rosai Hospital Comments: none Author's name: Tomoyuki Nagao Institution: Department of cardiology, Chubu Rosai Hospital Email: cyphernation@yahoo.co.jp Address: 10‐6 1‐Chome Komei, Minato‐ku, Nagoya, Japan
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "The randomization was done using a block‐randomization method."  Comment: however, the process of selecting the blocks was not specified.
Allocation concealment (selection bias)	Unclear risk	Methods of allocation concealment not reported.
Blinding of participants and personnel (performance bias) all outcomes	Unclear risk	Insufficient information to judge.
Blinding of outcome assessment (detection bias) all outcomes	Low risk	Quote: "The only blinded aspects is for MRI images that were analysed by an independent radiologist in a blinded fashion."
Incomplete outcome data (attrition bias) all outcomes	Unclear risk	Outcomes were reported on all enrolled participants (number enrolled = number analysed). However, no participant flow chart was provided to determine if there was any attrition.
Selective reporting (reporting bias)	Unclear risk	No protocol available. The text was vague and unclear with no explicit identification of which outcomes were primary and which were secondary.

Nakamura 2019.

Study characteristics
Methods	Study design: prospective, randomised, single‐centre study Study grouping: parallel group Total duration of the study: December 2015 to December 2017 Duration of follow‐up: within 30 days after the ablation Number of study centres and location: single centre in Japan
Participants	Inclusion criteria: people with non‐valvular AF who received oral anticoagulation with DOACs Exclusion criteria: aged < 20 years, presence of any intracardiac thrombi, any prosthetic heart valve, rheumatic mitral valve disease, severe renal insufficiency (CrCl < 30 mL/minute in people receiving dabigatran and < 15 mL/minute in people receiving rivaroxaban, apixaban, or edoxaban), and an allergy to heparin or history of heparin‐induced thrombocytopenia Total number of participants: 846 Number of randomised participants: 846 Number lost to follow‐up/withdrawn: 2 were ineligible for analysis Number of analysed participants: 844 Number of participants in each treatment group: interrupted DOAC: 424; uninterrupted DOAC: 422 Baseline characteristics Interrupted anticoagulation Age, years, mean: 65 (SD 10) Men, n (%): 298 (70.4) CHA2DS2‐VASc score, mean: 2.1 (SD 1.5) CHADS2 score, mean: 1.1 (SD 1.1) HAS‐BLED score, mean: 1.4 (SD 1.0) Type of AF, n (%): persistent: 128 (30.3) long‐standing persistent: 59 (13.9) paroxysmal: 236 (55.8) Comorbidities, n (%): diabetes mellitus: 65 (15.4) hypertension: 235 (55.6) stroke/TIA: 33 (7.8) heart failure: 41 (9.7) Concomitant medications, n (%): dabigatran: 38 (9.0) rivaroxaban: 151 (35.7) apixaban: 125 (29.6) edoxaban: 109 (25.8) Uninterrupted anticoagulation Age, years, mean: 65 (SD 10) Men, n (%): 298 (70.8) CHA2DS2‐VASc score, mean: 2.0 (SD 1.5) CHADS2 score, mean: 1.1 (SD 1.0) HAS‐BLED score, mean: 1.3 (SD 1.0) Type of AF, n (%): persistent: 126 (29.9) long‐standing persistent: 73 (17.3) paroxysmal: 222 (52.7) Comorbidities, n (%): diabetes mellitus: 66 (15.7) hypertension: 222 (52.7) stroke/TIA: 30 (7.1) heart failure: 39 (9.3) Concomitant medications, n (%): dabigatran: 27 (6.4) rivaroxaban: 160 (38.0) apixaban: 117 (27.8) edoxaban: 117 (27.8) Group differences: no significant differences between groups
Interventions	Periprocedural anticoagulation Anticoagulant used: dabigatran, rivaroxaban, apixaban, edoxaban Doses: dabigatran 110 mg or 150 mg twice daily, rivaroxaban 10 mg or 15 mg once daily in the morning, apixaban 2.5 mg or 5 mg twice daily, or edoxaban at 30 mg or 60 mg once daily in the morning Duration of anticoagulant therapy in trial: not reported Intensity of anticoagulation or dose adjustment: all participants received oral anticoagulation with the standard or reduced DOAC dosage approved in Japan Adherence to anticoagulant treatment: not reported Time of interruption of the anticoagulants prior to procedure: DOACs interrupted on the day of the procedure and reinitiation on the next morning after the procedure. Last dose before interruption of the twice‐daily DOACs (dabigatran and apixaban) was given in the evening on the day before the procedure, and that of the once‐daily DOACs (rivaroxaban and edoxaban) was given on the morning of the day before the procedure Heparin bridge therapy: none used preprocedure Time and strategy of resumption of interrupted OAC after procedure: re‐initiation of the DOACs on the next morning after the procedure. Postablation: all participants received a continuous infusion of unfractionated heparin 10,000 units per 24 hours until the DOACs were reinitiated on the next morning after the procedure TOE: performed in all participants on the day of the ablation procedure or the day before Ablation procedure Type of ablation, n (%): PVI alone: 374 (88.4); PVI plus additional left atrial ablation: 49 (11.6) Ablation energy source, n (%): irrigated radiofrequency ablation catheter: 395 (93.4); cryoballoon: 8 (1.9), hot balloon: 20 (4.7) Intraprocedural anticoagulant Initial heparin bolus 10,000 units IV, followed by a continuous and additional heparin bolus infusion to maintain an ACT 300–400 seconds. The ACT was measured every 10 minutes until the ACT value reached 300 seconds and thereafter every 10–30 minutes
Outcomes	Evaluated incidence of thromboembolic and haemorrhagic events within 30 days after the ablation procedure. Neurological assessments performed before and after procedure Primary outcome Composite of symptomatic thromboembolisms, including an ischaemic stroke, TIA, and systemic embolus, and major bleeding events Major bleeding: defined as cardiac tamponade or pericardial effusions requiring drainage, intracranial and gastrointestinal haemorrhages, haemothorax, retroperitoneal bleeding, any bleeding requiring a blood transfusion, and vascular access site complications requiring any intervention Minor bleeding: defined as bleeding other than major bleeding events, such as mild pericardial effusions, pseudoaneurysms, and groinhaematomas not requiring drainage or intervention, rebleeding at the vascular access sites, nasal bleeding, and haematuria Silent cerebral ischaemic lesions: defined as a hyperintense lesion on DWI, corresponding to a reduced apparent diffusion coefficient map on the next day after the ablation Secondary outcome Incidence of symptomatic thromboembolisms, major and minor bleeding events, and silent cerebral ischaemic lesions
Notes	Investigators' conflicts of interest: not reported Funding: not reported Country: Japan Setting: Gunma Prefectural Cardiovascular Center Comments: none Author's name: Kohki Nakamura Institution: Division of Cardiology, Gunma Prefectural Cardiovascular Center Email: kohkinakamura@yahoo.co.jp Address: Division of Cardiology, Gunma Prefectural Cardiovascular Center, 3‐12 Kameizumi‐machi, Maebashi City, Gunma 371‐0004, Japan
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "randomization was performed in a 1:1 fashion based on a computer generated list of random numbers, according to the uninterruption or interruption of the DOACs on the procedural day."
Allocation concealment (selection bias)	Unclear risk	Methods of allocation concealment not reported.
Blinding of participants and personnel (performance bias) all outcomes	High risk	Participants and study personnel were not blinded.
Blinding of outcome assessment (detection bias) all outcomes	Low risk	Quote: "all MR imaging studies were reviewed by experienced radiologists blinded to the patient characteristics."
Incomplete outcome data (attrition bias) all outcomes	Low risk	Flowchart of study participants was presented, reason for exclusion was reported. Reason for using modified intention‐to‐treat was explained.
Selective reporting (reporting bias)	Unclear risk	Study reported all expected outcomes; however, there is no protocol to compare.

Nogami 2019.

Study characteristics
Methods	Study design: prospective, randomised, open‐label, multicentre, clinical interventional superiority trial Study grouping: parallel group Total duration of the study: March 2014 to April 2019 Duration of follow‐up: all participants followed up for 12 months after ablation. Secondary outcomes assessed for 3 months Number of study centres and location: 28 treatment centres in Japan
Participants	Inclusion criteria: aged 20–85 years when informed consent was obtained were eligible for ABRIDGE‐J if they had paroxysmal or persistent NVAF with AF ablation planned, had documented AF, and were eligible for treatment with dabigatran or warfarin according to the prescribing guidelines in Japan Exclusion criteria: valvular AF, defined as the presence of a prosthetic heart valve (annuloplasty with or without a prosthetic ring, commissurotomy, or valvuloplasty (or a combination) were permitted), haemodynamically significant mitral valve stenosis, or rheumatic heart disease Total number of participants: 504 Number of randomised participants: 500; minimally interrupted dabigatran group: 249; uninterrupted warfarin group: 251 Number lost to follow‐up/withdrawn: 78 discontinued treatment prematurely Number of analysed participants: 442 Number of participants in each treatment group: minimally interrupted dabigatran group: 220; uninterrupted warfarin group: 222 Baseline characteristics Interrupted anticoagulation Age, years, median: 65.0 (IQR 59.0–71.0) Men, n (%): 171 (77.7) CHA2DS2‐VAScb score, median: 2 (IQR 1–3) HAS‐BLED score, median: 1 (IQR 0–2) Type of AF, n (%): paroxysmal: 138 (62.7) persistent: 52 (23.6) long‐standing persistent: 30 (13.6) Comorbidities, n (%): previous stroke: 15 (6.8) CAD: 7 (3.2) previous myocardial infarction: 2 (0.9) previous gastrointestinal tract bleeding: 0 CHF: 8 (3.6) previous gastric ulcer: 5 (2.3) renal dysfunction: 15 (6.8) liver dysfunction: 6 (2.7) alcohol abuse: 2 (0.9) diabetes: 36 (16.4) hypertension: 123 (55.9) cardiomyopathy: 5 (2.3) respiratory disease: 9 (4.1) Concomitant medications, n (%): warfarin: 67 (30.5) dabigatran: 39 (17.7) rivaroxaban: 34 (15.5) apixaban: 26 (11.8) aspirin: 13 (5.9) clopidogrel bisulphate: 3 (1.4) ticlopidine hydrochloride: 0 cilostazol: 0 eicosapentaenoic acid: 3 (1.4) NSAID: 1 (0.5) Uninterrupted anticoagulation Age, years, median: 66.0 (IQR 59.0–71.0) Men, n (%): 160 (72.1) CHA2DS2‐VAScb score, median: 2 (IQR 1–3) HAS‐BLED score, median: 1 (IQR 1–2) Type of AF, n (%): paroxysmal: 138 (62.2) persistent: 55 (24.8) long‐standing persistent: 29 (13.1) Comorbidities: previous stroke: 12 (5.4) CAD: 14 (6.3) previous myocardial infarction: 6 (2.7) previous gastrointestinal tract bleeding:3 (1.4) CHF: 14 (6.3) previous gastric ulcer: 9 (4.1) renal dysfunction: 16 (7.2) liver dysfunction: 12 (5.4) alcohol abuse: 4 (1.8) diabetes: 34 (15.3) hypertension: 126 (56.8) cardiomyopathy: 5 (2.3) respiratory disease: 8 (3.6) Concomitant medications: warfarin: 68 (30.6) dabigatran: 33 (14.9) rivaroxaban: 35 (15.8) apixaban: 24 (10.8) aspirin: 14 (6.3) clopidogrel bisulphate: 3 (1.4) ticlopidine hydrochloride: 0 cilostazol: 1 (0.5) eicosapentaenoic acid: 2 (0.9) NSAID: 7 (3.2) Group differences: demographic and clinical characteristics were well‐balanced between groups
Interventions	Periprocedural anticoagulation   Anticoagulant used: dabigatran and warfarin Dose: dabigatran 150 mg or 110 mg twice daily, warfarin to target PT/INR of 2.0–3.0 Duration of anticoagulant therapy in trial: for 4 weeks before the procedure Intensity of anticoagulation or dose adjustment: dabigatran 110 mg twice daily for participants with moderate renal disorders (CrCl 30–50 mL/minute, calculated using the Cockcroft–Gault formula), those concomitantly receiving p‐glycoprotein antagonists, or those with a high risk of bleeding (aged > 70 years, with history of gastrointestinal tract haemorrhage). Warfarin intensity (PT/INR) 1.6–2.6 for people aged ≥ 70 years Adherence to anticoagulant treatment: not reported Time of interruption of the anticoagulants prior to procedure: dabigatran therapy was interrupted before catheter ablation (1–2 doses were put on hold before ablation) Heparin therapy (intraprocedural or as bridge therapy): heparin bridging was recommended if dabigatran therapy was discontinued ≥ 24 hours before the ablation procedure based on the Japanese recommendations and guidelines Time and strategy of resumption of the interrupted OAC after procedure: dabigatran was resumed within 18 hours after the procedure. Anticoagulation was continued in both groups for 3 weeks after the procedure TOE: within 48 hours before ablation Ablation procedure Type of ablation, n (%) Interrupted arm: RFCA PVI = 176 (80.0), cryoballoon PVI = 34 (15.5), superior vena cava isolation = 40 (18.2), RFCA CTI = 124 (56.4), cryoablation CTI = 25 (11.4), linear ablation = 48 (21.8) Uninterrupted arm: RFCA PVI = 171 (77.0), cryoballoon PVI = 44 (19.8), superior vena cava isolation = 30 (13.5), RFCA CTI = 123 (55.4), cryoablation CTI = 25 (11.3), linear ablation = 47 (21.2) Ablation energy source: radiofrequency and cryoablation Intraprocedural anticoagulant ACT 300–400 seconds was to be achieved and maintained, where possible, during the ablation procedure
Outcomes	Primary outcome Incidence of embolism during the perioperative period and the existence or non‐existence of an atrial thrombus just before ablation, detected by TOE or intracardiac echocardiography. Primary outcome was assessed from the time of randomisation Secondary outcomes Assessed from the start of the ablation procedure until 3 months after ablation Incidence of adjudicated major bleeding events, which were defined according to the ISTH criteria Other secondary safety and efficacy consisted of a composite incidence of major bleeding events, thromboembolic events (stroke, systemic embolism, and TIA), all‐cause death, and a composite incidence of all bleeding events, including minor bleeding, thromboembolic events, and all‐cause death, during and within 3 months after ablation. Bleeding events that did not satisfy the ISTH criteria for major bleeding were considered minor bleeding events
Notes	Investigators' conflicts of interest: quote: "Dr Nogami reported personal fees from Boehringer Ingelheim and Daiichi Sankyo during the conduct of the study and grants from Medtronic and personal fees from Abbott, St Jude Medical, and Japan Lifeline outside the submitted work. Dr Harada reported grants from Nippon Boehringer Ingelheim Co, Ltd, during the conduct of the study. Dr Sekiguchi reported research grants from Abbott outside the submitted work. Dr Y. Yoshida reported grants from Boehringer Ingelheim during the conduct of the study and personal fees from Daiichi Sankyo, Pfizer, Bristol‐Myers Squibb, and Bayer outside the submitted work. Dr Goya reports honoraria from Medtronic, Johnson & Johnson, and St Jude Medical. Dr Origasa reported honoraria from Bayer Yakuhin, Ltd, and Daiichi Sankyo. Dr Hirao reported honoraria from Boehringer Ingelheim. Dr Aonuma reported research grants and/or honoraria from Boehringer Ingelheim Japan during the conduct of the study and from Boston Scientific Japan, Japan Life Line, Nihon Kohden, Biotronik Japan, Century Medical, Toray Industries, Medtronic, and Daiichi Sankyo. No other disclosures were reported" Funding: Boehringer Ingelheim Country: Japan Setting: treatment centres Comments: none Author's name: Kazutaka Aonuma Institution: Department of Cardiology, Faculty of Medicine, University of Tsukuba Email: kaonuma@md.tsukuba.ac.jp Address: 1‐1‐1 Tennodai, Tsukuba 305‐8575, Japan
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "used randomization registration system Mebix Inc to conduct random allocation sequence, participant enrollment, and assignment of participants to the interventions."
Allocation concealment (selection bias)	Low risk	Quote: "used randomization registration system Mebix Inc to conduct random allocation sequence, participant enrollment, and assignment of participants to the interventions."
Blinding of participants and personnel (performance bias) all outcomes	High risk	Open‐label study.
Blinding of outcome assessment (detection bias) all outcomes	Low risk	Quote: "all end points were adjudicated by a panel of experts in a blinded manner."
Incomplete outcome data (attrition bias) all outcomes	Low risk	Study flow chart was provided, reasons for not performing ablation were reported. All participants who underwent ablation were analysed.
Selective reporting (reporting bias)	Low risk	Study outcomes were reported as in the protocol. However, length of hospital stay (secondary outcome stated in the protocol) was not mentioned in results (we did not consider this major bias).

Reynolds 2018.

Study characteristics
Methods	Study design: prospective, multicentre, randomised, parallel‐group, open‐label clinical trial Study grouping: parallel group Total duration of the study: December 2015 to May 2018 Duration of follow‐up: completed follow‐up through either 30 days or the occurrence of a study endpoint Number of study centres and location: 18 study sites in the US
Participants	Inclusion criteria: for prospective apixaban cohort were aged ≥ 18 years and scheduled for catheter ablation for the treatment of non‐valvular AF, with a planned continuation of OAC for minimum 1 month after the procedure Exclusion criteria: people with mechanical heart valves, advanced hepatic or renal (CrCl < 15 mL/minute or on dialysis) dysfunction, ongoing or planned dual antiplatelet therapy, history of stroke or TIA within 6 months, history of prior intracranial bleeding, significant baseline anaemia or thrombocytopenia Total number of participants: 306 Number of randomised participants: 300 Number lost to follow‐up/withdrawn: 5 Number of analysed participants: 295 Number of participants in each treatment group: minimally interrupted apixaban: 145; uninterrupted apixaban: 150 Baseline characteristics Interrupted anticoagulation Age, years, mean: 64.3 (SD 10.3) Men (%): 66.9 CHA2DS2‐VASc score, mean: 2.4 (SD 1.6) HAS‐BLED score, mean: 1.1 (SD 0.8) Type of AF, (%): persistent < 1 year: 35.9 persistent ≥ 1 year: 1.4 paroxysmal: 62.8 Comorbidities, (%): diabetes mellitus: 23.5 hypertension: 70.3 prior stroke: 2.8 heart failure: 9.0 CAD: 17.3 chronic kidney disease: 5.6 valvular disease: 9.9 COPD: 10.4 sleep apnoea: 33.3 Concomitant medications, (%): aspirin: 17.2 other antiplatelet drug: 2.1 Uninterrupted anticoagulation Age, years, mean: 62.8 (SD 9.9) Men (%): 67.3 CHA2DS2‐VASc score, mean: 2.2 (SD 1.6) HAS‐BLED score, mean: 1.0 (SD 0.9) Type of AF, (%): persistent < 1 year: 31.3 persistent ≥ 1 year: 2.0 paroxysmal: 66.7 Comorbidities, (%): diabetes mellitus: 22.0 hypertension: 68.0 prior stroke: 4.0 heart failure: 14.1 CAD: 28.2 chronic kidney disease: 3.3 valvular disease: 10.1 COPD: 8.7 sleep apnoea: 30.8 Concomitant medications (%): aspirin: 28.0 other antiplatelet drug: 0.7 Group differences: no significant differences in demographic or clinical characteristics between groups
Interventions	Periprocedural anticoagulation Anticoagulants used: apixaban Dose: 5 mg twice daily or 2.5 mg twice daily Duration of anticoagulant therapy in trial: minimum of 21 days before the planned ablation procedure, taken either as pre‐existing therapy, or newly initiated upon study entry.  Participants randomised 3 days before procedure Intensity of anticoagulation or dose adjustment: apixaban 2.5 mg was also used Adherence to anticoagulant treatment: treatment compliance was assessed by participant self‐report Time of interruption of the anticoagulants prior to procedure: 1 dose of apixaban (morning dose) on the day of ablation (minimally interrupted) was held Heparin therapy (intra‐procedural or as bridge therapy): none used Time and strategy of resumption of the interrupted OAC after procedure: in the evening of the procedure day, participants resumed apixaban at their usual dose (usually 5 mg) if there had been no prohibitive complications. OAC was then continued for ≥ 1 month postprocedure, when the final study visit was scheduled TOE: reported that TOE was not required Ablation procedure Type of ablation: PVI Ablation energy source: cryoballoon ablation, or force‐sensing radiofrequency ablation Intraprocedural anticoagulant Sites were instructed to administer a heparin bolus before transseptal puncture and to maintain a target ACT > 300 seconds
Outcomes	Endpoints were assessed from the time of randomisation for 30 days Primary safety endpoint Clinically significant bleeding, defined as any event meeting BARC criteria ≥ 2 Primary efficacy endpoint Non‐haemorrhagic stroke or systemic embolism Secondary endpoints Composite of stroke or systemic embolism or major bleeding (BARC criteria ≥ 3) Composite of non‐haemorrhagic stroke or TIA Individual components of those composites Death and cardiovascular death
Notes	Investigators' conflicts of interest: not reported Funding: Baim Institute for Clinical Research with financial support from Bristol‐Myers Squibb and Pfizer Country: US Setting: 18 study sites in US Comments: none Author's name: Matthew R Reynolds Institution: Lahey Hospital & Medical Center, Burlington, Massachusetts Email: matthew.reynolds@baiminstitute.org Address: Baim Institute for Clinical Research, 930 Commonwealth Avenue, Boston, Massachusetts 02215
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Methods of randomisation not described.
Allocation concealment (selection bias)	Unclear risk	Method of allocation concealment not reported.
Blinding of participants and personnel (performance bias) all outcomes	High risk	Open‐label study.
Blinding of outcome assessment (detection bias) all outcomes	Low risk	Quote: "all potential endpoints were reviewed and adjudicated by an independent endpoints committee. Other procedure‐attributable adverse events not meeting endpoint criteria were reviewed by safety officers of the study sponsor as well as the principal investigators."
Incomplete outcome data (attrition bias) all outcomes	Low risk	Participants flow chart was provided. Attrition was explained and all endpoint analyses were conducted on the evaluable patient population, which included all intention‐to‐treat participants who were randomised.
Selective reporting (reporting bias)	Low risk	Outcomes reported matched published protocol (NCT02608099).

Tabish 2010.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Total duration of the study: January 2009 to August 2009 Duration of follow‐up: 3–6 months Number of study centres and location: single centre in China
Participants	Inclusion criteria: aged 65–75 years with AF undergoing catheter ablation  Exclusion criteria: not reported (abstract) Total number of participants: 31 Number of randomised participants: 31 Number lost to follow‐up/withdrawn: not reported Number of analysed participants: 31 Number of participants in each treatment group: low‐dose warfarin group: 16; normal‐dose warfarin group: 15 Baseline characteristics: not reported (abstract)
Interventions	Periprocedural anticoagulation Anticoagulants: warfarin Dose: warfarin to target standard or low‐intensity PT/INR Duration of anticoagulant therapy in trial: not reported Intensity of anticoagulation or dose adjustment: INR of low‐dose warfarin group was maintained at 1.5–2.0 and normal‐dose warfarin group at 2.1–2.5 in perioperative period Adherence to anticoagulant treatment: not reported Time of interruption of the anticoagulants prior to procedure: not reported Heparin bridge therapy: not reported Time and strategy of resumption of the interrupted OAC after the procedure: not reported in the abstract TOE: not reported Ablation procedure Type of ablation: not reported Ablation energy source: not reported Intraprocedural anticoagulant: not reported
Outcomes	Events regarding bleeding and thromboembolism, no further details reported
Notes	The corresponding author contacted as this was an abstract, but no further data obtained. Investigators' conflicts of interest: not reported Funding: not reported Country: China Setting: Department of Cardiology, Tongji Medical College Teaching Hospital Comments: none Author's name: Hussain Tabish Institution: Union Hospital, Tongji Medical College, Wuhan Email: not reported Address: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Could not be determined from the conference abstract.
Allocation concealment (selection bias)	Unclear risk	Could not be determined from the conference abstract.
Blinding of participants and personnel (performance bias) all outcomes	Unclear risk	Could not be determined from the conference abstract.
Blinding of outcome assessment (detection bias) all outcomes	Unclear risk	Could not be determined from the conference abstract.
Incomplete outcome data (attrition bias) all outcomes	Unclear risk	Could not be determined from the conference abstract.
Selective reporting (reporting bias)	Unclear risk	Could not be determined from the conference abstract.

Xing 2017.

Study characteristics
Methods	Study design: prospective, randomised, non‐blinded study Study grouping: parallel group Total duration of the study: January 2014 to September 2017 Duration of follow‐up: periprocedural period defined as the period from the day of ablation to the 7th day postablation Number of study centres and location: single centre in China
Participants	Inclusion criteria: people with AF, aged ≥ 70 years, who underwent the first‐time RFCA for AF and had indications to receive anticoagulation with warfarin according to AF guidelines Exclusion criteria: history of stroke or TIA, intracardiac thrombi detected by TOE, severe heart diseases (prosthetic heart valve/severe valvular heart diseases according to American College of Cardiology/American Heart Association guidelines/dilated cardiomyopathy/hypertrophic cardiomyopathy), or severe liver or renal dysfunction Total number of participants: 101 Number of randomised participants: 101 Number lost to follow‐up/withdrawn: 0 Number of analysed participants: 101 Number of participants in each treatment group: low‐intensity warfarin (group A): 52 standard‐intensity warfarin (group B): 49 Baseline characteristics Low‐intensity warfarin (group A) Age, years, mean: 73.8 (SD 3.3) Men, n (%): 33 (63.5) Type of AF, n (%): paroxysmal: 38 (73.0) persistent: 14 (27.0) Comorbidities, n (%): hypertension: 37 (71.2) diabetes: 14 (26.9) heart failure: 18 (34.6) coronary heart disease: 13 (25.0) CHA2DS2 score, mean: 1.7 (SD 0.9) CHA2DS2‐VASc score, mean: 3.2 (SD 1.1) 0, n (%): 0 (0) 1, n (%): 2 (3.8) ≥ 2, n (%): 50 (96.2) HASBLED score, mean: 2.1 (SD 0.7) Concomitant medications, n (%): antiplatelet drugs: 8 (15.4) Standard‐intensity warfarin (group B) Age, years, mean: 73.6 (SD 2.6) Men, n (%): 35 (71.4) Type of AF, n (%): paroxysmal: 39 (79.6) persistent: 10 (20.4) Comorbidities, n (%): hypertension: 32 (65.3) diabetes: 11 (22.4) heart failure: 16 (32.7) coronary heart disease: 12 (24.5) CHA2DS2 score, mean: 1.6 (SD 1.0) CHA2DS2‐VASc score, mean: 2.8 (SD 1.3) 0, n (%): 0 (0) 1, n (%): 4 (8.2) ≥ 2, n (%): 44 (91.8) HASBLED score, mean: 2.0 (SD 0.7) Concomitant medications, n (%): antiplatelet drugs: 6 (12.2) Group differences: no significant differences between groups
Interventions	Periprocedural anticoagulant Anticoagulant used: warfarin Dose: dose to achieve a standard INR of 2.0–3.0 or low‐intensity INR of 1.5–2.0 Duration of anticoagulant therapy in trial: warfarin given ≥ 4 weeks before procedure Intensity of anticoagulation or dose adjustment: warfarin adjusted to standard‐ or low‐intensity INR Adherence to anticoagulant treatment: not reported Time of interruption of anticoagulants prior to procedure: participants in low‐intensity INR group maintained an INR of 1.5–2.0 (minimally interrupted) Heparin bridge therapy: not used Time and strategy of resumption of the interrupted OAC after procedure: regardless of treatment groups, if the INRs were beyond 1.5–3.0 after procedure, warfarin dose had to be adjusted to maintain INRs at 1.5–3.0. Adjustments made to warfarin were decided by physicians TOE: performed on all cases within 3 days before ablation Ablation procedure Type of ablation: PVI Ablation energy source: radiofrequency Intraprocedural anticoagulant All participants received bolus heparin 100 U/kg. ACT was checked every 15 minutes, and continuous infusion of heparin adjusted (group A: 144 (SD 33) units/kg; group B: 132 (SD 29) units/kg) to maintain an ACT of 300 seconds (group A: 301 (SD 15) seconds; group B: 311 (SD 13) seconds)
Outcomes	Primary outcome Periprocedural thromboembolic complications and major bleeding. Thromboembolic complications included systemic thromboembolic events, ischaemic stroke, and TIA. Major bleeding was defined as bleeding requiring transfusion, invasive intervention, or anticoagulation therapy discontinuation, such as cerebral bleeding and pericardial effusions requiring drainage Secondary outcomes Periprocedural asymptomatic cerebral emboli (ACE) lesions defined as: focal hyperintense areas detected by the diffusion‐weighted sequence without any symptoms detected by MRI 7 days postablation Minor bleeding defined as those that could be cured by conservative treatment and did not require discontinuing warfarin. Periprocedural period defined as the period from the day of ablation to the 7th day postablation
Notes	Investigators' conflicts of interest: none Funding: authors received no financial support for the research, authorship, or publication of this article. Country: China Setting: Shaoxing Hospital of Zhejiang University Comments: none Author's name: Hangyuan Guo Institution: Department of Cardiology, Shaoxing People's Hospital Email: ghangyuan@hotmail.com Address: 568 Zhongxing North Road, Shaoxing, Zhejiang Province, 312000, PR China
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Method of randomisation not reported.
Allocation concealment (selection bias)	Unclear risk	Method of allocation not reported.
Blinding of participants and personnel (performance bias) all outcomes	High risk	Open‐label study.
Blinding of outcome assessment (detection bias) all outcomes	Unclear risk	No information to judge.
Incomplete outcome data (attrition bias) all outcomes	Unclear risk	No participant flow chart provided.
Selective reporting (reporting bias)	Unclear risk	Although they reported on expected outcomes, no protocol to compare it with.

Yamaji 2019.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Total duration of the study: January 2017 to October 2018 Duration of follow‐up: data obtained over 120 days (from 30 days before ablation to 90 days after ablation). Initial follow‐up visit was scheduled 2 weeks after AF ablation Number of study centres and location: single centre in Japan
Participants	Inclusion criteria: people with paroxysmal AF, persistent and long‐standing AF, or atrial tachycardia; people who underwent their first AF ablation between January 2017 and October 2018 Exclusion criteria: decreased renal function (CrCl rate < 30 mL/minute)   Total number of participants: 584 Number of randomised participants: 584 Number lost to follow‐up/withdrawn: 0 Number of analysed participants: 584 Number of participants in each treatment group: minimally interrupted DOAC: 307; uninterrupted DOAC: 277 Baseline characteristics Interrupted anticoagulation Age, years, mean: 65.0 (SD 10.5) Men, n (%): 212 (69) CHADS2 score, mean: 0.90 (SD 0.87) CHA2DS2‐VASc score, mean: 1.88 (SD 1.37) HAS‐BLED score, mean: 1.35 (SD 1.05) Type of AF (n): paroxysmal AF: 199 persistent AF: 61 long‐standing persistent AF: 40 Comorbidities, n (%): diabetes mellitus: 34 (11) hypertension: 144 (47) prior stroke or TIA: 13 (4) CHF: 22 (7) Concomitant medications, (n): antiplatelet agents: 10 dabigatran: 56 rivaroxaban: 109 apixaban: 59 edoxaban: 83 Uninterrupted anticoagulation Age, years, mean: 66.4 (SD 10.3) Men, n (%): 211 (76.17) CHADS2 score, mean: 0.97 (SD 0.97) CHA2DS2‐VASc score, mean: 1.88 (SD 1.37) HAS‐BLED score, mean: 1.42 (SD 1.04) Type of AF, (n): paroxysmal AF: 171 persistent AF: 65 long‐standing persistent AF: 38 Comorbidities, n (%): diabetes mellitus: 39 (14) hypertension: 134 (48) prior stroke or TIA: 16 (6) CHF: 6 (2) Concomitant medications, (n): antiplatelet agents: 7 dabigatran: 83 rivaroxaban: 65 apixaban: 58 edoxaban: 71 Group differences: no significant differences in clinical and echocardiogram parameters and thromboembolic and bleeding risk scores between groups, except for AF duration, which was about 1 year shorter in the minimally interrupted group than the uninterrupted group. Significant difference in proportion of participants using dabigatran and rivaroxaban between groups. More participants with heart failure in minimally interrupted group
Interventions	Procedural anticoagulation Anticoagulants used: dabigatran, rivaroxaban, edoxaban, and apixaban Dose: dabigatran and apixaban twice a day (morning and evening). Rivaroxaban and edoxaban once a day (morning) Duration of anticoagulant therapy in trial: initiated 30 days before ablation Intensity of anticoagulation or dose adjustment: not reported Adherence to anticoagulant treatment: rivaroxaban administered once a day in the morning, rather than in the evening, to maintain a sufficient adherence rate Time of interruption of the anticoagulants prior to procedure: holding the morning dose of DOAC on the day of ablation Heparin bridge therapy: not used Time and strategy of resumption of interrupted OAC after procedure: single dose of apixaban or dabigatran resumed in evening of the day of ablation or 4 hours after the completion of the postmeridiem ablation session, with confirmation of haemostasis. Rivaroxaban and edoxaban administrations resumed in the morning of the day after ablation. DOAC therapy continued for ≥ 3 months after TOE: not reported  Ablation procedure Type of ablation: superior vena cava isolation/cavotricuspid isthmus ablation Ablation energy source: not reported Intraprocedural anticoagulant Heparin bolus administered just before septal puncture, based on age, sex, and bodyweight. If pre‐ACT value 120–130 U/Kg for a pre‐ACT ≥ 150 seconds and 140–150 U/kg for a pre‐ACT < 150 seconds. IV infusion heparin 400 U/hour to maintain an ACT of 300–400 seconds
Outcomes	Primary outcome Pre‐ACT Primary safety outcome Composite of bleeding and thromboembolic complications (yielding the bleeding and thromboembolic risk score) Secondary outcome Thromboembolic and bleeding complications Thromboembolic complications: cerebrovascular accidents and TIAs once intracranial haemorrhage had been ruled out. Pulmonary embolism and deep venous embolism Major bleeding complications defined as cardiac tamponade, retroperitoneal bleeding, and groin haematoma requiring blood transfusion Cardiac tamponade defined by characteristic clinical features and the presence of a considerable pericardial effusion requiring drainage. Late cardiac tamponades were those occurring greater than 48 hours after the procedure Minor bleeding complications: pericardial effusion reduced haemoglobin without blood transfusion, and haematuria defined as minor complications. Pericardial effusion defined as an effusion identified in the pericardial space by routine follow‐up echocardiography, without haemodynamic disturbance (non‐tamponade)
Notes	Investigators' conflicts of interest: none Funding: not reported Country: Japan Setting: Okayama Heart Clinic Comments: selection and doses of DOACs were not randomised and were at the discretion of each treating physician, considering the participant's characteristics (including renal function) and drug manufacturer's directions Author's name: Hirosuke Yamaji Institution: Heart Rhythm Center, Okayama Heart Clinic, Okayama University Email: yamaji2@mac.com Address: Heart Rhythm Center, Okayama Heart Clinic, Okayama University, Takeda 54‐1, Naka‐Ku, Okayama 703‐8251, Japan
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Randomisation method not reported. Also, they mentioned occurrence of randomisation error. Quote: "although patients were randomly allocated to the min‐Int and Unint DOAC therapy groups, due to unintended randomized error, the number of patients was not completely equivalent between these two anticoagulation strategy groups."
Allocation concealment (selection bias)	Unclear risk	Methods of allocation concealment not reported.
Blinding of participants and personnel (performance bias) all outcomes	Unclear risk	No mention if participants and personnel were aware of intervention.
Blinding of outcome assessment (detection bias) all outcomes	Unclear risk	Methods of blinding of outcome assessors not reported.
Incomplete outcome data (attrition bias) all outcomes	Unclear risk	Participant flow chart is not reported, no mention of excluded participants. Insufficient information to judge.
Selective reporting (reporting bias)	Unclear risk	No protocol available.

Yoh 2019.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Total duration of the study: not provided Duration of follow‐up: 30 days after the procedure Number of study centres and location: not reported
Participants	Inclusion criteria: people with non‐valvular AF receiving DOAC and undergoing the first ablation Exclusion criteria: people undergoing the second or third session of RFCA for NVAF Total number of participants: 105 Number of randomised participants: 105 Number lost to follow‐up/withdrawn: 0 declared Number of analysed participants: 105 Number of participants in each treatment group: interrupted DOACs: 64; uninterrupted DOACs: 41 Baseline characteristics Interrupted anticoagulation Age, years, mean: 70 (SD 8) Men, n (%): 43 (67.2) CHA2DS2‐VASc score, mean: 2.6 (SD 1.5) HAS‐BLED score, mean: 1.3 (SD 0.9) CHADS2 score, mean: 1.5 (SD 1.1) Type of AF, n (%): paroxysmal: 34 (53.1) persistent: 16 (25.0) long‐standing persistent: 14 (21.9) Comorbidities, n (%): diabetes mellitus: 12 (18.8) hypertension: 48 (75.0) stroke: 4 (6.3) structural heart disease: 1 (1.6) heart failure: 8 (12.5) Type of DOACs, n (%): dabigatran: 8 (12.5) rivaroxaban: 14 (22.2) apixaban: 25 (39.1) edoxaban: 17 (26.6) Concomitant medications, n (%): antiplatelet drugs: 5 (7.8) Uninterrupted anticoagulation Age, years, mean: 65 (SD 12) Men, n (%): 25 (61.0) CHA2DS2‐VASc score, mean: 2.6 (SD 1.8) HAS‐BLED score, mean: 1.7 (SD 1.1) CHADS2 score, mean: 1.6 (SD 1.1) Type of AF, n (%): paroxysmal: 16 (39.0) persistent: 15 (36.6) long‐standing persistent: 10 (24.4) Comorbidities, n (%): diabetes mellitus: 11 (26.8) hypertension: 28 (68.3) stroke: 6 (14.6) structural heart disease: 0 (0.0) heart failure: 4 (9.8) Type of DOACs, n (%): dabigatran: 16 (39.0) rivaroxaban: 8 (19.5) apixaban: 14 (34.2) edoxaban: 3 (7.3) Concomitant medications, n (%): antiplatelet drugs: 4 (9.8) Group differences: age of enrolled participants was significantly higher in the interrupted DOAC group. In baseline and procedural characteristics of study population, there were significant differences in number of participants administrated dabigatran in the uninterrupted group and in those receiving edoxaban in the interrupted group
Interventions	Periprocedural anticoagulation Anticoagulant used: dabigatran, apixaban, edoxaban, and rivaroxaban Dose: not reported Duration of anticoagulant therapy in trial: not reported Intensity of anticoagulation or dose adjustment: not reported Adherence to anticoagulant treatment: not reported Time of interruption of the anticoagulants prior to procedure: interruption was given on morning of day before procedure for both twice‐daily DOACs (dabigatran and apixaban) and once‐daily DOACs (rivaroxaban and edoxaban) Heparin bridge therapy: not used Time and strategy of resumption of interrupted OAC after procedure: the twice‐daily DOACs (dabigatran and apixaban) were reinitiated on evening of day after the procedure, while the once‐daily DOACs (rivaroxaban and edoxaban) reinitiated on next morning after procedure TOE: not reported Ablation procedure Type of ablation: PVI (95%) and PVI with additional left atrial ablation (5%) Ablation energy source: radiofrequency Intraprocedural anticoagulant Initial heparin bolus 5000 units then continuous and additional heparin bolus infusion to maintain ACT at 300–400 seconds during procedure. IV heparin administration stopped at end of procedure and neutralised using protamine infusion
Outcomes	Symptomatic cerebral infarction: detected cerebral ischaemic lesion by brain MRI after the procedure, with symptoms of neurological defects such as hemiparesis Silent cerebral ischaemia: detected by brain MRI after procedure, without any symptoms Intracranial bleeding: detected by brain MRI after procedure Major and minor bleeding
Notes	Investigators' conflicts of interest: not reported Funding: not reported Country: Japan Setting: Kansai Medical university Comments: the paper is published as abstract, additional information was provided by the corresponding author Authors' names: Masue Yoh; Masahiko Takagi; Takuro Yoshio; Hiroki Takahashi; Ichiro Shiojima  Institution: Kansai Medical university Email: takagims@takii.kmu.ac.jp  Address: 10‐15 Fumizono‐cho, Moriguchi, Osaka
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No available data in the abstract.
Allocation concealment (selection bias)	Unclear risk	No available data in the abstract.
Blinding of participants and personnel (performance bias) all outcomes	Unclear risk	No available data in the abstract.
Blinding of outcome assessment (detection bias) all outcomes	Unclear risk	No available data in the abstract.
Incomplete outcome data (attrition bias) all outcomes	Unclear risk	did not declare the source of funding
Selective reporting (reporting bias)	Unclear risk	No available data in the abstract.

Yoshimura 2017.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: parallel group Total duration of the study: March 2013 to December 2014 Duration of follow‐up: asymptomatic cerebral microthromboembolism evaluated on the day after the procedure. New thromboembolism from 1 hour to 10 days after a thromboembolic event Number of study centres and location: single centre in Japan
Participants	Inclusion criteria: people with paroxysmal AF, persistent AF, and long‐standing persistent AF who underwent AF ablation. Included people who had received radiofrequency who had undergone previous ablation ≥ 6 months and required a second ablation Exclusion criteria: not reported  Total number of participants: 176 Number of randomised participants: 174 Number lost to follow‐up/withdrawn: 2 Number of analysed participants: 105 Number of participants in each treatment group: interrupted apixaban: 50; uninterrupted rivaroxaban: 55 Baseline characteristics Interrupted anticoagulation Age, years, mean: 58.7 (SD 8.9) Men, n (%): 41 (82.0) CHA2DS2‐VASc score, mean: 1.7 (SD 1.4) HAS‐BLED score, mean: 1.2 (SD 0.8) CHADS2 score, mean: 1.1 (SD 1.0) Type of AF, paroxysmal, n (%): 31 (62.0) Comorbidities, n (%): diabetes mellitus: 5 (10.0) hypertension: 32 (64.0) history of stroke: 8 (16.0) CHF: 5 (10.0) CAD: 7 (14.0) Uninterrupted anticoagulation Age, years, mean: 59.1 (SD 12.4) Men, n (%): 45 (81.8) CHA2DS2‐VASc score, mean: 1.7 (SD 1.4) HAS‐BLED score, mean: 1.4 (SD 1.2) CHADS2 score, mean: 1.1 (SD 1.1) Type of AF, paroxysmal, n (%): 33 (60.0) Comorbidities, n (%): diabetes mellitus: 7 (12.7) hypertension: 43 (78.2) history of stroke: 7 (12.7) CHF: 4 (7.3) CAD: 5 (9.1) Group differences: no significant differences in participants' characteristics between groups
Interventions	Periprocedural anticoagulation Anticoagulant used: apixaban, rivaroxaban Dose: not reported Duration of anticoagulant therapy in trial: not reported Intensity of anticoagulation or dose adjustment: not reported Adherence to anticoagulant treatment: not reported Time of interruption of the anticoagulants prior to procedure: apixaban interrupted on morning of procedure (minimally interrupted) Heparin bridge therapy: none given Time and strategy of resumption of interrupted OAC after procedure: not reported TOE: all participants underwent TOE on day or 1 day before procedure Ablation procedure Type of ablation: CFAE ablation with/without PVI Ablation energy source: radiofrequency power Intraprocedural anticoagulant Bolus heparin used to maintain an ACT > 300 seconds during procedure
Outcomes	Complications of cerebral thromboembolism and hemopericardium Asymptomatic cerebral microthromboembolism evaluated by MRI and diagnosed by blinded radiologist
Notes	Investigators' conflicts of interest: none Funding: Japan Society for the Promotion of Science (JSPS) KAKENHI Grant Number 26461077 Country: Japan Setting: Kagoshima University Hospital Comments: none Author's name: Naoya Oketani Institution: Department of Cardiovascular Medicine and Hypertension, Research Field in Medicine and Health Sciences, Kagoshima University Email: oketani@m.kufm.kagoshima‐u.ac.jp Address: 8‐35‐1 Sakuragaoka, Kagoshima 890‐8520, Japan
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "randomization was stratified by type of AF and sex using the table of random numbers."
Allocation concealment (selection bias)	Unclear risk	Methods of allocation concealment not reported.
Blinding of participants and personnel (performance bias) all outcomes	High risk	No sufficient information available to judge.
Blinding of outcome assessment (detection bias) all outcomes	Low risk	Quote: "all of the patients underwent cerebral MRI including DW‐ [diffusion‐weighted] and T2W‐MRI [T2‐weighted MRI] on the day after the ablation procedure and were diagnosed with or without new or old thromboembolism by radiologists who were blinded to this study."
Incomplete outcome data (attrition bias) all outcomes	Low risk	Participant flow chart was provided. Total randomised were analysed.
Selective reporting (reporting bias)	Unclear risk	No protocol available.

Yu 2019.

Study characteristics
Methods	Study design: randomised, open‐label, multicentre trial Study grouping: parallel group Total duration of the study: June 2015 to May 2019 Duration of follow‐up: 1 month after ablation Number of study centres and location: 3 tertiary hospitals in Korea
Participants	Inclusion criteria: people with AF; aged 20–80 years; people who had undergone catheter ablation of AF due to symptomatic, drug‐refractory AF, and people who could receive NOAC; consented to study  Exclusion criteria: aged < 20 years or > 80 years; valvular AF; significant structural heart disease other than left ventricular hypertrophy; left atrial diameter ≥ 60 mm; CrCl < 30mL/minutes; history of previous AF ablation or cardiac surgery Total number of participants: 533 Number of randomised participants: 326 Number lost to follow‐up/withdrawn: 0 Number of analysed participants: 326 Number of participants in each treatment group: interrupted single dose skipped NOACs (SDS): 110; interrupted 24‐hour skipped NOACs (24S): 110; uninterrupted NOACs (UI): 106 Baseline characteristics Interrupted anticoagulation Age, years, mean: SDS: 57.9 (SD 11.1) 24S: 58.4 (SD 11.3) Men, n (%): SDS: 79 (71.8) 24S: 83 (75.5) CHA2DS2‐VASc score, mean: SDS: 1.7 (SD 1.5) 24S: 1.6 (SD 1.4) Type of AF, paroxysmal, n (%): SDS: 74 (67.3) 24S: 61 (55.5) Comorbidities, n (%): diabetes mellitus: SDS: 16 (14.5) 24S: 21 (19.1) hypertension: SDS: 45 (40.9) 24S: 50 (45.5) stroke/TIA: SDS: 17 (15.5) 24S: 11 (10.0) heart failure: SDS: 16 (14.5) 24S: 16 (14.5) vascular disease: SDS: 11 (10.0) 24S: 6 (5.5) NOAC type, n (%): dabigatran: SDS: 36 (32.7) 24S: 37 (33.6) rivaroxaban: SDS: 36 (32.7) 24S: 36 (32.7) apixaban: SDS: 38 (34.5) 24S: 37 (33.6) CrCl (mL/minute), mean: SDS: 90.9 (SD 31.5) 24S: 91.4 (SD 37.0) NOAC dosing, n (%): underdosing, n (%): SDS: 16 (14.5) 24S: 19 (17.3) labelled use, n (%): SDS:93 (84.5) 24S: 91 (82.7) overdosing, n (%): SDS: 1 (0.9) 24S: 0 Uninterrupted anticoagulation Age, years, mean: 58.6 (SD 11.7) Men, n (%): 81 (76.4) CHA2DS2‐VASc score, mean: 1.6 (SD 1.4) Type of AF, paroxysmal, n (%): 67 (63.2) Comorbidities, n (%): diabetes mellitus: 10 (9.4) hypertension: 45 (42.5) stroke/TIA: 12 (11.3) heart failure: 11 (10.4) vascular disease: 8 (7.6) NOAC type, n (%): dabigatran: 35 (33.0) rivaroxaban: 32 (30.2) apixaban: 39 (36.8) CrCl (mL/minute), mean: 97.5 (SD 54.1) NOAC dosing, n (%): underdosing: 21 (19.8) labelled use: 85 (80.2) overdosing: 0 Group differences: no differences between groups
Interventions	Periprocedural anticoagulation Anticoagulant used: dabigatran, apixaban, rivaroxaban Dose: not reported Duration of anticoagulant therapy in trial: preprocedural anticoagulation maintained for ≥ 3 weeks before ablation Intensity of anticoagulation or dose adjustment: not reported Adherence to anticoagulant treatment: not reported Time of interruption of the anticoagulants prior to procedure: single dose skipped (SDS), or 24‐hour skipped (24S) NOACs Heparin bridge therapy: bridging with low molecular weight heparin carried out in participants with persistent AF who were assigned to the 24S group Time and strategy of resumption of the interrupted OAC after procedure: participants resumed anticoagulation at their usual dose on evening of procedure day if there were no prohibitive complications. Anticoagulation continued for ≥ 8 weeks after procedure TOE: performed before all ablation procedures Ablation procedure Type of ablation: circumferential PVI and creation of cavotricuspid isthmus block Ablation energy source: RFCA Intraprocedural anticoagulation All participants received intraprocedural heparin during the catheter ablation procedure. The ACT was maintained at 350–400 seconds during ablation procedure
Outcomes	Primary endpoint Incidence of major bleeding events as defined by the ISTH. Major bleeding events were considered from the first femoral puncture to 1 month after the RFCA Secondary endpoint Thromboembolic and other procedure‐related complications, such as vascular complications or minor bleeding events. Minor bleeding events defined as clinical bleeding events that did not fulfil the ISTH criteria for major bleeding events
Notes	Investigators' conflicts of interest: not reported Funding: governmental (Korea Health 21R&D Project, Ministry of Health and Welfare) Country: Korea Setting: Severance Cardiovascular Hospital, Yonsei University Health System Comments: none Author's name: Hui‐Nam Pak Institution: Yonsei University Health System Email: hnpak@yuhs.ac Address: 50‐1 Yonsei‐ro, Seodaemun‐gu, Seoul 03722, Republic of Korea
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "a central randomization strategy using computer‐generated random permutation sequences was conducted."
Allocation concealment (selection bias)	Low risk	Central randomisation algorithm used.
Blinding of participants and personnel (performance bias) all outcomes	High risk	Open‐label study.
Blinding of outcome assessment (detection bias) all outcomes	Low risk	Quote: "all outcome events were adjudicated by an independent committee in a blinded manner."
Incomplete outcome data (attrition bias) all outcomes	Low risk	Authors mentioned exactly how many patients were screened, randomised, and excluded (n = 207), and the reason for exclusion. They specified treatment group numbers and related outcome measure.
Selective reporting (reporting bias)	Low risk	Study matched the published protocol. Primary outcomes were reported as planned.

ACE: angiotensin‐converting enzyme; ACT: activation clotting time; AF: atrial fibrillation; ARB: angiotensin receptor blocker; BARC: Bleeding Academic Research Consortium; CAD: coronary artery disease; CFAE: complex fractionated atrial electrogram; CHF: congestive heart failure; COPD: chronic obstructive pulmonary disease; CrCl: creatinine clearance; CT: computer tomography; CTI: cavotricuspid isthmus; DOAC: direct oral anticoagulant; DWI: diffusion‐weighted imaging; H2RA: histamine 2 receptor antagonist; HAS‐BLED: Hypertension. Abnormal renal and liver function. Stroke. Bleeding; INR: international normalised ratio; IQR: interquartile range; ISTH: International Society of Thrombosis and Hemostasis; IV: intravenous; MRI: magnetic resonance imaging; n: number; NOAC: novel oral anticoagulant; NSAID: non‐steroidal anti‐inflammatory drug; NVAF: non‐valvular atrial fibrillation; OAC: oral anticoagulant; PPI: proton pump inhibitor; PT: prothrombin time; PVI: pulmonary vein isolation; RFCA: radiofrequency catheter ablation; SD: standard deviation; TOE: transoesophageal echocardiography; TIA: transient ischaemic attack.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Abhishek 2011	Not a randomised comparison.
Aoyama 2019	Not a randomised comparison.
Baltogiannis 2016	Not a randomised comparison.
Brinkmeier 2018	Not a randomised comparison.
Calkins 2019	No interrupted anticoagulation as a comparison.
Cavalli 2019	Not a randomised comparison.
ChiCTR‐OPN‐15006584	Not a randomised comparison.
Di 2017	Not a randomised comparison.
Di Biase 2014a	Not a randomised comparison.
Di Biase 2014b	Not a randomised comparison.
Efremidis 2015	Not a randomised comparison.
EUCTR2012‐001484‐79‐DE	No interrupted arm.
EUCTR2016‐003069‐25‐HU	No interrupted arm.
Finlay 2010	Not a randomised comparison.
Gunawardene 2017	Not a randomised comparison.
Hohnloser 2019	No interrupted anticoagulation as a comparison.
jRCTs031180249	Not the intervention of interest.
Kim 2016	Not the intervention of interest.
Kirchhof 2018	Not the intervention of interest.
Konduru 2012	Not a randomised comparison.
Kuwahara 2013	Not a randomised comparison.
Lane 2018	Not a randomised comparison.
Muller 2016	Not a randomised comparison.
NCT01729871	No interrupted anticoagulation as a comparison.
NCT02504177	No interrupted anticoagulation as a comparison.
Oh 2013	Not the intervention of interest.
Okumura 2016	Not the intervention of interest.
Page 2011	Not a randomised comparison.
Page 2014	Not a randomised comparison.
Saad 2011	Not a randomised comparison.
Sagawa 2018	Not a randomised comparison.
Sakamoto 2019	Not the intervention of interest.
Steffel 2017	Not intervention of interest.
Stepanyan 2014	Not a randomised comparison.
Tscholl 2017	Not a randomised comparison.
UMIN000013341	No interrupted anticoagulation as a comparison.
UMIN000028892	Not the intervention of interest.
UMIN000029693	Not a randomised comparison.
Wakamatsu 2020	Not a randomised comparison.
Wazni 2007	Not a randomised comparison.
Xing 2018	Not a randomised comparison.
Yamaji 2013	Not a randomised comparison.
Yamaji 2018	Not the intervention of interest.
Yoshimoto 2020	No interrupted anticoagulation as a comparison.

Differences between protocol and review

We made the following changes from the protocol (Bawazeer 2019).

We added the outcome asymptomatic thromboembolism as a secondary outcome. Literature showed that the incidence of such a complication in people undergoing ablation ranges from 1.7% to 38.9% with VKA as the main anticoagulant (Forleo 2016). The nature of catheter ablation (the type of ablation and the source of energy), intraprocedural cardioversion during ablation, and the management of anticoagulation around the time of the procedure are important factors that may increase the risk of asymptomatic thromboembolism (Petzl 2020). Despite variable terms used by included studies, we believe that it is clinically important to assess such complications in the setting of different interruption strategies and with different anticoagulant drugs (VKS and DOAC). Deneke 2015 defined the silent cerebral event as "an acute new MRI [magnetic resonance imaging]‐detected brain lesion typical to cerebral ischemia in a patient without clinically apparent neurological deficit." In that clinical review, the author related silent cerebral events and silent cerebral lesions to cerebral ischaemic infarcts due to embolic "fingerprint" specific to the type of ablation procedure. Therefore, we considered that despite the different terms used, they reflect a relatively similar outcome. 

We added the outcome of minor bleeding in the summary of findings table.

We decided to conduct a subgroup analysis for the outcome of thromboembolism despite the fact that only 6/12 studies reported events. We thought that such subgroup analysis may provide additional meaningful insights and may explain the heterogeneity identified in this outcome, hence better use of evidence in the clinical practice.

We did not contact researchers and pharmaceutical companies who produced anticoagulant drugs to request information on any unpublished trials, but will do so in an update of the review.

Contributions of authors

GAB: lead author of review; co‐ordinated review writing; provided methodological perspective; resolved conflicts; interpreted analyses; wrote results and discussion; drafted final review; and responded to all editorial comments.

HAK: co‐ordinated review writing; provided methodological perspective; resolved any conflicts; wrote effects of the intervention, summary of findings table, and discussion; performed GRADE assessment; and drafted final review.

AAA: performed title and abstract screening; full‐text reviewing; data extraction; wrote description of included studies; assessed risk of bias; contributed to writing methods and abstract; and entered text into Review Manager 5 and all references.

NOB: performed title and abstract screening, full‐text reviewing, data extraction, and description of included studies; assessed risk of bias; and contributed to writing method and abstract sections.

AMA: performed title and abstract screening, full‐text reviewing, data extraction, and description of included studies; assessed risk of bias; and contributed to writing method and abstract sections.

TSK: provided general advice and reviewed protocol (Bawazeer 2019); provided a clinical perspective; and contributed to interpretation of analyses and conclusion of review.

MM: provided general advice on results and reviewed analysis section; provided a methodological perspective; and contributed to interpretation of results.

KMA: provided general advice on protocol (Bawazeer 2019); provided methodological perspective; and performed critical revision of draft review.

LAA: provided general advice on protocol (Bawazeer 2019); provided methodological perspective; contributed to interpretation of analyses; contributed to discussion; and provided critical revision of draft review.

All authors approved final manuscript.

Sources of support

Internal sources

• No sources of support provided

External sources

• Research Center of the Female Scientific and Medical Colleges, Saudi Arabia

  This research project was supported by a grant from the “Research Center of the Female Scientific and Medical Colleges”, Deanship of Scientific Research, King Saud University

• NIHR, UK

  This project was supported by the NIHR via Cochrane Infrastructure funding to the Heart Group. The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the Systematic Reviews Programme, NIHR, NHS or the Department of Health and Social Care.

Declarations of interest

GAB: none.

HAK: none.

AAA: none.

NOB: none.

AMA: none.

TSK: none.

MM: none.

KMA: none.

LAA: none.

Edited (no change to conclusions)