Antiplatelet and anticoagulation for patients with prosthetic heart valves

Abstract

Background

Patients with prosthetic heart valves are at increased risk for valve thrombosis and arterial thromboembolism. Oral anticoagulation alone, or the addition of antiplatelet drugs, has been used to minimise this risk. An important issue is the effectiveness and safety of the latter strategy.

Objectives

This is an update of our previous review; the goal was to create a valid synthesis of all available, methodologically sound data to further assess the safety and efficacy of combined oral anticoagulant and antiplatelet therapy versus oral anticoagulant monotherapy in patients with prosthetic heart valves.

Search methods

We updated the previous searches from 2003 and 2010 on 16 January 2013 and searched the Cochrane Central Register of Controlled Trials (CENTRAL) on The Cochrane Library (2012, Issue 12), MEDLINE (OVID, 1946 to January Week 1 2013), and EMBASE (OVID, 1980 to 2013 Week 02). We have also looked at reference lists of individual reports, review articles, meta‐analyses, and consensus statements. We included reports published in any language or in abstract form.

Selection criteria

All reports of randomised controlled trials comparing standard‐dose oral anticoagulation to standard‐dose oral anticoagulation and antiplatelet therapy in patients with one or more prosthetic heart valves.

Data collection and analysis

Two review authors independently performed the search strategy, assessed trials for inclusion and study quality, and extracted data. We collected adverse effects information from the trials.

Main results

One new study has been identified and included in this update. In total, 13 studies involving 4122 participants were included in this review update. Years of publication ranged from 1971 to 2011. Compared with anticoagulation alone, the addition of an antiplatelet agent reduced the risk of thromboembolic events (odds ratio (OR) 0.43, 95% confidence interval (CI) 0.32 to 0.59; P < 0.00001) and total mortality (OR 0.57, 95% CI 0.42 to 0.78; P = 0.0004). Aspirin and dipyridamole reduced these events similarly. The risk of major bleeding was increased when antiplatelet agents were added to oral anticoagulants (OR 1.58, 95% CI 1.14 to 2.18; P = 0.006).

For major bleeding, there was no evidence of heterogeneity between aspirin and dipyridamole and in the comparison of trials performed before and after 1990, around the time when anticoagulation standardisation with the international normalised ratio was being implemented. A lower daily dose of aspirin (< 100 mg) may be associated with a lower major bleeding risk than higher doses.

Authors' conclusions

Adding antiplatelet therapy, either dipyridamole or low‐dose aspirin, to oral anticoagulation decreases the risk of systemic embolism or death among patients with prosthetic heart valves. The risk of major bleeding is increased with antiplatelet therapy. These results apply to patients with mechanical prosthetic valves or those with biological valves and indicators of high risk such as atrial fibrillation or prior thromboembolic events. The effectiveness and safety of low‐dose aspirin (100 mg daily) appears to be similar to higher‐dose aspirin and dipyridamole. In general, the quality of the included trials tended to be low, possibly reflecting the era when the majority of the trials were conducted (1970s and 1980s when trial methodology was less advanced).

Plain language summary

Oral anticoagulation plus an antiplatelet drug is better than anticoagulants alone for reducing death or blood clots after heart valve replacement

After heart valve replacement oral anticoagulation (a blood thinner) is frequently used to keep blood clots from forming on the valve. These blood clots can block the flow of blood through the valve or break off and cause a stroke. Blood thinners, such as coumadin, reduce the risk of these outcomes and require careful monitoring. Drugs that affect the platelets, such as aspirin, are not effective alone but may add benefit to the blood thinners. This updated review of 13 trials found that oral anticoagulation and antiplatelet drugs were more effective than anticoagulation alone. The addition of antiplatelet drugs to anticoagulants increases the risk of bleeding by about 50%. Low‐dose aspirin (less than 100 mg daily) may be associated with the lowest risk of bleeding. However, in general the quality of the included trials tended to be low, possibly reflecting the era when the majority of the trials were conducted (1970s and 1980s when trial methodology was less advanced).

Summary of findings

Summary of findings 1. Antiplatelet and oral anticoagulation against oral anticoagulation alone for prosthetic heart valves.

Antiplatelet and oral anticoagulation against oral anticoagulation alone for prosthetic heart valves
Patient or population: patients with prosthetic heart valves Settings: Outpatients Intervention: Antiplatelet and oral anticoagulation against oral anticoagulation alone
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of Participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Antiplatelet and oral anticoagulation against oral anticoagulation alone
Thromboembolism Follow‐up: 1 to 2.5 years	Study population		OR 0.43 (0.32 to 0.59)	4122 (13 studies)	⊕⊕⊕⊕ high1,2,3,4,5,6
	67 per 1000	30 per 1000 (22 to 40)
	Moderate
	100 per 1000	46 per 1000 (34 to 62)
Mortality Follow‐up: 1 to 2.5 years	Study population		OR 0.57 (0.42 to 0.78)	4122 (13 studies)	⊕⊕⊕⊝ moderate1,2,3,4,5
	55 per 1000	32 per 1000 (24 to 43)
	Moderate
	82 per 1000	48 per 1000 (36 to 65)
Major Bleeding Follow‐up: 1 to 2.5 years	Study population		OR 1.58 (1.14 to 2.18)	3856 (11 studies)	⊕⊕⊕⊝ moderate1,2,3,4,5
	34 per 1000	53 per 1000 (39 to 71)
	Moderate
	39 per 1000	60 per 1000 (44 to 81)
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (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; OR: Odds ratio;
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

¹ None of the trials provided an adequate description of allocation concealment.
² Four trials described the generation of the allocation sequence (Sullivan 1971; PACTE 1978; Turpie 1993; Meschengieser 1997). 
³ Three trials were blinded to participant and health care provider (Sullivan 1971; Dale 1977; Turpie 1993); all were placebo controlled. 
⁴ Only three trials (PACTE 1978; Turpie 1993; LIWACAP 2007) used independent adjudication of outcome events. 
⁵ A description of follow‐up was adequate for all but one (Starkman 1982) of the thirteen trials. Loss to follow‐up in the Sullivan trial was about 10%.
⁶ Large reduction in risk of thromboembolism

Background

Patients with prosthetic heart valves are at increased risk for both valve thrombosis and arterial thromboembolic events, including stroke (Chesebro 1986; Stein 2001). Consequently, anticoagulation therapy is used to lessen the thromboembolic risk, albeit at the expense of increased anticoagulation‐associated haemorrhage. Recently, several systematic reviews have attempted to clarify the current best evidence for prosthetic valve management (Cannegieter 1994; Loewen 1998). As such, current recommendations tend to be very specific and are tailored to several clinical features including prosthetic valve location and type, presence of atrial fibrillation, and prior history of thromboembolism (Loewen 1998; Stein 2001). Unfortunately, the literature supporting these recommendations is often difficult to interpret due to small numbers of patients, lack of consistent control groups, and older studies with anticoagulation monitoring that predates the international normalised ratio (INR) but uses the prothrombin time ratio. Monitoring of oral anticoagulation with the prothrombin time was, however, imprecise because of differences in thromboplastin responsiveness. The INR was developed in the early 1980s as a means of calibrating the various thromboplastins used for measuring the prothrombin time. In essence the INR is the prothrombin time ratio that would be obtained if the World Health Organization international reference preparation for thromboplastin had been used to obtain the test (Hirsh 2001). It was through the early 1990s that the INR standard was widely implemented (CAPCSS 1997).

As a means of improving the efficacy of antithrombotic therapy following prosthetic valve implantation, oral anticoagulation has been augmented with antiplatelet drugs. Two such drugs, aspirin and dipyridamole, have been tested in this setting. Aspirin is the most widely investigated antiplatelet drug and has been shown to be effective and safe for many cardiac and other vascular indications (Patrano 2001; Antiplatelet 2002). Dipyridamole, which has both vasodilator and antiplatelet properties, is no more effective than aspirin; the mechanism of action remains controversial (Patrano 2001). We are unaware of any similar trials underway using the newer thienopyridine compounds such as ticlopidine or clopidogrel.

Although some of the trials demonstrated improved effectiveness of combination oral anticoagulation and antiplatelet drugs over anticoagulation alone, with no substantial increase in bleeding risk, the results are far from consistent. Previous meta‐analyses addressing the efficacy and safety of combined antiplatelet drugs and oral anticoagulation in patients with prosthetic valves were potentially limited, having reviewed either English language trials only, or trials limited to aspirin or dipyridamole (Fiore 1993; Cappelleri 1995; Pouleur 1995).

Despite the widespread use of dual antiplatelet therapy in patients with ischaemic heart disease and the development of novel oral anticoagulants, which have begun to replace warfarin for stroke prevention in some patients with atrial fibrillation, these strategies are only now being tested in patients with mechanical heart valves (Eikelboom 2012; Hoffman 2012). It is has been argued that the validity of the historical data for current recommendations can be questioned in view of the changes in valve prostheses, the patient population, and antithrombotic therapies (Hermans 2013). Large randomised controlled trials of novel oral anticoagulants with or without antiplatelet drugs in contemporary patients with mechanical heart valves are eagerly awaited.

Objectives

This is an update of our previous review; the goal of the review was to create a valid synthesis of all available, methodologically sound data to further assess the safety and efficacy of combined oral anticoagulant (OAC) and antiplatelet therapy versus OAC monotherapy in patients with prosthetic heart valves.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCT).

Types of participants

Patients of any age with at least one prosthetic heart valve were enrolled immediately (within two weeks) following valve surgery.

Types of interventions

Patients were randomised to either OAC and antiplatelet therapy or OAC monotherapy. Minimum follow‐up time was six months.

Types of outcome measures

Two authors independently extracted data on three major outcomes: 1) rates of thromboembolism; 2) total mortality; and 3) major haemorrhagic complications. The primary author's definitions for the above were accepted.

Search methods for identification of studies

Electronic searches

We updated the searches from 2003 and 2010 on 16 January 2013 and searched the Cochrane Central Register of Controlled Trials (CENTRAL) on The Cochrane Library (2012, Issue 12), MEDLINE (OVID, 1946 to January Week 1 2013), and EMBASE (OVID, 1980 to 2013 Week 02).

The RCT filter for MEDLINE and EMBASE has been updated from Dickersin 1994 for MEDLINE and Lefebvre 1996 for EMBASE to Lefebvre 2011.

The latest search strategies are in Appendix 1; the previous strategies are in Appendix 2 (2010) and Appendix 3 (2003).

We applied no language restrictions.

Searching other resources

We also carried out a manual search of reference lists from the individual reports, review articles, meta‐analyses, and consensus statements.

Data collection and analysis

Selection of studies

Two authors independently assessed the titles and abstracts of the references identified from the search strategy for eligibility for the review. Differences of opinion were resolved by consensus.

Assessment of risk of bias

The risk of bias tool as outlined in the Cochrane Handbook was applied. (Higgins 2011)Two authors independently assessed each trial for internal validity and extracted information on: generation of allocation sequence, concealment of allocation, participant and health care provider blinding (double‐blind), adequacy of description of follow‐up, and whether there was independent adjudication of outcome events. Differences were resolved by consensus.

Data extraction and management

We extracted the following data from each study: duration of follow‐up, target international normalised ratio or prothrombin time (PT) ratio, antiplatelet type and dose, and outcome events. Prosthetic valve type and position were not specifically reviewed as the original publications did not consistently examine the outcomes using these variables.

Data sythesis

We entered the data onto RevMan5.

Statistical analysis

We performed all analyses on an intention‐to‐treat basis. We calculated a typical odds ratio (OR) along with 95% confidence intervals (CI) for individual trials and for the summary results. For the primary analyses we used a fixed‐effect model. We considered a P value less than 0.05 (two‐sided) to be statistically significant. We used Chi² tests of heterogeneity to assess the validity of combining trials. As these tests have low sensitivity for detecting heterogeneity we assumed a more liberal level of statistical significance (P < 0.1) (Fleiss 1981). We also applied the I² statistic to quantify heterogeneity (Higgins 2011). To investigate for the possibility of publication bias we constructed funnel plots (Higgins 2011)

Sensitivity analyses

We performed sensitivity analyses on all three outcome events. We assessed the impact of the following variables (pre‐defined): the type and dose of antiplatelet agent used (aspirin versus dipyridamole); whether the study was published before or after 1990 (around the time when anticoagulation standardisation with the international normalised ratio was increasingly accepted); studies published in English versus other languages; abstract versus full publication; whether the studies were double‐blind (participant and health care provider); the type of statistical model used (fixed‐effect or random‐effects); and with the Meschengieser 1997 and LIWACAP 2007 trials excluded (see below).

Results

Description of studies

Results of the search

Our search retrieved 81 potentially relevant articles, of which 24 appeared to meet our inclusion criteria. Independent scrutiny of the complete articles and, when necessary an English translation, revealed that seven of the 24 were studies that did not meet our inclusion criteria (see Characteristics of excluded studies) (Chesebro 1983; Saitoh 1988; Hayashi 1994; Toyohira 1995; Voith 1997; Moriyama 1998; Hassouna 2000). A total of 13 studies, reported in 17 papers, were included in this review update (see Characteristics of included studies). Studies were excluded primarily because of non‐randomised designs or where the level of OAC between treatment arms was not standardised.

Included studies

A total of 13 RCTs with 4122 participants were included. One trial was identified from our updated search and included in this review update (Dong 2011). See Characteristics of included studies table for details. A summary table provides additional information on age, sex, location and type of prosthetic heart valves, atrial fibrillation, degree of anticoagulation, antiplatelet drugs, and length of follow‐up (Table 2).

1. Summary of characteristics of included studies.

Study	N	Age (years)	Sex (% male)	% Atrial fibrillation	% Mitral valve	% Aortic valve	% Multiple positions	% Mechanical valves	INR	Antiplatelet drugi	Follow‐up (years)
Sullivan 1971	163	a	50	39	51	38	11	100	3.0 to 4.5	D	1
Altman 1976	122	NR	75	NR	26	74	0	100	1.8 to 2.3	A	2
Dale 1977	148	Mean 51	75	14 b	100	0	0	100	2.0 to 2.2	A	1
Kasahara 1977	78	NR	NR	NR	NA	NA	NA	NA	d	D	~ 2.5
PACTE 1978	290	Mean 49	57	NR	41	39	21	100	e	D	1
Bran 1980	101	Mean 52	52	NR	54	34	12	100	f	D	2
Rajah 1980	165	NR	NR	NR	NR	NR	NR	NR	1.9 to 3.0	D	2.5
Starkman 1982	259	NR	NR	NR	NR	NR	NR	NR	g	D	1
Turpie 1993	370	Mean 58	53	45	44	46	10	76	3.0 to 4.5	A j	2.5
Meschengieser 1997	503	Median 53	58	18	29	66	4	100	2.5 to 3.5	A j	2
Laffort 2000	229	Mean 63	50	49 c	100	0	40 k	100	2.5 to 3.5	A	1
LIWACAP 2007	198	Mean 60	47	28	27	63	11	100	h	A j	0.5
Dong 2011	1496	Mean 34 to 35	40	40	83	43	16	100	1.8 to 2.5	Aj	2

N = number of study patients; NA = not available; NR = not reported

^aThe ages ranged from 20 to 79 years, 65% were between 40 and 59 years.
^bDescribed as continuous arrhythmia.
^cDescribed as non‐sinus.
^dProthrombin time ratio 1.7.
^eProthrombin time ratio 1.3 to 1.6.
^fNot specified.
^gQuick time 25% to 35% of control.
^hLow‐intensity (with aspirin) target 2 to 3 and high‐intensity 3 to 4.5.
ⁱA = aspirin, D = dipyridamole.
^jLow‐dose aspirin < 100 mg.
^k40% had a concomitant aortic valve replacement.

Previous meta‐analyses asking a similar question included only three (Fiore 1993) or four (Cappelleri 1995) of the English language published trials or only those trials in which dipyridamole (Pouleur 1995) was used. Two French language (PACTE 1978; Starkman 1982) and one Japanese language (Kasahara 1977) publication were identified. One trial was published in abstract form only (Rajah 1980). This latter study was identified among the reference lists from previous publications.

Of the trials included, six involved the antiplatelet agent dipyridamole at daily doses of 400 mg (Sullivan 1971; Kasahara 1977), 225 to 400 mg (Rajah 1980), 375 mg (PACTE 1978; Starkman 1982), or up to 5 mg/kg (Bran 1980). The remaining seven trials involved aspirin at doses of 500 mg daily (Altman 1976), 1000 mg daily (Dale 1977), 100 mg daily (Turpie 1993; Meschengieser 1997; LIWACAP 2007), 200 mg daily (Laffort 2000), and 75 to 100 mg daily (Dong 2011).

Trials conducted before the use of the international normalised ratio reported target anticoagulation control as follows: elevated prothrombin times at twice normal (Sullivan 1971), 1.9 to 3.0 times normal (Rajah 1980), 1.8 to 2.3 times normal (Kasahara 1977), 25% to 35% greater than normal (PACTE 1978); Quick time 25% to 35% of control (Starkman 1982); as thrombotest equivalents (10% of normal) (Dale 1977); or unspecified (Bran 1980) Quick time or thrombotest equivalents. The five trials published after 1990 reported a therapeutic anticoagulation goal international normalised ratio of 3.0 to 4.5 (Turpie 1993; LIWACAP 2007), 2.5 to 3.5 (Meschengieser 1997; Laffort 2000), and 1.8 to 2.5 (Dong 2011).

Two trials used different intensities of OAC and were included in this meta‐analysis. The Meschengieser 1997 trial was a comparison of high‐intensity OAC (international normalised ratio 3.5 to 4.5) versus low‐intensity OAC (2.5 to 3.5) and 100 mg aspirin. The LIWACAP 2007 trial was a comparison of standard intensity OAC (INR between 3.0 and 4.5, target 3.7) versus low‐intensity OAC (INR between 2.0 and 3.0, target 2.5) and 100 mg aspirin. One might anticipate that the combination might be less effective than high‐intensity anticoagulation and, as such, provide a conservative estimate of effectiveness. Similarly, it is not unreasonable to expect that the risk of bleeding would be less than if the intensity of anticoagulation were similar in both groups. In essence, the inclusion of these trials should push the odds ratios for both effectiveness and bleeding towards the null. We planned sensitivity analyses to assess the impact of these trials on the overall results.

Study outcomes

Of the three main study outcomes, an arterial thromboembolic event was well defined and the primary endpoint of all included trials. Definitions were similar and involved either transient or permanent cerebral ischaemic injury or ultrasound and/or surgically confirmed other systemic arterial embolism. One trial (Laffort 2000) reported prosthetic non‐obstructive thrombi and transient ischaemic attacks (TIA) as minor embolic events. For this meta‐analysis, the reported TIAs were classified as major thromboembolic events.

The reported data for major haemorrhagic complications of anticoagulant therapy, with or without antiplatelet therapy, were less consistent. Three of the original publications provided no data on bleeding (Kasahara 1977; Bran 1980; Rajah 1980). Data on bleeding for the Kasahara trial were abstracted from the dipyridamole meta‐analysis (Pouleur 1995) that included unpublished data obtained from the registration file reviewed by the US Food and Drug Administration. Data from the Rajah article were also available from the Pouleur meta‐analysis but excluded from ours as the denominator from the active treatment groups was reported as 68 rather than 78 as in the original abstract (rendering the results somewhat suspect). The Meschengieser 1997 trial defined significant bleeding as that causing death, requiring transfusion or hospitalisation. In the Turpie 1993 trial major bleeding was defined as overt haemorrhage associated with > 20 g/L drop in the haemoglobin level, the requirement for transfusion of more than two units of blood, or any intracranial, intraocular, intraarticular, or retroperitoneal bleeding. Minor bleeding constituted either epistaxis, genitourinary bleeding, or easy bruising. In the LIWACAP 2007 trial major bleeding comprised intracranial bleeding at computerised tomography (CT) scan, retroperitoneal bleeding at CT scan, ocular bleeding with blindness, articular bleeding, bleeding that reduced haemoglobin concentration by two g/dL or more or requiring the transfusion of two or more units of blood, or bleeding that required surgical intervention. Two trials (Altman 1976; Dale 1977) did not distinguish between major and minor bleeding events; for this analysis, any intracerebral or gastroenteric bleeding event or episode of haemoptysis was taken to represent a significant haemorrhage. One trial (Sullivan 1971) reported only three non‐fatal gastrointestinal bleeding events that were presumed to be clinically severe. One trial (Dong 2011) did not include a definition of major versus minor bleeding. A table outlined all types of haemorrhages; for this analysis major bleeding events included the cerebral haemorrhages.

Mortality data were explicit for 12 trials and not provided in the original publication of the Kasahara 1977 trial; these data were abstracted from the dipyridamole meta‐analysis (Pouleur 1995) that included unpublished data. The LIWACAP 2007 trial reported vascular deaths and no other deaths were reported.

Risk of bias in included studies

Studies were included in the meta‐analysis if: 1) patients with prosthetic heart valves were enrolled; 2) there was a comparison of the addition of an antiplatelet agent to OAC; 3) treatment groups were assigned through random allocation; and 4) objective methods were used to assess for the development of major clinical outcomes or adverse consequences. Abstracts were not excluded. Review authors were not blinded as to the author, journal, or type of publication.

Funnel plots for thromboembolic events, mortality, and major bleeding did not show evidence of publication bias (Figure 1; Figure 2; Figure 3). A 'Risk of bias' summary based on our judgements about each risk of bias item for each included study is shown in Figure 4.

1.

Funnel plot of comparison: 1 Antiplatelet and oral anticoagulation against oral anticoagulation alone, outcome: 1.1 Thromboembolism.

2.

Funnel plot of comparison: 1 Antiplatelet and oral anticoagulation against oral anticoagulation alone, outcome: 1.2 Mortality.

3.

Funnel plot of comparison: 1 Antiplatelet and oral anticoagulation against oral anticoagulation alone, outcome: 1.3 Major bleeding.

4.

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

Except for Turpie 1993, the trials tended to be of lower methodological quality. None of the trials provided an adequate description of allocation concealment. Four trials described the generation of the allocation sequence (Sullivan 1971; PACTE 1978; Turpie 1993; Meschengieser 1997). Three trials were blinded to participant and health care provider (Sullivan 1971; Dale 1977; Turpie 1993); all were placebo‐controlled. Only three trials (PACTE 1978; Turpie 1993; LIWACAP 2007) used independent adjudication of outcome events. A description of follow‐up was adequate for all but one (Starkman 1982) of the 13 trials.

Effects of interventions

See: Table 1

See Summary of findings table 1.

Thromboembolic events

The addition of an antiplatelet agent to OAC resulted in a clinically important and statistically significant reduction in thromboembolic events. Compared with OAC alone the combination of OAC and an antiplatelet agent reduced the risk of thromboembolic events with an odds ratio (OR) of 0.43 (95% confidence interval (CI) 0.32 to 0.59; P < 0.00001; Analysis 1.1). The test for heterogeneity was not significant (P = 0.83, I²= 0%). This treatment effect corresponds to a relative risk of 0.45 (95% CI 0.34 to 0.61). Trials were grouped according to study era (performed pre‐1990 or later) and the antiplatelet agent used (dipyridamole or aspirin; Analysis 2.1). Results were consistent across time and between the available agents. The results were not influenced by the meta‐analysis technique used, whether a fixed‐effect or random‐effects model; all were statistically significant.

1.1. Analysis.

Comparison 1: Antiplatelet and oral anticoagulation against oral anticoagulation alone, Outcome 1: Thromboembolism

2.1. Analysis.

Comparison 2: Subgroup analyses, Outcome 1: Thromboembolic events (aspirin vs dipyridamole)

Mortality

Total mortality was significantly reduced when antiplatelet agents were added to OAC with an OR of 0.57 (95% CI 0.42 to 0.78; P = 0.0004; Analysis 1.2). There was no evidence of important heterogeneity (P = 0.15, I² = 29%). The effect on mortality corresponds to a relative risk of 0.59 (95% CI 0.44 to 0.79). Both aspirin and dipyridamole trials (Analysis 2.2; Analysis 2.3) and trials that were published before or after 1990 reduced mortality similarly.

1.2. Analysis.

Comparison 1: Antiplatelet and oral anticoagulation against oral anticoagulation alone, Outcome 2: Mortality

2.2. Analysis.

Comparison 2: Subgroup analyses, Outcome 2: Mortality (aspirin vs dipyridamole)

2.3. Analysis.

Comparison 2: Subgroup analyses, Outcome 3: Mortality (aspirin only trials, by dosage)

Major bleeding

Data on major bleeding were available for 12 trials. Major bleeding was increased with combination therapy, with an OR of 1.58 (95% CI 1.14 to 2.18; P = 0.006; Analysis 1.3). The test of heterogeneity was not significant (P = 0.25, I² = 20%). The Meschengieser 1997and LIWACAP 2007 trials were a comparison of high‐intensity OAC versus the combination of low‐dose aspirin and low‐intensity OAC; therefore, different degrees of anticoagulation were targeted. In those trials the risk of bleeding was not shown to be different between groups (OR 0.76, 95% CI 0.33 to 1.74, P = 0.51; Analysis 3.4). When the Meschengieser 1997 and LIWACAP 2007 trials were excluded from the analysis, there was stronger evidence that the risk of bleeding was increased with the addition of an antiplatelet agent (OR 1.80, 95% CI 1.26 to 2.57, P = 0.001; Analysis 3.3). Unpublished data on bleeding from the Kasahara 1977 trial were used in a previous meta‐analysis (Pouleur 1995) and included in our analysis. Repeating the analysis with these data excluded did not alter the results. Similarly, the inclusion of the small number of events from the unpublished bleeding data from the Rajah 1980 trial did not influence the results or conclusions.

1.3. Analysis.

Comparison 1: Antiplatelet and oral anticoagulation against oral anticoagulation alone, Outcome 3: Major bleeding

3.4. Analysis.

Comparison 3: Selected sensitivity analyses, Outcome 4: Major bleeding (low‐intensity OAC versus higher‐intensity OAC)

3.3. Analysis.

Comparison 3: Selected sensitivity analyses, Outcome 3: Major bleeding (excluding Meschengieser and LIWACAP)

There was no evidence of statistical interaction for the risk of major bleeding between the trials using dipyridamole or aspirin (OR 2.22 versus 1.44; P = 0.29; Analysis 2.4) or whether the studies were published before or after 1990 (OR 2.34 versus 1.26; P = 0.08). We explored an aspirin dose effect. There did not appear to be an excess risk of bleeding among the four low‐dose aspirin trials (OR 0.96, 95% CI 0.60 to 1.55, P = 0.87; Analysis 2.5). On the other hand, higher‐dose aspirin was associated with a significant excess risk of bleeding (OR 2.58, 95% CI 1.43 to 4.66, P = 0.002; Analysis 2.5). There was evidence of statistical interaction between high and low‐dose aspirin (P = 0.01) and this remained when the Meschengieser 1997 and LIWACAP 2007 trials were excluded (P = 0.04).

2.4. Analysis.

Comparison 2: Subgroup analyses, Outcome 4: Major bleeding (aspirin vs dipyridamole)

2.5. Analysis.

Comparison 2: Subgroup analyses, Outcome 5: Major bleeding (aspirin only trials, by dosage)

Additional sensitivity analyses

We explored the impact of excluding the Meschengieser 1997 and LIWACAP 2007 trials on the endpoints of death and thromboembolic events (Analysis 3.1; Analysis 3.2); the results were unchanged. There were no differences for any of the endpoints among trials published in English as compared with other languages, whether published as an abstract or manuscript, or if double‐blind methodology was used or not.

3.1. Analysis.

Comparison 3: Selected sensitivity analyses, Outcome 1: Thromboembolism (excluding Meschengieser and LIWACAP))

3.2. Analysis.

Comparison 3: Selected sensitivity analyses, Outcome 2: Mortality (excluding Meschengieser and LIWACAP)

Discussion

The main conclusion of our meta‐analysis is that the addition of an antiplatelet agent, either aspirin or dipyridamole, to warfarin in patients with prosthetic heart valves reduces the risk of death and systemic thromboembolic events. Our analysis showed that dipyridamole and aspirin reduced the risks of death and thromboembolism similarly. The risk of major bleeding is increased with both dipyridamole and aspirin. Although the point estimate of bleeding risk seemed to favour aspirin over dipyridamole and for trials performed after 1990 rather than before, there was no evidence of statistical heterogeneity. Lower‐dose (100 mg or less) aspirin may have the lowest bleeding risk.

Chesebro and colleagues performed a randomised trial comparing warfarin and dipyridamole (400 mg daily) to warfarin and aspirin (500 mg daily) in patients with a prosthetic heart valve replacement (Chesebro 1983). The risk of a thromboembolic event was slightly lower, but not statistically significant, among those allocated dipyridamole compared with aspirin (0.5 versus 1.8 per 100 patient‐years). Bleeding rates were higher among those receiving concomitant aspirin as compared to dipyridamole (6.6 versus 1.6 per 100 patient‐years, P < 0.001). These results are discordant with ours and may reflect the dose of aspirin chosen; 500 mg in the Chesebro study (Chesebro 1983) as compared to the lower risk of bleeding when doses of aspirin of 100 mg daily (Turpie 1993) are used.

The Antiplatelet Trialists' Collaboration found that aspirin or other antiplatelet drugs were protective against vascular events in high‐risk patients. Furthermore, they felt that the available evidence supports the use of low‐dose aspirin (75 to 150 mg daily) as an effective antiplatelet regimen for long‐term use (Antiplatelet 2002). In our meta‐analysis the relative effectiveness and safety of aspirin may reflect patient selection, the target intensity of anticoagulation (target international normalised ratio (INR)), or the dose of aspirin used. Among the six aspirin trials there was some evidence of statistical heterogeneity for total mortality (P = 0.09) but not risk of thromboembolism (P = 0.48) or major bleeding (P = 0.15). One possible explanation is the dose of aspirin used: 100 mg daily in one trial (Turpie 1993) compared to 200 to 1000 mg daily in the other aspirin trials (Altman 1976; Dale 1977; Laffort 2000). The risk of death was lower in the low‐dose aspirin trial (Turpie 1993) (odds ratio (OR) 0.37, 95% confidence interval (CI) 0.17 to 0.84; P = 0.014) compared with the higher‐dose aspirin trials (OR 1.15, 95% CI 0.53 to 2.49; P = 0.72) and the test of interaction for the differences between these subgroups was conventionally statistically significant (P = 0.05). For the low‐dose aspirin trial (Turpie 1993) the risk of major bleeding (OR 1.29, 95% CI 0.68 to 2.44) was not increased compared with warfarin alone. Although the risk of major bleeding was increased for the higher‐dose aspirin trials (OR 2.58) and statistically significantly greater than OAC alone (P = 0.002), there was no statistical evidence of interaction based on dose of aspirin (P = 0.13). It must be stressed that these subgroup analyses, although pre‐defined, are based on a limited number of events in each subgroup and, as such, are potentially unstable. They should be considered hypothesis generating.

Nonetheless, more detailed discussion of three of the recent trials may provide further insight into the potential role of aspirin. Of the 12 trials, Turpie 1993 had the highest methodology score (Characteristics of included studies). It was a double‐blind, randomised controlled trial where 186 patients were assigned to aspirin (100 mg/day sustained release) plus warfarin and 184 to placebo plus warfarin (Turpie 1993). Patients were included if they had a mechanical prosthetic valve or were those with tissue valves and atrial fibrillation or a history of thromboembolism. The target INR was 3.0 to 4.5. The primary endpoint (major embolism or death) was reduced among those assigned to aspirin (1.9% versus 8.5% per year; P < 0.001). The stroke rate (1.3% versus 4.2% per year; P = 0.027) and overall mortality (2.8% versus 7.4%; P = 0.01) was reduced with aspirin. Furthermore, a composite outcome that could reflect net clinical benefit (major systemic embolism, nonfatal intracranial haemorrhage, death due to haemorrhage, and vascular deaths) was also reduced with aspirin (3.9% versus 9.9% per year; P = 0.005). Although the risk of bleeding was increased with aspirin this was primarily due to minor bleeding including bruising, epistaxis, and haematuria. Importantly, the risk of major haemorrhagic events did not differ significantly between groups (8.5% aspirin versus 6.6% placebo; P = 0.43).

In the Meschengieser 1997 trial patients were randomised to either a high target INR (3.5 to 4.5; mean achieved 3.98) or a lower target INR (2.5 to 3.5; mean achieved 3.11) plus aspirin 100 mg daily. The primary outcome events were rates of thromboembolism and bleeding. The rates of thromboembolism were similar at 2.8% and 2.7%, respectively. The risk of major bleeding (4.5% warfarin alone versus 2.3% warfarin plus aspirin) and minor bleeding (17% warfarin alone versus 14% warfarin plus aspirin) did not differ between groups but tended to favour the combination of low‐dose aspirin and lower target level of anticoagulation. Three intracranial haemorrhages occurred in the warfarin alone arm; none were seen in the combination arm. Therefore, the addition of low‐dose aspirin with a lower level of anticoagulation was as effective, and possibly safer, when compared with a higher level of anticoagulation. Similar results were seen in the LIWACAP 2007 trial in which patients were randomised to standard‐intensity OAC (INR between 3.0 and 4.5, target 3.7) versus low‐intensity OAC (INR between 2.0 and 3.0, target 2.5) and 100 mg aspirin. Follow‐up was only for six months and there were few events in this pilot study. There were no differences in thromboembolic or major bleeding events although the trial was severely underpowered.

These results are consistent with the randomised trial by Altman 1991 who compared the effect of a low (INR 2.0 to 3.0) or high (INR 3.0 to 4.3) degree of anticoagulation in combination with dipyridamole (150 mg/day) and aspirin (660 mg/day) in patients with heart valve replacement. The rates of thromboembolic events were similar between the low and high INR groups (1.92 versus 4.94 per 100 patient‐years, respectively), although there were very few events overall. The risk of bleeding, however, was less with the lower target INR (3.8 versus 24.7 per 100 patient‐years, P < 0.02). They concluded that a lower INR (2.0 to 3.0) used conjointly with platelet inhibitors was effective and safer than a higher target INR (Altman 1991).

The most recent trial (Dong 2011) included young patients (mean age of 35 years) with primarily rheumatic heart disease who underwent mechanical valve replacement. The risk of major bleeding was only 0.4%. In addition, they used low‐dose aspirin (75 to 100 mg daily) and a target INR of 1.8 to 2.5. In this study the rate of reliable anticoagulation was only 33% to 36%. Thromboembolism rates favoured the combination of OAC and low‐dose aspirin at 2.1% versus 3.6% with OAC alone (OR 0.59, 95% CI 0.32 to 1.09).

Improvements over previous meta‐analyses

The first overview comparing rates of valve thrombosis, major embolism, and bleeding was unable to show an advantage of the combination of aspirin and anticoagulation over anticoagulation alone; bleeding risk was increased, however (Cannegieter 1994). The studies included in that overview were not comparative randomised trials. The meta‐analyses performed by Fiore and colleagues (Fiore 1993) and Cappelleri and colleagues (Cappelleri 1995) were subject to bias through inclusion of English language only trials published as full manuscripts. The Fiore meta‐analysis, published in abstract form only, included four trials comparing the use of aspirin as an adjunct to oral anticoagulation. The Cappelleri meta‐analysis included five trials; four involving aspirin and one involving dipyridamole. Moreover, the study by Chesebro and colleagues (Chesebro 1983) was included in both these meta‐analyses, but was excluded from ours, as patients treated with warfarin were only randomised to receive either aspirin or dipyridamole. Although a control group receiving warfarin alone was subsequently included in the analysis, patients were not randomised to warfarin alone or warfarin and either of the two antiplatelet regimens. The Pouleur et al meta‐analysis comprised trials of dipyridamole only (Pouleur 1995). It also included previously unpublished data, updated from the original publications, and included in a submission to the US Food and Drug Administration (FDA). For the endpoints of thromboembolic events and death the results of our meta‐analysis are in accord with theirs and sensitivity analyses show that inclusion or exclusion of the unpublished data from published trials does not materially change our results or conclusions. The one discrepancy is for the endpoint of bleeding. Our meta‐analysis suggests that bleeding risk is increased with dipyridamole but theirs does not (OR 2.22 versus 1.001, respectively). The test of heterogeneity between the analyses is conventionally statistically significant (P = 0.046). This may reflect our emphasis on major bleeding, whereas they included data on any haemorrhagic events, whether fatal or not (Pouleur 1995). The differences may also reflect the use of unpublished data. In this regard, the bleeding risks for two of the studies (Sullivan 1971; PACTE 1978) were qualitatively different in the Pouleur meta‐analysis as compared with the original publications. In both original studies the bleeding risk was slightly higher with dipyridamole; in the data used in the meta‐analysis the bleeding risk was lower. As such, our results may represent a more conservative estimate of bleeding risk.

In comparison with the above, our meta‐analysis is more powerful, including data on 4122 participants from 13 trials and is less subject to bias through our inclusion of trials in any language, using dipyridamole or aspirin, and publications as a manuscript or abstract. Furthermore, the robustness of our conclusions is reinforced through extensive sensitivity analyses which, among other things, included the impact of unpublished data, English language or other publications, and some measures of individual trial quality. None of these methodological factors impacted on our estimate of the effectiveness and safety of adjunctive antiplatelet therapy.

Potential limitations

The results and conclusions derived from a meta‐analysis are governed by the internal validity of the individual studies included. We found no evidence of publication bias (Figure 1; Figure 2; Figure 3). In general, the quality of the included trials tended to be low (Figure 4), possibly reflecting the era when the majority of the trials were conducted (1970s and 1980s when trial methodology was less advanced). It has also been shown that incorporation of studies of low methodological quality tends to show an increased estimate of benefit (Moher 1998). However, our sensitivity analyses did not detect any effect of exclusion of individual trials (some of which were of lesser quality, for example Rajah 1980 published as an abstract only) or a difference among double‐blinded trials on the rates of thromboembolism or death.

We did not use blinded techniques during study selection or data abstraction as such procedures have not been found to be necessary (Berlin 1997). The possibility of publication bias in medical research is important, but difficult to eliminate when performing a meta‐analysis (Lau 1997). Currently, use of unpublished data in meta‐analyses remains controversial, but such data should not be systematically excluded (Cook 1993). Our analysis included all published data from the included randomised trials as well as some unpublished data obtained for an FDA submission and used in a previous meta‐analysis (Pouleur 1995). We were unable to verify these unpublished data. However, sensitivity analyses excluding the unpublished data did not alter our conclusions. We included trials which claimed to have used proper randomisation techniques but did not seek to authenticate the veracity of these claims (Clarke 1995). Furthermore, although improper concealment of treatment allocation has been shown to be an important source of bias in randomised controlled trials (and therefore in meta‐analyses based upon them) none of the included trials provided adequate information on this (Schultz 1995). Given the methodological quality of the included studies our conclusions would be of moderate grade.

Authors' conclusions

Implications for practice.

Our results suggest that antiplatelet drugs, whether aspirin or dipyridamole, could be safely added to anticoagulation with an acceptable risk of bleeding and with the expectation that rates of death and thromboembolism would be reduced. These results apply to patients with mechanical prosthetic valves or those with biological valves and indicators of high risk such as atrial fibrillation or prior thromboembolic events.

As to choice of antiplatelet agent, our data do not allow us to make strong recommendations. The effectiveness of low‐dose aspirin (100 mg daily) appears similar to higher‐dose aspirin and dipyridamole. Our updated meta‐analysis suggests that a daily aspirin dose of 100 mg or less seems to be associated with a lower risk of major bleeding. The Antiplatelet Trialists' Collaboration (Antiplatelet 2002) concluded that low‐dose aspirin (75 mg daily) is an effective antiplatelet regimen for long‐term use and is at least as effective as higher daily doses. Dipyridamole was no more effective than aspirin. Moreover, Stein 2001 and colleagues (Antithrombotic Therapy in Patients with Mechanical and Biological Prosthetic Heart Valves) stated: "Data are insufficient to recommend dipyridamole over low dose aspirin in combination with warfarin" and further "...in view of the advantageous effects of low‐dose aspirin in combination with oral anticoagulants, the indications for use of dipyridamole require further investigation". Therefore, there appears to be a consensus view that low‐dose aspirin is the preferred antiplatelet drug.

It is quite possible that outcomes would be improved with a more widespread use of antiplatelet drugs as compared with existing guidelines (Stein 2001). A survey performed in 1997 showed that low‐dose aspirin was underused by North American cardiac surgeons. Only 21% of respondents routinely used aspirin in conjunction with anticoagulants among patients with prosthetic heart valves. The two most common reasons for not using aspirin were the perceived increase in bleeding risk (49% of non‐users) or lack of proven benefit (23% of non‐users) (Ray 1997). These concerns are unfounded and not in accord with the available trial and meta‐analytic evidence.

Implications for research.

Prosthetic valve type and position were not specifically reviewed as the original publications did not consistently examine the outcomes using these variables. Certainly, the benefits of combined oral anticoagulant (OAC) and antiplatelet drugs would be expected to be greatest in patients with high‐risk characteristics for thromboembolism. For example, patients with a mechanical mitral valve with atrial fibrillation, combined mitral and aortic valve prostheses, or any valve type with prior thromboembolic events would be considered higher risk than a single mechanical aortic or mitral prothesis and no high‐risk indicators. Trials of OAC and OAC with various antiplatelet drugs, assessing the balance of thromboembolic event prevention versus increased risk of bleeding, in these relatively lower‐risk subgroups would be clinically important.

Although aspirin appears to be the preferred antiplatelet drug, the recent widespread use of adenosine diphosphate (ADP) receptor blockers (e.g. clopidogrel (Sudlow 2009), prasugrel (TRITON TIMI‐38 2009), or ticagrelor (Wallentin 2009)) and favourable results with novel cyclo‐oxygenase inhibitors (triflusal, TRAC 2005) raises the issue of whether aspirin is indeed the best antiplatelet to combine with OAC in the population of patients with prosthetic heart valves. A contemporary, well‐designed randomised controlled trial comparing OAC with low‐dose aspirin against OAC with one of the newer antiplatelet agents would also likely be clinically important.

What's new

Date	Event	Description
10 March 2021	Review declared as stable	We are not aware of any new studies since the publication of this review and this review is therefore not considered for updating.

History

Protocol first published: Issue 1, 2002
Review first published: Issue 4, 2003

Date	Event	Description
6 March 2013	New citation required and conclusions have changed	The addition of a new study provides some evidence that low‐dose aspirin (< 100 mg daily) may be associated with a lower risk of major bleeding.
26 February 2013	New search has been performed	The search was updated in January 2013. One new study was found for inclusion.
15 February 2011	Feedback has been incorporated	Added a link to the Cochrane Editorial Unit's report on feedback on anticoagulant reviews in the 'Published notes' section.
15 February 2010	New search has been performed	The search was updated to January 2010. One new study was found and added to the review. Conclusions not changed.
8 September 2008	Amended	Converted to new review format.
9 June 2003	New citation required and conclusions have changed	Substantive amendment.

Notes

Feedback received on this review, and other reviews and protocols on anticoagulants, is available on the Cochrane Editorial Unit website at http://www.editorial-unit.cochrane.org/anticoagulants-feedback.

Appendices

Appendix 1. Search strategies 2013

CENTRAL on The Cochrane Library

#1 MeSH descriptor: [Heart Valve Prosthesis] this term only
#2 MeSH descriptor: [Heart Valve Prosthesis Implantation] this term only
#3 (valv* near/3 prosthe*)
#4 (valv* near/3 bioprosthe*)
#5 (artificial near/3 valv*)
#6 (valv* near/3 replac*)
#7 (mechanical near/3 valv*)
#8 (#1 or #2 or #3 or #4 or #5 or #6 or #7)
#9 MeSH descriptor: [Anticoagulants] explode all trees
#10 anticoagulant*
#11 anti‐coagulant*
#12 "antivitamin k"
#13 warfarin
#14 coumadin*
#15 (vitamin near/3 antagonist*)
#16 acenocoumarol
#17 phenprocoumon
#18 coumarin*
#19 (#9 or #10 or #11 or #12 or #13 or #14 or #15 or #16 or #17 or #18)
#20 MeSH descriptor: [Platelet Aggregation Inhibitors] explode all trees
#21 antiplatelet*
#22 anti‐platelet*
#23 antithrombocytic*
#24 (platelet* near/3 inhibitor*)
#25 (platelet* near/3 antagonist*)
#26 antiaggregant*
#27 aspirin*
#28 dipyridamole
#29 ticlopidine
#30 clopidogrel
#31 trifusal
#32 thienopyridine*
#33 (#20 or #21 or #22 or #23 or #24 or #25 or #26 or #27 or #28 or #29 or #30 or #31 or #32)
#34 (#8 and #19 and #33)

MEDLINE

1. exp Heart Valve Prosthesis/
2. (valv$ adj3 prosthe$).tw.
3. (valv$ adj3 bioprosthe$).tw.
4. (artificial adj3 valv$).tw.
5. exp Heart Valve Prosthesis Implantation/
6. (artificial adj3 valv$).tw.
7. (valv$ adj3 replac$).tw.
8. (mechanical adj3 valv$).tw.
9. or/1‐8
10. exp Anticoagulants/
11. anticoagulant$.tw.
12. anti‐coagulant$.tw.
13. "antivitamin k".tw.
14. warfarin.tw.
15. coumadin.tw.
16. acenocoumarol.tw.
17. phenprocoumon.tw.
18. coumarin$.tw.
19. or/10‐18
20. exp Platelet Aggregation Inhibitors/
21. antiplatelet$.tw.
22. anti‐platelet$.tw.
23. antiaggregant$.tw.
24. antithrombocytic$.tw.
25. platelet inhibitor$.tw.
26. aspirin$.tw.
27. dipyridamole.tw.
28. ticlopidine.tw.
29. trifusal.tw.
30. thienopyridine$.tw.
31. platelet antagonist$.tw.
32. or/20‐31
33. 9 and 19 and 32
34. randomized controlled trial.pt.
35. controlled clinical trial.pt.
36. randomized.ab.
37. placebo.ab.
38. drug therapy.fs.
39. randomly.ab.
40. trial.ab.
41. groups.ab.
42. 34 or 35 or 36 or 37 or 38 or 39 or 40 or 41
43. exp animals/ not humans.sh.
44. 42 not 43
45. 33 and 44
46. (2010* or 2011* or 2012* or 2013*).ed.
47. 45 and 46

EMBASE

1. exp heart valve prosthesis/
2. (valv$ adj3 prosthe$).tw.
3. (valv$ adj3 bioprosthe$).tw.
4. (artificial adj3 valv$).tw.
5. heart valve replacement/
6. (artificial adj3 valv$).tw.
7. (valv$ adj3 replac$).tw.
8. (mechanical adj3 valv$).tw.
9. or/1‐8
10. exp anticoagulant agent/
11. anticoagulant$.tw.
12. anti‐coagulant$.tw.
13. "antivitamin k".tw.
14. warfarin.tw.
15. coumadin.tw.
16. acenocoumarol.tw.
17. phenprocoumon.tw.
18. coumarin$.tw.
19. or/10‐18
20. exp antithrombocytic agent/
21. antiplatelet$.tw.
22. anti‐platelet$.tw.
23. antiaggregant$.tw.
24. antithrombocytic$.tw.
25. platelet inhibitor$.tw.
26. platelet antagonist$.tw.
27. aspirin$.tw.
28. dipyridamole.tw.
29. ticlopidine.tw.
30. trifusal.tw.
31. thienopyridine$.tw.
32. or/20‐31
33. 9 and 19 and 32
34. random$.tw.
35. factorial$.tw.
36. crossover$.tw.
37. cross over$.tw.
38. cross‐over$.tw.
39. placebo$.tw.
40. (doubl$ adj blind$).tw.
41. (singl$ adj blind$).tw.
42. assign$.tw.
43. allocat$.tw.
44. volunteer$.tw.
45. crossover procedure/
46. double blind procedure/
47. randomized controlled trial/
48. single blind procedure/
49. 34 or 35 or 36 or 37 or 38 or 39 or 40 or 41 or 42 or 43 or 44 or 45 or 46 or 47 or 48
50. (animal/ or nonhuman/) not human/
51. 49 not 50
52. 33 and 51
53. (2010* or 2011* or 2012* or 2013*).em.
54. 52 and 53

Appendix 2. Search strategies 2010

CENTRAL on The Cochrane Library

#1 MeSH descriptor heart valve prosthesis this term only
#2 MeSH descriptor Heart Valve Prosthesis Implantation this term only
#3 (valv* in All Text near/3 prosthe* in All Text)
#4 (valv* in All Text near/3 bioprosthe* in All Text)
#5 (artificial in All Text near/3 valv* in All Text)
#6 (valv* in All Text near/3 replac* in All Text)
#7 (mechanical in All Text near/3 valv* in All Text)
#8 (#1 or #2 or #3 or #4 or #5 or #6 or #7)
#9 MeSH descriptor anticoagulants explode all trees
#10 anticoagulant* in All Text
#11 anti‐coagulant* in All Text
#12 "antivitamin k" in All Text
#13 warfarin in All Text
#14 coumadin* in All Text
#15 (vitamin in All Text near/3 antagonist* in All Text)
#16 Acenocoumarol in All Text
#17 phenprocoumon in All Text
#18 coumarin* in All Text
#19 (#9 or #10 or #11 or #12 or #13 or #14 or #15 or #16 or #17 or #18)
#20 MeSH descriptor Platelet Aggregation Inhibitors explode all trees
#21 antiplatelet* in All Text
#22 anti‐platelet* in All Text
#23 antithrombocytic* in All Text
#24 (platelet* in All Text near/3 inhibitor* in All Text)
#25 (platelet* in All Text near/3 antagonist* in All Text)
#26 antiaggregant* in All Text
#27 aspirin* in All Text
#28 dipyridamole in All Text
#29 ticlopidine in All Text
#30 clopidogrel in All Text
#31 trifusal in All Text
#32 thienopyridine* in All Text
#33 (#20 or #21 or #22 or #23 or #24 or #25 or #26 or #27 or #28 or #29 or #30 or #31 or #32)
#34 (#8 and #19 and #33)

MEDLINE (on Ovid)

1 exp Heart valve prosthesis/
2 (valv$ adj3 prosthe$).tw.
3 (valv$ adj3 bioprosthe$).tw.
4 (artificial adj3 valv$).tw.
5 exp Heart valve prosthesis implantation/
6 (artificial adj3 valv$).tw.
7 (valv$ adj3 replac$).tw.
8 (mechanical adj3 valv$).tw.
9 or/1‐8
10 exp Anticoagulants/
11 anticoagulant$.tw.
12 anti‐coagulant$.tw.
13 "antivitamin k".tw.
14 warfarin.tw.
15 coumadin.tw.
16 Acenocoumarol.tw.
17 phenprocoumon.tw.
18 coumarin$.tw.
19 or/10‐18
20 exp Platelet aggregation inhibitors/
21 antiplatelet$.tw.
22 anti‐platelet$.tw.
23 antiaggregant$.tw.
24 antithrombocytic$.tw.
25 platelet inhibitor$.tw.
26 platelet antagonist$.tw.
27 aspirin$.tw.
28 dipyridamole.tw.
29 ticlopidine.tw.
30 trifusal.tw.
31 thienopyridine$.tw.
32 or/20‐31
33 9 and 19 and 32
34 randomized controlled trial.pt.
35 controlled clinical trial.pt.
36 randomized.ab.
37 placebo.ab.
38 exp Clinical Trials as Topic/
39 randomly.ab.
40 trial.ti.
41 or/34‐40
42 exp animal/ not humans/
43 41 not 42
44 33 and 43

EMBASE (on Ovid) (to 2010 Week 02)

1 exp Heart valve prosthesis/
2 (valv$ adj3 prosthe$).tw.
3 (valv$ adj3 bioprosthe$).tw.
4 (artificial adj3 valv$).tw.
5 heart valve replacement/
6 (artificial adj3 valv$).tw.
7 (valv$ adj3 replac$).tw.
8 (mechanical adj3 valv$).tw.
9 or/1‐8
10 exp anticoagulant agent/
11 anticoagulant$.tw.
12 anti‐coagulant$.tw.
13 "antivitamin k".tw.
14 warfarin.tw.
15 coumadin.tw.
16 Acenocoumarol.tw.
17 phenprocoumon.tw.
18 coumarin$.tw.
19 or/10‐18
20 exp antithrombocytic agent/
21 antiplatelet$.tw.
22 anti‐platelet$.tw.
23 antiaggregant$.tw.
24 antithrombocytic$.tw.
25 platelet inhibitor$.tw.
26 platelet antagonist$.tw.
27 aspirin$.tw.
28 dipyridamole.tw.
29 ticlopidine.tw.
30 trifusal.tw.
31 thienopyridine$.tw.
32 or/20‐31
33 9 and 19 and 32
34 random$.tw.
35 factorial$.tw.
36 (crossover$ or cross‐over$).tw.
37 placebo$.tw.
38 (doubl$ adj blind$).tw.
39 (singl$ adj blind$).tw.
40 assign$.tw.
41 allocat$.tw.
42 volunteer$.tw.
43 Crossover Procedure/
44 Double‐blind Procedure/
45 Randomized Controlled Trial/
46 Single‐blind Procedure/
47 or/34‐46
48 (animal/ or nonhuman/) not human/
49 47 not 48
50 33 and 49

Appendix 3. Search strategy 2003

CENTRAL on The Cochrane Library

#1 HEART‐VALVE‐PROSTHESIS*:ME
#2 HEART‐VALVE‐PROSTHESIS‐IMPLANTATION*:ME
#3 (VALVE near PROSTHE*)
#4 (VALVE near BIOPROSTHE*)
#5 (ARTIFICIAL near VALVE)
#6 (VALVE near REPLACEMENT*)
#7 (((((#1 or #2) or #3) or #4) or #5) or #6)
#8 ANTICOAGULANTS*:ME
#9 ANTICOAGULANT*
#10 (VITAMIN near ANTAGONIST*)
#11 ANTIVITAMIN*
#12 WARFARIN
#13 COUMADIN
#14 (((((#8 or #9) or #10) or #11) or #12) or #13)
#15 PLATELET‐AGGREGATION‐INHIBITORS*:ME
#16 ANTIPLATELET*
#17 ANTITHROMBOCYTIC*
#18 (PLATELET* near INHIBITOR*)
#19 (PLATELET* near ANTAGONIST*)
#20 ANTIAGGREGANT*
#21 ASPIRIN
#22 DIPYRIDAMOLE
#23 TICLOPIDINE
#24 ((((((((#15 or #16) or #17) or #18) or #19) or #20) or #21) or #22) or #23)
#25 ((#7 and #14) and #24)

Data and analyses

Comparison 1. Antiplatelet and oral anticoagulation against oral anticoagulation alone.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Thromboembolism	13	4122	Odds Ratio (M‐H, Fixed, 95% CI)	0.43 [0.32, 0.59]
1.2 Mortality	13	4122	Odds Ratio (M‐H, Fixed, 95% CI)	0.57 [0.42, 0.78]
1.3 Major bleeding	11	3856	Odds Ratio (M‐H, Fixed, 95% CI)	1.58 [1.14, 2.18]

Comparison 2. Subgroup analyses.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Thromboembolic events (aspirin vs dipyridamole)	13		Odds Ratio (M‐H, Fixed, 95% CI)	Subtotals only
2.1.1 Aspirin	7	3066	Odds Ratio (M‐H, Fixed, 95% CI)	0.45 [0.31, 0.67]
2.1.2 Dipyridamole	6	1056	Odds Ratio (M‐H, Fixed, 95% CI)	0.40 [0.24, 0.66]
2.2 Mortality (aspirin vs dipyridamole)	13		Odds Ratio (M‐H, Fixed, 95% CI)	Subtotals only
2.2.1 Aspirin	7	3066	Odds Ratio (M‐H, Fixed, 95% CI)	0.58 [0.38, 0.89]
2.2.2 Dipyridamole	6	1056	Odds Ratio (M‐H, Fixed, 95% CI)	0.56 [0.35, 0.89]
2.3 Mortality (aspirin only trials, by dosage)	7		Odds Ratio (M‐H, Fixed, 95% CI)	Subtotals only
2.3.1 Aspirin ‐ higher dose	3	499	Odds Ratio (M‐H, Fixed, 95% CI)	1.15 [0.53, 2.49]
2.3.2 Aspirin ‐ 100 mg	4	2267	Odds Ratio (M‐H, Fixed, 95% CI)	0.45 [0.27, 0.76]
2.4 Major bleeding (aspirin vs dipyridamole)	11		Odds Ratio (M‐H, Fixed, 95% CI)	Subtotals only
2.4.1 Aspirin	7	3066	Odds Ratio (M‐H, Fixed, 95% CI)	1.44 [1.00, 2.07]
2.4.2 Dipyridamole	4	790	Odds Ratio (M‐H, Fixed, 95% CI)	2.22 [1.08, 4.56]
2.5 Major bleeding (aspirin only trials, by dosage)	7		Odds Ratio (M‐H, Fixed, 95% CI)	Subtotals only
2.5.1 Aspirin ‐ higher dose	3	499	Odds Ratio (M‐H, Fixed, 95% CI)	2.58 [1.43, 4.66]
2.5.2 Aspirin ‐ 100 mg	4	2567	Odds Ratio (M‐H, Fixed, 95% CI)	0.96 [0.60, 1.55]
2.6 Major bleeding (pre‐1990 and after)	11	3856	Odds Ratio (M‐H, Fixed, 95% CI)	1.58 [1.14, 2.18]
2.6.1 Pre‐1990	6	1060	Odds Ratio (M‐H, Fixed, 95% CI)	2.34 [1.34, 4.08]
2.6.2 After 1990	5	2796	Odds Ratio (M‐H, Fixed, 95% CI)	1.26 [0.84, 1.89]

2.6. Analysis.

Comparison 2: Subgroup analyses, Outcome 6: Major bleeding (pre‐1990 and after)

Comparison 3. Selected sensitivity analyses.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Thromboembolism (excluding Meschengieser and LIWACAP))	11	3421	Odds Ratio (M‐H, Fixed, 95% CI)	0.40 [0.29, 0.56]
3.2 Mortality (excluding Meschengieser and LIWACAP)	10	1925	Odds Ratio (M‐H, Fixed, 95% CI)	0.60 [0.42, 0.85]
3.3 Major bleeding (excluding Meschengieser and LIWACAP)	9	3155	Odds Ratio (M‐H, Fixed, 95% CI)	1.80 [1.26, 2.57]
3.4 Major bleeding (low‐intensity OAC versus higher‐intensity OAC)	2	701	Odds Ratio (M‐H, Fixed, 95% CI)	0.76 [0.33, 1.74]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Altman 1976.

Study characteristics
Methods	RCT
Participants	122 participants; 92 M, 30 F with a prosthetic aortic or mitral valve
Interventions	Randomised to OAC (PT ratio 1.8 to 2.3) or OAC and aspirin (500 mg daily) Average duration of follow‐up ~ 2 years
Outcomes	Thromboembolic events; haemorrhagic events; mortality
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Generation of allocation sequence not described.
Allocation concealment (selection bias)	High risk	Outcome events adjudication not described
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not double‐blind
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not described
Blinding of outcome assessment (detection bias) All outcomes	High risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Not described

Bran 1980.

Study characteristics
Methods	RCT
Participants	108 participants with mechanical valve randomised (7 died "immediately after operation" and 1 lost to follow‐up) 101 participants; 53 M, 48 F, followed for average of 22 months
Interventions	Randomised to OAC or OAC and dipyridamole (5 mg/kg/day) Degree of anticoagulation not specified Duration of follow‐up ~ 2 years
Outcomes	Primary outcome: thromboembolic events; haemorrhage; mortality
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	High risk	Not described
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not double‐blind
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not described
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not described
Incomplete outcome data (attrition bias) All outcomes	High risk	7 of 108 not included in the analysis

Dale 1977.

Study characteristics
Methods	RCT
Participants	148 participants; 111 M, 37 F with Starr‐Edwards aortic ball valves
Interventions	Anticoagulation with OAC (thrombotest activity level > 10% above normal, equivalent to INR 2.0 to 2.2) and randomised to either aspirin 500 mg bid or placebo Duration of follow‐up ~ 1 year
Outcomes	Primary outcome: thromboembolic events Secondary outcomes: mortality and complications
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	High risk	Not described
Blinding (performance bias and detection bias) All outcomes	Low risk	Double‐blind (participant and health care provider)
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Double‐blind
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Low risk	Not described

Dong 2011.

Study characteristics
Methods	RCT
Participants	1496 participants with 1) mechanical heart valve replacement, 2) no history of cerebrovascular disease or peripheral vascular disease Mean age 34 years, 92% male
Interventions	Randomised to OAC (INR 1.8 to 2.5) plus aspirin (75 to 100 mg) or OAC and matching placebo
Outcomes	Primary outcome not defined Outcomes: death; thromboembolism (valvular and non‐valvular thrombosis); bleeding
Notes	Definitions for thromboembolism and major bleeding not discussed. Adjudication of outcomes not described.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	High risk	Not described
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Double‐blind
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not described
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Low risk	Not described

Kasahara 1977.

Study characteristics
Methods	RCT
Participants	78 participants. Data on age and sex not available.
Interventions	Randomised to OAC (target PT ratio 1.7) or OAC and dipyridamole 400 mg/day Duration of follow‐up ~ 2.5 years
Outcomes	Primary outcome: thromboembolic events and haemorrhage
Notes	Published in Japanese Mortality data not included in original publication, but extracted from dipyridamole meta‐analysis (Pouleur 1995)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details
Allocation concealment (selection bias)	High risk	No details
Blinding (performance bias and detection bias) All outcomes	Unclear risk	No detail
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not described
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	No details

Laffort 2000.

Study characteristics
Methods	RCT
Participants	229 total; 115 M, 114 F, mean age 63 years 100% MVR, 40% concomitant AVR, and 32% CABG
Interventions	Randomised to OAC (INR 2.5 to 3.5) or OAC and aspirin 200 mg daily Duration of follow‐up ~ 1 year
Outcomes	Primary outcome: composite triple endpoint of death, major thromboembolic events, or major haemorrhage at 1 year Secondary endpoint was transthoracic echocardiographic or TEE evidence of intracardiac or perivalvular strands or thrombi
Notes	Authors concluded that aspirin did not offer any overall benefit. Not described.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	High risk	Not described
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not double blinded
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not described
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Not described

LIWACAP 2007.

Study characteristics
Methods	RCT
Participants	198 participants; 93 M, 105 F with mechanical prosthetic valves Excluded for the need for adjunctive antiplatelet therapy, allergy or intolerance to aspirin, combined bypass surgery, or emergency surgery
Interventions	Randomised to low‐intensity OAC (INR between 2.0 and 3.0, target 2.5) and aspirin 100 mg daily or standard intensity OAC (INR between 3.0 and 4.5, target 3.7) Follow‐up 6 months (the duration of aspirin use)
Outcomes	Vascular death; thromboembolism; major haemorrhage
Notes	A comparison of low‐dose aspirin and less intense oral anticoagulation against standard‐intensity oral anticoagulation
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	High risk	Not described
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not double blinded
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not described
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Adjudication
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Not described

Meschengieser 1997.

Study characteristics
Methods	RCT
Participants	503 participants; 292 M, 211 F with mechanical prosthetic valves Excluded for previous GI bleed or thromboembolism
Interventions	Randomised to OAC (INR 3.5 to 4.5) or OAC (INR 2.5 to 3.5) and aspirin 100 mg/day Median duration of follow‐up of ~ 2 years
Outcomes	Thromboembolism; major haemorrhage
Notes	A comparison of low‐dose aspirin and less intense oral anticoagulation against standard‐intensity oral anticoagulation
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Not described
Allocation concealment (selection bias)	High risk	Not described
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not double‐blind
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not described
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Not described

PACTE 1978.

Study characteristics
Methods	RCT
Participants	290 participants; 166 M, 134 F
Interventions	Randomised to OAC (PT ratio 1.3 to 1.6) or OAC and dipyridamole 375 mg/day Duration of follow‐up ~ 1 year
Outcomes	Thromboembolic events, total mortality, bleeding
Notes	Published in French 389 participants recruited, 345 entered the study, and 290 adhered to the protocol
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Adequate
Allocation concealment (selection bias)	High risk	Not described
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not double blinded, independent adjudication of outcome events
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not described
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Adjudication
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Not described

Rajah 1980.

Study characteristics
Methods	RCT
Participants	165 participants; reported as well matched for age, sex, and type and site of valve replaced
Interventions	Randomised to OAC (INR 1.9 to 3.0) or OAC and dipyridamole 225 to 400 mg/day Average follow‐up 30 and 26 months, respectively
Outcomes	Arterial thromboembolism; mortality
Notes	Published in abstract form only
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details
Allocation concealment (selection bias)	High risk	No details
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not described
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No details
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Not described

Starkman 1982.

Study characteristics
Methods	RCT
Participants	259 participants randomised, all with prosthetic heart valves. Information on age and sex was not provided.
Interventions	Randomised to OAC (Quick time 25% to 35% of control) or OAC and dipyridamole 375 mg/day Duration of follow‐up ~ 1 year
Outcomes	Systemic embolisation; major haemorrhage; mortality
Notes	Published in French
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Adequate
Allocation concealment (selection bias)	High risk	Not described
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not double‐blind
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not described
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Not described

Sullivan 1971.

Study characteristics
Methods	RCT
Participants	163 participants; 82 M, 81F with prosthetic mitral or aortic valves, or both
Interventions	Anticoagulation with OAC (INR 3.0 to 4.5) and randomised to dipyridamole 400 mg/day or placebo Average duration of follow‐up ~ 1 year
Outcomes	Treatment success: patient living, no emboli, and on drug Treatment failure: death, embolus, drug discontinued
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Adequate
Allocation concealment (selection bias)	High risk	Not described
Blinding (performance bias and detection bias) All outcomes	Low risk	Double‐blind (participant and health care provider)
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not described
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not described
Incomplete outcome data (attrition bias) All outcomes	High risk	About 10% excluded due to loss to follow‐up

Turpie 1993.

Study characteristics
Methods	RCT
Participants	370 participants; 187 M, 183 F with mechanical (76%) and tissue valves and felt to be at high risk for thromboembolic events because of atrial fibrillation or history of thromboembolism
Interventions	Anticoagulation with OAC (INR 3.0 to 4.5) and randomised to aspirin 100 mg/day or placebo Average duration of follow‐up ~ 2.5 years
Outcomes	Thromboembolism; major haemorrhage; mortality
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Described
Allocation concealment (selection bias)	High risk	Not described
Blinding (performance bias and detection bias) All outcomes	Low risk	Double‐blind (participant and health care provider)
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Independent adjudication of outcome events
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Adjudication
Incomplete outcome data (attrition bias) All outcomes	Low risk	Described

AVR = aortic valve replacement; bid = twice a day; CABG = coronary bypass graft surgery; D = dipyridamole; F = female; GI = gastrointestinal; INR = international normalised ratio; M = male; MVR = mitral valve replacement; OAC = oral anticoagulant; PT = prothrombin time; Quick time = antiquated way of a measuring degree of anticoagulation; RCT = randomised controlled trial; TEE = transesophageal echocardiogram

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Chesebro 1983	Excluded because it compares 2 treatment groups both getting OAC and either ASA or dipyridamole; there is no OAC alone treatment arm
Hassouna 2000	Comparison of standard‐dose coumadin against low‐dose coumadin and dipyridamole; there is no standard level of OAC between treatment arms
Hayashi 1994	Non‐randomised design
Moriyama 1998	Non‐randomised design; no comparison to OAC alone
Saitoh 1988	Non‐randomised study of biochemical endpoints; no comparison to OAC alone
Toyohira 1995	Non‐randomised; no clinical endpoints
Voith 1997	The basis of 'randomisation' between coumarin and coumarin plus aspirin was the agreement, or lack thereof, between patients and their treating physicians; this study is not truly randomised

ASA = acetylsalicylic acid; OAC = oral anticoagulant

Contributions of authors

Dr D Massel:

Conceived, designed, and co‐ordinated the review
Assisted with preparation of the protocol
Screened retrieved papers against inclusion criteria
Appraised quality of papers
Abstracted data from papers
Analysed and interpreted the data
Participated in writing the review
Provided a methodological and clinical perspective

Dr S Little:

Prepared and designed the protocol
Developed the search strategy
Undertook searches
Appraised quality of papers
Abstracted data from papers
Data collection for the review
Screened search results
Organised retrieval of papers
Data management for the review
Entered data into RevMan (RevMan 2012)
Participated in writing the review

Sources of support

Internal sources

• Department of Medicine, University of Western Ontario, Canada

• The Methodist DeBakey Heart & Vascular Center, Houston, TX., USA

External sources

• No sources of support supplied

Declarations of interest

None known.

Stable (no update expected for reasons given in 'What's new')