Combined anticoagulation and antiplatelet therapy for high-risk patients with atrial fibrillation: a systematic review

Antithrombotic management of atrial fibrillation

The management of AF consists of a rate and/or rhythm control strategy in combination with antithrombotic therapy (ATT). The aim of the former is to control the heart rate without attempting to restore the heart's normal rhythm (sinus rhythm), whereas the latter attempts to re-establish and maintain sinus rhythm. Regardless of which strategy is implemented, all patients should be assessed for individual stroke risk and receive appropriate ATT. Clinical guidelines2,24 recommend oral anticoagulant for patients who are at high risk of stroke, and either oral anticoagulation or antiplatelet(s) for those deemed to be at intermediate risk, although the European Society of Cardiology (ESC) guidelines2 prefer oral anticoagulation over antiplatelet(s) therapy in this group. Among those patients who are at low risk of stroke (those < 65 years of age with no stroke risk factors), the National Institute for Health and Care Excellence (NICE)24 recommends antiplatelet therapy (APT), whereas the ESC guidelines2 recommend APT or no treatment, with a preference for no therapy. 2,24

In order to determine the most appropriate ATT for each patient, his or her individual risk of stroke should be assessed. The main risk factors for stroke among patients with AF are described above (see Risk of stroke), but include previous stroke or TIA, age ≥ 65 years, HF, hypertension and diabetes mellitus, which together constitute the widely used stroke risk assessment tool, the CHADS₂ (Congestive heart failure, Hypertension, Age ≥ 75 years, Diabetes mellitus, and prior Stroke or TIA or thromboembolism) score,25 although there are numerous other stroke risk stratification schemas available19 (Table 1).

In the UK, the NICE guidelines24 currently recommend aspirin 75–300 mg daily (unless contraindicated) for patients aged < 65 years with no moderate- or high-risk factors and who, thus, are deemed to be at low risk (≤ 1% annual risk) of stroke. For patients at moderate risk (4% annual risk), namely those aged < 75 years with hypertension, diabetes mellitus or vascular disease (CAD or peripheral artery disease) and those ≥ 65 years without any high-risk factors, NICE24 suggests anticoagulation or aspirin. Among patients at high risk (12% annual risk) of stroke, i.e. those with a previous stroke/TIA or thromboembolism (TE), clinical evidence of valve disease, HF, or impaired left ventricular (LV) function on echocardiography, or aged ≥ 75 years with hypertension, diabetes mellitus or vascular disease, NICE24 recommends anticoagulation with warfarin. The ESC guidelines2 have adopted a risk factor-based approach to determine appropriate thromboprophylaxis (Figure 1 and Table 1) and these guidelines have superseded the NICE recommendations in clinical practice in the UK. 2

FIGURE 1.

Clinical flow chart for the use of ATT in patients with AF. Redrawn from the ESC guidelines. 2 a, Congestive HF, hypertension, age ≥ 75 years, diabetes mellitus (1 point for each), stroke/TIA/TE (2 points); b, other clinically relevant non-major risk factors: age 65–74 years, female sex, vascular disease.

In patients with AF who have no risk factors for stroke, the ESC guidelines2 recommend either aspirin 75–325 mg daily or no ATT, with a preference for no treatment over aspirin. 2 For those with one ‘clinically relevant non-major’ risk factor [HF or left ventricular ejection fraction (LVEF) ≤ 40%, hypertension, diabetes mellitus, vascular disease, age 65–74 years, female sex], the ESC advises that oral anticoagulation or aspirin (75–325 mg) should be administered, with an oral anticoagulant (OAC) preferred over aspirin. Among those patients with one ‘major’ (previous stroke/TIA/TE or aged ≥ 75 years) or two or more ‘clinically relevant non-major’ risk factors, a OAC is recommended. Where a OAC is recommended, this includes adjusted-dose warfarin (INR 2.0–3.0) or one of the new anticoagulant drugs (see Description of technology under assessment).

In addition, CAD is also increasing in prevalence as a consequence of the improvements in survival due to advances in medical therapy and the ageing population. 30 Between 30% and 40% of patients with AF have concomitant CAD,11 and some of these patients may also require percutaneous coronary intervention (PCI) with stent implantation. Patients with AF and CAD are at increased risk of both stroke and further coronary events. An increasingly common management problem arises when faced with an anticoagulated patient with AF who presents with acute coronary syndrome (ACS) or those who require PCI with stent implantation. 32

Current guidelines for antithrombotic therapy in atrial fibrillation patients with acute coronary syndrome or undergoing percutaneous coronary intervention or stenting

The joint American College of Cardiology (ACC)/American Heart Association (AHA)/ESC 2006 guidelines on the management of AF recommend that following PCI or revascularisation surgery in patients with AF, low-dose aspirin (< 100 mg/day) and/or clopidogrel (75 mg/day) may be given concurrently with anticoagulation to prevent myocardial ischaemic events,1 although it is acknowledged that these strategies have not been thoroughly evaluated and are associated with an increased risk of bleeding. The 2006 ACC/AHA/ESC guidelines also suggest that clopidogrel should be given for a minimum of 1 month after implantation of a bare-metal stent, ≥ 3 months for a sirolimus (CYPHER™, Cordis)-eluting coronary stent-P020026, ≥ 6 months for a paclitaxel (ION™, Boston Scientific)-eluting coronary stent system-P100023, and ≥ 12 months in selected patients, following which warfarin may be continued as monotherapy in the absence of a subsequent coronary event. 1 Broadly similar recommendations are made in the eighth ACCP guidelines,33 which suggest that a low dose of aspirin (< 100 mg per day) or clopidogrel (75 mg per day) may be given with anticoagulation, although the risk of bleeding may be increased, particularly in elderly patients. The UK NICE guidelines24 do not address this topic, although acknowledging that adding aspirin to warfarin increases bleeding, and that it is a matter for individual assessment of the risk–benefit ratio in prescribing aspirin plus warfarin in patients with associated CAD.

Furthermore, all of the published guidelines do not address the issue of a presentation with ACS (where PCI is often performed) and bleeding risk. Given the need to balance stroke prevention, recurrent cardiac ischaemia and/or stent thrombosis, two more recent consensus documents,34,35 based on systematic reviews of patients on OAC undergoing PCI and stenting, advocate initial triple therapy (with OAC, aspirin and clopidogrel) in such patients, and the use of bare-metal stents (owing to the need for prolonged multiple-drug ATT with drug-eluting stents). However, triple ATT is associated with a higher risk of major bleeding and this risk must be considered before treatment initiation. 34,36 Therefore, the ESC Working Group on Thrombosis consensus guidelines35 recommend limiting triple ATT to 2–4 weeks in patients who are at high risk of haemorrhage (Table 2).

Search strategies

The following resources were searched for relevant studies:

• Bibliographic databases: The Cochrane Library [Cochrane Central Register of Controlled Trials (CENTRAL)] 2010 Issue 3; MEDLINE (Ovid) 1950 to September week 1 2010; MEDLINE In-Process and Other Non-Indexed Citations from inception to 27 September 2010; and EMBASE (Ovid) 1980 to September 2010. Searches were based on index and text words that encompassed the population: atrial fibrillation and the interventions; combined anticoagulation and antiplatelet therapy.

• Ongoing trials were sought in ClinicalTrials.gov, National Institute for Health Research (NIHR) Clinical Research Network Portfolio, Current Controlled Trials (CCT) and the WHO International Clinical Trials Registry Platform (ICTRP).

• Reference lists from identified systematic reviews were checked.

• Citations of relevant studies were examined.

• Further information was sought from clinical experts.

All study types were sought. Searches were not limited by language or date and were carried out during September 2010 by an information specialist.

Search strategies used in the bibliographic databases can be found in Appendix 2.

Scoping searches were undertaken to identify completed and ongoing systematic reviews from the following resources: The Cochrane Library [Cochrane Database of Systematic Reviews (CDSR), Database of Abstracts of Reviews of Effects (DARE), Health Technology Assessment (HTA) database, CENTRAL and NHS Economic Evaluation Database (NHS EED)], Aggressive Research Intelligence Facility (ARIF) database of reviews, HTAi portal, MEDLINE (Ovid) 1950 onwards and EMBASE (Ovid) 1980 onwards. The systematic reviews were used to check if there were additional relevant studies.

Study selection

All records identified in the searches were imported into a Reference Manager database (Reference Manager v.11, Thomson ResearchSoft, San Francisco, CA, USA). Duplicate entries were allowed to be removed by the inbuilt feature in Reference Manager and also removed when encountered by reviewers.

Owing to the number of retrieved records and the complexity of the publications, a three-stage process was used to select the studies for review.

Data extraction

Data were extracted into a standard form in Microsoft Excel 2007 v.12 (Microsoft Corporation, Redmond, WA, USA) from the main and supporting publications (where relevant) of all included primary studies by one reviewer. A second reviewer checked the accuracy of extracted information. Disagreements were resolved by consensus or by referral to a third reviewer if necessary.

Information regarding study design (including intervention/comparators) and characteristics of study participants was extracted. This included antithrombotic regimens used [anticoagulant ± antiplatelet(s) or placebo], type of ATT used and dose, target INR values used, indication for ATT (e.g. AF ± ACS or stent implantation), study setting (country), study design, sample size, patient inclusion and exclusion criteria, patient characteristics (e.g. age, sex, type and duration of AF, anticoagulant naive or experienced), comparability of patients between different arms (for RCTs and non-randomised trials), primary outcome measures, secondary outcome measures, length of follow-up, statistical methods used, effect sizes and uncertainty.

Data on the following outcomes were sought from included studies.

Quality assessment

The quality of included studies was assessed by one reviewer. A second reviewer checked the accuracy of extracted information. Disagreements were resolved by consensus or by referral to a third reviewer if necessary.

The methodological quality of RCTs was assessed in terms of the randomisation process, allocation concealment (adequate, unclear, inadequate or not used), degree of blinding, particularly of the outcome assessors, and patient attrition rate, using the Cochrane Collaboration risk of bias assessment tool. 37

The quality assessment of studies undertaking non-randomised comparisons was undertaken using the Centre for Reviews and Dissemination (CRD)'s checklist for cohort studies. 38 Information on the following was captured: method of outcome measurement, blinding of assessors, whether or not outcome definitions were clearly explained, and which parts of the study were prospective. In addition, the following topic-specific data that were considered relevant to the quality of the studies were assessed: ‘Were the indications for use of APT given?’ and ‘Was it clear whether patients were on APT at the start or commenced such therapy during the observation period?’

Data from randomised studies that were obtained from non-randomised comparisons were classed and treated as non-randomised data. For example, when data from a subset of patients in two or more arms of a RCT were combined to compare with data from another subset of patients obtained from these or other arms of the same study.

From non-randomised comparisons the potential for confounding by indication was ever present; whereby APT was added to ACT, based on clinical judgement of a potential risk of adverse outcomes in some patients if such therapy was not given. Conversely, in those without such perceived risk APT may not have been given. Thus, the patients receiving anticoagulation alone would differ from those receiving the combined therapy, and thus any comparison between the two would be confounded.

Data analysis/synthesis

Assessment of publication bias

The number of relevant studies for a given comparison was too small to allow formal assessment of publication bias.

Between-study differences

The subsequent sections will report the event rates for each outcome. Methodological heterogeneity exists between the included studies that may explain any differences in the event rates reported. Rather than repeat these methodological differences for each and every outcome of interest, the reader will be referred to the following discussion of these differences. Where specific differences in the methodology between the included studies are apparent, which are important to highlight and/or only pertinent to that particular outcome, these differences will be specified under that outcome.

The differences in the event rates reported by the included studies may reflect differences in the population risk profile, with some studies including high-risk AF populations (three RCTS39,41,43 and seven non-randomised comparisons50,55,58,64,65,72,73) and/or intermediate-risk patients with AF (one RCT39), whereas other studies did not report the risk profile of included patients (two RCTs40,42) and 11 other non-randomised comparisons. 51–54,56,57,59–63

The sample size also varied considerably between included studies, from 43 participants in one RCT40 to 118,606 in a large non-randomised comparison. 63 As a result of the overall sample size, the number of patients receiving combined ACT and APT and the comparator also varied considerably, with only 34 patients receiving the combination therapy in Bover et al. ,54 between 21 and 36 patients receiving the various permutations of ACT plus APT in the PETRO study,73 and 76 patients receiving combination therapy and 81 receiving ACT alone in the FFAACS (Fluindione, Fibrillation Auriculaire, Aspirin et Contraste Spontané) trial,41 which will have influenced the reported event rates for each outcome.

Further, the included studies comprise both randomised and non-randomised data. Among non-randomised comparisons there is the potential for confounding by indication with the use of APT, as this was often given at the discretion of the treating physician, with patients at high risk of a vascular event and/or those less likely to bleed receiving combination therapy. Indeed, Bover et al. 54 reported that the patients receiving combined therapy were at a higher risk of stroke than those who were administered adjusted-dose acenocoumarol (INR 2.0–3.0) alone. Moreover, the number of patients with previous experience of an anticoagulant agent or APT in each study may also affect the event rate, for example patients who can tolerate either ACT or APT will continue on such therapy and therefore may be less likely to bleed on treatment than those who experience a bleed and therefore discontinue such therapy – ‘ATT survivor’.

The included studies also compared different types of anticoagulant and APT in various permutations, which makes comparison of event rates across studies using different interventions and comparators difficult. Studies compared a VKA, either warfarin,40,42,43,50–53,55–57,63–65,72 acenocoumarol,39,54 or fluindione41 in combination with either aspirin41–43,50,51,53–58,60–65,72,73 or other antiplatelet agents, such as triflusal,39,54 clopidogrel,40,63 or dual APT of aspirin plus clopidogrel. 63 Furthermore, two other studies compared an ODTI (anticoagulant) – either ximelagatran64,65 or dabigatran73 – in combination with aspirin (in different doses) or alone.

Among those studies comparing VKAs plus aspirin to a VKA alone,39–43,50–54,54–57,59–65,67,69,72 different VKA regimes were used in the combination therapy arm, either fixed dose (1.25 mg42) or adjusted dose to maintain a target INR range [e.g. INR 1.2–1.5,43 INR 2.0–3.0,40,64,65 INR 1.9–2.5,54 INR 2.0–2.6,41 INR 1.4–2.4 (high risk) and INR 1.25–2.0 (intermediate risk)39]. Of the included RCTs, therapies were administered in either an open-label42,43 or in a double-blind fashion. 41 In addition, the APT also varied (aspirin, triflusal, clopidogrel, and aspirin plus clopidogrel). In the studies reporting randomised comparisons, aspirin was utilised in different doses (300 mg,42 325 mg43 and 100 mg41), and also in non-randomised comparisons (≤ 100 mg,64,65,72 100 mg,61 81 or 325 mg73 and dose not specified in others51,53,56–58,62). Similarly, other antiplatelets were used in different doses such as triflusal (600 mg,39 600 mg and 300 mg54), clopidogrel (75 mg:40 dose not specified63) and dual APT of aspirin plus clopidogrel (dose not specified50,63), which makes direct comparison between studies difficult.

In addition, some randomised studies used the same target INR range in both the intervention and comparator arm (RCTs40,41 besides non-randomised comparisons54,51,53, 56,57,59,61,62,64,65,72), whereas others did not (RCTs39,42,43 and non-randomised comparisons54,55), again making difficult the direct comparison between the intervention and comparator arms within the studies. However, the majority of studies did use the standard therapeutic INR target of 2.0–3.0 in the comparator arm40,41,51,53–55,57,59,61,62,64,65,72 whereas others did not, although only four studies39,40,43,54 reported time in therapeutic range (TTR).

There were also differences across studies in the definitions of the outcomes of interest used and these differences are discussed, where relevant, under each outcome.

Furthermore, the considerable variation in the length of follow-up (e.g. 22 days40 to 4.92 years54) in each of the included studies may have influenced event rates. The combination of a short duration of follow-up for outcomes that are not particularly common together with a small sample size may have resulted in studies being underpowered. Of note the AFASAK II study (Second Copenhagen Atrial Fibrillation, Aspirin and Anticoagulation Study)42 was prematurely terminated when results of the SAAF III (Stroke Prevention in Atrial Fibrillation) trial,43 demonstrating the superiority of adjusted-dose warfarin (INR 2.0–3.0) alone, over combination of adjusted-dose warfarin (INR 1.2–2.5) and aspirin 325 mg in preventing stroke or SE, were published. Further, the FFAACS study41 was also terminated early due to poor recruitment. It should also be noted, that Bover et al. 54 was a non-randomised comparison that followed up of a proportion of the patients enrolled in the NASPEAF (National Study for Prevention of Embolism in Atrial Fibrillation) study39 (although it is not clear how many patients from NASPEAF were included in Bover et al. , within each arm of the latter study), with addition of newly recruited participants, over a longer period of time.

Moreover, the temporal changes in the management of AF over the last 20 years may have influenced the event rate reported in studies enrolling patients in the early 1990s (AFASAK II42 and SPAF III43) compared with those from 2000 onwards. 39,40,41,54,63,64,65,73

Outcome 1: stroke

Thirteen articles yielded outcome data for stroke. 42–45,47–50,54,55,63,66,69 Of these, three studies in seven articles42–45,47–49 reported randomised comparisons. The findings of these are reported in Table 9. The remaining five articles50,54,55,63,66,69 reported non-randomised comparisons, of which four were primary studies,50,54,55,63 and two were secondary analyses66,69 of the SPORTIF III and SPORTIF V studies. Table 10 presents the findings of these studies.

The study by Hansen et al. 63 reported stroke outcomes for a large number of patients with AF (118,606) over a long follow-up (3.3 years); however, neither the number of stroke events nor the details of the antiplatelet and ACT were reported. Therefore, this study is not reported in this section. Of the studies that reported non-randomised comparisons, Lopes et al. ,50 Teitelbaum et al. 66 and Hart et al. 55 are not mentioned further in this section. The reasons for these have been reported in Appendix 7. The characteristics of these studies have been reported previously (see Table 6).

Stroke events were reported either on their own (stroke alone) or in conjunction with other events such as embolism or bleeding in the included studies. In those studies that reported stroke alone, strokes were frequently classified as non-fatal, fatal, haemorrhagic, ischaemic or disabling. A precise definition of these groupings or subclassifications of stroke was not always supplied in the study reports and/or the definitions may have varied between studies for the same subclassification.

The findings of the included studies for each of these composite and/or subclassifications of stroke are detailed below.

Stroke: all

One randomised comparison42 and two non-randomised comparisons54,69 compared a VKA plus aspirin, to a VKA alone. The pooled analysis of the SPORTIF III and V trials69 and the longitudinal follow-up study by Bover et al. ,54 add data to the randomised comparisons on the risk of stroke for patients receiving VKA plus aspirin compared with VKA alone.

The AFASAK II study42 and the pooled analysis of SPORTIF trials by Flaker et al. 69 defined stroke as an acute onset of focal neurological deficit lasting ≥ 24 hours. Bover et al. 54 did not report a precise a definition of stroke in their study.

The AFASAK II42 study compared combined fixed-dose warfarin (1.25 mg) plus aspirin (300 mg daily), with either adjusted-dose warfarin (target INR 2.0–3.0) alone or with fixed-dose warfarin (1.25 mg daily) alone. The findings of this study42 have been reported in Table 9. The risk profile of the patients enrolled in this study was not specified. There were no significant differences in the rate of stroke between patients receiving the combination of fixed-dose warfarin and aspirin, and either those receiving fixed-dose warfarin alone [11/171 (6.4%) vs 13/167 (7.8%), respectively] with a RR of 0.83 (95% CI 0.38 to 1.79), or those receiving adjusted-dose warfarin alone [11/171 (6.4%) vs 10/170 (5.9%), respectively], RR 1.09 (95% CI 0.48 to 2.51), over a mean follow-up period of 3.5 years. 42

The pooled analysis of the SPORTIF studies by Flaker et al. 69 compared adjusted-dose warfarin (INR 2.0–3.0) plus aspirin (100 mg) with adjusted-dose warfarin (INR 2.0–3.0) alone, over a mean follow-up period of 16.5 months. The rate of stroke was similar in patients in the combined therapy group compared with those on adjusted-dose warfarin (INR 2.0–3.0 alone) [11/481 (2.3%) vs 67/3172 (2.1%)], respectively. The rate of stroke was much higher in the AFASAK II study42 for patients receiving combination fixed-dose warfarin plus aspirin than for those with adjusted-dose warfarin plus aspirin in the SPORTIF studies;69 11 out of 171 (6.4%) compared with 11 out of 481 (2.3%), respectively. The stroke rate was also much higher in patients receiving either adjusted-dose, 10 out of 170 (5.9%) or fixed-dose, 13 out of 167 (7.8%) warfarin alone in the AFASAK II42 study than for those receiving adjusted-dose warfarin alone in the SPORTIF studies,69 67 out of 3172 (2.1%).

Bover et al. 54 compared adjusted-dose acenocoumarol (INR target 1.9–2.5) plus aspirin (100 mg) with adjusted-dose acenocoumarol (INR 2.0–3.0) alone, over a mean follow-up period of 4.92 years. The combination of acenocoumarol with aspirin demonstrated fewer stroke events [1/34 (2.9%)] than with acenocoumarol alone [15/265 (5.7%)].

Bover et al. 54 also compared combination adjusted-dose acenocoumarol (INR 1.9–2.5) plus two different regimes of triflusal (600 and 300 mg) with adjusted-dose acenocoumarol (INR 2.0–3.0) alone. Fewer strokes were observed with the combination of acenocoumarol plus triflusal 600 mg than for acenocoumarol alone, 5 out of 155 (3.2%) compared with 15 out of 265 (5.7%), respectively. However, stroke rates were higher in those receiving acenocoumarol plus triflusal 300 mg than in those receiving with acenocoumarol alone, 8 out of 120 (6.7%) and with 15 out of 265 (5.7%), respectively. However, there were population complexities in this non-randomised study (see Between-study differences, above).

The pooled analysis of the SPORTIF trials by Flaker et al. 69 also compared ximelagatran (36 mg twice daily) plus additional aspirin (100 mg) with ximelagatran (36 mg) alone, over a mean follow-up period of 16.5 months. A higher rate of stroke was observed in patients on combined therapy group than in those on ximelagatran alone [11/531 (2.1%) vs 50/3120 (1.6%), respectively]. However, it is to be noted that aspirin use was based on clinical need and, thus, the comparison may be confounded by indication. 65,68

Fatal stroke

The AFASAK II42 and SPAF III43 studies reported randomised comparisons for the outcome of fatal stroke comparing different regimes of combined warfarin plus aspirin, with warfarin alone, whereas Bover et al. ,54 reported non-randomised data comparing acenocoumarol plus aspirin with acenocoumarol alone. The findings of these studies are reported in Tables 9 and 10, respectively.

In both studies reporting randomised comparisons (AFASAK II42 and SPAF III43), stroke was defined as a focal neurological deficit of presumed vascular genesis lasting more than 24 hours, where stroke assessment was undertaken using neuroimaging. However, Bover et al. 54 did not report a precise definition of stroke in their study.

The AFASAK II study42 compared fixed-dose warfarin (1.25 mg) plus aspirin (300 mg daily), with either adjusted-dose warfarin (target INR 2.0–3.0) alone or with fixed-dose warfarin (1.25 mg daily) alone (see Table 9). The risk profile of the patients enrolled in this study was not specified. No fatal strokes were reported among patients receiving either combined warfarin and aspirin or those receiving warfarin alone. However, two fatal strokes were reported in patients receiving fixed-dose warfarin alone [2/167 (1.2%) vs 0/171 (0%), respectively, RR 0.20 (95% CI 0.01 to 4.04)] over a mean follow-up period of 3.5 years. 42

The SPAF III study43 compared adjusted-dose warfarin (INR 1.2–1.5) plus aspirin (325 mg) with adjusted-dose warfarin (target INR 2.0–3.0) alone in high-risk patients with AF. A non-significant but higher incidence of fatal stroke was observed in the combined therapy arm than in those treated with warfarin alone [5/521 (0.9%) vs 1/523 (0.2%), respectively, RR 5.02 (95% CI 0.59 to 42.81)] over a mean follow-up period of 1.1 years. 43

Only eight fatal strokes occurred in these two RCTs. Among those receiving combined therapy, the rate of fatal stroke was 0.9% (5/521) in the SPAF III43 study compared with 0% (0/171) in the AFASAK II study. 42 The rate of fatal stroke was similar but numerically higher among those receiving adjusted-dose warfarin alone in the SPAF III study,43 1/523 (0.2%), than 0/170 (0%) in the AFASAK II study,42 and higher among those receiving fixed-dose warfarin alone in the AFASAK II42 study [2/167 (1.2%) vs 1/523 (0.2%), respectively].

Bover et al. 54 reported a non-randomised comparison for the incidence of fatal stroke comparing adjusted-dose acenocoumarol (INR 1.9–2.5) plus an antiplatelet in three different regimes (triflusal 600 mg, triflusal 300 mg and aspirin 100 mg) with adjusted-dose acenocoumarol (INR 2.0–3.0) alone. The combination of acenocoumarol plus aspirin and acenocoumarol plus triflusal 300 mg demonstrated a higher proportion of fatal strokes than acenocoumarol alone [1/34 (2.9%), 3/120 (2.5%) vs 4/265 (1.5%), respectively] during a mean follow-up of 4.92 years. Rates of fatal stroke were similar among those receiving combination acenocoumarol plus triflusal 600 mg and acenocoumarol alone [2/155 (1.3%) vs 4/265 (1.5%), respectively].

Non-fatal stroke

One study (NASPEAF39) reported a randomised comparison for non-fatal stroke comparing adjusted-dose acenocoumarol (INR 1.4–2.4) plus triflusal (600 mg) with acenocoumarol alone (INR 2.0–3.0) in high-risk patients, and a combination of adjusted-dose acenocoumarol (INR 1.2–2.0) plus additional triflusal (600 mg) with acenocoumarol alone (INR 2.0–3.0) in intermediate-risk patients. Table 9 presents the findings of this study. Stroke was defined as a focal neurological deficit lasting more than 24 hours, where neuroimaging was used to define the ischaemic or intracranial aetiology. 39

Similar rates of non-fatal stroke occurred with combination therapy and anticoagulation alone in the high-risk patients [6/223 (2.7%) vs 6/247 (2.4%), respectively], RR 1.11 (0.36–3.38), during a median follow-up of 2.95 years. Analogous rates were observed in the intermediate-risk group in both the combination therapy and anticoagulation alone group [3/222 (1.4%) vs 3/232 (1.3%), respectively], RR 1.05 (95% CI 0.21 to 5.12), after a median follow-up of 2.6 years.

There was no non-randomised evidence identified for non-fatal stroke.

Haemorrhagic stroke

The AFASAK II study42 reported randomised data, and Bover et al. 54 reported a non-randomised comparison for the outcome of haemorrhagic stroke.

The AFASAK II study42 compared fixed-dose warfarin (1.25 mg) plus aspirin (300 mg) with fixed-dose warfarin (1.25 mg) or adjusted-dose warfarin (INR 2.0–3.0) alone. The risk profile of the patients enrolled in this study was not specified. No haemorrhagic strokes were reported in either those patients on combination therapy or in those receiving fixed-dose warfarin alone, over a mean follow-up period of 3.5 years. One haemorrhagic stroke occurred in a patient receiving adjusted-dose warfarin [1/170 (0.6%); RR 0.33 (95% CI 0.01 to 8.08)42 compared with combination therapy] (see Table 9).

Bover et al. 54 compared adjusted-dose acenocoumarol (INR 1.9–2.5) plus an antiplatelet in three different regimes (triflusal 600 mg, triflusal 300 mg, aspirin 100 mg) with adjusted-dose acenocoumarol (INR 2.0–3.0) alone, over a mean follow-up period of 4.92 years.

Fewer haemorrhagic strokes were observed in patients in all three combination therapy arms (triflusal 600 mg, triflusal 300 mg, aspirin 100 mg) than in those patients receiving acenocoumarol alone, 1/155 (0.6%), 0/120 (0%), 1/34 (2.9%) versus 5/265 (1.9%), respectively. 54

Ischaemic stroke

The AFASAK II42 and SPAF III43 studies reported randomised comparisons for the outcome of ischaemic stroke. The findings of these studies have been reported in Table 9. There was no non-randomised evidence available for the outcome of ischaemic stroke.

The AFASAK II study,42 comparing fixed-dose warfarin (1.25 mg) plus aspirin (300 mg) with adjusted-dose (INR 2.0–3.0) or fixed-dose (1.25 mg) warfarin alone, reported a non-significant but higher incidence of ischaemic stroke in the combined therapy arm [8/171 (4.7%) compared with either adjusted dose 3/170 (1.8%), RR 2.65 (95% CI 0.72 to 9.82)] or fixed-dose [5/167 (2.9%); RR 1.56 (95% CI 0.52 to 4.68)] warfarin alone. The SPAF III study43 reported significantly higher rates of ischaemic stroke in the combined therapy arm [adjusted-dose warfarin (INR 1.2–1.5) plus aspirin 325 mg) than in those with adjusted-dose warfarin (INR 2.0–3.0) alone [43/521 (8.3%) vs 11/523 (2.1%), respectively, RR 3.92 (95% CI 2.05 to 7.52)] in high-risk patients with AF over a mean follow-up period of 1.1 years.

The rate of ischaemic stroke varied between these two RCTs. 42,43 In patients receiving combination therapy, the risk of ischaemic stroke was much higher in the SPAF III43 study than in the AFASAK II42 study [43/521 (8.3%) vs 8/171 (4.7%), respectively]. Among those receiving dose-adjusted warfarin (INR 2.0–3.0), the rate of ischaemic stroke was similar in both SPAF III43 and AFASAK II studies42 [11/523 (2.1%) vs 3/170 (1.8%), respectively]. The rate of ischaemic stroke was higher in those receiving fixed-dose warfarin in the AFASAK II42 study than in those receiving dose-adjusted warfarin in either AFASAK II42 or SPAF III43 study [5/167 (2.9%) vs 3/170 (1.8%), 11/523 (2.1%), respectively]. The differences in the rates may reflect the heterogeneity between the included studies (see Between-study differences).

Disabling stroke

The AFASAK II42 and SPAF III43 studies reported randomised comparisons for the outcome of disabling stroke. The findings of these studies are reported in Table 9. There was no non-randomised evidence available for this outcome.

The SPAF III study43 defined disabling stroke as stroke that was graded ≥ 2 on the modified Rankin scoring system, whereas the AFASAK II study42 did not specify a definition for disabling stroke.

The AFASAK II study42 reported a non-significant but higher incidence of disabling stroke in the combined therapy arm [fixed-dose (1.25 mg) warfarin plus aspirin 300 mg] than in either adjusted-dose warfarin (target INR 2.0–3.0) alone [4/171 (2.3%) vs 3/170 (1.8%), respectively, RR 1.33 (95% CI 0.30 to 5.83)] or fixed-dose warfarin (1.25 mg) alone [4/171 (2.3%) vs 2/167 (1.2%), respectively] (see Table 9) over a mean follow-up period of 3.5 years. The risk profile of the patients enrolled in this study42 was not specified.

The SPAF III study43 reported significantly higher rates of disabling stroke in the combined therapy arm [adjusted-dose warfarin (INR 1.2–1.5) plus aspirin 325 mg] than in the adjusted-dose warfarin (INR 2.0–3.0) alone group [31/521 (5.9%) vs 10/523 (1.9%) respectively, RR 2.83 (95% CI 1.44 to 5.57)] in high-risk patients with AF over a mean follow-up period of 1.1 years. 43

The rate of disabling strokes was much higher in patients receiving combination therapy in the SPAF III43 study than in the AFASAK II42 study [31/521 (5.9%) vs 4/171 (2.3%), respectively]. Similar rates of disabling stroke were evident in patients receiving adjusted-dose warfarin alone in both SPAF III43 and AFASAK II42 studies [10/523 (1.9%) vs 3/170 (1.8%), respectively], and those receiving fixed-dose warfarin alone in the AFASAK II42 study [2/167 (1.2%)]. Such differences reflect significant heterogeneity between the included studies (see Between-study differences, above).

Other stroke definitions

The AFASAK II study42 also reported the incidence of minor, non-disabling and non-infarct strokes. The findings of this study are reported in Table 9. The definitions of these subclassifications have not been reported in the study. 42 Fixed-dose warfarin (1.25 mg) plus aspirin (300 mg) demonstrated a non-significant but higher risk of minor stroke than with either adjusted-dose warfarin (INR 2.0–3.0) alone [4/171 (2.3%) vs 0/170 (0%), respectively, RR 8.95 (95% CI 0.49 to 164.92)] or fixed-dose warfarin alone [4/171 (2.3%) vs 3/167 (1.8%), respectively, RR 1.30 (95% CI 0.30 to 5.73)]. 42

This study also demonstrated similar rates of non-disabling stroke among those receiving combination therapy [3/171 (1.8%)], adjusted-dose warfarin alone [4/170 (2.4%)] and fixed-dose warfarin alone [4/167 (2.4%)]. 42 The rate of non-infarct stroke was the same among those receiving combination therapy and adjusted-dose warfarin alone [3/171 (1.8%) vs 3/170 (1.8%), respectively] but was twice as high in those receiving fixed-dose warfarin alone [6/167 (3.6%)]42 (see Table 6).

There was no non-randomised evidence available for these three subclassifications of stroke.

The differences in stroke outcomes reported in the included studies may reflect the methodological differences between these studies discussed above (see Between-study differences). In addition, although four studies39,42,43,69 used the same definition of stroke, one non-randomised study54 did not provide a specific definition of stroke, and the stroke subtypes reported varied and were not always clearly defined by each study, which may account for variation in the reported event rates. The likelihood of stroke is increased when INR is < 2.0 and, therefore, it is possible that studies using INR targets of < 2.0 in the combination therapy arm may have experienced higher rates of stroke than those using standard INR targets (2.0–3.0), particularly in high-risk populations. Furthermore, only three studies (two randomised39,43 and one non-randomised54) reported TTR for ACT plus APT and ACT alone. TTR is associated with the incidence of stroke events; when TTR is good (≥ 58%), the likelihood of adverse events (ischaemic and haemorrhagic strokes) is reduced. 92 Therefore, differences in the TTR may help to explain differences in the event rates reported.

Outcome 2: transient ischaemic attack

Three studies, reported in five articles39,42,43,45,47 yielded outcome data for TIA (Table 11). Of these, all three reported randomised comparisons,39,42,43 supported by two subgroup analyses. 45,47 No non-randomised comparisons reported TIA separately as an outcome.

Transient ischaemic attack was similarly defined in the NASPEAF39 and AFASAK II42 studies as an acute onset of focal neurological deficit of presumed vascular genesis lasting < 24 hours, regardless of computerised tomography (CT)/magnetic resonance imaging (MRI) findings (AFASAK II) or confirmed by neurological imaging (NASPEAF). The SPAF III43 study did not define TIA.

Both the AFASAK II42 and SPAF III43 studies compared warfarin plus aspirin with warfarin alone, but the warfarin and aspirin regimes differed between the studies.

The AFASAK II42 study compared fixed-dose warfarin (1.25 mg daily) plus aspirin (300 mg daily) with adjusted-dose warfarin (INR 2.0–3.0) or fixed-dose warfarin (1.25 mg daily). The risk profile of the patients enrolled in this study was not specified. The rate of TIA among patients receiving the combination of warfarin plus aspirin was twice that of patients receiving adjusted-dose warfarin (INR 2.0–3.0) alone [2/171 (1.2%) vs 1/170 (0.6%), respectively, RR 1.99 (95% CI 0.18 to 21.72)] and half that of patients receiving fixed-dose warfarin (1.25 mg daily) alone [2/171 (1.2%) vs 4/167 (2.4%), respectively, RR 0.49 (95% CI 0.09 to 2.63)] over a mean 3.5-year follow-up period.

The SPAF III43 study compared adjusted-dose warfarin (INR 1.2–1.5) plus aspirin (325 mg daily) with adjusted-dose warfarin (INR 2.0–3.0) alone in high-risk patients with AF. A non-significant but numerically higher number of TIAs were observed in the combined therapy arm than in those receiving adjusted-dose warfarin alone [23/251 (4.4%) vs 15/523 (2.9%), respectively, RR 1.54 (95% CI 0.81 to 2.92)] over a mean 1.1-year follow-up period.

The TIA event rate was different in these two randomised comparisons. In the combination of adjusted-dose warfarin (INR 1.2–1.5) plus aspirin (325 mg) arm of the SPAF III43 study, the rate of TIA was 4.4% (23/521) compared with 1.2% (2/171) among those receiving combination fixed-dose warfarin (1.25 mg) plus aspirin (300 mg) in the AFASAK II42 study. The rate of TIA was also higher in those receiving adjusted-dose warfarin (INR 2.0–3.0) alone in the SPAF III43 study [15/523 (2.9%)] than in those receiving either adjusted- (INR 2.0–3.0) or fixed-dose warfarin (1.25 mg) alone in the AFASAK II42 study [1/170 (0.6%) and 4/167 (2.4%), respectively].

The NASPEAF39 randomised comparison compared adjusted-dose acenocoumarol (INR 1.4–2.4) in combination with triflusal 600 mg with adjusted-dose acenocoumarol (INR 2.0–3.0) alone in high-risk patients and adjusted-dose acenocoumarol (INR 1.2–2.0) and triflusal 600 mg in combination compared with adjusted-dose acenocoumarol (INR 2.0–3.0) alone in intermediate-risk patients.

In the high-risk population, a similar rate of TIA was observed with adjusted-dose acenocoumarol (INR 1.4–2.4) in combination with triflusal 600 mg compared with adjusted-dose acenocoumarol (INR 2.0–3.0) alone [2/223 (0.9%) vs 3/247 (1.2%), respectively, RR 0.74 (95% CI 0.12, 4.38)] after a median follow-up of 2.95 years. No TIAs occurred during the median 2.6 years' follow-up in the intermediate-risk patients. 39

Two further articles (AFASAK II,47 NASPEAF45) provided subgroup analyses on the AFASAK II42 and NASPEAF39 studies; however, these articles simply reported duplicate data from the original studies.

No studies of non-randomised comparisons provided further evidence on TIA.

The differences in TIA outcomes reported in the included studies may reflect the methodological differences between these studies discussed in detail above (see Between-study differences).

Outcome 3: stroke and systemic embolism

Five studies, reported in 10 articles,39,42–45,47,67–69,71 yielded outcome data for the combination of stroke and SE. Of these, three studies in six articles,39,42–44,45,47 reported randomised comparisons (Table 12). Two studies in four articles67–69,71 reported pooled analyses of non-randomised comparisons using data from two randomised studies (SPORTIF III and V). The characteristics of the randomised and non-randomised comparison studies have been presented previously in Tables 4 and 6, respectively.

A precise definition of stroke was given in all the study reports, but the definitions of stroke that were used varied between the studies. Although the three randomised comparisons39,42,43 and two pooled analyses of the SPORTIF III and V trials67,69 defined stroke as an acute onset of focal neurological deficit lasting ≥ 24 hours, NASPEAF39 also included TIA, AFASAK II42 included fatal strokes, SPAF III43 included only ischaemic strokes, whereas the SPORTIF III and V trials67,69 included both ischaemic strokes and intracranial haemorrhage (ICH) in their definition. Three studies, NASPEAF,39 SPAF III43 and SPORTIF,67,69 defined SE as an abrupt vascular insufficiency related to arterial occlusion, without previous clinical symptoms (NASPEAF39) or previous evidence of obstructive disease (SPAF III43); SPORTIF III and V67,69 required clinical and radiological evidence of arterial occlusion in the absence of another possible mechanism, and in the presence of atherosclerotic peripheral vascular disease, diagnosis of embolism required angiographic demonstration of acute arterial occlusion. The AFASAK II42 study did not define SE, but specified the sites of the event and required verification using angiography, surgery, scintigraphy or autopsy. From a clinical perspective, it was assumed that these different definitions of embolism were broadly similar and considered the same for the purposes of this review.

For the purpose of this review we are considering SE and TE as the same. It is assumed from the definitions of outcomes provided by the studies that TE refers to arterial TE not venous TE. From this point onwards the term systemic embolism (SE) will be used, but the original terms reported by the studies will be retained in the tables.

Two randomised comparisons42,43 and the two pooled non-randomised comparisons67,69 (Table 13) compared warfarin plus aspirin with warfarin alone in different regimes. The AFASAK II42 study compared fixed-dose warfarin (1.25 mg daily) plus aspirin (300 mg daily) with adjusted-dose warfarin (INR 2.0–3.0) or fixed-dose warfarin (1.25 mg daily). The risk profile of the patients enrolled in the AFASAK II study42 was not specified. The rate of stroke and systemic embolism (including fatal strokes) was the same among patients receiving the combination therapy and patients receiving adjusted-dose warfarin (INR 2.0–3.0) alone [12/171 (7.0%) vs 12/170 (7.1%), respectively, RR 0.99 (95% CI 0.46 to 2.15)] after a median follow-up period of 3.5 years. 42 A non-significant but numerically lower number of people experienced a stroke and systemic embolism among those receiving combination therapy than in those receiving fixed-dose warfarin alone [12/171 (7.0%) vs 14/167 (8.4%), respectively, RR 0.84 (95% CI 0.40 to 1.76)] during the median 3.5-year follow-up period.

The SPAF III43 study compared adjusted-dose warfarin (INR 1.2–1.5) plus aspirin (325 mg daily) with adjusted-dose warfarin (INR 2.0–3.0) alone in high-risk patients with AF. The study43 reported significantly more ischaemic strokes and systemic emboli among those receiving combination therapy than in those receiving adjusted-dose warfarin (INR 2.0–3.0) alone [44/521 (8.4%) vs 11/523 (2.1%), respectively; RR 4.02 (95% CI 2.10 to 7.69)] over a mean 1.1-year follow-up period.

The pooled analyses of the SPORTIF trials67,69 compared combination adjusted-dose warfarin (INR 2.0–3.0) plus aspirin ≤ 100 mg with adjusted-dose warfarin (INR 2.0–3.0) alone, in a pooled analysis of SPORTIF III and V69 and in a subgroup analysis of the pooled SPORTIF III and V67 among those who had experienced an embolic event prior to enrolment. For the whole cohort, the rate of stroke and systemic embolism was very similar in patients receiving the combination therapy to those receiving adjusted-dose warfarin (INR 2.0–3.0) alone, 11 out of 481 (2.3%) versus 69 out of 3172 (2.2%), respectively, during the mean 16.5-month follow-up period. 69

In the pooled analysis restricted to those patients with a previous embolic event prior to randomisation, the rate of stroke and systemic embolism was higher, but not significantly so, among those receiving combination therapy than in those receiving adjusted-dose warfarin (INR 2.0–3.0) alone [13/186 (6.9%) vs 23/567 (4.1%)] during the mean 16.6-month follow-up period. 67

The rate of stroke and systemic embolism was much higher in the AFASAK II42 and SPAF III43 studies than in the pooled analysis of the SPORTIF trials69 for those receiving combination therapy compared with warfarin alone. In the AFASAK II42 and SPAF III43 studies the rate of stroke and systemic embolism were 12 out of 171 (7.0%) and 44 out of 521 (8.4%), respectively, compared with 11 out of 481 (2.3%) in the pooled analysis of SPORTIF. 69

The rate of stroke and systemic embolism was very similar among those receiving adjusted-dose warfarin alone in the SPAF III43 study and the pooled analysis of the SPORTIF69 trial [11/523 (2.1%) and 69/3172 (2.2%), respectively]. However, the rate of stroke and systemic embolism was much higher in the AFASAK II42 study for those receiving adjusted-dose warfarin (INR 2.0–3.0) alone or fixed-dose warfarin (1.25 mg) alone [12/170 (7.1%) and 14/167 (8.4%), respectively] compared with SPAF III43 and the pooled analysis of SPORTIF. 69

The rate of stroke and systemic embolism was very similar in AFASAK II42 but higher in SPAF III43 when compared with the pooled subgroup analysis of SPORTIF III and V restricted to patients with a previous embolic event,67 for those receiving combination warfarin plus aspirin [12/171 (7.0%), 44/521 (8.4%) and 13/186 (6.9%), respectively]. The rate of stroke and systemic embolism was much higher among patients receiving either fixed or adjusted-dose warfarin alone in AFASAK II,42 14/167 (8.4%) and 12/170 (7.1%), respectively, during a median 3.5 year follow-up, and lower in SPAF III43 for those receiving warfarin alone, 11/523 (2.1%) compared with those receiving warfarin alone in the pooled subgroup analysis of SPORTIF,67 23/567 (4.1%) during a mean/median 16.6 month follow-up. The variations in the rates may reflect the heterogeneity between included studies, as discussed above (see Between-study differences, above).

One randomised comparison (NASPEAF39) compared adjusted-dose acenocoumarol (INR 1.4–2.4) in combination with triflusal (600 mg) with adjusted-dose acenocoumarol (INR 2.0–3.0) alone in high-risk patients, and adjusted-dose acenocoumarol (INR 1.25–2.0) and triflusal (600 mg) in combination compared with adjusted-dose acenocoumarol (INR 2.0–3.0) alone in intermediate-risk patients. 39

In the high-risk population, adjusted-dose acenocoumarol (INR 1.4–2.4) in combination with triflusal 600 mg was associated with a non-significant but numerically lower number of stroke and systemic embolism than adjusted-dose acenocoumarol (INR 2.0–3.0) alone [12/223 (5.4%) vs 20/247 (8.1%), respectively, RR 0.66 (95% CI 0.33 to 1.33)], after a median follow-up of 2.95 years. 39 Similarly, when analyses involved only stroke and fatal systemic embolism, adjusted-dose acenocoumarol (INR 1.4–2.4) in combination with triflusal 600 mg was associated with a non-significant but numerically lower number of stroke and systemic emboli than adjusted-dose acenocoumarol (INR 2.0–3.0) alone [4/223 (1.8%) vs 8/247 (3.2%), respectively, RR 0.55 (95% CI 0.17 to 1.81)]. 39

In the intermediate-risk population, adjusted-dose acenocoumarol (INR 1.25–2.0) in combination with triflusal 600 mg was also associated with a non-significant but numerically lower number of stroke and systemic embolism than adjusted-dose acenocoumarol (INR 2.0–3.0) alone [3/222 (1.4%) vs 7/232 (3.0%), respectively, RR 0.45 (95% CI 0.12 to 1.71)] after a median follow-up of 2.6 years. 39 Similarly, when analyses involved only stroke and fatal systemic embolism, adjusted-dose acenocoumarol (INR 1.25–2.0) in combination with triflusal 600 mg was associated with a non-significant but numerically lower number of stroke and systemic emboli than adjusted-dose acenocoumarol (INR 2.0–3.0) alone [0/222 (0%) vs 3/232 (1.3%), respectively, RR 0.15 (95% CI 0.01 to 2.87)]. 39

Three further articles provided post hoc analyses on the NASPEAF44,45 and AFASAK II47studies; however, these papers simply reported duplicate data from the original studies.

In addition to the data on warfarin plus aspirin compared with warfarin alone, the pooled analyses of the SPORTIF III and V studies67,69 also provide data on the risk of stroke and systemic embolism for patients receiving ximelagatran 36 mg given twice daily plus aspirin ≤ 100 mg compared with ximelagatran 36 mg alone. 67,69

In the pooled analyses including all SPORTIF patients,69 combination therapy yielded a slightly higher, but non-significant, rate of stroke and systemic embolism than in those receiving ximelagatran alone, 12/531 (2.3%) versus 58/3120 (1.9%), respectively, during the 16.5-month follow-up period. 69

In just those patients with a previous embolic event, combination therapy yielded a rate of stroke and systemic embolism that was twice that of those receiving ximelagatran alone [11/157 (7.0%) vs 22/629 (3.5%), respectively], during a median 16.6-month follow-up, although this difference was not significant (RR 2.00, 95% CI 0.99 to 4.04). 67

Other pooled analyses of the SPORTIF III and V68,71 trials are not presented in the table to avoid duplication of data.

The differences in stroke and systemic embolism outcomes reported in the included studies may reflect the methodological differences between these studies discussed in detail above (see Between-study differences).

Outcome 4: systemic embolism

Eight studies, reported in 11 articles39–45,47,52,54,61,73 yielded outcome data for SE alone. Of these, four studies39–43 reported randomised comparisons (Table 14), supported by three subgroup analyses. 44,45,47 However, these subgroup analyses did not provide additional data for this outcome and, thus, are not considered further in this section (see Appendix 7).

Four studies52,54,61,73 reported non-randomised comparisons; however, data from Blich et al. 61 and Toda et al. 52 are not reported further in this section (Table 15). The reasons for this can be found in Appendix 7.

A precise definition of SE was not always given in the study reports and/or the definitions vary between studies. The NASPEAF,39 FFAACS41 and SPAF III43 studies defined SE as an abrupt vascular insufficiency related to arterial occlusion, without previous clinical symptoms39 or previous evidence of obstructive disease,43 with one specifying the site of occlusion as affecting the mesenteric, renal, splenic or limb arteries. 41 The AFASAK II42 study did not define a systemic embolic event, but specified the sites of the event and required verification using angiography, surgery, scintigraphy or autopsy. Of the two non-randomised comparisons, PETRO73 defined a SE as an acute non-intracerebral or non-coronary vascular event, whereas Bover et al. 54 did not define SE.

Four studies,41–43,54 three randomised comparisons41,42,43 and one non-randomised comparison54 compared a VKA plus aspirin with a VKA alone. The AFASAK II42 and SPAF III43 studies both compared warfarin plus aspirin with warfarin alone, although the warfarin and aspirin regimes differed between the studies. The FFAACS41 study compared fluindione plus aspirin to fluindione alone, whereas one non-randomised comparison54 compared acenocoumarol plus aspirin with acenocoumarol alone.

The AFASAK II42 study compared fixed-dose warfarin (1.25 mg daily) plus aspirin (300 mg daily) with adjusted-dose warfarin (INR 2.0–3.0) or fixed-dose warfarin (1.25 mg daily). The risk profile of the patients enrolled in this study was not specified. The rates of SE were very small and there were no differences between groups during the median 3.5 years of follow-up; combination therapy compared with adjusted-dose warfarin (INR 2.0–3.0) alone [1/171 (0.6%) vs 2/170 (1.2%), respectively, RR 0.50 (95% CI 0.05 to 5.43)] and compared with fixed-dose (1.25 mg) warfarin alone [1/171 (0.6%) vs 1/167 (0.6%), RR 0.98 (95% CI 0.06 to 15.49)]. The rates of fatal SE were also presented, but given the very low rates of all SE these do not add anything meaningful. 42

The SPAF III43 study compared adjusted-dose warfarin (INR 1.2–1.5) plus aspirin (325 mg daily) with adjusted-dose warfarin (INR 2.0–3.0) alone in high-risk patients with AF. One patient receiving combination therapy experienced a SE compared with no patients who received warfarin alone [(1/521 (0.2%) vs 0/523 (0%), respectively, RR 3.01 (95% CI 0.12 to 73.75)] during the mean 1.1-year follow-up period. 43

The FFAACS41 study compared adjusted-dose fluindione (INR 2.0–2.6) plus aspirin 100 mg with adjusted-dose fluindione (INR 2.0–2.6) alone. The rate of SE among patients receiving combination therapy was twice that of patients receiving fluindione alone [2/76 (2.6%) vs 1/81 (1.2%), respectively, RR 2.13 (95% CI 0.20 to 23.03)] during a mean 0.84-year follow-up, although this difference was not significant.

The non-randomised study by Bover et al. 54 provided additional data on the effect of a VKA plus aspirin compared with a VKA alone. This study compared adjusted-dose acenocoumarol (INR 1.9–2.5) in combination with aspirin (100 mg) with adjusted-dose acenocoumarol (INR 2.0–3.0). There were fewer systemic emboli during a mean 4.92-year follow-up in those receiving combination therapy than in those receiving acenocoumarol alone [0/34 (0%) vs 7/265 (2.6%), respectively; RR 0.51 (95% CI 0.03 to 8.68)], but the difference was not significant. 54

In each study there were very few systemic embolic events. The rate was similar between the four studies41–43,54 and between those receiving combination VKA plus aspirin and those receiving VKA therapy alone,41–43,54 despite methodological and clinical differences between these studies (see Between-study differences).

The NASPEAF39 randomised comparison compared adjusted-dose acenocoumarol (INR 1.4–2.4) in combination with triflusal 600 mg with adjusted-dose acenocoumarol (INR 2.0–3.0) alone in high-risk patients, and adjusted-dose acenocoumarol (INR 1.2–2.0) and triflusal 600 mg in combination compared with adjusted-dose acenocoumarol (INR 2.0–3.0) alone in intermediate-risk patients.

In both comparisons, no systemic embolic events occurred in patients receiving acenocoumarol in combination with triflusal, but a small number of patients in both the high- and intermediate-risk groups experienced a systemic embolic event with acenocoumarol alone [3/247 (1.2%) vs 1/232 (0.4%), respectively]. There were no statistically significant differences between combination therapy and anticoagulation treatment alone in either the high-risk (RR 0.16; 95% CI 0.01 to 3.05) or intermediate-risk (RR 0.35; 95% CI 0.01 to 8.50) populations after a median of 2.95 and 2.6 years of follow-up, respectively. 39

One non-randomised study54 compared adjusted-dose acenocoumarol (INR 1.9–2.5) in combination with triflusal (600 mg) with adjusted-dose acenocoumarol (INR 2.0–3.0) alone, and adjusted-dose acenocoumarol (INR 1.9–2.5) and triflusal (300 mg) in combination compared with adjusted-dose acenocoumarol (INR 2.0–3.0) alone. This study adds data to the randomised comparison in the NASPEAF39 trial above.

Combination acenocoumarol (INR 1.9–2.5) with either triflusal 600 mg or triflusal 300 mg was associated with lower rates of SE, 0 out of 155 (0%) and 2 out of 120 (1.7%), respectively, than acenocoumarol alone, 7 out of 265 (2.6%), after a mean 4.92-year follow-up. 54

One additional study, PETRO,73 reported non-randomised comparisons for the outcome of SE.

The PETRO study73 contained three comparisons: (1) dabigatran 50 mg (twice daily) plus aspirin (either 81 mg or 325 mg daily) compared with dabigatran 50 mg twice daily; (2) dabigatran 150 mg (twice daily) plus aspirin (either 81 mg or 325 mg daily) compared with dabigatran 150 mg twice daily; and (3) dabigatran 300 mg (twice daily) plus aspirin (either 81 mg or 325 mg daily) compared with dabigatran 300 mg twice daily.

Systemic emboli occurred only in patients receiving combination dabigatran 50 mg (once/twice daily) plus aspirin 81 mg and dabigatran 50 mg twice daily alone. The proportion experiencing a SE was higher in patients receiving the combination therapy than in those receiving dabigatran alone [1/21 (4.8) vs 1/59 (1.7), respectively] after a 22-week follow-up period. 73

The differences in SE outcomes reported in the included studies may reflect the methodological differences between these studies discussed in detail above (see Between-study differences).

Outcome 5: acute myocardial infarction

Five studies reported in nine articles39,42–45,47,54,68,69 yielded outcome data for acute myocardial infarction (AMI) (or ACS). Of these, three studies in six articles39,42–45,47 reported randomised comparisons. The key characteristics of these studies have been previously reported previously in Table 4.

The remaining three articles54,68,69 reported non-randomised comparisons; one a primary study by Bover et al. 54 and two secondary analyses of the SPORTIF III and SPORTIF V studies by White et al. 68 and Flaker et al. 69 The characteristics of the studies reporting non-randomised comparisons have been reported previously in Table 6.

Only data from five of the included studies39,42,43,54,69 have been reported in this section. Reasons for non-inclusion of data from other studies have been reported in Appendix 7.

The findings of the studies that report randomised comparisons are shown in Table 16 and non-randomised comparisons in Table 17.

A precise definition of AMI and its subclassification was not always supplied in the study reports and/or the definitions varied between the studies. Among the included studies the AFASAK II trial42 and the analysis of the SPORTIF III and SPORTIF V studies by Flaker et al. ,69 defined AMI by presence of any two assessment criteria, i.e. history of typical chest pain, serial creatine kinase MB isozyme changes typical of AMI, or electrocardiogram changes typical of AMI. The NASPEAF trial39 reported data for non-fatal AMI. Definition of AMI was not specified in the SPAF III trial,43 NASPEAF study39 or in the study by Bover et al. 54

The AFASAK II42 and SPAF III43 studies reported randomised comparisons for different regimes of combined warfarin plus additional aspirin compared with warfarin alone. The findings of these studies have been reported in Table 16.

The AFASAK II42 study reported no AMI events among patients receiving the combination of fixed-dose warfarin (1.25 mg) plus aspirin (300 mg). The AMI event rate was lower but not significantly so among those receiving combination therapy than in those receiving either fixed-dose warfarin (1.25 mg) alone [0/171 (0%) vs 6/167 (3.6%), RR 0.08 (95% CI 0.00 to 1.32)] or adjusted-dose warfarin alone (INR 2.0–3.0) [0/171 (0%) vs 4/170 (2.4%), RR 0.11 (95% CI 0.01 to 2.04)] over a mean follow-up period of 3.5 years. The risk profile of the patients enrolled in this study42 was not specified.

The SPAF III43 study reported a non-significant but higher incidence of AMI events in the combined therapy group than in those receiving adjusted-dose warfarin (INR 2.0–3.0) alone [10/521 (1.9%) vs 5/523 (1.0%), respectively], RR 2.01 (95% CI 0.69 to 5.83), in high-risk patients over a mean follow-up period of 1.1 years. 43

The AMI rate was different in these two RCTs. Rates of AMI were higher in the combined therapy arm of the SPAF III study than those receiving combination therapy in the AFASAK II42 study [1.9% (10/521) vs 0% (0/171), respectively]. However, the AMI rates were lower in those receiving adjusted-dose warfarin (INR 2.0–3.0) alone in the SPAF III study43 [5/523 (1.0%)] than in those receiving either adjusted-dose warfarin alone or fixed-dose warfarin alone [(4/170 (2.4%) and 6/167 (3.6%), respectively] in the AFASAK II study. 42

Flaker et al. ,69 in their post hoc analysis of non-randomised comparisons from the SPORTIF III and V studies, reported fewer AMI events in the combined therapy than in those on adjusted-dose warfarin (INR 2.0–3.0) alone [4/481 (0.8%) vs 46/3172 (1.5%), respectively] over a mean follow-up period of 16.5 months. However, aspirin was indicated in patients with previous CAD in the SPORTIF studies. 64,65

Bover et al. 54 compared adjusted-dose acenocoumarol (INR 1.9–2.5) plus aspirin (100 mg) with adjusted-dose acenocoumarol alone (INR 2.0–3.0) over a mean follow-up period of 4.92 years. This study also compared combination acenocoumarol and two different regimes of triflusal, which will be discussed in a subsequent section. The findings of this study for the outcome of AMI are reported in Table 17.

No AMIs occurred in the 34 patients receiving combination acenocoumarol and aspirin compared with 5 out of 265 (1.9%) AMIs in those receiving acenocoumarol alone. 54 The rate of AMIs was lower in all three combination therapy arms than in the arm with adjusted-dose acenocoumarol alone. No AMIs occurred in those receiving acenocoumarol plus triflusal 600 mg or acenocoumarol plus aspirin 100 mg, and one patient receiving acenocoumarol plus triflusal 300 mg experienced an AMI [1/120 (0.8%)] compared with 5 out of 265 (1.9%) patients receiving adjusted-dose warfarin alone.

Flaker et al. 69 also reported non-randomised comparisons for ximelagatran (36 mg) plus aspirin (100 mg) with ximelagatran (36 mg) alone, over a mean follow-up period of 16.5 months. A slightly higher rate of AMIs was observed in patients on combined therapy than in those on ximelagatran alone [10/531 (1.9%) vs 40/3120 (1.3%), respectively]. However, it is to be noted that aspirin use was indicated in patients with previous CAD in the original SPORTIF studies. 65,68

No studies were identified that reported randomised comparisons for AMI outcome comparing ximelagatran in combination with aspirin with ximelagatran alone.

The NASPEAF study39 compared adjusted-dose acenocoumarol (INR 1.4–2.4) plus triflusal (600 mg) with adjusted-dose acenocoumarol alone (INR 2.0–3.0) in high-risk patients during a median follow-up of 2.95 years, and combination adjusted-dose acenocoumarol (INR 1.2–2.0) plus triflusal (600 mg) with adjusted-dose acenocoumarol alone (INR 2.0–3.0) in intermediate-risk patients during a median follow-up of 2.6 years. This study specified outcomes for non-fatal AMIs. No non-fatal AMIs occurred in the NASPEAF study. 39

Bover et al. 54 reported non-randomised AMI outcome data comparing adjusted-dose acenocoumarol (INR 1.9–2.5) plus triflusal in two different regimes (600 mg or 300 mg) or aspirin (100 mg) with adjusted-dose acenocoumarol (INR 2.0–3.0) alone, over a mean follow-up period of 4.92 years.

The rate of AMI was lower in all three combination therapy arms than for adjusted-dose acenocoumarol alone. No AMIs occurred in those receiving acenocoumarol plus triflusal 600 mg or acenocoumarol plus aspirin 100 mg, and one patient receiving acenocoumarol plus triflusal 300 mg experienced an AMI [1/120 (0.8%)] compared with 5 out of 265 (1.9%) patients receiving adjusted-dose acenocoumarol alone.

The combination of adjusted-dose acenocoumarol (target INR 1.9–2.5) with either triflusal 600 mg or triflusal 300 mg or aspirin (100 mg) demonstrated fewer events of AMI [1/155 (0%), 1/120 (0.8%) and 0/34 (0%), respectively] than acenocoumarol given alone in adjusted dose alone with target INR of 2.0–3.0 [5/265 (1.9%)].

The differences in AMI outcomes reported in the included studies may reflect the methodological differences between these studies discussed in detail above (see Between-study differences). In addition, only two studies,42,69 one randomised42 and one non-randomised69 provided a specific definition for AMI, whereas three others39,43,54 (two randomised39,43 and one non-randomised54) did not. Both the AFASAK II42 and SPORTIF III and V69 studies used the same standard definition of AMI. Four studies42,43,54,69 (two randomised42,43 and two non-randomised54,69) reported all AMIs, whereas one randomised study39 reported only non-fatal AMI events. Of note here for the non-randomised comparisons54,69 is the potential confounding of the addition of APT to ACT at physicians' discretion, which may have resulted in patients at risk of an AMI being given APT, which may account for variation in the reported event rates.

Outcome 6: in-stent thrombosis

No studies were identified that reported in-stent thrombosis outcome data comparing ACT plus APT with anticoagulant alone in an AF population.

Outcome 7: vascular death

Four studies, reported in seven articles39,41–45,47 yielded outcome data for vascular death. Of these, all four studies reported randomised comparisons39,41–43 (Table 18) supported by three subgroup analyses. 44,45,47 No non-randomised comparisons reported vascular death as an outcome.

Vascular death was defined as sudden or any other death occurring within 30 days after a vascular event or progressive HF in the NASPEAF study. 39 The FFAACS study41 reported vascular death as one due to any of the following reasons: ischaemic or haemorrhagic stroke (Rankin score between 4 and 5 followed by death), an AMI, sudden, fatal SE, fatal haemorrhage, arterial aneurysm rupture, gangrene secondary to severe ischaemia and/or pulmonary embolism.41 Vascular death was not defined separately in the AFASAK II study42 or the SPAF III study. 43 The definitions were considered broadly similar for the purposes of this review.

Three randomised comparisons41–43 compared a VKA plus aspirin with a VKA alone. The AFASAK II42 and SPAF III43 studies both compared warfarin plus aspirin with warfarin alone, although the warfarin and aspirin regimes differed between the studies. The FFAACS41 study compared fluindione plus aspirin with fluindione alone.

The AFASAK II42 study compared fixed-dose warfarin (1.25 mg daily) plus aspirin (300 mg daily) with adjusted-dose warfarin (INR 2.0–3.0) or fixed-dose warfarin (1.25 mg daily). The risk profile of the patients enrolled in this study was not specified. The rates of vascular death were low and there were no significant differences in the rate of vascular death between the treatment groups during the median 3.5 years of follow-up: combination therapy compared with adjusted-dose warfarin (INR 2.0–3.0) alone [3/171 (1.8%) vs 5/170 (2.9%), respectively, RR 0.60 (95% CI 0.14 to 2.46)] and compared with fixed-dose (1.25 mg) warfarin alone [3/171 (1.8%) vs 2/167 (1.2%), respectively, RR 1.46 (95% CI 0.25 to 8.66)]. 42

The SPAF III43 study compared adjusted-dose warfarin (INR 1.2–1.5) plus aspirin (325 mg daily) with adjusted-dose warfarin (INR 2.0–3.0) alone in high-risk patients with AF. The rate of vascular death was the same in both the combination therapy and warfarin-alone arms [27/521 (5.2%) vs 27/523 (5.2%), respectively, RR 1.00 (95% CI 0.6 to 1.69)] during the mean 1.1-year follow-up period. 43

The FFAACS41 study compared adjusted-dose fluindione (INR 2.0–2.6) plus aspirin 100 mg with adjusted-dose fluindione (INR 2.0–2.6) alone. The number of vascular deaths in both groups was small and the difference was not significant [3/76 (3.9%) vs 2/81 (2.5%), respectively, RR 1.60 (95% CI 0.27 to 9.31)] during a mean 0.84-year follow-up.

The rate of vascular death differed between the studies. Among those patients receiving combination therapy, the rate of vascular death was highest in the SPAF III43 study: 27 out of 521 patients (5.2%) compared with 3 out of 171 patients (1.8%) in the AFASAK II42 study and 3 out of 76 patients (3.9%) in the FFAACS study. 41 Among those receiving anticoagulation alone, again the rate of vascular death was highest in the SPAF III study,43 27 out of 523 (5.2%) patients, with rates of 1.2% (2/167) and 2.9% (5/170) among those fixed- and adjusted-dose warfarin in the AFASAK II study, respectively, and 2.5% (2/81) in those patients receiving fluindione in the FFAACSs. 41 The NASPEAF39 randomised comparison compared adjusted-dose acenocoumarol (INR 1.4–2.4) in combination with triflusal 600 mg with adjusted-dose acenocoumarol (INR 2.0–3.0) alone in high-risk patients and adjusted-dose acenocoumarol (INR 1.2–2.0) and triflusal 600 mg in combination compared with adjusted-dose acenocoumarol (INR 2.0–3.0) alone in intermediate-risk patients.

Fewer vascular deaths occurred in patients receiving combination therapy than in those receiving acenocoumarol alone in both the high-risk [6/223 (2.7%) vs 17/247 (6.9%), respectively] and intermediate-risk [2/222 (0.9%) vs 11/232 (4.7%), respectively] groups, but these differences were not significant: RR 0.39 (95% CI 0.16 to 0.97) and RR 0.19 (95% CI 0.04 to 0.85), respectively.

No studies reported non-randomised comparisons of ACT plus APT compared with ACT alone for the outcome of vascular death.

The differences in vascular mortality reported in the included studies may reflect the methodological differences between these studies discussed in detail above (see Between-study differences). Of the four randomised studies,39,41–43 only two provided a specific definition of vascular death,39,41 which may reflect the variation in vascular mortality reported between the included studies.

Outcome 8: all-cause mortality

Ten articles39,41–43,47,50,54,58,68,69 yielded outcome data for all-cause mortality. Of these, four studies in five articles39,41–43,47 reported randomised comparisons. The remaining five articles50,54,58,68,69 reported non-randomised comparisons, of which three were primary studies,54,54,58 and two were secondary analyses of the SPORTIF III and SPORTIF V studies by White et al. 68 and Flaker et al. 69

Of the studies that reported non-randomised comparisons, those by Lopes et al. ,50 Stenestrand et al. 58 and White et al. 68 are not mentioned further in this section because two of these50,68 did not furnish details of number of patients (denominator) in either therapy group and one did not report the number of events. 58 The reasons for non-inclusion of their data have been reported in Appendix 7. The characteristics of these studies have been reported previously (see Table 6).

All-cause mortality was frequently classified as death from non-vascular, indeterminant, unknown or sudden causes. A precise definition of these groupings or subclassifications of mortality was not always supplied in the study reports and/or the definitions may vary between studies for the same subclassification.

The findings of the included studies for each of these composites and/or subclassifications of all-cause mortality are detailed in Tables 19 and 20.

All-cause mortality

Two randomised comparisons (AFASAK II42 and SPAF III43) and one non-randomised comparison69 compared the combination of warfarin plus aspirin with warfarin alone.

The AFASAK II42 randomised comparison compared combined fixed-dose warfarin (1.25 mg) plus aspirin (300 mg daily) with either adjusted-dose warfarin (target INR 2.0–3.0) alone or with fixed-dose warfarin (1.25 mg daily) alone. The risk profile of the patients enrolled in this study was not specified. The rate of all-cause mortality was lower among those patients receiving combined therapy than in those receiving fixed-dose warfarin [9/171 (5.3%) vs 17/170 (10%), respectively, RR 0.53 (95% CI 0.24 to 1.15)] and higher than those patients receiving adjusted-dose warfarin [9/171 (5.3%) vs 6/167 (3.6%), respectively, RR 1.46 (95% CI 0.53 to 4.03)] over a mean follow-up period of 3.5 years, although these differences were not significant. 42 The SPAF III study43 compared adjusted-dose warfarin (INR 1.2–1.5) plus aspirin (325 mg) with adjusted-dose warfarin (target INR 2.0–3.0) alone in high-risk patients with AF. The study43 demonstrated similar rates of all-cause mortality for patients treated with adjusted-dose warfarin (INR1.2–1.5) in combination with aspirin (325 mg) compared with those treated with adjusted-dose warfarin (INR 2.0–3.0) alone [42/521 (8.1%) vs 35/523 (6.7%), respectively, RR 1.20 (95% CI 0.78 to 1.86)] over a mean follow-up period of 1.1 years.

There were small differences in the rate of all-cause mortality in these two RCTs. 42,43 In the combination therapy arm of the SPAF III study43 the mortality rate was slightly higher at 8.1% (42/521) than 5.3% (9/171) in the combined therapy arm in the AFASAK II study. 42 Among those patients receiving adjusted-dose warfarin alone, the rate of all-cause mortality was also higher in the SPAF III43 study [35/523 (6.7%)] than in the AFASAK II42 study [6/167 (3.6%)], but lower than those receiving fixed-dose warfarin alone in the AFASAK II study42 [35/523 (6.7%) vs 17/170 (10%), respectively].

The pooled analysis of the SPORTIF studies by Flaker et al. 69 compared adjusted-dose warfarin (INR 2.0–3.0) plus aspirin (100 mg) with adjusted-dose warfarin (INR 2.0–3.0) alone, over a mean follow-up period of 16.5 months. The rate of all-cause mortality was the same in patients receiving combined therapy or warfarin alone [17/481 (3.5%) vs 112/3172 (3.5%), respectively].

The mortality rate was much higher in the combined therapy arm of the AFASAK II42 and SPAF III43 studies than in the combined therapy arm in the SPORTIF III and V69 studies [9/171 (5.3%), 42/521 (8.1%) and 17/481 (3.5%), respectively]. The mortality rate was also much higher in patients receiving adjusted-dose warfarin alone in the SPAF III43 study [35/523(6.7%)] and fixed-dose warfarin alone in the AFASAK II42 study [17/170 (10%)], but similar among those patients receiving adjusted-dose warfarin alone in the AFASAK II42 and SPORTIF III and V studies [6/167 (3.6%) and 112/3172 (3.5%), respectively].

One study (NASPEAF39) reported randomised comparisons on all-cause mortality comparing adjusted-dose acenocoumarol (INR 1.4–2.4) plus triflusal (600 mg) with acenocoumarol alone (INR 2.0–3.0) in high-risk patients, and the combination of adjusted-dose acenocoumarol (INR 1.2–2.0) plus triflusal (600 mg) with acenocoumarol alone (INR 2.0–3.0) in intermediate-risk patients. The findings of this study are presented in Table 19.

The study demonstrated lower rates of all-cause mortality with combined therapy than acenocoumarol alone in the high-risk group [12/223 (5.4%) vs 23/247 (9.3%), respectively; RR 0.58 (95% CI 0.29 to 1.13)] over a median 2.95 year follow-up period, as well as in the intermediate-risk group [6/222 (2.7%) vs 20/232(8.6%), respectively; RR 0.31 (95% CI 0.13 to 0.77), over a median follow-up of 2.6 years, although these differences were not significant.

There was no non-randomised evidence available for all-cause mortality for this comparison.

The FFAACS41 study demonstrated very similar rates of all-cause mortality for patients treated with adjusted-dose fluindione (INR 2.0–2.6) in combination with aspirin (100 mg) to those with adjusted-dose fluindione (INR 2.0–2.6) plus placebo [3/76 (3.9%) vs 3/81 (3.7%), respectively; RR 1.07 (95% CI 0.22 to 5.12], over a mean follow-up period of 0.84 years.

There was no non-randomised evidence available for all-cause mortality for this comparison.

The pooled analysis of SPORTIF trials by Flaker et al. 69 reported non-randomised comparisons for all-cause mortality comparing ximelagatran (36 mg) plus aspirin (100 mg) with ximelagatran (36 mg) alone, over a mean follow-up period of 16.5 months. Fewer deaths were observed in patients on combined therapy than in those on ximelagatran alone [3/531 (0.6%) vs 95/3120 (3.0%), respectively]. However, it is to be noted that aspirin use was based on clinical need and thus the comparison may be confounded by indication. 65,68

There was no randomised evidence available for all-cause mortality for this comparison.

Mortality due to non-vascular causes

The AFASAK II42 and SPAF III43 studies reported randomised comparisons for mortality due to non-vascular causes comparing combinations of different regimes of warfarin plus aspirin to warfarin alone. There were no non-randomised comparisons identified for this outcome.

The AFASAK II42 RCT demonstrated similar rates of mortality due to non-vascular causes in patients receiving the combination of fixed-dose warfarin (1.25 mg) and aspirin (300 mg) compared with those receiving fixed-dose warfarin (1.25 mg) alone [1/171 (0.6%) vs 2/170 (1.2%), respectively); RR 0.50 (95% CI 0.05 to 5.43)] over a mean follow-up period of 3.5 years. No non-vascular deaths occurred in patients receiving adjusted-dose warfarin (INR 2.0–3.0) alone [1/171 (0.6%) vs 0/167 (0%), respectively); RR 2.93 (95% CI 0.12 to 71.42)]. The stroke risk of this population was not specified. 42

The SPAF III43 study demonstrated similar rates of mortality due to non-vascular causes in high-risk patients treated with adjusted-dose warfarin (INR 1.2–1.5) plus aspirin (325 mg) compared with those treated with adjusted-dose warfarin (INR 2.0–3.0) alone [12/521 (2.3%) vs 8/523(1.5%), respectively); RR 1.51 (95% CI 0.62 to 3.65)] over a mean follow-up period of 1.1 years. 43

There were very few non-vascular deaths in these two RCTs. 42,43 In the combination therapy arms, the event rate was higher in the SPAF III43 study at 2.3% (12/521) compared with 0.6% (1/171) in the AFASAK II study. 42 Rates of non-vascular mortality were similar in those receiving adjusted-dose warfarin alone in the SPAF III43 study [8/523 (1.5%)] and fixed-dose warfarin in the AFASAK II study [2/170 (1.2%)]. No non-vascular deaths occurred in the AFASAK II42 study among patients receiving adjusted-dose warfarin. The differences might reflect the methodological heterogeneity between studies as explained previously (see Between-study differences).

The NASPEAF39 study reported randomised comparisons on non-vascular cause mortality comparing adjusted-dose acenocoumarol (INR 1.4–2.4) plus triflusal (600 mg) with acenocoumarol alone (INR 2.0–3.0) in high-risk patients, and the combination of adjusted-dose acenocoumarol (INR 1.2–2.0) plus triflusal (600 mg) with acenocoumarol alone (INR 2.0–3.0) in intermediate-risk patients. The findings of this study are presented in Table 19.

The study demonstrated similar rates of non-vascular death when combined therapy was compared with acenocoumarol alone in the high-risk group [6/223 (2.7%) vs 6/247(2.4%), respectively; RR 1.11 (95% CI 0.36 to 3.38)] and lower but non-significant non-vascular mortality rates in the intermediate-risk group on combined therapy compared with those on acenocoumarol alone [4/222 (1.8%) vs 9/232 (0.48%), respectively; RR 0.46 (95% CI 0.15 to 10.49].

There were no non-randomised comparisons identified for this outcome.

Mortality due to indeterminant or unknown cause

The AFASAK II42 and SPAF III43 studies reported randomised comparisons for mortality due to unknown causes comparing combinations of different regimes of warfarin plus aspirin with warfarin alone. There were no non-randomised comparisons identified for this outcome.

The AFASAK II study42 demonstrated similar rates of mortality from unknown causes across all arms (see Table 19). The event rate was 1.2% in patients receiving combination therapy (2/171) and those receiving adjusted-dose warfarin alone [2/167); RR 0.98 (95% CI 0.14 to 6.85)] and similar in those receiving fixed-dose warfarin alone [3/170 (1.8%); RR 0.66 (95% CI 0.11 to 3.92)] over a mean follow-up period of 3.5 years. The stroke risk of this population was not specified. 42

The SPAF III43 study demonstrated a higher but statistically non-significant rate of mortality owing to indeterminant causes in high-risk patients treated with adjusted-dose warfarin (INR 1.2–1.5) plus aspirin (325 mg) than in those treated with adjusted-dose warfarin (INR 2.0–3.0) alone [3/521 (0.6%) vs 0/523 (0%), respectively; RR 7.00 (95% CI 0.36 to 135.18)] over a mean follow-up period of 1.1 years. 43

The rates of indeterminate mortality were slightly lower in the SPAF III study43 than in the AFASAK II study42 in both the combined therapy group as well as those receiving warfarin alone, despite the methodological differences between these two randomised comparisons. 42,43

Other definitions

Bover et al. 54 reported non-randomised comparisons for non-cardiac and sudden mortality comparing adjusted-dose acenocoumarol (INR 1.9–2.5) plus three different antiplatelet regimes (triflusal 600 mg, triflusal 300 mg or aspirin 100 mg) with adjusted-dose acenocoumarol (INR 2.0–3.0) alone, over a mean follow-up period of 4.92 years. A specific definition for either outcome was not specified. There were no randomised comparisons identified for this outcome.

More non-cardiac deaths were observed in patients receiving any of the combined therapy regimes (triflusal 600 mg, triflusal 300 mg or aspirin 100 mg) [6/155 (3.9%, 3/120 (2.5%) and 1/34 (2.9%), respectively] than those receiving adjusted-dose acenocoumarol alone [3/265 (1.1%)]. 54

The study reported a higher proportion of sudden deaths in patients receiving a combination of either acenocoumarol plus triflusal 600 mg or acenocoumarol plus aspirin 100 mg than with acenocoumarol alone [4/155 (2.6%), 1/34 (2.9%) vs 3/265 (1.1%), respectively] and a lower rate in those receiving combined acenocoumarol plus triflusal 300 mg than in those receiving acenocoumarol alone [0/120 (0%) vs 3/265 (1.1%), respectively].

The differences in all-cause mortality reported in the included studies may reflect the methodological differences between these studies discussed above (see Between-study differences). In addition, although all-cause mortality was frequently classified as death from non-vascular, indeterminant, unknown or sudden causes, a precise definition of these groupings or subclassifications of mortality was not always supplied in the study reports and/or the definitions may vary between studies for the same subclassification, which may account for some variation in the reported event rates.

Overall summary for mortality (excluding vascular death)

Five studies (three randomised39,41,42 and two non-randomised54,69) demonstrated that there is no evidence that combination therapy with ACT plus APT significantly reduces the risk of all-cause39,41,42,54,69, non-vascular,39,42 or non-cardiac54 mortality, mortality from unknown causes,42,43 and sudden death54 compared with ACT alone.

Outcome 9: bleeding

Twenty-seven articles yielded outcome data for bleeding. 39–45,47,51,53,54,56,57,59–65,72,73 Five of these studies in eight articles reported randomised comparisons. 39–45,47 The remaining 19 articles reported non-randomised comparisons of which 14 were primary studies51,53,54,56,57,59–65,72,73 and five were secondary analyses of the SPORTIF III and SPORTIF V studies. 66–70 Of those that reported non-randomised comparisons, data from four articles are reported in this section,54,63,69,73 as the others do not report any further relevant data. These other studies are reported in Appendix 7 except for the study by Akins et al. ,67 which has been reported elsewhere in the results section of the report; however, for the outcome of bleeding it does not report the number of bleeding events by therapy group.

Bleeding events were reported either on their own or in conjunction with other events such as embolism and mortality. In those studies that reported bleeding alone, bleeding was classified as major, minor or non-severe, and intracranial. A precise definition of these subclassifications was not always supplied in the study reports and/or the definitions may vary between studies for the same subclassification. The findings of the included studies for each of these subclassifications of bleeding are detailed in Tables 21 and 22.

All bleeding outcomes

Three studies36,41,73 reported all bleeding outcomes, one randomised41 and two non-randomised36,73 comparisons.

One randomised comparison41 in high-risk patients compared adjusted-dose fluindione (INR 2.0–2.6) plus aspirin 100 mg with adjusted-dose fluindione (INR 2.0–2.6) plus placebo. There were significantly more bleeding events in patients receiving combined therapy than in those on fluindione plus placebo [13/76 (17.1%) vs 2/81 (2.5%), respectively; RR 6.93 (95% CI 1.62 to 29.69] during the mean 0.84-year follow-up.

One non-randomised comparison (PETRO73) compared combinations of different doses of dabigatran (50, 150 and 300 mg) plus different regimes of aspirin (81 and 325 mg) with dabigatran alone (50 mg, 150 mg, 300 mg). Higher proportions of bleeding were found in patients receiving combination therapy at all doses of dabigatran plus aspirin than in those receiving dabigatran alone (see Table 22). A higher proportion of bleeding events were observed in patients on the combination therapy of dabigatran 50 mg plus either aspirin 81 mg or 325 mg [2/21 (9.5%) and 3/27 (11.1%), respectively] than in those receiving dabigatran 50 mg alone [2/59 (3.4%)]. A higher proportion of events were observed in patients on the combined therapy of dabigatran 150 mg plus either aspirin 81 or 325 mg [8/36 (22.2%), 7/33 (21.2%), respectively] than in those receiving dabigatran 150 mg alone [15/100 (15%)]. A higher proportion of patients on the combined therapy of dabigatran 300 mg plus either aspirin 81 or 325 mg [11/34 (32.4%), 14/30 (46.7%), respectively] suffered a bleeding event than in those receiving dabigatran 300 mg alone [14/105 (13.3%)] during a mean follow-up period of 22 weeks. Randomised comparisons for dabigatran plus an antiplatelet agent compared with dabigatran alone were not identified.

Hansen et al. 63 reported registry data comparing warfarin (INR target not stated) in combination with either aspirin (dose not stated) or clopidogrel (dose not stated) or both clopidogrel and aspirin (dose not stated), with warfarin alone (dose not stated). The rate of bleeding was similar among patients receiving warfarin plus aspirin or warfarin alone [1209/18,345 (6.6%) vs 3642/50,919 (7.2%], respectively), although the rate of bleeding was slightly lower in patients receiving either warfarin plus clopidogrel (69/1430 (4.8%)] or triple therapy [64/1261 (5.1%)] than in those receiving warfarin alone [3642/50,919 (7.2%)]. However, the use of an antiplatelet agent is confounded by indication and given that bleeding is a contraindication to ATT-only patients felt to be at low risk of bleeding may have been given combination therapy in this non-randomised comparison.

Major (or severe) haemorrhage

Four randomised comparisons39–43 and three non-randomised comparisons54,69,73 reported data on major (or severe) haemorrhage.

The AFASAK II42 study defined major haemorrhage as fatal, life-threatening, or potentially life-threatening, requiring surgical treatment or blood transfusion. All life-threatening bleeds were confirmed from hospital records. The SPAF III43 study defined major haemorrhage according to the Landfeld criteria, i.e. overt bleeding that was fatal, life-threatening, potentially life-threatening, or acute or subacute leading to reoperation or moderate or severe blood loss. 93 The NASPEAF39 study defined severe haemorrhage as requiring hospital admission, blood transfusion, or surgery. The FFAACS study defined severe haemorrhage as needing treatment (including transfusion) or hospitalisation. 41 These definitions are broadly comparable and are considered equivalent for the purposes of this review.

The AFASAK II42 and SPAF III43 studies reported randomised comparisons for major haemorrhage comparing different regimes of combined warfarin plus aspirin with warfarin alone. The findings of these studies are reported in Table 21.

The AFASAK II42 study reported very low event rates with a non-significant difference in rates of major bleeding between combined fixed-dose warfarin (1.25 mg daily) plus aspirin (300 mg daily) and adjusted-dose warfarin (INR 2.0–3.0) alone [1/171 (0.6%) vs 4/170 (2.4%), respectively, RR 0.25, 95% CI 0.03 to 2.20] or fixed-dose warfarin (1.25 mg daily) alone [1/171 (0.6%) vs 3/167 (1.8%), respectively, RR 0.33, 95% CI 0.03 to 3.09] during the mean 3.5 years of follow-up. The risk profile of the patients enrolled in this study42 was not specified.

The SPAF III43 study reported very similar rates of major bleeding in patients on either adjusted-dose warfarin (INR 1.2–1.5) plus aspirin (325 mg daily) or adjusted-dose warfarin (INR 2.0–3.0) alone in high-risk patients with AF [13/521 (2.5%) vs 12/523 (2.3%), respectively, RR 1.08, 95% CI 0.5 to 2.36] during the mean 1.1-year follow-up period. 43

Flaker et al. 69 reported non-randomised data on a pooled analysis of the SPORTIF III and V studies comparing combined adjusted-dose warfarin (INR 2.0–3.0) plus aspirin (100 mg) with adjusted-dose warfarin alone (INR 2.0–3.0) (Table 22). Higher rates of major bleeding were reported in the combined therapy group than in the warfarin alone group [25/481 (5.2%) vs 100/3172 (3.2%), respectively] during the mean 16.5-month follow-up.

There were small differences in the event rates of major bleeding in the two RCTs. In the combination therapy arms of the randomised comparisons, the rate of major bleeding was higher in the SPAF III43 study than in the AFASAK II42 study [13/521 (2.5%) vs 1/171 (0.6%), respectively], and much lower than the rate of major bleeding with combination warfarin and aspirin therapy in the SPORTIF III and V69 studies [25/481 (5.2%)]. However, rates were similar in those receiving warfarin alone in the SPAF III study,43 12 out of 523 patients (2.3%) to those on either adjusted-dose warfarin (INR 2.0–3.0) alone or those receiving fixed-dose warfarin (1.25 mg) alone [4/170 (2.4%) and 3/167 (1.8%), respectively] in the AFASAK II study,42 and adjusted-dose warfarin alone in SPORTIF III and V69 studies [100/3172 (3.2%)]. Of note is the fact that the SPAF III43 and AFASAK II42 studies included intracerebral haemorrhage events in their definitions of major bleeding; however, the SPORTIF studies69 include both ICH as well as fatal bleed in the total rate of major haemorrhage. This might also explain the differences in the event rates between these studies in addition to the methodological heterogeneity discussed in detail above (see Between-study differences).

The NASPEAF trial39 reported very similar major bleeding event rates in patients on combined adjusted-dose acenocoumarol (INR 1.4–2.4) plus triflusal (600 mg) and those on adjusted-dose acenocoumarol (INR 2.0–3.0) alone in high-risk patients during the median 2.6-year follow-up [12/223 (5.4%) vs 13/247 (5.3%), respectively, RR 1.02 95% CI 0.47 to 2.19]. The rate of major bleeding was lower, but not significantly so, among intermediate-risk patients receiving combined adjusted-dose acenocoumarol (INR 1.2–2.0) and triflusal (600 mg) than in those receiving adjusted-dose acenocoumarol (INR 2.0–3.0) alone during a median 2.9-year follow-up [5/222 (2.3%) vs 10/232 (4.3%), respectively, RR 0.52, 95% CI 0.18 to 1.50]. 39

Bover et al. 54 reported data comparing combined adjusted-dose acenocoumarol (INR 1.9–2.5) in combination with three antiplatelet regimes (triflusal 600 mg and 300 mg, aspirin 100 mg) to adjusted-dose acenocoumarol (INR 2.0–3.0) alone. Higher rates of bleeding were observed in patients on combined acenocoumarol plus aspirin [7/34 (20.6%)] and lower rates in those on combined acenocoumarol plus triflusal 600 mg [10/155 (6.5%)] or combined acenocoumarol plus triflusal 300 mg [6/120 (5.0%)] than in those on acenocoumarol alone [35/265 (12.1%)] during the mean 4.92 years of follow-up. 54 However, the population in this study was derived from a cohort of another RCT (see Between-study differences).

The FFAACS study41 reported higher, but not significantly different, rates of major bleeding with combined adjusted-dose fluindione (INR 2.0–2.6) plus aspirin (100 mg) than with adjusted-dose fluindione (INR 2.0–2.6) plus placebo in high-risk patients during the mean 0.84-year follow-up [3/76 (3.9%) vs 1/81 (1.2%), respectively, RR 3.19, 95% CI 0.34 to 30.07]. 41

There was no non-randomised evidence for this comparison identified for major bleeding.

Flaker et al. 69 reported data on a pooled analysis of the SPORTIF III and V studies comparing combined ximelagatran (36 mg twice daily) and aspirin (≤ 100 mg) with ximelagatran (36 mg twice daily) alone. Lower rates of major haemorrhage were reported in patients on combination therapy than in those on ximelagatran alone [2/531 (0.4%) vs 78/3120 (2.5%), respectively] during the 16.5-month follow-up.

The PETRO study73 reported no major bleeding events in patients on dabigatran 50 mg or 150 mg (in combination with aspirin or given alone). However, a higher proportion of patients on combined therapy of dabigatran 300 mg plus either aspirin 81 mg or 325 mg [1/34 (2.9%), 3/30 (10%) respectively] suffered a major bleeding event than those on dabigatran 300 mg alone [0/105 (0%)] during a mean follow-up period of 22 weeks.

Intracranial haemorrhage

Three randomised comparisons39,42,43 and no non-randomised comparisons reported data on ICH. None of the studies included a definition of ICH.

The AFASAK II42 and SPAF III43 studies reported randomised comparisons for ICH comparing different regimes of combined warfarin plus aspirin to warfarin alone. The findings of these studies are reported in Table 21.

The AFASAK II42 study reported very low event rates with a non-significant difference in rates of intracranial bleeding between combined fixed-dose warfarin (1.25 mg daily) plus aspirin (300 mg daily) and adjusted-dose warfarin (INR 2.0–3.0) alone [0/171 (0%) vs 2/170 (1.2%), respectively, RR 0.19, 95% CI 0.01 to 4.11] or fixed-dose warfarin (1.25 mg daily) alone [0/171, (0%) vs 1/167 (0.6%), respectively, RR 0.33, 95% CI 0.13 to 7.94] during the median 3.5 years of follow-up. The risk profile of the patients enrolled in this study was not specified. 42

The SPAF III43 study reported very similar rates of ICH in patients on either adjusted-dose warfarin (INR 1.2–1.5) plus aspirin (325 mg daily) or adjusted-dose warfarin (INR 2.0–3.0) alone in high-risk patients with AF [5/521 (0.9%) vs 3/523 (0.6%), respectively, RR 1.67, 95% CI 0.4 to 6.96] during the mean 1.1-year follow-up period. 43

The rate of ICH was very low and similar in both of these RCTs. In the combined therapy arm, the rate of ICH was 0.9% (5/521) in the SPAF III43 study compared with 0% in the AFASAK II study. 42 Rates of ICH were similar in those receiving either fixed- or adjusted-dose warfarin in the AFASAK II42 study [2/170 (1.2%) and 1/167 (0.6%), respectively] and adjusted-dose warfarin in the SPAF III study [3/523 (0.6%)]. The difference in the rates may be explained by methodological heterogeneity between the included studies (see Between-study differences).

The NASPEAF trial39 reported low event rates with non-significant differences in rates of ICH between combined adjusted-dose acenocoumarol (INR 1.4–2.4) plus triflusal (600 mg), and adjusted-dose acenocoumarol (INR 2.0–3.0) alone [2/223 (0.9%) vs 5/247 (2.0%), respectively, RR 0.44, 95% CI 0.09 to 2.26] in high-risk patients during the median 2.6-year follow-up, or combined adjusted-dose acenocoumarol (INR 1.2–2.0) plus triflusal (600 mg) compared with adjusted-dose acenocoumarol (INR 2.0–3.0) alone [1/222 (0.5%) vs 4/232 (1.7%), respectively, RR 0.48, 95% CI 0.05 to 4.21] in intermediate-risk patients during a median 2.9-year follow-up. 39

Minor (or non-severe) bleeding

Five randomised comparisons39–43 and no non-randomised comparisons reported data on minor or non-severe bleeding. Definitions for minor bleeding were not clearly specified in these studies.

The AFASAK II42 and SPAF III43 studies reported randomised comparisons for minor bleeding comparing different regimes of combined warfarin plus aspirin with warfarin alone. The findings of these studies are reported in Table 21.

The AFASAK II42 study reported a non-significant difference in rates of minor bleeding when combined fixed-dose warfarin (1.25 mg daily) plus aspirin (300 mg daily) was compared with either adjusted-dose warfarin (INR 2.0–3.0) alone [28/171 (16.4%) vs 42/170 (24.7%), respectively, RR 0.66, 95% CI 0.43 to 1.02] or fixed-dose warfarin (1.25 mg daily) alone [28/171 (16.4%) vs 21/167 (12.6%), respectively, RR 1.30, 95% CI 0.77 to 2.19] during the median 3.5 years of follow-up. The risk profile of the patients enrolled in this study42 was not specified.

The SPAF III43 study also reported similar rates of minor haemorrhage in patients on either adjusted-dose warfarin (INR 1.2–1.5) plus aspirin (325 mg daily) or adjusted-dose warfarin (INR 2.0–3.0) alone in high-risk patients with AF [6/521 (1.2%) vs 4/523 (0.8%), respectively, RR 1.5 95% CI 0.43 to 5.30] during the mean 1.1-year follow-up period. 43

The rates of minor bleeding were much higher in the AFASAK II42 study than in the SPAF III43 study for both the combination therapy [28/171 (16.4%) vs 6/521 (1.2%), respectively] and warfarin-alone arms [adjusted-dose warfarin alone 42/170 (24.7%) vs 4/523 (0.8%), respectively] and 21/167 (12.6%) for fixed-dose warfarin alone in the AFASAK II42 study arms.

There was no non-randomised evidence reported for this comparison/outcome combination.

Lidell et al. 40 reported a non-significant difference in rates of minor bleeding between patients on either combined adjusted-dose warfarin (INR 2.0–3.0) plus clopidogrel (75 mg), or adjusted-dose warfarin (2.0–3.0) plus placebo [0/20 (0%) vs 5/23 (21.8%), respectively, RR 0.10, 95% CI 0.01 to 1.17] during the mean follow-up of 22 days.

There was no non-randomised evidence for this comparison identified for minor bleeding.

The NASPEAF study39 reported non-significant differences in rates of non-severe haemorrhage between combined adjusted-dose acenocoumarol (INR 1.4–2.4) plus triflusal (600 mg) and adjusted-dose acenocoumarol (INR 2.0–3.0) alone in high-risk patients during the median 2.6-year follow-up [20/223 (8.9%) vs 18/247 (7.3%), respectively, RR 1.23, 95% CI 0.67 to 2.27] or combined adjusted-dose acenocoumarol (INR 1.2–2.0) plus triflusal (600 mg) compared with adjusted-dose acenocoumarol (INR 2.0–3.0) alone [16/222 (7.2%) vs 15/232 (6.5%), respectively, RR 1.11, 95% CI 0.56 to 2.20] in intermediate-risk patients during a median follow-up of 2.9 years.

There was no non-randomised evidence for this comparison identified for minor bleeding.

The FFAACS trial41 reported a significant difference in rates of non-severe bleeding, with more events in patients on combined adjusted-dose fluindione (INR 2.0–2.6) plus aspirin (100 mg), than in those on adjusted-dose fluindione (INR 2.0–2.6) plus placebo in high-risk patients during the mean 0.84-year follow-up [10/76 (13.2%) vs 1/81 (1.2%), respectively, RR 10.66, 95% CI 1.39 to 81.28].

There was no non-randomised evidence for this comparison identified for minor bleeding.

The differences in bleeding outcomes reported in the included studies may reflect the methodological differences between these studies, which are discussed in detail above (Between-study differences). Various definitions of major bleeding were used across included studies (although these were considered broadly comparable for the purposes of this review), and subclassifications of bleeding varied between studies and were not always clearly defined. In addition, the likelihood of bleeding is reduced when the INR is < 3.0 and, therefore, studies using INR targets < 3.039,42,43,54 in either the intervention and/or comparator arms may have resulted in few bleeding events. Furthermore, only four studies39,40,43,54 (three randomised39,40,43 and one non-randomised54) reported TTR for ACT plus APT and ACT alone. TTR is associated with the incidence of bleeding events; when TTR is better (≥ 70%) the likelihood of adverse bleeding events is significantly reduced. 94 Therefore, differences in the TTR may help to explain differences in the bleeding event rates reported. Moreover, in the combined therapy group in the non-randomised studies, those patients with a high risk of bleeding may not have received additional APT and, therefore, potential confounding by indication may also account for differences in the bleeding rates reported.

Outcome 10: patient quality of life

Of the included studies, no study was identified that reported quality-of-life outcome for the comparisons of interest.

Outcome 11: major adverse events (all-cause mortality, non-fatal myocardial infarction and stroke) and other composite outcomes

No study was identified that reported major adverse events comprising all-cause mortality, non-fatal MI and stroke. Six articles39,41,43 –45,54 reported other composite events, which included combined end points consisting of two or more previously reported outcomes. Three studies (in five articles39,41,43 –45) reported randomised comparisons, and one study54 reported non-randomised comparisons for various composite end points. The findings of these studies are reported in Table 23 and 24, respectively.

Severe bleeding, non-fatal stroke, transient ischaemic attack, systemic embolism and vascular death

The NASPEAF study reported a randomised comparison on the composite outcome of severe bleeding, non-fatal stroke, TIA, SE and vascular death comparing adjusted-dose acenocoumarol (INR 1.4–2.4) plus triflusal (600 mg) with acenocoumarol alone (INR 2.0–3.0) in high-risk patients, and the combination of adjusted-dose acenocoumarol (INR 1.2–2.0) plus triflusal (600 mg) with acenocoumarol alone (INR 2.0–3.0) in intermediate-risk patients. A lower but statistically non-significant rate of the composite end point occurred in the combined therapy group than in those receiving anticoagulant alone in the high-risk patients [22/223 (9.9%) vs 34/247 (13.8%), respectively, RR 0.72 (95% CI 0.43 to 1.19)] during a median follow-up of 2.95 years. A similar trend was observed in the intermediate-risk group, for which the combination therapy arm demonstrated a lower composite event rate than the acenocoumarol-alone arm [8/222 (3.6%) vs 21/232 (9.1%) respectively, RR 0.40 (95% CI 0.18 to 0.88)] after a median follow-up of 2.6 years (see Table 23).

No other study was identified that evaluated this composite outcome.

Embolism, stroke, acute myocardial infarction and vascular death

The NASPEAF study39 reported a randomised comparison on the composite outcome of embolism, stroke, AMI and vascular death.

A lower but statistically non-significant rate of the composite end point was observed in patients receiving combined therapy than in those on anticoagulant alone, in both the high-risk patients [13/223 (5.8%) vs 25/247 (10.1%), respectively, RR 0.58 (95% CI 0.30 to 1.10)] during a median follow-up of 2.95 years, as well as the intermediate-risk patients [4/222 (1.8%) vs 8/232 (3.4%), respectively, RR 0.52 (95% CI 0.16 to 1.71)] after a median follow-up of 2.6 years (see Table 23). No other study reporting a composite end point of embolism, stroke, AMI and vascular death was identified.

Non-fatal stroke, transient ischaemic attack, systemic embolism and vascular death

The NASPEAF study39 reported a lower rate of non-fatal stroke, TIA, SE and vascular death as a composite end point in patients receiving combined therapy than in those on anticoagulant alone, in both the high-risk patients [14/223 (6.3%) vs 29/247 (11.7%), respectively, RR 0.53 (95% CI 0.29, 0.99)] during a median follow-up of 2.95 years, as well as the intermediate-risk patients [5/222 (2.3%) vs 15/232 (16.5%), respectively, RR 0.35 (95% CI 0.13 to 0.94)] after a median follow-up of 2.6 years (see Table 23).

No other study reporting a composite end point of embolism, stroke, AMI and vascular death was identified.

Systemic embolism and death

The FFAACS41 study reported randomised data comparing adjusted-dose fluindione (INR 2.0–2.6) plus aspirin 100 mg to adjusted-dose fluindione (INR 2.0–2.6) alone. Although not significantly different, composite events of SE and death were reported among patients receiving combination therapy compared with patients receiving fluindione alone [5/76 (6.6%) vs 2/81 (2.5%), respectively, RR 2.66 (95% CI 0.53 to 13.33)] during a mean 0.84-year follow-up (see Table 23).

No other study reporting the composite end point of SE and death was identified.

Stroke, systemic embolism and vascular death

The SPAF III study43 reported a randomised comparison for rates of the composite outcome of stroke, SE and vascular death comparing adjusted-dose warfarin (INR 1.2–1.5) plus aspirin (325 mg) with adjusted-dose warfarin (target INR 2.0–3.0) alone in high-risk patients with AF. A significantly higher incidence of the composite end point was observed in the combined therapy arm than in those receiving warfarin alone [66/521 (12.7%) vs 37/523 (7.1%), respectively, RR 1.79 (95% CI 1.22 to 2.63)] over a mean follow-up period of 1.1 years. 43

No other studies reporting data on this composite end point were identified.

Ischaemic events (all)

Bover et al. 54 reported non-randomised data on the composite outcome of all ischaemic events comparing adjusted-dose acenocoumarol (INR 1.9–2.5) plus three different regimes of APT (triflusal 600 mg or 300 mg, aspirin 100 mg) with adjusted-dose acenocoumarol alone (INR 2.0–3.0) over a mean follow-up period of 4.92 years (see Table 24).

A combination of adjusted-dose acenocoumarol (target INR 1.9–2.5) with triflusal 600 mg or aspirin 100 mg demonstrated fewer ischaemic events [4/155 (2.6%) and 0/34 (0%), respectively] than acenocoumarol alone [22/265 (8.3%)]. However, patients receiving acenocoumarol plus triflusal 300 mg demonstrated more ischaemic events than those on acenocoumarol alone [11/120 (9.2%) vs 22/265 (8.3%), respectively] (see Table 23).

There were no randomised comparisons identified that reported a composite end point of all ischaemic events.

Stroke, systemic/coronary ischaemic events, acute myocardial infarction and mortality

Bover et al. 54 reported lower rates of the composite end point of stroke, systemic/coronary ischaemic events, AMI and mortality in patients on combined therapy of acenocoumarol with either triflusal 600 mg, triflusal 300 mg or aspirin 100 mg [9/155 (5.8%), 12/120 (10%) and 3/34 (8.8%), respectively] than those on acenocoumarol alone [37/265 (13.9%)] over a mean follow-up period of 4.92 years. Of note is the fact that this study consisted of the majority of patients enrolled from another RCT (see Between-study differences).

There were no randomised comparisons identified that reported the composite end point of stroke, systemic/coronary ischaemic events, AMI and mortality.

The differences in major adverse event outcomes reported in the included studies may reflect the methodological differences between these studies discussed in detail above (Between-study differences). Different combinations of major adverse events were examined in composite events in each of the included studies and, therefore, it is not possible to compare across studies.

Outcome 12: revascularisation procedures

No studies were identified that reported the outcome of revascularisation procedures comparing combined anticoagulant plus APT with ACT alone.

Outcome 13: percentage time in therapeutic international normalised ratio range

Four studies39,40,42,54 reported in four articles provided outcome data on percentage time in therapeutic INR range (TTR) for ACT in both the intervention (combined anticoagulation plus APT) and comparator (ACT-alone) arms. Of these, three studies39,40,42 reported randomised comparisons and one study54 reported non-randomised comparisons. The characteristics of these studies have been reported previously in Tables 4 and 6, respectively, and the findings of these studies are reported in Tables 25 and 26, respectively.

Lidell et al. 40 and the SPAF III study43 reported TTR for warfarin plus clopidogrel40 or warfarin plus aspirin43 and warfarin alone. 40,43 In the study by Lidell et al. ,40 TTR was reported to be 100% in both therapy arms,40 whereas the SPAF III43 study reported TTR to be 54% in the combined therapy arm and 61% in the warfarin-alone arm. 43 It should be noted that the SPAF III study43 consisted of a longer follow-up period of a mean of 1.1 years, whereas Lidell et al. 40 followed up only 43 patients over a mean follow-up period of 22 days. Furthermore, the SPAF III43 study used multiple centres utilising testing reagents with multiple sensitivities, whereas Lidell et al. 40 report a central assessment laboratory for all samples.

The NASPEAF study39 reported TTR for acenocoumarol plus triflusal and acenocoumarol alone in high- and intermediate-risk groups. A TTR of 73% was reported in patients receiving combination therapy and 67% in those receiving acenocoumarol alone in the high-risk category. TTR was similar in both therapy arms in the intermediate-risk group (66% in combination therapy arm and 65% in acenocoumarol alone).

The non-randomised comparison by Bover et al. 54 reported a lower TTR in the patients receiving combination acenocoumarol plus triflusal 600 mg (54.2%) and those receiving combination acenocoumarol plus aspirin 100 mg (53%) than in those receiving adjusted-dose acenocoumarol alone (62%). TTR was similar in patients receiving combination acenocoumarol plus triflusal 300 mg to those receiving acenocoumarol alone (59.1% vs 62%, respectively).

The TTR varied markedly between the studies. The study by Lidell et al. 40 achieved 100% TTR in both treatment groups, probably as a result of the small sample size and the relatively short follow-up period. In the combined therapy arms of the other two randomised comparisons, TTR was higher in NASPEAF39 in both the high- and intermediate-risk groups than in the SPAF III43 study (73% and 66% vs 54%, respectively). TTR was lower in all three combined therapy arms of the non-randomised comparison54 than in the combined therapy arms in two of the RCTs,39,40 which may be a reflection of the tighter INR control undertaken in RCTs than in non-RCTs settings but similar to TTR in the SPAF III study. 43 TTR was similar in the anticoagulation-alone arms of NASPEAF39 (67% and 65% in high- and intermediate-risk patients, respectively), the SPAF III43 study (61%) and Bover et al. 54 (62%).

Vitamin K antagonist plus antiplatelet therapy compared with vitamin K antagonist alone

Warfarin, acenocoumarol and fluindione were the VKAs investigated in the included studies. A summary of their findings according to the intervention and comparator are detailed as follows, and Forrest plots (without summary estimates) are available in Appendix 8.

Warfarin plus aspirin compared with warfarin alone

This comparison was investigated in five articles. 42,43,63,68,69 Of these, two studies reported randomised comparisons42,43 and the remaining three were non-randomised comparisons. 63,68,69Table 27 presents the outcomes of these studies. Warfarin and aspirin dosage differed across the studies, along with significant population heterogeneity.

In both RCTs, AFASAK II42 and SPAF III,43 event rates for all categories of stroke were low and similar in patients on combined warfarin [fixed dose42 or adjusted dose (INR 1.2–1.5)43] plus aspirin (30042 or 325 mg43) to those on warfarin [fixed dose42 or adjusted dose (INR 1.2–1.5)42,43] alone, except for ischaemic strokes, for which both studies rates were higher with combination therapy than ACT alone, but not significantly so.

Of the non-randomised comparisons, only Flaker et al. 69 reported outcome data for stroke comparing adjusted-dose warfarin (INR 2.0–3.0) plus ≤ 100 mg aspirin with adjusted-dose warfarin (INR 2.0–3.0) alone, indicating a similar rate of strokes across the two arms.

Differences in the rates of TIA and SE outcomes between the study arms was not significantly different in both the AFASAK II42 and SPAF III studies. 43 No non-randomised study was identified that reported TIA and SE outcome for warfarin plus aspirin compared with warfarin alone.

The rate of the combined end point of stroke and SE was similar across the study arms in the AFASAK II study,42 whereas the SPAF III43 study reported higher rates in patients on combined warfarin plus aspirin than in those with adjusted-dose warfarin (INR 2.0–3.0) alone. The non-randomised comparison by Flaker et al. 69 demonstrated similar rates across the study arms. A subgroup of patients from this cohort with a history of previous embolism was analysed by Akins et al. ,68 demonstrating a higher proportion of patients in the combined therapy arm suffering the end point of stroke or SE than in those on warfarin alone.

The SPAF III43 and AFASAK II42 RCTs did not demonstrate a significant difference in the event rates of AMI between combination therapy and warfarin alone. Flaker et al. ,69 in their non-randomised comparison also demonstrated similar events of AMI in patients on combined therapy compared with those on warfarin alone.

Similar rates of vascular mortality were observed across the study arms in the two RCTs. 42,43 No non-randomised comparisons were identified that reported vascular mortality comparing combined warfarin plus aspirin with warfarin alone.

The AFASAK II42 and SPAF III43 studies demonstrated no significant difference in the rates of all-cause mortality in the combined therapy arms compared with adjusted-dose warfarin (INR 2.0–3.0) alone. Flaker et al. 69 reported a similar proportion of all-cause mortality across arms in a non-randomised comparison.

Similar rates of haemorrhage (intracranial, major and minor) were reported in the combined therapy group compared with warfarin alone in both the AFASAK II42 and SPAF III43 studies. Of the non-randomised comparisons, Hansen et al. 63 reported a smaller proportion of patients suffering a haemorrhagic event in patients on combined therapy than in those on warfarin alone, in a large non-randomised cohort of patients with AF (n = 118,606), followed up over a period of 3.3 years. The study by Flaker et al. ,69 however, demonstrated a higher proportion of patients experiencing haemorrhage in the combined therapy group than in those on warfarin alone over a period of 16.5 months.

Significantly higher rates of the composite end point of stroke, SE and vascular death were reported in patients on combined warfarin plus aspirin than in those on warfarin alone in the SPAF III study. 43 No other study reported outcomes for this comparison.

The SPAF III43 RCT reported TTRs that were within the therapeutic range (in this case between INR 1.5–2.5) for patients on combined therapy for 54% of the time and those on warfarin alone were reported to be within therapeutic range (INR 2.0–3.0) for 61% of the time.

Of the studies that reported randomised comparisons, the AFASAK II study42 was prematurely terminated when results of the SPAF-III trial42 were published, demonstrating the superiority of adjusted-dose warfarin (INR 2.0–3.0) alone, over the combination of adjusted-dose warfarin (INR 1.2–2.5) and aspirin 325 mg, in preventing stroke or SE. 42 Both of these comparisons used different open-label warfarin regimes in the combination and comparator arm, and different doses of aspirin (300 mg AFASAK II42 and 325 mg SPAF III43), and had varying lengths of follow-up (mean 3.5 years in the AFASAK II42 study and mean 1.1 years in the SPAF III43 study). The SPAF III43 study did not consider diabetes mellitus a stroke risk factor, which could have introduced patients at lower risk of stroke into the study, whereas the AFASAK II42 study did not specify stroke risk. Of the non-randomised studies, aspirin was administered at the physician's discretion. 63,69 One study was conducted on hospitalised patients in whom the dosage of warfarin and aspirin was not reported. 63 These factors make it potentially difficult to infer a clear effect of combined therapy on vascular events in a high-risk AF population.

Warfarin plus clopidogrel compared with warfarin alone

This comparison was investigated in two studies, of which one was a randomised comparison40 (the other reported a non-randomised comparison63). Table 27 presents the outcomes of these studies. Of note is the dearth of studies conducted on a group of patients with AF at a specified high risk of stroke randomised to combined therapy of adjusted-dose warfarin (INR of 2.0–3.0) plus clopidogrel and adjusted-dose warfarin (INR 2.0–3.0) alone.

Data were available only for rates of haemorrhage in these two studies. Lidell et al. 40 reported very low event rates for minor haemorrhage in a randomised comparison of a small, predominantly male, sample size (n = 43), followed up over a very short period of time (22 days). Furthermore, Hansen et al. 63 reported a higher proportion of patients suffering from haemorrhage in the warfarin group than in the combined therapy group in a large sample size (n = 118,606) of hospitalised patients followed up over a period of 3.3 years. Clopidogrel was administered according to physician's discretion in this study. Furthermore, the dosage of both warfarin or clopidogrel was unknown in this study. 63 Therefore, from the available evidence, it is difficult to determine the effect of combined therapy on vascular events.

Warfarin plus aspirin plus clopidogrel (triple therapy) compared with warfarin alone

One non-randomised study63 investigated this comparison. 63

Data were available only for rates of haemorrhage for this comparison. Hansen et al. 63 reported a higher proportion of patients suffering from haemorrhage in the warfarin-only group than in the triple therapy group. Although the study was conducted on a large sample size (n = 118,606) over a mean of 3.3 years of follow-up, the dosage of warfarin, aspirin or clopidogrel was not reported. Furthermore, APT was administered at physician's discretion. The evidence is, therefore, insufficient to determine the benefit of combined therapy over warfarin alone for vascular events.

Fluindione plus aspirin compared with fluindione alone

This comparison was investigated in one randomised study41 comparing fluindione (INR 2.0–2.6) plus aspirin (100 mg) with fluindione (INR 2.0–2.6) plus placebo in high-risk patients with AF over a mean follow-up period of 0.84 years. Non-randomised evidence was not identified for this comparison.

The study41 reported very low event rates of SE, vascular death, all-cause mortality, and the composite end point of non-fatal SE and vascular death, with non-significant differences between combined therapy and fluindione plus placebo. However, a significantly higher rate of haemorrhage was observed in patients on combination therapy than in those on fluindione plus placebo. The study was conducted on a small sample size (n = 157) over a mean follow-up period of 0.84 years on a high-risk AF population, 85% of whom were anticoagulant experienced at entry. Of note is the low event rate and premature termination of the trial because of a low enrolment rate. All of these factors render it difficult to meaningfully evaluate the benefit of combination therapy over anticoagulant alone for this combination.

Acenocoumarol plus aspirin compared with acenocoumarol alone

This comparison was investigated in one non-randomised comparison by Bover et al. ,54 comparing adjusted-dose acenocoumarol targeting an INR range of 1.9–2.5 plus aspirin 100 mg with adjusted-dose acenocoumarol (INR 2.0–3.0) alone. The study54 also compared the combination of acenocoumarol plus different regimes of triflusal (300 and 600 mg) with acenocoumarol alone. These comparisons have been reported in previous sections. Many of the patients in the study had been participants in the NASPEAF RCT;39 however, it was difficult to identify which patients these were, what – if any –subsequent treatment they received and, thus, their influence on the findings of this non-randomised comparison. 54

The study54 reported a very small number of outcome events, with fewer events of strokes (total), SE and AMI in the combined therapy group than in the acenocoumarol-alone group. The study54 also reported the composite end points of ischaemic events, stroke, AMI and mortality with no significant differences in events in patients on combined therapy compared with those on acenocoumarol alone. However, patients on combination therapy demonstrated more non-cardiac and sudden deaths, along with a greater prevalence of severe, fatal, and non-GI bleeding than those on acenocoumarol alone.

Of note is, the considerably greater prevalence of stroke risk factors in the patients on acenocoumarol plus aspirin (embolism or age > 75 years, males, HF, diabetes mellitus, dyslipidaemia, coronary disease, smokers) than in those on acenocoumarol alone. 54 There were very few patients in the combined therapy group (n = 34) compared with those on acenocoumarol alone (n = 265). Therefore, it is difficult to conclude the benefit of combined therapy over acenocoumarol alone.

Acenocoumarol plus triflusal compared with acenocoumarol alone

This comparison was investigated in two studies: one reporting a randomised comparison39 and one a non-randomised comparison. 54 No study was identified with a clearly specified group of patients with AF, at a high-risk of stroke, randomised to combination therapy of adjusted-dose acenocoumarol targeting an INR of 2.0–3.0 plus triflusal and adjusted-dose acenocoumarol (INR 2.0–3.0) alone.

Acenocoumarol and triflusal dosage differed between the studies. The NASPEAF study39 compared adjusted-dose acenocoumarol in different regimes (INR 1.4–2.4 and INR 1.25–2.0) plus triflusal 600 mg, with adjusted-dose acenocoumarol (INR 2.0–3.0) alone in patients at a high risk and intermediate risk of stroke. Bover et al. 54 compared adjusted-dose acenocoumarol (INR 1.9–2.5) combined with different regimes of triflusal (600 mg, 300 mg) with adjusted-dose acenocoumarol (INR 2.0–3.0) alone, wherein most patients consisted of previously randomised patients in the NASPEAF RCT. 39 As mentioned previously it was difficult to identify the specific distribution of these patients. 54 It is also important to note that patients on combined therapy had more stroke risk factors than patients on acenocoumarol alone (combination with triflusal 600 mg; greater percentage of patients with previous embolism and dyslipidaemia; combination therapy with triflusal 300 mg consisted of more patients with previous embolism or age > 75 years, diabetes mellitus, dyslipidaemia).

The NASPEAF RCT39 reported no difference in rates of non-fatal stroke between combination ACT plus APT and ACT alone in either a high- or intermediate-risk population. 39 Bover et al. 54 reported a higher proportion of patients on acenocoumarol alone suffering a stroke than in those on combination therapy with acenocoumarol plus triflusal 600 mg, whereas a similar number of events were reported in patients on combined acenocoumarol and triflusal 300 mg than in those receiving acenocoumarol alone (see Table 27).

Similar rates of TIA were observed in both treatment arms in the high- and intermediate-risk population in the NASPEAF RCT. 39 No non-randomised evidence was identified reporting TIA for this comparison.

Very few events of non-fatal SE were observed in the NASPEAF study. 39 The non-randomised comparison study by Bover et al. 54 demonstrated fewer events of SE in patients on combined therapy (with either triflusal 600 mg or 300 mg) than in those on acenocoumarol alone.

Rates of the combined end point of stroke and SE were similar across the arms in both the high-risk group as well as the intermediate-risk group in the NASPEAF study. 39 Non-randomised comparisons were not identified for this end point.

No AMI events were reported in the NASPEAF study. 39 However, Bover et al. 54 demonstrated slightly fewer AMI events in patients on combined therapy (with either triflusal 600 mg or 300 mg) than in those on acenocoumarol alone. 54

The NASPEAF study39 demonstrated significantly lower rates of vascular mortality in patients on combined acenocoumarol plus triflusal 600 mg than in those on acenocoumarol alone in both the high- and intermediate-risk groups.

A non-significant lower rate of all-cause mortality was reported in the high-risk group in the NASPEAF study39 for the combination therapy. This difference was more pronounced in intermediate-risk patients and reached statistical significance. A lower rate of all-cause mortality was reported in patients on combined therapy than in those on acenocoumarol alone in the intermediate-risk group. 39 Furthermore, Bover et al. ,54 in their non-randomised comparison, reported a higher proportion of non-cardiac deaths in patients on combined therapy (acenocoumarol plus either triflusal 600 mg or triflusal 300 mg) than in those on acenocoumarol alone.

No significant differences in the rates of intracranial, severe, non-severe or gastrointestinal (GI) haemorrhage were reported in the randomised NASPEAF study39 comparing the combination of acenocoumarol plus triflusal with acenocoumarol alone in high- and intermediate-risk patients. 39 Bover et al. 54 reported a smaller proportion of patients suffering a severe, fatal or a non-GI haemorrhage in the combined therapy group(s) (acenocoumarol plus either triflusal 600 mg or triflusal 300 mg) than in the acenocoumarol-alone group. However, more patients in the combination therapy group(s) demonstrated GI bleeding than those on acenocoumarol alone.

The rate of the combined end points of non-fatal stroke, TIA, SE and vascular death was significantly lower in patients on combined acenocoumarol plus additional triflusal 600 mg than in those on acenocoumarol alone in both high- and intermediate-risk groups. 39 A similar trend was observed for the combined end point of severe bleeding, non-fatal stroke, TIA, SE and vascular death in the intermediate-risk group in the NASPEAF study. 39 Furthermore, Bover et al. 54 reported fewer events of composite end points (ischaemic events, and stroke, systemic events, AMI and mortality) in the combined therapy group(s) than in the acenocoumarol-alone group.

The NASPEAF study39 reported slightly better TTR of acenocoumarol in the combination therapy arm than in the acenocoumarol-alone arm in the high-risk group. However, in a non-randomised comparison, Bover et al. 54 reported slightly better TTR in patients on acenocoumarol alone than in those on combination therapy of acenocoumarol plus either triflusal 300 mg or 600 mg.

Overall, there seem to be fewer negative events in the combined therapy arms, with statistically significant differences in the mortality rates and composite end points, than in the acenocoumarol-alone arms in either the high- or the intermediate-risk groups in the randomised NASPEAF study. 39 However, there seems to be no statistically significant difference in rate of haemorrhage between the two arms in either risk group. 39 A similar trend was demonstrated in the non-randomised comparison by Bover et al. ,54 with fewer patients suffering stroke, SE, bleeding and composite end points in the combined therapy group than in the acenocoumarol-alone group.

Dabigatran plus aspirin compared with dabigatran alone

This comparison was investigated in the PETRO73 study comparing different regimes of dabigatran (50/150/300 mg twice daily) plus aspirin 81 mg or 325 mg doses with adjusted-dose dabigatran (50/150/300 mg twice daily) alone.

This PETRO study73 reported none or very small number of systemic embolic events in each therapy group. A higher proportion of patients on combined dabigatran (300/150/50 mg twice daily) plus aspirin (81 or 325 mg) experienced a haemorrhagic event than in those on dabigatran (300/150/50 mg twice daily) alone.

The PETRO study73 was conducted on a sample of 502 antithrombotic-experienced patients with AF (82% males) at a high risk of stroke over a follow-up period of 12 weeks. 73 However, after entry of about half of the patients, the requirement for patients to have a history of CAD was removed to facilitate inclusion, which could have allowed inclusion of lower-risk patients as well. The numerical distribution of patients in each group (dabigatran and dabigatran plus aspirin) was uneven, and it was not clear if aspirin was administered at random or conditionally. Therefore, the benefit or harm of combined therapy over anticoagulant alone is not clear for this comparison.

Ximelagatran plus aspirin compared with ximelagatran alone

This comparison was investigated in the pooled analyses of the SPORTIF trials (SPORTIF III64 and SPORTIF V65) by Flaker et al. ,69 and Akins et al. ,67 comparing ximelagatran (36 mg twice daily) plus aspirin (100 mg), with ximelagatran (36 mg twice daily) alone. The study by Flaker et al. 69 demonstrated that a higher proportion of patients on combined ximelagatran plus aspirin suffered a stroke, AMI, haemorrhage, or combined end point of stroke and SE than those on ximelagatran alone. Akins et al. 67 conducted an analysis on a high-risk subgroup (those with history of embolism) for the same cohort, and demonstrated a similar trend for the combined end point of stroke and SE (Table 27). Furthermore, Flaker et al. 69 demonstrated fewer major bleeding events and lower all-cause mortality in the combination arm than in those on ximelagatran alone. The SPORTIF trials64,65 were conducted on patients with AF with at least one risk factor for stroke over a mean follow-up of 16.5 months. Aspirin was indicated for patients with CAD, and there were significant baseline differences between patients administered combined therapy and those on anticoagulant alone. 69 There was no randomised evidence identified for this comparison. Therefore, the benefit of combination therapy over ximelagatran alone is difficult to evaluate from the available evidence.

Population

The review aimed to assess the clinical benefit of combined therapy with ACT plus APT over ACT alone on vascular events in a population at high risk of stroke, with high risk determined either by history of AMI with PCI with or without stent, or having a CHADS₂ score of ≥ 2. However, not all studies identified such a population.

Intervention and comparator

The review was aimed at investigating combined ACT plus APT in comparison with ACT alone (plus placebo), with the dose of ACT adjusted to target the recommended INR range of 2.0–3.0 in both study arms.

Outcomes

The review aimed to assess the benefit of combined therapy over ACT alone on vascular events in a high-risk AF population.

The primary outcome measures assessed were stroke, TIA, SE, the composite end point of SE and stroke, MI, vascular death along with secondary outcomes of all-cause mortality, bleeding events and composite end point consisting of various primary outcomes. Composite end points of stroke and SE were not specified in the review protocol; however, it was considered clinically relevant and reported in a considerable number of included studies and, therefore, was agreed to be reported in the review. The review protocol also specified in-stent thrombosis, revascularisation procedures and quality-of-life outcome measures; however, none of the included studies was found to report these events.

Strengths of the assessment

• Studies included in the assessment consisted of both randomised and non-randomised comparisons in an attempt to investigate all the available evidence around the subject.

• A comprehensive search strategy was undertaken encompassing all relevant databases.

• Robust review methodology was used.

Limitations of the assessment

• It was originally intended that an IPD analysis would be undertaken to specifically address the effect of APT added to ACT (compared with ACT alone) on various outcomes (including time to first vascular event/first major haemorrhage or clinically relevant bleeding/death and time within therapeutic INR range) (see Appendix 1 and Chapter 4, Changes to protocol). Predefined subgroup analyses were to be developed to possibly include stent type; warfarin-naive versus warfarin-experienced subjects; short- and long-term outcomes; patients with diabetes mellitus; and CHADS₂ score ≥ 2 and < 2. However, it became clear from the range of included studies that the methodological heterogeneity between studies, the clinical heterogeneity within and between studies, and the relatively small number of events was against such analyses being able to appreciably add to the findings of the review. To aid explanation, we draw the reader's attention to Table 27. This table summarises the key features and findings of the included studies grouped by similar intervention and comparator. Examining the section of the table containing the five studies that investigated combination ACT plus APT compared with ACT alone reveals that the intervention and comparator regimens were heterogeneous for both elements and, furthermore, the study designs were a mix of randomised and non-randomised comparisons. As with aggregate patient data meta-analysis, clinical and methodological study homogeneity are still overriding considerations prior to undertaking IPD meta-analyses and, thus, it was not an option to pool data across all studies in this case. Only two of the studies67,69 had similar intervention/comparator characteristics and these were the same non-randomised comparison where aspirin was added to warfarin therapy based on clinical indication. Thus, as mentioned previously in this report, the treatment comparison in these studies was confounded by indication and IPD meta-analyses would therefore also be confounded. Where IPD analysis might have been beneficial is in possibly revealing data on outcomes previously unreported for a given study. In the current example, the greatest potential for this was with the two non-randomised comparisons with similar intervention/comparisons or for the outcome of TTR for all warfarin/aspirin studies. However, the utility of this was limited given the aforementioned limitation of combining data across studies. Similar issues also affected the value of IPD analyses for other intervention/comparator combinations (see Table 27). Thus, although the benefits of an IPD approach are well recognised97 in the current report, the approach offered limited advantage. These issues were discussed with the NIHR and with their agreement it was decided not to undertake the planned the IPD analysis. As such, some aspirational aspects of the current work could not be achieved.

• Individual participant data analysis could not be undertaken for various reasons. Included studies reported low event rates, with methodological heterogeneity and ambiguity along with the fact that it was very difficult to identify studies with similar study designs, population characteristics, intervention and comparator therapies, and outcome measures. There is paucity of directly relevant randomised evidence to undertake a meta-analysis for the merits of one technology over another.

• The evidence was such that three stages of study selection were required, with one of these stages being unforeseen. With hindsight, this process might have been more efficiently achieved.

• Although the review initially aimed to identify high-risk patients, none of the included studies specified a high-risk group as per the established stroke risk assessment criteria. Studies were also included if stroke risk was not specified. This might have introduced studies with patients at a lower risk of stroke.

• It was intended to include only those studies that reported data for patients on combined ACT plus APT with a target INR range for ACT of between 2.0 and 3.0. However, this criterion could have been too restrictive; therefore, those studies in which no INR range was specified were also included, as it could not be ruled out that the appropriate INR was utilised.