Types of studies

Randomised clinical trials that directly compare the effects of long‐term (more than four weeks) oral anticoagulant therapy with antiplatelet therapy for the prevention of cerebral and systemic embolisms in patients with non‐valvular AF.

Types of participants

Patients with non‐valvular AF, who are eligible for long‐term treatment with oral anticoagulants or antiplatelet therapy in order to reduce the risk of cerebral and systemic embolism. We will include both patients with and without prior stroke or TIA.

Types of interventions

Treatment with an anticoagulant agent (e.g. vitamin K antagonist (VKA), oral direct factor Xa inhibitor or direct thrombin inhibitor) versus treatment with an agent whose primary antithrombotic action is presumed to be inhibition of platelet aggregation (e.g. aspirin, clopidogrel, ticlopidine, dipyridamole, indobufen or triflusal).

Types of outcome measures

Electronic searches

We will search the trials registers of the Cochrane Stroke Group and the Cochrane Heart Group.  In addition, we will search the following electronic databases and trials registers:

1. Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, latest issue);

2. MEDLINE (from 1950) (Appendix 1);

3. EMBASE (from 1980);

4. Stroke Trials Directory (http://www.strokecenter.org/trials);

5. ClinicalTrials.gov (http://www.clinicaltrials.gov); and

6. Current Controlled Trials (http://www.controlled‐trials.com).

We will modify the MEDLINE search strategy to search the other databases.

Searching other resources

In an effort to identify further published, unpublished, ongoing and planned trials we will:

1. screen reference lists of relevant trials;

2. contact relevant pharmaceutical companies;

3. contact authors, colleagues and researchers active in the field;

4. identify and handsearch the proceedings of relevant conferences;

5. search Google Scholar; and

6. use Science Citation Index Cited Reference Search for forward tracking of relevant references.

We will not apply any language restrictions to our searches, and we will attempt to obtain translations of potentially relevant non‐English language papers.

Selection of studies

Three review authors will independently screen titles and abstracts of references identified by the searches and exclude obviously irrelevant citations. We will then obtain the full paper copy of the remaining articles. The same review authors will then independently assess these for inclusion. We will resolve any disagreements regarding which trials are eligible for inclusion by discussion, with input from an external expert when needed. If a trial is excluded, we will keep a record of both the report and the reason for exclusion.

We will not use a scoring system to assess the quality of each trial, but for each included trial we will collect information about:

1. the method of randomisation (including concealment of allocation);

2. blinding (care provider, patient, outcome assessment);

3. the number of patients lost to follow‐up; and

4. whether the trial data were analysed according to the principle of 'intention‐to‐treat' (ITT).

Data extraction and management

Three review authors will independently extract data from the report of each eligible trial and record the information on a specially designed data extraction form. The review authors will not be blinded to journal or institution. We will extract the following data from each report:

• inclusion and exclusion criteria;

• method of randomisation;

• masked versus open‐label intervention;

• diagnostic criteria used for the assessment of major vascular events: stroke (both ischaemic and haemorrhagic), vascular death (including fatal haemorrhages), myocardial infarction or systemic embolism;

• number of patients in each treatment group with outcome events;

• generic name and used dose(s) of oral anticoagulant and antiplatelet agents;

• duration of anticoagulant therapy in the trial, the intensity of anticoagulation dose‐adjusted using the prothrombin time ratio (PTR) or International Normalised Ratio (INR) and adherence to anticoagulant treatment;

• concomitant treatment of any non‐steroidal anti‐inflammatory drugs (NSAIDs);

• relevant baseline characteristics (e.g. sex, age, CHADS2 score) and

• outcomes (as listed above).

One review author (KBS) will enter the data into the Cochrane Review Manager software, RevMan 5.1 (RevMan 2011). Other review authors (EB, RS) will check these data against the hard copy data extraction forms to correct any clerical data entry errors. If any relevant data are missing from the available publications, we will make direct contact with the principal investigators concerned.

Assessment of risk of bias in included studies

We will make an assessment of the risk of selection bias by appraisal of the random sequence and allocation concealment of the intervention, the risk of performance bias by appraisal of the blinding of the investigator, or patient, or both and information bias by blinding of the outcome assessor. We will appraise attrition bias by the number of patients lost to follow‐up and patients excluded from the study, and also appraise any reporting biases in the studies. We will score these data as low risk of bias, unclear or high risk of bias.

Measures of treatment effect

For dichotomous outcomes, we will calculate a weighted estimate of the treatment effects across trials (odds ratio (OR) with a 95% confidence interval (CI)).

Dealing with missing data

If the published information does not allow ITT analysis we will contact the authors to get as complete follow‐up data as possible on all randomised patients for the originally proposed period of follow‐up.

Assessment of heterogeneity

We will test for heterogeneity between trial results with Cochran's Q statistic and I² statistic (percentage of total variation across studies due to heterogeneity).

Assessment of reporting biases

We will use funnel plots to assess reporting bias. We will also assess funnel plots qualitatively.

Data synthesis

We will use the random‐effects model to assess the overall treatment effects.

Subgroup analysis and investigation of heterogeneity

Where possible, we will do subgroup analyses for the following.

1. Primary versus secondary prevention. Effect of intervention among those with and those without prior stroke or TIA and comparison of the effect between the two subgroups.

2. Patients who have received VKA treatment with a sufficient time‐in‐therapeutic range (TTR): 60% or more of the INR measurements within the therapeutic range 2.0 to 3.0 versus less than 60% (ACTIVE 2008).

3. Patients aged more than 75 years versus those of 75 years or less.

4. Race.

5. Sex.

6. Baseline stroke risk factors (CHADS2 score of 0, 1 and ≥ 2; scores of 0 and 1 will be combined in case the group of patients scoring 0 is very small).

7. Renal function (normal versus reduced renal function).

We will use the method described by Deeks for performing subgroup analyses (Deeks 2001).

Sensitivity analysis

If we find evidence of heterogeneity we intend to conduct sensitivity analyses by stratifying for trial quality and for blinded versus open label studies.