BACKGROUND Non-valvular atrial fibrillation (AF) is associated with an increased risk of stroke mediated by embolism of stasis-precipitated thrombi from the left atrial appendage. OBJECTIVE The objective is to characterize the efficacy and safety of oral anticoagulants (OACs) for the primary prevention of stroke in patients with chronic AF. METHODS We searched the Cochrane Stroke Group Trials Register (last searched in June 2004). In addition, we searched the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, Issue 4, 2004), MEDLINE (1966 to June 2004), and the reference lists of recent review articles. We also contacted the Atrial Fibrillation Collaboration and experts working in the field to identify unpublished and ongoing trials. METHODS All randomized controlled trials comparing OACs with control in patients with chronic non-valvular atrial fibrillation and no history of transient ischemic attack (TIA) or stroke. METHODS Trials for inclusion were independently selected by two authors who also extracted each outcome and double-checked the data. The Peto method was used for combining odds ratios. All analysis were, as far as possible, intention-to-treat. Since the published results of four trials included 3% to 8% of participants with prior stroke or TIA, unpublished results excluding these participants were obtained from the Atrial Fibrillation Investigators. RESULTS Of 2313 participants without prior cerebral ischemia from five randomized trials, the mean age was 69 years. Participant features and study quality were similar between trials: the OAC in all five trials was warfarin. About half of participants (n = 1154) were randomized to adjusted-dose warfarin with mean achieved INRs ranging between 2.0 to 2.6. During 1.5 years mean follow up, warfarin was associated with large, highly statistically significant reductions in all strokes (odds ratio (OR) 0.39, 95% confidence interval (CI) 0.26 to 0.59), ischemic stroke (OR 0.34, 95% CI 0.23 to 0.52), all disabling or fatal stroke (OR 0.47, 95% CI 0.28 to 0.80), death (OR 0.69, 95% CI 0.50 to 0.94) and the combined endpoint of all stroke, myocardial infarction or vascular death (OR 0.56, 95% CI 0.42 to 0.76). The observed rates of intracranial and extracranial hemorrhage were not significantly increased by OAC therapy, but the confidence intervals were wide. CONCLUSIONS Treatment with adjusted-dose warfarin to achieved INRs of 2 to 3 reduces stroke, disabling or fatal stroke, and death for patients with non-valvular AF. The benefits were not substantially offset by increased bleeding among these participants in randomized clinical trials. Limitations include relatively short follow up and imprecise estimates of bleeding risks from the selected participants enrolled in the trials. For primary prevention of stroke in AF patients, about 25 strokes and about 12 disabling or fatal strokes would be prevented yearly for every 1000 atrial fibrillation patients given OACs.