To characterize the efficacy and safety of anticoagulants and antiplatelet agents for prevention of stroke in patients with atrial fibrillation. Randomized trials identified by using the search strategy developed by the Cochrane Collaboration Stroke Review Group. All published randomized trials testing antithrombotic agents to prevent stroke in patients with atrial fibrillation. Data on interventions, number of participants, duration of exposure and occurrence of all stroke (ischemic and hemorrhagic), major extracranial bleeding, and death were extracted independently by two investigators. Sixteen trials included a total of 9874 participants (mean follow-up, 1.7 years). Adjusted-dose warfarin (six trials, 2900 participants) reduced stroke by 62% (95% CI, 48% to 72%); absolute risk reductions were 2.7% per year for primary prevention and 8.4% per year for secondary prevention. Major extracranial bleeding was increased by warfarin therapy (absolute risk increase, 0.3% per year). Aspirin (six trials, 3119 participants) reduced stroke by 22% (CI, 2% to 38%); absolute risk reductions were 1.5% per year for primary prevention and 2.5% per year for secondary prevention. Adjusted-dose warfarin (five trials, 2837 participants) was more efficacious than aspirin (relative risk reduction, 36% [CI, 14% to 52%]). Other randomized comparisons yielded inconclusive results. Adjusted-dose warfarin and aspirin reduce stroke in patients with atrial fibrillation, and warfarin is substantially more efficacious than aspirin. The benefit of antithrombotic therapy was not offset by the occurrence of major hemorrhage among participants in randomized trials. Judicious use of antithrombotic therapy, tailored according to the inherent risk for stroke, importantly reduces stroke in patients with atrial fibrillation.