Warfarin is an established treatment for prevention of ischaemic stroke in patients with atrial fibrillation, but the value of this agent relative to aspirin in unclear. In the first Stroke Prevention in Atrial Fibrillation (SPAF-I) study, direct comparison of warfarin with aspirin was limited by the small number of thromboembolic events. SPAF-II aims to address this issue and also to assess the differential effects of the two treatments according to age. We compared warfarin (prothrombin time ratio 1.3-1.8, international normalised ratio 2.0-4.5) with aspirin 325 mg daily for prevention of ischaemic stroke and systemic embolism (primary events) in two parallel randomised trials involving 715 patients aged 75 years or less and 385 patients older than 75; we sought reductions in the absolute rate of primary events by warfarin compared with aspirin of 2% per year and 4% per year, respectively. In the younger patients, warfarin decreased the absolute rate of primary events by 0.7% per year (95% CI-0.4 to 1.7). The primary event rate per year was 1.3% with warfarin and 1.9% with aspirin (relative risk [RR] 0.67, p = 0.24). The absolute rate of primary events in low-risk younger patients (without hypertension, recent heart failure, or previous thromboembolism) on aspirin was 0.5% per year (95% CI 0.1 to 1.9). Among older patients, warfarin decreased the absolute rate of primary events by 1.2% per year (95% CI-1.7 to 4.1). The primary event rate per year was 3.6% with warfarin and 4.8% with aspirin (RR 0.73, p = 0.39). In this older group, the rate of all stroke with residual deficit (ischaemic or haemorrhagic) was 4.3% per year with aspirin and 4.6% per year with warfarin (RR 1.1). Warfarin may be more effective than aspirin for prevention of ischaemic stroke in patients with atrial fibrillation, but the absolute reduction in stroke rate by warfarin is small. Younger patients without risk factors had a low rate of stroke when treated with aspirin. In older patients the rate of stroke (ischaemic and haemorrhagic) was substantial, irrespective of which agent was given. Patient age and the inherent risk of thromboembolism should be considered in the choice of antithrombotic prophylaxis for patients with atrial fibrillation.