In patients with nonvalvular atrial fibrillation, warfarin prevents ischemic stroke, but dose adjustment, coagulation monitoring, and bleeding limit its use. To compare the efficacy of the oral direct thrombin inhibitor ximelagatran with warfarin for prevention of stroke and systemic embolism. Double-blind, randomized, multicenter trial (2000-2001) conducted at 409 North American sites, involving 3922 patients with nonvalvular atrial fibrillation and additional stroke risk factors. Adjusted-dose warfarin (aiming for an international normalized ratio [INR] 2.0 to 3.0) or fixed-dose oral ximelagatran, 36 mg twice daily. The primary end point was all strokes (ischemic or hemorrhagic) and systemic embolic events. The primary analysis was based on demonstrating noninferiority within an absolute margin of 2.0% per year according to the intention-to-treat model. During 6405 patient-years (mean 20 months) of follow-up, 88 patients experienced primary events. The mean (SD) INR with warfarin (2.4 [0.8]) was within target during 68% of the treatment period. The primary event rate with ximelagatran was 1.6% per year and with warfarin was 1.2% per year (absolute difference, 0.45% per year; 95% confidence interval, -0.13% to 1.03% per year; P<.001 for the predefined noninferiority hypothesis). When all-cause mortality was included in addition to stroke and systemic embolic events, the rate difference was 0.10% per year (95% confidence interval, -0.97% to 1.2% per year; P = .86). There was no difference between treatment groups in rates of major bleeding, but total bleeding (major and minor) was lower with ximelagatran (37% vs 47% per year; 95% confidence interval for the difference, -14% to -6.0% per year; P<.001). Serum alanine aminotransferase levels rose to greater than 3 times the upper limit of normal in 6.0% of patients treated with ximelagatran, usually within 6 months and typically declined whether or not treatment continued; however, one case of documented fatal liver disease and one other suggestive case occurred. The results establish the efficacy of fixed-dose oral ximelagatran without coagulation monitoring compared with well-controlled warfarin for prevention of thromboembolism in patients with atrial fibrillation requiring chronic anticoagulant therapy, but the potential for hepatotoxicity requires further investigation.