The primary objective of this study was to compare the safety of four fixed-dose regimens of edoxaban with warfarin in patients with non-valvular atrial fibrillation (AF). In this 12-week, parallel-group, multicentre, multinational study, 1,146 patients with AF and risk of stroke were randomised to edoxaban 30 mg qd, 30 mg bid, 60 mg qd, or 60 mg bid or warfarin dose-adjusted to a target international normalised ratio of 2.0-3.0. The study was double-blind to edoxaban dose, but open-label to warfarin. Primary outcomes were occurrence of major and/or clinically relevant non-major bleeding and elevated hepatic enzymes and/or bilirubin. Mean age was 65 +/- 8.7 years and 64.4% were warfarin-naïve. Whereas major plus clinically relevant non-major bleeding occurred in 3.2% of patients randomised to warfarin, the incidence of bleeding was significantly higher with the edoxaban 60 mg bid (10.6%; p=0.002) and 30 mg bid regimens (7.8%; p=0.029), but not with the edoxaban 60 mg qd (3.8%) or 30 mg qd regimens (3.0%). For the same total daily dose of 60 mg, both bleeding frequency and trough edoxaban concentrations were higher in the 30-mg bid group than in the 60-mg qd group. There were no significant differences in hepatic enzyme elevations or bilirubin values among the groups. The safety profiles of edoxaban 30 and 60 mg qd in patients with AF were similar to warfarin. In contrast, the edoxaban bid regimens were associated with more bleeding than warfarin. These results suggest that in this three-month study, edoxaban 30 or 60 mg qd are safe and well-tolerated.