This study aimed to investigate the effects of anticoagulants on ultra-aged patients with nonvalvular atrial fibrillation (AF). We retrospectively studied 320 consecutive patients with AF (median age, 91 years; range 90-100.1 years). Patients were categorized as follows: patients taking direct oral anticoagulant (DOAC group, n = 93), those taking warfarin (warfarin group, n = 147), and those not taking oral anticoagulants (non-OAC group, n = 80). During the follow-up periods (median 3.00 years; first and fourth quantiles, 1.13 and 4.56 years, respectively), in thromboembolic events, the DOAC, warfarin, and non-OAC groups showed the lowest (0%, 0/93; 0%/year), intermediate (4.7%, 7/149; 1.43%/year), and highest (5%, 4/80; 2.65%/year) incidence rates, respectively. In major bleeding events, the DOAC, warfarin, and non-OAC groups showed the highest (9.67%, 9/96; 5.00%/year), intermediate (8.1%, 12/149; 2.46%/year), and lowest (0%, 0/80; 0%/year) incidence rates, respectively. These differences in the relationships of the 3 groups were statistically significant. Confounding factors did not affect these results. Bruises associated with impairment of motor function with aging caused major bleeding in approximately 60% of major bleeding cases. The Cox proportional hazards model revealed that warfarin decreased mortality, whereas antiplatelet drugs increased mortality. In conclusion, DOACs had considerably high incidence of major bleeding events, whereas absence of OAC treatment was associated with substantially high thromboembolic events. Warfarin showed acceptable incidence ratios of both events. At present, warfarin is thus believed to be adequate for ultra-aged (≥90 years) patients with nonvalvular AF. Avoidance of bruises was important to prevent major bleeding events. Antiplatelet drugs were suggested not to be adequate for these patients.