Aims: To compare the outcomes of real-world Japanese patients with nonvalvular atrial fibrillation who are ineligible for phase III trials of direct oral anticoagulants with those of eligible patients. Methods: In retrospective cohort design, consecutively registered patients with nonvalvular atrial fibrillation who had taken warfarin were followed up and assessed eligibility of patients for phase III trials of direct oral anticoagulants. The effects of the ineligibility of patients on outcomes were estimated using Cox proportional hazards models to calculate the hazard ratio (HR) and 95% confidence interval. Results: We registered 7826 Japanese patients with nonvalvular atrial fibrillation from 71 hospitals. Nearly half (48.2%, n=3772) of these patients were ineligible for phase III trials of direct oral anticoagulants, mainly because of low CHADS2 scores (26.4%), renal dysfunction (9.5%), anemia (6.4%), and chronic treatment with nonsteroidal anti-inflammatories (4.0%). After excluding patients with a CHADS2 score <2 (n=2064, 26.4%) from total ineligible patients, the remaining ineligible patients (n=1708) exhibited significantly greater risks of major bleeding (unadjusted hazard ratio 2.00, 95% confidence interval 1.63–2.44, p<0.0001), stroke/systemic embolism (unadjusted hazard ratio 1.53, 95% confidence interval 1.17–1.98, p=0.0016), and all-cause mortality (unadjusted hazard ratio 2.84, 95% confidence interval 2.36–3.43, p<0.0001) compared to the eligible patients. Conclusions: The benefits and risks of direct oral anticoagulants suggested by phase III trials may not necessarily apply to patients ineligible for Phase III trials. This gap between evidence and practice is an issue in the real-world safety and efficacy of anticoagulants.