BACKGROUND: Direct oral anticoagulants (DOACs) have been proposed as a more convenient alternative to vitamin K antagonists (VKAs), which are commonly associated with poor treatment persistence in nonvalvular atrial fibrillation (nv-AF). METHODS: Using data from the French National Healthcare databases (Regime General, 50 million beneficiaries), a cohort study was conducted to compare the 1-year non-persistence rates in nv-AF patients initiating dabigatran (N=11,141) or rivaroxaban (N=11,126) versus VKA (N=11,998). Treatment discontinuation was defined as a switch between oral anticoagulant (OAC) classes or a 60-day gap with no medication coverage, with the additional criterion of no reimbursement for international normalized ratio monitoring during this gap for VKA patients. Considering death as a competing risk, differences between 1-year discontinuation rates were used to compare each DOAC versus VKA. 95% confidence intervals [CIs] were estimated via bootstrapping. Baseline patient characteristics were adjusted using inverse probability of treatment weighting. Subgroup analyses considered DOAC dose at initiation, age, risk of stroke, and bleeding. RESULTS: Adjusted 1-year discontinuation rates were higher for dabigatran than for VKA new users (36.8% vs 30.2%; difference: 6.6% [95% CI, 5.5 to 7.6]) and for rivaroxaban versus VKA new users (33.4% vs 30.4%; 3.0% [1.9 to 4.1]). Similar differences were found in all subgroup analyses, except in dabigatran and rivaroxaban patients <75 y (dabigatran vs VKA: 0.3% [-1.4 to 1.8]; rivaroxaban vs VKA: -2.6% [-4.3 to -0.9]) and dabigatran 150 mg new users (-1.1% [-3.1 to 0.7]). Consistent results were obtained when considering both switches between OAC classes and death as competing risks of treatment discontinuation. CONCLUSION: Results from this nationwide cohort study showed high non-persistence levels with all OACs and suggest that persistence with both dabigatran and rivaroxaban therapy is not better than persistence with VKA therapy. Hospitalizations for bleeding among non-persistent patients were unlikely to explain these high non-persistence rates. This article is protected by copyright. All rights reserved.Read less <