Objectives: To compare the benefit-risk and medical costs of rivaroxaban low dosage (15mg) versus vitamin-K antagonists (VKA) for non-valvular (NV) atrial fibrillation (AF) in real-life setting. MethOds: All new users of anticoagulant for NVAF in 2013 or 2014 were identified and followed for 1 year in the French SNIIRAM nationwide claims database. NFAV patients were those with long-term disease registration, hospitalisation, or procedure for AF, without rheumatic valve disease or valve replacement. Patients with rivaroxaban 15mg were 1:1 matched with those with VKA, on gender, age, date of the first drug dispensing, and high-dimensional propensity score, including arterial thrombosis and bleeding risk factors. Relative risk (RR) of the composite criterion (hospitalisation with primary diagnosis for stroke and systemic embolism, major bleeding, and death) during drug exposure was estimated using Cox proportional hazard risk model. Medical costs were calculated according to the collective perspective for the same period. Results: Of 220,011 incident anticoagulant users treated for NVAF in 2013 or 2014, 23,356 patients with rivaroxaban 15mg were matched with the same number of VKA patients. The risk of the composite was significantly lower with rivaroxaban 15mg than VKA (RR: 0.89 [CI95%: 0.84 to 0.94]). The mean cost per patient was higher for anticoagulants and drugs for AF (€ 688 vs € 97), but lower for lab tests (€ 196 vs € 464), transports (€ 216 vs € 289), nursing acts (€ 677 vs 962€), medical visits (€ 848 vs € 952), specific AF hospitalisations (€ 892 vs € 1,066), and total medical cost (€ 8,337 vs € 10,010). cOnclusiOns: The study shows that rivaroxaban 15mg for NVAF is cost-saving compared to VKA with a better benefit-risk in real-life setting and a 17% lower medical cost for the French collective perspective.