OBJECTIVES: To compare costs and effectiveness of dabigatran  versus factor Xa inhibitors (apixaban and rivaroxaban) in patients with non-valvular atrial fibrillation (NVAF) from a private and public health care system perspective in Brazil.  METHODS: The cost-effectiveness analysis was based on Markov modeling, and estimated the costs and clinical outcomes, in a lifetime horizon, associated to dabigatran, apixaban and rivaroxaban in patients with NVAF. Epidemiological and efficacy data derived from international literature and a modified Delphi panel with Brazilian experts (local clinical practice pattern on the management of NVAF patients). The model estimated the number of events (ischaemic strokes, systemic embolisms, transient ischaemic attacks) associated with the respective treatments. To each clinical event costs, disabilities and/or reduction in quality of life, and risk of death were assigned. Only direct medical costs were considered and obtained from Brazilian official databases. Costs and benefits were discounted at 5% yearly, according to Brazilian Health Technologies Assessment guidelines. Sensitivity analysis was designed to assess uncertainty.   RESULTS: Considering 100 patients, base case analysis showed that dabigatran was associated with additional 4.6 life years (LY), additional 4.3 QALYs, and demonstrated a lower incidence of events (1,4 events avoided) versus apixaban. Compared to rivaroxaban, dabigatran was also associated with additional 17.8 life years (LY), additional 19.3 QALYs, and demonstrated a lower incidence of events (2.8 events avoided). Under both perspectives and versus both comparators, dabigatran was associated with lower costs. Sensitivity analyses confirmed the favorable results of the base case.  CONCLUSIONS: Findings suggest that dabigatran is a dominant option for stroke prevention when used instead of direct factor Xa inhibitors (apixaban or rivaroxaban) in NVAF patients from the Brazilian public and private health care system perspective.