OBJECTIVES: To evaluate the cost-effectiveness of new oral anticoagulants (apixaban, dabigatran and rivaroxaban) compared relative to each other and relative to warfarin for stroke prevention in patients with atrial fibrillation. METHODS: We developed a decision analytic model, designed as a probabilistic Markov model containing 200 different probability distributions. The model included eight health states; atrial fibrillation (AF), heart failure, moderate stroke sequela, severe stroke sequela, atrial fibrillation with previous AMI,  atrial fibrillation with previous stroke, atrial fibrillation with major gastrointestinal bleeding and dead. Efficacy data was collected from a recently published report from the Canadian Health Technology Assessment Agency (CADTH). Epidemiological input data was gathered from registries. Data on Quality of Life were based on published EQ-5D data and costs were mainly based on national tariffs. The analysis was stratified according to risk of stroke as measured by CHA2DS2-VASc and risk of bleeding as measured by HAS-BLED. RESULTS: When the new drugs were compared relative to each other for medium risk patients, simulations indicated a 41% probability that apixaban is cost-effective, with 34% and 24% for dabigatran and rivaroxaban, respectively. For high risk patients, simulations indicated a 56% probability that dabigatran is cost-effective, with 29% and 15% for apixaban and rivaroxaban, respectively. When comparing each of the new anticoagulants with the old (warfarin), analyses indicate a probability between 21% and 60% that warfarin is still cost-effective. In separate analyses, we explored the influence of the price of apixaban. Conclusions are clearly, highly dependent on the assumed price of apixaban for this indication. CONCLUSIONS: The results indicate that there is great uncertainty regarding which treatment is the cost-effective alternative. Given current evidence, it may be too early to conclude that the new oral anticoagulants are cost-effective compared to warfarin.