OBJECTIVE: To develop a health and economic simulation model of antithrombosis therapy that would serve as a general-purpose tool for research planning, clinical trial analysis, and education, and to demonstrate its use in a research planning exercise. METHODS: The Atrial Fibrillation Antithrombosis Model (AFAM) provides estimates of health and cost outcomes for up to four competing, user-defined antithrombosis (AT) strategies that may be used for individuals with atrial fibrillation (AF). These include conventional anticoagulation (AC) or enhanced AT such as improved management of a conventional coumarin anticoagulant (e.g., warfarin provided through AC services) or a new agent with some clinical advantage. Each strategy is characterized by: (1) patient eligibility pattern, (2) efficacy, (3) impact on quality of life; (4) discontinuation rate; and (5) initial and recurrent cost. AFAM is based on the Duke Stroke Policy Model (SPM), a validated natural history simulation that projects the health and cost outcomes of individuals at risk for stroke and those who have experienced stroke. As a demonstration, we applied the model to a research planning exercise. Using AFAM we calculated the threshold efficacy that would lead to an EAT that is cost-effective (i.e., incremental CE ratio of < $50,000.) RESULTS: Assuming an EAT costs 50% greater than AC, and the relative risk of stroke for AC is 0.6, then EAT would be cost-effective if the relative risk for stroke is no less than 0.95. In absolute terms, a cost-effective EAT would need to reduce annual stroke risk from 3% down to 2.85%. CONCLUSIONS: The AFAM takes advantage of a previously developed, validated natural history simulation model. With relative ease this can be used to facilitate research planning, trial analysis and education. Its present application serves to demonstrate that treatments that are only modestly superior to existing therapies for stroke prevention can be cost-effective.