SummaryDespite evidence from 6 major clinical trials that warfarin effectively prevents strokes in atrial fibrillation, clinicians and health care managers may remain reluctant to support anticoagulant prophylaxis because of its perceived costs. Yet, doing nothing also has a price. To assess this, we carried out a pharmacoe-conomic analysis of warfarin use in atrial fibrillation. The course of the disease, including the occurrence of cerebral and systemic emboli, intracranial and other major bleeding events, was modeled and a meta-analysis of the clinical trials and other relevant literature was carried out to estimate the required probabilities with and without warfarin use. The cost of managing each event, including acute and subsequent care, home care equipment and MD costs, was derived by estimating the cost per resource unit, the proportion consuming each resource and the volume of use. Unit costs and volumes of use were determined from established US government databases, all charges were adjusted using cost-to-charge ratios, and a 3% discount rate was applied to costs incurred beyond the first year. The proportions of patients consuming each resource were estimated by fitting a joint distribution to the clinical trial data, stroke outcome data from a recent Swedish study and aggregate ICD-9 specific, Massachusetts discharge data. If nothing is done, 3.2% more patients will suffer serious emboli annually and the expected annual cost of managing a patient will increase by DM 2,544 (1996 German Marks), from DM 4,366 to DM 6,910. Extensive multiway sensitivity analyses revealed that the higher price of doing nothing persists except for very extreme combinations of inputs unsupported by literature or clinical standards. The price of doing nothing is thus so high, both in health and economic terms, that cost-consciousness as well as clinical considerations mandate warfarin prophylaxis in atrial fibrillation.