INTRODUCTION: Initiation of warfarin therapy is complicated by its narrow therapeutic index and inter-patient dose-effect variability. A '10-mg nomogram' warfarin initiation protocol permits safe therapeutic anticoagulation in outpatients started on warfarin. We aimed to develop a safe and effective warfarin maintenance dose prediction tool in these patients. METHODS: Baseline potential predictor variables were collected on a retrospective cohort of outpatients initiated on warfarin for venous thromboembolism treatment. The primary outcome was the warfarin maintenance dose, defined as mean warfarin dose over the last 10 days of the first month of warfarin treatment. Univariate and multivariate analyses were performed to determine which baseline variables were warfarin maintenance dose predictors. An independent cohort of patients validated the derived warfarin maintenance dose prediction rule. RESULTS: Patient's age and weight, cumulative dose of warfarin over the first week of induction and international normalized ratio (INR) on days 3, 5 and 8 were statistically significant predictors of the warfarin maintenance dose. Our final prediction rule reads: maintenance dose (in mg) = 2.5 + 10% of the first week cumulative dose - INR value at day 8 + 1.5 if INR was below 2.0 at day 5. In the validation cohort, the predicted dose was strongly correlated with the actual maintenance dose (r = 0.88, P < 0.0001). The mean difference between observed and predicted dose was not clinically significant: -0.1 +/- 1.1 mg. CONCLUSION: In outpatients initiated on warfarin using a '10-mg nomogram', a simple prediction rule can accurately predict warfarin maintenance dose. Prospective studies employing the rule are indicated.