Study objectives: To identify the risk of thromboembolism after withholding or reversing the effect of warfarin therapy following a major hemorrhage. Design: Retrospective medical record review. Setting: Tertiary-care hospital. Patients: Twenty-eight patients with prosthetic heart valves receiving warfarin were hospitalized for major hemorrhage from 1990 to 1997. The mean +/- SD age was 61 +/- 11 years (15 men and 13 women). Twenty patients had St. Jude valves, 4 patients had Carpentier-Edwards bioprosthetic valves, 2 patients had Starr Edwards valves, and 2 patients had Bjork-Shiley valves. Valves were in the mitral position in 12 patients, the aortic position in 12 patients, and both mitral and aortic positions in 4 patients. The average interval from valve surgery to index bleeding wa 7 years. Twenty-five patients had GI or retroperitoneal hemorrhage, 2 patients had an intracranial hemorrhage, and 1 patient had a subdural hematoma. Interventions: Vitamin K was administered to five patients and fresh frozen plasma was given to seven patients to reverse anticoagulation. The mean duration of anticoagulation withholding was 15 +/- 4 days. Measurements and results: None of the patients had thromboembolic complications. There were four in-hospital deaths. Twenty-two of the 24 hospital survivors resumed warfarin therapy at hospital discharge. At 6-month follow-up, 10 of 19 patients remaining on warfarin therapy had recurrent GI bleeding. Conclusions: Thromboembolic risk is low in prosthetic heart valve patients hospitalized with major hemorrhage when their warfarin therapy is reversed or withheld. Recurrent bleeding within 6 months of the resumption of anticoagulation is common, and aggressive treatment of the bleeding source and the risk-benefit ratio of continued anticoagulation need to be considered.