From May 1979 until July 1992, 1184 patients, 2 to 86 years of age (median = 59 years), received 1367 St. Jude Medical prostheses (578 aortic, 440 mitral, 3 tricuspid, and 170 multiple). Overall early mortality was 4% with 2.4%, 4.3%, and 8.2% after aortic, mitral and multiple valve replacement, respectively. Follow-up was 100% complete (4936 patient-years). Actuarial survival at 10 years (including both early and late deaths) was 71% +/- 4.1% after aortic valve replacement, 58.8% +/- 5.1% after mitral valve replacement, and 58.9% +/- 9.6% after multiple valve replacement. Multivariate analysis identified age, previous operations, diabetes mellitus, extent of coronary artery disease, preoperative New York Heart Association class, and additional procedures as independent prognostic factors for overall survival. All patients including 20 children (2 to 18 years of age) received sodium warfarin. The linearized risk per 100 patient-years for all embolic events (major and minor) was 2.4, 4.4, and 5.0 after aortic, mitral, and multiple valve replacement, respectively; when only major events were considered, the linearized risks were 0.33, 1.17, 1.54, respectively. Freedom from major systemic embolism at 10 years was 96% +/- 2% after aortic valve replacement, 88% +/- 4% after mitral valve replacement, and 89% +/- 3% after multiple valve replacement. Five patients had valve thrombosis, four of whom definitely received inadequate anticoagulation therapy. Ninety-seven patients had 107 episodes of anticoagulant-related hemorrhage (2.2/100 patient-years) accounting for 17 late deaths; 14 of the deaths involved an international normalized ratio in excess of 3.5. To reduce the rate of thromboembolism without increasing the rate of anticoagulant-related hemorrhage, we propose that the international normalized ratio should be kept between 2.5 and 3.0. No recorded structural failure or significant hemolysis was found in the absence of periprosthetic leak. This experience encourages us to continue using the St. Jude Medical prosthesis.