The intermediate clinical results of 289 patients undergoing isolated mitral valve replacement with three different low-profile mechanical prostheses have been retrospectively analyzed and compared. Between June, 1980, and September, 1983, 70 patients received the Omniscience prosthesis, 159 patients the Medtronic-Hall valve, and 60 patients the convexo-concave 70 degree Björk-Shiley prosthesis. Hospital mortality was 15% for the Björk-Shiley group, 4.4% for the Medtronic-Hall group, and 7.1% for the Omniscience group. Cumulative follow-up was 88 years (mean 1.7 years) for the Björk-Shiley, 229 years (mean 1.5 years) for the Medtronic-Hall, and 223 years (mean 3.3 years) for the Omniscience group. All patients were placed on a program of anticoagulant therapy (dicumarol) postoperatively. Actuarial survival rates (+/-SE) 2 years postoperatively were comparable for the three groups: Björk-Shiley, 90% +/- 4.7%; Medtronic-Hall, 93% +/- 2.2%; and Omniscience, 88% +/- 4.1% (p = NS). Late mortality, expressed at linearized rates (percent patient-year +/-SE), was 3.4% +/- 1.9% for the Björk-Shiley group, 1.7% +/- 0.8% for the Medtronic-Hall group, and 3.6% +/- 1.2% for the Omniscience group (p = NS). Actuarially determined rates of freedom from thromboembolic complications (systemic embolism and valvular thrombosis) 2 years postoperatively were 97% +/- 2.2% for the Björk-Shiley group, 94% +/- 2.1% for the Medtronic-Hall, and 84% +/- 4.7% for the Omniscience group (p = 0.05, Omniscience versus Medtronic-Hall; p = 0.02, Omniscience versus Björk-Shiley) The actuarial probability of being free from reoperation 2 years postoperatively was 92 +/- 3.5 for the Björk-Shiley group, 92 +/- 2.9 for the Medtronic-Hall group, and 82 +/- 3.9 for the Omniscience group (p = 0.04). The major cause for reoperation in the Omniscience group was valve thrombosis (seven patients), yielding a linearized incidence (+/-SE) of 3.1 +/- 1.1 (p = 0.01). No statistically significant differences were obtained regarding the incidence of prosthetic infective endocarditis or perivalvular leak. Overall rates of anticoagulant-related hemorrhage were comparable for the three groups.