During a 5 year period (January 1979-December 1983) 357 patients were submitted to mitral valve replacement. These were performed by the same surgeon and were randomized in 2 groups: Group A consisted of 179 patients who received a St Jude Medical (SJM) prosthesis in the mitral position. Group B comprised 178 patients with a Björk-Shiley valve (BSM) initially (113 patients from 1979 to December 1981 matched with 111 SJM) and later a Starr-Edwards 6120 valve prosthesis (65 patients matched with 63 SJM). Analysis of 21 preoperative clinical, hemodynamic data and operative variables showed the groups to be well randomized. All patients were anticoagulated postoperatively. A follow-up study was performed each year postop: at the end of 1986 there was a 35 to 95 months follow-up with a mean of 64.7 months (1596 patient years follow-up). Fifteen patients were lost to follow-up. There were 8.4% deaths related to the prosthesis in group A and 20.2% in group B (p less than 0.001). The difference was due mainly to deaths from thromboembolic complications and sudden deaths. The rate of peripheral arterial embolic complications was 2.3% in group A and 4.3% in group B per patient year (NS). The difference between the 2 groups is significant for all thromboembolic events including sudden deaths: 3.1% in group A and 7.9% per patient year in group B (p less than 0.001). There were no statistical differences in the rates of endocarditis per patient year (0.3% in group A, 0.9% in group B), reoperation (0.75% in group A, 0.89% in group B), or anticoagulant related hemorrhage (1.6% in group A, 2.4% in group B). Actuarial survival rate, including all postoperative deaths, is significantly different (p less than 0.05) at 5 years, 87.6% +/- 4.5 (group A) versus 77.4% +/- 6 (group B) and at 7 years follow-up, 83.4% +/- 6.5 (group A) versus 73.2% +/- 7.2 (group B). The probability of freedom from death and complications related to the prosthesis is significantly different (p less than 0.001) at 5 years postoperatively: 79% +/- 6.5 for group A versus 54% +/- 7.5 for group B and at 7 years: 72% +/- 7.5 (group A) versus 46% +/- 8.5 (group B). Comparison of subgroups, 113 BSM versus 111 SJM (1979-81) and 65 SE 6120 versus 63 SJM (1982-83) showed similar significant differences in the results: however there were more early deaths, valve thrombosis, valve dysfunctions and sudden late deaths in the BSM group and more peripheral arterial emboli in the SE 6120 group.(ABSTRACT TRUNCATED AT 400 WORDS)