A prospective evaluation of 333 consecutive patients undergoing isolated mitral valve replacement between 1982 and 1985 was performed to identify the predictors of survival and valve failure. Follow-up between 2 and 6 years postoperatively (mean, 32 +/- 17 months) was 98% complete. Four prostheses were inserted to permit a prospective evaluation of alternative valves: Björk-Shiley mechanical (n = 118), Ionescu-Shiley pericardial (n = 146), Carpentier-Edwards porcine (n = 38), and Hancock pericardial (n = 31). Hospital mortality was 6%, and actuarial survival at 5 years was 74% +/- 5%. Multivariate Cox regression analysis identified advancing age (less than 40 years, 88% +/- 7%; greater than 70 years, 50% +/- 14%) and poor left ventricular function (ejection fraction less than 0.20, 62% +/- 17%; ejection fraction greater than 0.60, 80% +/- 7%) as independent predictors of postoperative survival. Freedom from structural valve dysfunction, prosthetic valve endocarditis, reoperation, and valve-related mortality and morbidity were 86% +/- 4%, 91% +/- 4%, 81% +/- 4%, and 72% +/- 5%, respectively, at 5 years. The actuarial incidence of valve failure was inordinately high with the Hancock pericardial valve (p less than 0.05). Freedom from thromboembolic events (78% +/- 8% at 5 years) was significantly lower in patients with poor ventricular function (ejection fraction (less than 0.20, 54% +/- 20%; ejection fraction greater than 0.60, 73% +/- 11%; p less than 0.05). Survival after mitral valve replacement was determined by age and left ventricular function. Premature failure of the Hancock pericardial valve resulted in an unacceptable rate of valve-related complications.