This report examines and compares the causes of late non-surgical death in 146 of 690 (21%) patients undergoing isolated aortic valve replacement (AVR) and in 79 of 273 (29%) patients undergoing mitral valve replacement (MVR) over a five year follow up period. The distribution of valve related, cardiac but not valve-related and non-cardiac deaths was 43%, 23% and 34% respectively for AVR and 65%, 29% and 6% respectively for MVR; the difference between these distributions was statistically significant. The specific causes of valve related deaths included bleeding (11% vs. 5% for MVR vs. AVR), systemic embolization (6% vs. 4% for MVR vs. AVR), endocarditis (14% vs. 8% for MVR vs. AVR), valve regurgitation (8% vs. 5% for MVR vs. AVR) and valve obstruction (3% vs. 5% for MVR vs. AVR). Sudden death (less than one hour from the onset of acute symptoms) accounted for 23% of deaths for MVR and 16% for AVR. The deaths due to congestive heart failure with normal prosthetic valve function were 13% and 8% for MVR and AVR respectively. Non-cardiac causes accounted for only 6% of MVR deaths but 34% of AVR deaths (p < 0.001). There was no significant difference in the late mortality between mechanical and bioprosthetic valves in the aortic position (24% vs. 22%), but the cumulative rate of late deaths was higher in patients with the Björk-Shiley than with the Hancock valve in the mitral position (41% vs. 25%, p < 0.02). In conclusion, about one quarter of patients surviving either aortic or mitral valve replacement died within five years.(ABSTRACT TRUNCATED AT 250 WORDS)