We reviewed the charts of 1,598 patients undergoing valve replacement at the National Institutes of Health, Bethesda, Md, from 1956 through 1981. Retrospective analysis disclosed that 43 patients had prosthetic valve endocarditis (PVE). Twelve patients had early (less than 60 days after operation) and 31 patients had late (greater than 60 days after operation) endocarditis. The cumulative risk was 3% at five years and 5% at ten years. We also calculated the interval risk of PVE. The high risk of early PVE development peaked 15 days after operation. The peak risk was 45 episodes per 100,000 patient days. The risk then declined rapidly and from 150 days to 20 years remained stable at approximately one episode per 100,000 patient days. Nine hundred fifty-two patients had valve replacement with a Starr-Edwards prosthesis and 363 patients had valve replacement with a bioprosthetic valve; there was no significant difference in the risk of PVE in either group. Neither the valve make, position, model, nor the number of valves implanted affected the frequency of PVE or the mortality. Actuarial techniques disclosed the high early risk of PVE, the prolonged risk of PVE up to 150 days after operation, and the low but persistent risk late after operation. There was no significant difference in the risk of PVE in patients with bioprosthetic v mechanical valves.