A total of 108 patients hospitalized with active (acute) endocarditis on either a native aortic valve (n = 66) or a previously inserted replacement device (n = 42) underwent aortic valve replacement because they were too ill for hospital discharge. A nonstented aortic allograft valve was used in 78 patients and prosthetic (mechanical or bioprosthetic) valves in 30 patients. The survival rate was 82% at 1 months, 73% at 1 year, 64% at 5 years, and 36% at 15 years. It was better in patients with native valve endocarditis than prosthetic valve endocarditis. The incremental risk factors for death in the early phase postoperatively were older age at operation, higher New York Heart Association functional class, and a larger number of previous aortic valve procedures. There were 13 episodes of recurrent endocarditis, giving an actuarial freedom of 80% at 10 years. The hazard function for recurrent endocarditis had only a low constant phase when allograft valves were used, which contrasted with the existence of a high peaking early phase (in addition to the constant phase) when prosthetic devices were used. No risk factors for recurrent endocarditis were found in patients receiving a prosthesis, and "localized" versus "extensive" endocarditis was the only risk factor when an allograft was used. Reoperation was performed in 24 patients for a variety of reasons, and freedom from reoperation was 61% at 10 years. It is concluded that the allograft valve is the valve of choice when aortic valve replacement is required for active endocarditis.