For the determination of the risk factors associated with the development of and death caused by prosthetic valve endocarditis, data were reviewed from 66 patients who were prospectively entered into the Veterans Affairs Cooperative Study on Valvular Heart Disease and in whom prosthetic valve endocarditis subsequently developed. Data were recorded at 13 medical centers between October 1977 and September 1982 in patients randomized to receive a mechanical valve (Bjork-Shiley spherical disc, n = 510 patients) or a bioprosthetic valve (Hancock porcine heterograft, n = 522 patients). The average rate of prosthetic valve endocarditis development was 0.8% per year over an average follow-up period of 7.7 years. Of the 66 patients in whom prosthetic valve endocarditis developed (5.8%), 15 cases occurred within 2 months of operation (early) and 51 occurred after operation (late). The most significant preoperative predictor of prosthetic valve endocarditis was active endocarditis at the time of operation (7.4% versus 0.9%) (p = 0.001). Early prosthetic valve endocarditis occurred more frequently in patients who underwent operation for multivalvular disease (p = 0.023). Significantly related perioperative variables were coma, prolonged mechanical ventilation, deep postoperative wound infection, postoperative jaundice, ventricular tachycardia, ventricular fibrillation, and replacement of more than one valve (p < 0.05). Multivariate predictors were hypoxia (p = 0.001), preoperative endocarditis (p = 0.003), preoperative valve lesion (p = 0.020), and resident surgeon (p = 0.05). Significant preoperative variables predictive of late prosthetic valve endocarditis were mitral stenosis and mixed mitral stenosis-regurgitation. The only multivariate predictor of late prosthetic valve endocarditis was superficial wound infection (p = 0.004). Of deaths attributable to prosthetic valve endocarditis, 41% occurred in patients treated with antibiotics alone, 48% occurred in patients treated with surgical intervention and antibiotics, and death resulted in both patients who received no treatment. No difference was found in the risk of early or late postoperative prosthetic valve endocarditis developing in patients receiving the mechanical valve versus those receiving the bioprosthetic valve.