OBJECTIVE Identifying pretreatment and posttreatment predictors of time to prostate cancer specific death (PCSD) following external beam radiation therapy (RT) is the subject of this study. METHODS A Cox regression analysis was used to evaluate the ability of the pretreatment risk group to predict time to PCSD for 381 patients who underwent RT for clinically localized prostate cancer. Posttreatment factors analyzed for the 94 patients who experienced prostate specific antigen (PSA) failure included the time to failure, posttreatment PSA doubling time (DT) and timing of salvage hormonal therapy. RESULTS Despite a median age of 73 at diagnosis, 45% of patients with high risk disease were estimated to die of prostate cancer within 10 years following RT compared to 0% (p = 0.004) and 6% (p = 0.05) of those with low or intermediate risk disease, respectively. Predictors of time to PCSD following PSA failure included PSA DT (p = 0.01) and delayed use of hormonal therapy (p <or=0.002). Nearly identical estimates of PCSD and all cause death following PSA failure were noted for patients with a short PSA DT (ie 12 months or less). CONCLUSIONS Prostate cancer was a major cause of death during the first decade following RT for patients with clinically localized but high risk disease, and the cause of death for patients with a short PSA DT following RT was nearly always prostate cancer. These data provide evidence to propose the hypothesis that a short posttreatment PSA DT may serve as a possible surrogate for PCSD. Prospective validation is needed.