Assessment of outcome prediction models for patients with localized prostate carcinoma managed with radical prostatectomy or external beam radiation therapy

BACKGROUND. A clinical staging system for localized prostate carcinoma that provides reliable information on which management decisions regarding an individual patient can be based is lacking. This study compared the abilities of all published proposed clinical staging systems to predict time to prostate specific antigen (PSA) failure after radical prostatectomy or external beam radiation therapy for clinically localized prostate carcinoma. METHODS. A total of 1441 clinically localized prostate carcinoma patients who were managed with radical prostatectomy at the University of Pennsylvania in Philadelphia (n = 688) or the Brigham and Women's Hospital in Boston (n = 288) or with external beam radiation therapy at the Joint Center for Radiation Therapy in Boston (n = 465) were entered into this study. Patients who received adjuvant or neoadjuvant hormonal or radiation therapy were excluded. Akaike's Information Criterion (AIC) and Schwartz Bayesian Criterion (SBC) estimates, which are comparative measures, were calculated for each clinical staging system. Pairwise comparisons of the AIC and SBC estimates for the most predictive clinical staging systems were performed using a formal bootstrap technique with 2000 replications. RESULTS. Both the staging system based on the risk score and the staging system based on the calculated volume of prostate carcinoma and PSA (cV(Ca)-PSA) optimized the prediction of time to posttreatment PSA failure. The cV(Ca)-PSA system, however, provided a more clinically useful stratification of outcome. CONCLUSIONS. Improved clinical staging for patients with localized prostate carcinoma may be possible with parameters obtained during routine evaluation. Validation by other investigators is underway. (C) 1998 American Cancer Society.