BIOSTATISTICAL MODELING USING TRADITIONAL PREOPERATIVE AND PATHOLOGICAL PROGNOSTIC VARIABLES IN THE SELECTION OF MEN AT HIGH-RISK FOR DISEASE RECURRENCE AFTER RADICAL PROSTATECTOMY FOR PROSTATE-CANCER

Purpose: Biostatistical models predicting the risk of recurrence after radical prostatectomy for clinically localized prostate cancer are ne cessary. Identifying these high risk patients shortly after surgery, w hile tumor burden is minimal, makes them candidates for possible adjuv ant therapy and/or investigational phase II clinical trials. This stud y builds on previously proposed models that predict the likelihood of early recurrence after radical prostatectomy. Materials and Methods: I n our analysis we evaluate age, race, prostatic acid phosphatase and n uclear grade with the established prognostic variables of pretreatment prostate specific antigen, postoperative Gleason sum and pathological stage. Results: After multivariable Cox regression analysis using onl y statistically significant variables that predicted recurrence we dev eloped an equation that calculates the relative risk of recurrence (Rr ) as: Rr = exp[(0.51 x Race) + (0.12 x PSA(ST)) + (0.25 x Postop Gleas on sum) + (0.89 x Organ Conf.). These cases are then categorized into 3 distinct risk groups of relative risk of recurrence of low (<10.0), intermediate (10.0 to 30.0) and high (>30.0). Kaplan-Meier survival an alysis of these 3 risk groups reveals that each category has significa ntly different risks of recurrence (p <0.05). This model is validated with an independent cohort of radical prostatectomy patients treated a t a different medical center by multiple primary surgeons. Conclusions : This model suggests that race, preoperative prostate specific antige n, postoperative Gleason sum and pathological stage are important inde pendent prognosticators of recurrence after radical prostatectomy for clinically localized prostate cancer. Race should be considered in fut ure models that attempt to predict the likelihood of recurrence after surgery.