A MULTIVARIABLE ANALYSIS OF CLINICAL FACTORS PREDICTING FOR PATHOLOGICAL FEATURES ASSOCIATED WITH LOCAL FAILURE AFTER RADICAL PROSTATECTOMYFOR PROSTATE-CANCER

Purpose: A multivariate analysis is used to determine the predictive v alue of pretreatment clinical indicators on pathologic features associ ated with local failure after radical prostatectomy in patients with p rostate cancer. Methods and Materials: A retrospective review of the p athologic findings of 235 patients with adenocarcinoma of the prostate treated between 1990 and 1993 with a radical retropubic prostatectomy was performed. The preoperative clinical data including the serum pro state specific antigen, clinical stage, Gleason sum, and endorectal ma gnetic resonance scan findings are used to identify patients prior to definitive treatment who would be at high risk for having pathologic f eatures associated with local failure at radical prostatectomy. Outcom e prediction curves are constructed from a logistic regression multiva riate analysis displaying the probability of pathologic involvement of the seminal vesicle, extracapsular disease, or positive surgical marg ins as a function of the preoperative prostate specific antigen and Gl eason sum for the cases when the endorectal magnetic resonance scan is positive, negative, or not included in the multivariate analysis. Res ults: Factors identified on multivariate analysis as significant predi ctors of seminal vesicle invasion include endorectal magnetic resonanc e scan findings (p < 0.0001), and preoperative prostate specific antig en (p = 0.017). Endorectal magnetic resonance scan findings (p = 0.001 6), preoperative prostate specific antigen (p = 0.0002), and Gleason s um (p < 0.0001) were significant predictors of extracapsular extension and preoperative prostate specific antigen (p < 0.0001) and Gleason s um (p = 0.03) were significant predictors of disease extending to the margins of resection. Clinical stage was not a significant predictor ( p > 0.05) of pathologic features associated with local failure on mult ivariate analysis. As a single modality, endorectal surface coil magne tic resonance imaging was accurate 93%, 69%, and 72% of the time for p redicting seminal vesicle invasion, transcapsular disease, and final p athologic stage, respectively. Failure to recognize microscopic penetr ation of the capsule found at the time of pathologic evaluation in a p rostate gland with a grossly intact capsule accounts for the majority (70%) of the staging inaccuracies. Conclusions: The use of the endorec tal surface coil magnetic resonance scan findings in conjunction with both the serum prostate specific antigen and Gleason sum improves the clinical accuracy of predicting those patients at high risk for clinic ally unsuspected extraprostatic disease. In particular, for the subgro up of patients with moderately elevated prostate specific antigen (> 1 0-20 ng/mL) and intermediate grade clinically organ confined prostate cancer [Gleason sum: 5-7] where the specificity of these tests to pred ict for occult extraprostatic disease is suboptimal, the additional in formation obtained from the endorectal coil magnetic resonance scan al lows the physician to definitively subgroup these patients into low an d high risk for seminal vesicle invasion or transcapsular disease.