TRANSITION ZONE CANCERS UNDERMINE THE PREDICTIVE ACCURACY OF PARTIN TABLE STAGE PREDICTIONS

Abstract

Purpose: The Partin tables represent the most widely used predictor of pathological stage in men with localized prostate cancer (PCa). The accuracy and performance of the tables have been tested across different populations. However, to our knowledge the potential limitations that may stem from differences between transition zone (TZ) and peripheral zone (PZ) prostate cancers has not been explored. We tested the predictive accuracy and performance of the Partin tables according to TZ vs PZ tumor predominance. Materials and Methods: Preoperative serum prostate specific antigen, clinical stage and biopsy Gleason sum data on 1,990 patients treated with radical retropubic prostatectomy were used to define the 2001 Partin probabilities of organ confinement and seminal vesicle invasion (SVI). Data on 1,320 patients who underwent staging pelvic lymphadenectomy and radical retropubic prostatectomy were used to define the probabilities of lymph node invasion (LNI) and organ confined disease (OC). ROC area under the curve was used to assess the predictive accuracy of the 2001 Partin tables relative to observed extracapsular extension (ECE), SVI, LNI and OC. Performance characteristics for each prediction were explored graphically with local regression, nonparametric smoothing plots. Results were compared between 222 TZ cancers and 1,768 PZ cancers. Results: The 1,990 radical retropubic prostatectomy specimens demonstrated ECE in 689 cases (34.6%) (TZ in 58 or 27.1% and PZ in 631 or 35.8%) and SVI in 224 (TZ in 13 or 6.1% and PZ in 211 or 11.9%). The 1,320 lymphadenectomy specimens demonstrated LNI in 56 cases (TZ in 2 or 0.9% and PZ in 54 or 4.6%). OC was found in 784 cases (59.4%) (TZ in 95 or 69.9% and PZ in 689 or 58.2%). Predictive accuracy was for ECE 76.4% (TZ 69.0% and PZ 77.2%), 78.0% for SVI (TZ 73.5% and PZ 78.3%), 78.6% for LNI (TZ 44.5% and PZ 79.9%) and 79.4% for OC (TZ 73.8% and PZ 80.0%). Conclusions: The biological tumor characteristics of TZ PCa differ from those of PZ PCa. These differences appear to undermine the accuracy of pathological stage predictions.