A NEW MODEL FOR PRE-OPERATIVELY PREDICTING EXTRA-PROSTATIC EXTENSION OF CLINICALLY LOCALIZED PROSTATE CANCER.
Major Paul A. Friedrichs, MC, USAF, Scott Optenberg, Ph.D., COL Ian M. Thompson Jr., MC. USA, LTC Judd Moul, MC, USA, Eric Klein, M.D., and David Crawford, M.D. Wilford Hall USAF Medical Center, Lackland AFB, TX, Brooke Army Medical Center, Ft. Sam Houston, TX, Walter Reed Army Medical Center, Washington, D.C., Cleveland Clinic, Cleveland, OH, and University Of Colorado, Denver, CO, Presentation to be made by Dr. Friedrichs.

   Introduction and Objectives: To develop a mathematical model to predict the presence of extra prostatic disease pre-operatively using a patient’s age, race, PSA, biopsy grade, clinical stage and whether or not neoadjuvant hormonal therapy (NHT) was used.

   Materials and Methods: Our model was derived from logistic regression analysis, receiver-operating-characteristic (ROC) curves and Hosmer Lemoshow tests (HLT) of the outcomes of 1,026 men who underwent radical prostatectomy (RP) at three institutions (two military and one civilian). Results were validated using the outcomes of 400 men who underwent RP at an independent civilian site.

   Results: Our model’s performance was equal, or improved when applied to the external population. Its predictive capacity significantly surpassed that of recently published nomograms using the HLT. NHT was not a predictor of a lower risk of extra-prostatic disease. Controlling for grade, clinical stage, age, and P/s/a. Hispanic patients had a 321% greater risk of extra-prostatic disease (ROC 71% HLT:n.s.) and 247% greater risk of seminal vesicle involvement (ROC: 72.7%, HLT:n.s.), and 980% greater risk of nodal metastases (ROC 78.2%, HLT:n.s.) compared to white patients. Black patients had an intermediate risk of extra-prostatic disease, compared to these two groups.

   Conclusions: 1) It is possible to accurately predict a patient’s risk of extra-prostatic disease. 2) Recently published nomograms do not adequately predict observed results in mixed race populations from multiple institutions. 3) Models will be made available for clinical use in counseling patients.