Predicting Prostate-Cancer-Specific Death: Is It In The Cancer?

4500 Background: Few resources exist with the ability to study tissue biomarkers and predict prostate cancer specific mortality. In this study we use a retrospective cohort with full clinicopathologic characterization, over 15 years of follow-up and biomarker studies performed over 10 years. Methods: From this cohort we built a tissue microarray (TMA) (640 patients) and immunostained markers of important prostate cancer (PCa) pathways. Analysis was perfomed in PCa, non-neoplastic and benign prostatic hyperplasia tissue using a standardized method. Slides were digitized, and quantified using an expression index (Intensity*Percentage) for nuclear/cytoplasmic expression and phosphorylated versus inactive forms. The list of biomarkers included: markers of proliferation (Ki-67) and apoptosis (TUNNEL and Caspase-3); survival pathways (AKT, SKP-2, GSK, FHKR, p27, PTEN, PIM2, C-MYC, FRAT, PKC and NFkB); hormonal regulation: DAX1, AR, ER, REG Gamma and SRC 1, 2, and 3. Stroma was represented through reactive stroma grading (RSG), perineural invasion (PNI) diameter, stromal caveolin, FGF4, and ps20. Other biomarkers were also included. A total of over 250 variables were analyzed. Results: An initial multivariate model of clinico-pathologic parameters showed that seminal vesicle invasion and Gleason score were significant predictors of PCa specific death. When all other biomarker variables were introduced into the model only 3 were independently predictive of PCa specific death: reactive stroma grading, PNI diameter percentage and REG gamma in non neoplastic tissues. No cancer markers were significant. Conclusions: This is the first study to introduce large number of tissue-based biomarkers in the study of prostate cancer specific death. Results suggest that the tumor microenvironment is critical in the determination of patients that will die of prostate cancer. It substantiates previous basic studies that indicate that the microenvironment, either as stroma or nerves, is pro-tumorigenic. We hypothesize that this might represent genetic variability in the host, which might regulate the ability of host response. The ability of predicting prostate cancer specific death might lie in the study of the host and its response, and not the cancer. No significant financial relationships to disclose.