Prediction of locoregional extension and metastatic disease in prostate cancer: a review

Despite efforts to enhance the accuracy of prediction of extraprostatic dis ease, approximately 40% of the men undergoing radical prostatectomy are fou nd at surgery to have non-organ-confined cancer. Predictive algorithms base d on multivariate regression analysis and neural networks are widely availa ble and are superior to our standard empirical methods of clinical staging. These algorithms have been validated in diverse and well-characterized pat ient groups. For enhancement of the predictive value, data input must be st andardized and improved input variables must be incorporated. In addition t o the three "classic" staging parameters, i.e., pretreatment prostate-speci fic antigen (PSA), biopsy pathology, and digital rectal examination, new va riables now show promise in predicting disease extent and may be integrated in future predictive models. This review focuses on our present methods fo r prediction of locoregional spread and distant metastases in men with clin ically localized prostate cancer.