Defining the window of opportunity in screening for prostate cancer: Validation of a predictive tumor classification model

BACKGROUND. Subdividing cancers according to the natural course of disease, both at the time of diagnosis and after radical prostatectomy, may influen ce management decisions of patients with prostate cancer. We investigated w hether categorization of prostate cancers into different prognostic subgrou ps is feasible. METHODS. In 218 screened participants of a randomized study, conventional p ost-operative tumor features were assessed fur their accuracy in predicting PSA relapse after radical prostatectomy using Cox regression analysis. Ind ependent prognostic tumor features were combined to identify subsets of can cers with similar biological potential. A cancer was defined that may be cu rable after its detection by screening tests, though is destined to progres s to clinically manifest disease and cancer-related mortality in the absenc e of screening. RESULTS. After a median follow-up of 33.0 months, pathological stage (P = 0 .03), tumor volume (P = 0.04), and margin status (P = 0.01) each independen tly predicted PSA relapse after surgery. The proportion of poorly different iated cancer proved highly superior to the Gleason score and most strongly predicted PSA relapse (P <0.0001). Based on combined independent prognostic tumor features, a turner classification model powerfully predicted PSA rel apse. CONCLUSIONS. Based on tumor characteristics, possibly harmless, and convers ely, possibly non-curable disease, may be distinguished from cancers that a re likely to show clinical progression in the absence of screening and trea tment. Prediction of these subclasses prior to treatment may eventually lea d to proper patient management. Prostate 46:154-162, 2001. (C) 2001 Wiley-L iss, Inc.