Background: The time course to PCSM is highly variable for men with high risk after prostatectomy (RP); a majority will die of other causes. Accurately stratifying PCSM risk can improve therapy decisions. We validated ability of the 22 gene Decipher genomic classifier (GC) to improve upon clinical variables to predict PCSM in men with high risk after RP. Methods: Men with adverse pathologic features (APF: pT3N0/N1, positive margins, or RP Gleason > 7) who underwent RP in 1988-2010 at Johns Hopkins, Cleveland Clinic, and Mayo Clinic (excluded men used to develop GC). Primary outcome: PCSM within 10 years of RP (PCSM10). We also analyzed men at very high PCSM risk: men with biochemical recurrence in < 2 years (BCR2), and men who developed metastasis after RP (MET). Logistic regression was used with GC dichotomized as < 0.6 (low-intermediate) vs. > 0.6 (high), adjusted for PSA, RP Gleason, and RP stage, or for CAPRA-S. AUC measured incremental improvement when adding GC to model of RP pathology. Results: 511 men (110 with PCSM10; 22%) had median follow-up 13.0 years. Risk of PCSM10 increased nearly 4-fold for high GC score (p < .0001), and increased AUC over RP pathology model from 0.767 to 0.804 (Table). GC score stratified PCSM10 risk from 11-46%. In men with BCR2 or MET, high GC increased PCSM10 risk 2-3 fold (p = .0007 or .005), with AUC of 0.743 and 0.671. GC similarly improved AUC when added to CAPRA-S. Conclusions: In the largest and longest followed sample to date, Decipher GC demonstrated clinically important prediction of PCSM at 10 years, independent of RP pathology features, in men at high risk due to APF, BCR2 or MET after RP.