A 17-gene biopsy-based reverse transcription polymerase chain reaction assay, which provides a Genomic Prostate Score (GPS-scale 0-100), has been validated as an independent predictor of adverse pathology and biochemical recurrence after radical prostatectomy (RP) in men with low- and intermediate-risk prostate cancer (PCa). To evaluate GPS as a predictor of PCa metastasis and PCa-specific death (PCD) in a large cohort of men with localized PCa and long-term follow-up. A retrospective study using a stratified cohort sampling design was performed in a cohort of men treated with RP within Kaiser Permanente Northern California. RNA from archival diagnostic biopsies was assayed to generate GPS results. We assessed the association between GPS and time to metastasis and PCD in prespecified uni- and multivariable statistical analyses, based on Cox proportional hazard models accounting for sampling weights. The final study population consisted of 279 men with low-, intermediate-, and high-risk PCa between 1995 and 2010 (median follow-up 9.8 yr), and included 64 PCD and 79 metastases. Valid GPS results were obtained for 259 (93%). In univariable analysis, GPS was strongly associated with time to PCD, hazard ratio (HR)/20 GPS units=3.23 (95% confidence interval [CI] 1.84-5.65; p<0.001), and time to metastasis, HR/20 units=2.75 (95% CI 1.63-4.63; p<0.001). The association between GPS and both end points remained significant after adjusting for National Comprehensive Cancer Network, American Urological Association, and Cancer of the Prostate Risk Assessment (CAPRA) risks (p<0.001). No patient with low- or intermediate-risk disease and a GPS of<20 developed metastases or PCD (n=31). In receiver operating characteristic analysis of PCD at 10 yr, GPS improved the c-statistic from 0.78 (CAPRA alone) to 0.84 (GPS+CAPRA; p<0.001). A limitation of the study was that patients were treated during an era when definitive treatment was standard of care with little adoption of active surveillance. GPS is a strong independent predictor of long-term outcomes in clinically localized PCa in men treated with RP and may improve risk stratification for men with newly diagnosed disease. Many prostate cancers are slow growing and unlikely to spread or threaten a man's life, while others are more aggressive and require treatment. Increasingly, doctors are using new molecular tests, such as the17-gene Genomic Prostate Score (GPS), which can be performed at the time of initial diagnosis to help determine how aggressive a given patient's cancer may be. In this study, performed in a large community-based healthcare network, GPS was shown to be a strong predictor as to whether a man's prostate cancer will spread and threaten his life after surgery, providing information that may help patients and their doctors decide on the best course of management of their disease.