Predicting risk of prostate specific antigen recurrence after radical prostatectomy with the Center for Prostate Disease Research and Cancer of the Prostate Strategic Urologic Research Endeavor databases

Abstract

Purpose: Biostatistical models to predict stage or outcome in patients with clinically localized prostate cancer with pretreatment prostate specific antigen (PSA), Gleason sum on biopsy or prostatectomy specimen, clinical or pathological stage and other variables, including ethnicity, have been developed. However, to date models have relied on small subsets from academic centers or military populations that may not be representative. Our study validates and updates a model published previously with the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE, UCSF, Urology Outcomes Research Group and TAP Pharmaceutical Products, Inc.), a large multicenter, community based prostate cancer database and Center for Prostate Disease Research (CPDR), a large military database.

Materials and methods: We validated a biostatistical model that includes pretreatment PSA, highest Gleason sum on prostatectomy specimen, prostatectomy organ confinement status and ethnicity, including white and black patients. We then revised it with the Cox regression analysis of the combined 503 PSA era surgical cases from the CPDR prospective cancer database and 1,012 from the CaPSURE prostate cancer outcomes database.

Results: The original equation with 3 risk groups stratified CaPSURE cases into distinct categories with 7-year disease-free survival rates of 72%, 42.1% and 27.6% for low, intermediate and high risk men, respectively. Parameter estimates obtained from a Cox regression analysis provided a revised model equation that calculated the relative risk of recurrence as: exponent (exp)[(0.54 x Race) + (0.05 x sigmoidal transformation of PSA [PSA(ST)]) + (0.23 x Postop Gleason) + (0.69 x Pathologic stage). The relative risk of recurrence, as calculated by the aforementioned equation, was used to stratify the cases into 4 risk groups. Very low-4.7 or less, low-4.7 to 7.1, high-7.1 to 16.7 and very high-greater than 16.7, and patients at risk had 7-year disease-free survival rates of 85.4%, 66.0%, 50.6% and 21.3%, respectively.

Conclusions: With a broad cohort of community based, academic and military cases, we developed an equation that stratifies men into 4 discrete risk groups of recurrence after radical prostatectomy and confirmed use of a prior 3 risk group model. Although the variables of ethnicity, pretreatment PSA, highest Gleason sum on prostatectomy specimen and organ confinement status on surgical pathology upon which the model is based are easily obtained, more refined modeling with additional variables are needed to improve prediction of intermediate risk in individuals.