External Validation of the Prostate Cancer Specific Comorbidity Index: A Claims Based Tool for the Prediction of Life Expectancy in Men with Prostate Cancer.

Abstract

Accurate assessment of life expectancy is critical to treatment decision making in men with prostate cancer. We sought to externally validate the PCCI (Prostate Cancer Comorbidity Index) to predict long-term mortality in men with prostate cancer and make it operational using claims data. We performed an observational study of 181,009 men with prostate cancer in the Veterans Affairs Health System who were diagnosed from 2000 to 2013. Overall mortality across the PCCI scores was analyzed using Kaplan-Meier and Cox proportional hazards analysis. Discrimination and calibration were measured using the C-index and the mean prediction error, respectively. Among men with a PCCI score of 0, 1-2, 3-4, 5-6, 7-9 and 10 or greater the 10-year overall mortality rate was 15%, 26%, 36%, 41%, 52% and 69%, respectively. Multivariable Cox analysis showed an increasing hazard of mortality with higher PCCI scores, including 1.22 (95% CI 1.18-1.27), 1.69 (95% CI 1.61-1.76), 2.08 (95% CI 2.00-2.17), 2.88 (95% CI 2.76-3.00) and 4.50 (95% CI 4.32-4.69) for a score of 1 to 2, 3 to 4, 5 to 6, 7 to 9 and 10 or greater, respectively. The C-index to predict overall mortality was 0.773. The mean absolute error to predict 10-year overall mortality was 0.032. Of the men with clinically localized disease, Gleason 6 or less with less than 10-year life expectancy and Gleason 7 or less with life expectancy less than 5 years as defined by the PCCI 3,999 of 12,185 (33%) and 1,038 of 3,930 (26%), respectively, underwent definitive local treatment. The PCCI is a claims based, externally validated tool to predict mortality in men with prostate cancer. Integrating the PCCI into clinical pathways may improve prostate cancer management through more accurate assessment of life expectancy.