Effect of Comorbidity on Prostate Cancer–Specific Mortality: A Prospective Observational Study

PURPOSE: To determine the effect of comorbidity on prostate cancer (PCa)–specific mortality across treatment types. PATIENTS AND METHODS: These are the results of a population-based observational study in Sweden from 1998 to 2012 of 118,543 men who were diagnosed with PCa with a median follow-up of 8.3 years (interquartile range, 5.2 to 11.5 years) until death from PCa or other causes. Patients were categorized by patient characteristics (marital status, educational level) and tumor characteristics (serum prostate-specific antigen, tumor grade and clinical stage) and by treatment type (radical prostatectomy, radical radiotherapy, androgen deprivation therapy, and watchful waiting). Data were stratified by Charlson comorbidity index (0, 1, 2, or ≥ 3). Mortality from PCa and other causes and after stabilized inverse probability weighting adjustments for clinical patient and tumor characteristics and treatment type was determined. Kaplan-Meier estimates and Cox proportional hazards regression models were used to calculate hazard ratios. RESULTS: In the complete unadjusted data set, we observed an effect of increased comorbidity on PCa-specific and other-cause mortality. After adjustments for patient and tumor characteristics, the effect of comorbidity on PCa-specific mortality was lost but maintained for other-cause mortality. After additional adjustment for treatment type, we again failed to observe an effect for comorbidity on PCa-specific mortality, although it was maintained for other-cause mortality. CONCLUSION: This large observational study suggests that comorbidity affects other cause–mortality but not PCa-specific– mortality after accounting for patient and tumor characteristics and treatment type. Regardless of radical treatment type (radical prostatectomy or radical radiotherapy), increasing comorbidity does not seem to significantly affect the risk of dying from PCa. Consequently, differences in oncologic outcomes that were observed in population-based comparative effectiveness studies of PCa treatments may not be a result of the varying distribution of comorbidity among treatment groups.