Abstract
CONTEXT: The widespread use of prostate-specific antigen (PSA) screening has led to the detection of more indolent prostate cancer (PCa) in healthy men. PCa treatment and screening must therefore balance the potential for life gained against the potential for harm. Fundamental to this balance is physician awareness of a patient's estimated life expectancy (LE). OBJECTIVE: To review the evidence on LE differences between men diagnosed with PCa and the general population. To examine clinician- and model-predicted LE and publicly available LE calculators. EVIDENCE ACQUISITION: A comprehensive search of the PubMed database between 1990 and September 2014 was performed according to Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines. Free text protocols of the following search terms were used "life expectancy prostate cancer", "life expectancy non-cancer", "non-cancer mortality prostate", and "comorbidity-adjusted life expectancy". Two internet search engines were queried daily for 1 mo for the search term "life expectancy calculator", and the top 20 results were examined. EVIDENCE SYNTHESIS: Of 992 articles and 32 websites screened, 17 articles and nine websites were selected for inclusion. Men with non-screening-detected PCa and distant disease at diagnosis were found to have shorter LE than age-matched peers, whereas men with localized PCa had prolonged LE. In general, clinician-predicted 10-yr LE was pessimistic and of limited accuracy; however, model-predicted LE provided only modest improvements in accuracy (c-index of models 0.65-0.84). Online LE calculators provide consistent LE estimates, but government life tables provide LE estimates near the mean for all calculators examined. CONCLUSIONS: The accuracy of clinician-predicted survival is limited, and while available statistical models offer improvement in discrimination, it is unclear whether they provide advantages over freely available government life tables. PATIENT SUMMARY: We examined differences in life expectancy between men diagnosed with prostate cancer and the general population, and ways of predicting life expectancy to help guide treatment decisions. We found that current models for predicting life expectancy specific to prostate cancer might not be any better than government life tables or simple rules of thumb.