Improved Prediction of Long-Term, Other Cause Mortality in Men With Prostate Cancer

Purpose: Comorbidity assessment is essential to triage of care for men with prostate cancer. We identified long-term risks of other cause mortality associated with comorbidities in the Charlson index and applied these to the creation of a prostate cancer specific comorbidity index. Materials and Methods: We sampled 1,598 cases of prostate cancer diagnosed in 1997 to 2004 at the Greater Los Angeles and Long Beach Veterans Affairs Medical Centers. We used Cox proportional hazards modeling to determine the risks of other cause mortality associated with comorbidities and used these hazard ratios to re-weight the Charlson index. We then compared the ability of each index to predict other cause mortality. Results: Cox modeling showed that moderate to severe liver disease, metastatic solid tumor, lymphoma and leukemia carried the highest risk (HR greater than 5) for other cause mortality, followed by moderate to severe chronic obstructive pulmonary disease, moderate to severe renal disease, dementia, hemiplegia and congestive heart failure (HR 2.5 to less than 3.5). The revised and original Charlson indices performed similarly in predicting other cause mortality across all patients (c-index 0.816 vs 0.802). However, in survival analysis our revised index identified 137 men with a greater than 90% probability of other cause mortality within 10 years while the original Charlson identified only 51. In multivariate modeling the odds of 5-year other cause mortality for men with original Charlson scores 1, 2, 3 and 4+ were 2.9, 6.0, 9.2 and 29.8, respectively, compared with 3.9, 6.2, 12.8 and 84.2 for the revised index. Conclusions: Re-weighting the Charlson index allowed for more accurate identification of men at highest risk for other cause mortality. Our revised index may be used to aid medical decision making for men with prostate cancer.