Competing risks of mortality in prostate cancer

Purpose: We have observed that in prostate cancer older age, black race, high stage and no treatment are independently associated with higher disease specific and overall mortality. Because disease nonspecific mortality was not examined one cannot reasonably infer that higher overall mortality is largely due to higher disease specific mortality. To understand better the interactions of independent prognostic variables with overall mortality we jointly evaluated their effects on type of death and time to death. Materials and Methods: Using the prostate cancer patients of the 1973 to 1990 public use tape of the Surveillance, Epidemiology and End Results program, we performed competing risks analysis with multivariate accelerated failure time model to examine if the prognostic factors were associated with type of death and time to death. Results: Older age, black race and no treatment were independently associated with higher relative risk of disease specific and nonspecific mortality. Localized stage was associated with lower and higher disease specific and nonspecific mortality than regional stage, respectively. The relative risks of disease specific and nonspecific mortality were significantly different from each other for all prognostic factors. Conclusions: To our knowledge, a competing risks analysis in prostate cancer has not been done previously. While all prognostic factors were associated with disease specific and nonspecific mortality, the relative contributions of disease specific and nonspecific mortality to overall mortality varied by prognostic factor, with some prognostic factors associated with relatively high disease specific mortality while others were associated with relatively high disease nonspecific mortality. The similarity of the associations of age, race and treatment with disease specific and nonspecific mortality suggests that they mark an unidentified factor(s) that affects patient health nonspecifically. Such a factor(s) may partly or completely explain the association of treatment with lower disease specific mortality.