Analysis of Cause-Specific Events in Competing Risks Survival Data

Publisher Summary There has been long-standing interest in competing risks, as this situation occurs frequently in a large variety of areas, including industrial engineering, demography, econometrics, and a long history of applications to biology and health. The focus is on biomedical applications, and in particular, on the analysis of individuals under treatment for cancer, who may be subject to a number of subsequent events following diagnosis and initial treatment. The chapter discusses the consideration of functions that may be estimated from competing risks data without making assumptions that cannot be verified by these data. Specifically, it considers inferences concerning cause-specific hazard functions and cumulative incidence functions, to include a discussion of hypothesis tests for comparing the cause-specific events between two groups. To present these methods, some mathematical definitions and notation are introduced. The unique feature of competing-risk problems in survival analysis is that there are both competing risks that may be correlated with the outcome of interest and a censoring mechanism that often can be assumed to be independent of the outcome of interest and, generally, of the competing risks as well. Competing-risk methods based on estimable quantities are generally straightforward and yield additional important information when multiple failure types are present.