ABSTRACT
In studies where the outcome is death, individuals may die from one of a number of dierent causes. For example, in a study to compare two or more treatments for prostatic cancer, patients may succumb to a stroke, myocardial infarction or the cancer itself. In some cases, an analysis of death from all causes may be appropriate, and standard methods for survival analysis can be used. More commonly, there will be interest in how the hazard of death from dierent causes depends on treatment eects and other explanatory variables. Of course, death from any one of a number of possible causes precludes its occurrence from any other cause, and this feature has implications for the analysis of data of this kind. In this chapter, we review methods for summarising data on survival times for dierent causes of death, and describe models for cause-specic survival data.
