Statistical Methods for Competing Risks Model
“Competing Risks” refers to the study of the time to event where there is more than one type of failure event. The distinct problem can be vital, since not only it can inform the patients what risks they are facing, but also it helps to select appropriate treatment for a particular patient. In Chapter 2 we introduce two methods, cause- specific hazard model and cumulative incidence function, to deal with the competing risks problem. In Chapter 3, we study the prognosis of different patterns of cancer recurrences using data from 209 patients who had surgical resection of pancreatic cancer at the Johns Hopkins Hospital between 1998 and 2007. We analyze different types of tumor recurrences and death as competing risks. We first apply Cox’s pro- portional hazard model to analyze the time from surgery to the composite endpoint of recurrence or death. We then analyze the nonparametric cumulative incidence func- tion under competing risks setting. The conditional cumulative incidence function given each event type will be presented to investigate whether the competing risks have different distribution patterns. Then, the cause-specific hazard model is applied to evaluate the effect of risk factors on the cause-specific hazards, and the results are compared with the conventional survival analysis that ignores the recurrence types. Finally, we use Cox’s proportional hazard model with time-dependent covariates to analyze the time from surgery to death. At last, we discuss implications of data analysis and future research.