AN ADDITIVE SUBDISTRIBUTION HAZARDS MODEL FOR COMPETING RISKS DATA

Abstract

Competing risk failure time data occur frequently in medical a number of methods have been proposed for the analysis of these data. The classic approach is to model all cause-specific hazards and then estimate the cumulative incidence curve based on these cause-specific hazards. Unfortunately, the cause-specific hazard function does not have a direct interpretation in terms of survival probabilities for the particular failure type. In this paper, we consider a more flexible model for the subdistribution. It is a combination of the additive model and the Cox model and allows one to perform a more detailed study of covariate effects. One advantage of this approach is that our regression modeling allows for non-proportional hazards. This leads to a new simple goodness-of-fit procedure for the proportional subdistribution hazards assumption that is very easy to use. We applied this method to melanoma data and estimated the cumulative death rate for those who died from melanoma after surgical removal of the tumor. It was found that two covariates had a time-varying effect and two other covariates had a constant effect in predicting the cumulative incidence curve in patients who died of melanoma following tumor removal surgery.

Keywords

• additive hazard model
• competing risk
• cumulative incidence function
• subdistribution hazard