Competing Risk Regression Models

Vaishali Halani, Manish Thaker

Abstract: Competing risks are common in clinical research, as patients are subject to multiple potential failure events, both diseases related and otherwise. Competing risks methodology is being increasingly applied to cause of mortality data as a way of obtaining real world probabilities of mortality broken down by specific causes. For example, cancer patients with cardiovascular and other comorbidities are at concurrent risk of multiple adverse events. Regression models are employed to understand and exploit the relationship between the lifetime variable and the covariates. The most widely used regression models in competing risks are proportional because specific hazard model and proportional sub distribution hazard model. These models are frequently used in literatures and many authors have tried to differentiate and interpret both the models in different way. Several modeling approaches are available to evaluate the relationship of covariates to cause-specific failures with competing risk. Depending on which model is used, a distinctly different picture of the relationship of covariates to outcomes may be seen. It is important to choose a modeling approach that addresses the question of interest and subsequently interpret the results accordingly. We compared cause specific hazard model and sub distribution hazard model with flexible regression model to analyze and predict competing risk data in clinical trial applications using R software. These models are useful for a detailed analysis of how covariate effects predict the cumulative incidence, and allows for a time-varying effect of the covariates. From the above comparison, we can say that the choice of method for competing risk data in clinical trial should be guided by the scientific question.

Keywords: survival analysis, competing risks, regression model, flexible regression