Interpretation of exposure effect in competing risks setting under accelerated failure time models

In survival studies, incidence of competing risks causes that the time of event of interest to be unknown. Analysis of competing risk data, often implemented using hazard-based method under proportional hazard assumption. In this study, we interpreted covariate effect under accelerated failure time model and cause-specific survival function. Methods & Materials: We considered weibull hazard and survival function as cause-specific hazard and survival function and explored the relation between these function. Estimation of parameters performed using Bayesian methods with non-informative priors that implemented in R2WinBUGS package of R software. Simulation study showed that, the relation between hazard and survival parameters for weibull distribution is also established between parameters of cause-specific hazard and cause-specific survival function. This relation also verified in PBC data set for logarithm of serum bilirubin and D-penicillamine effect. Although in competing risk studies, most of the analysis performed under PH assumption, analysis based on AFT models will also be applicable for these data. In these setting, coefficients can be interpreted as effects of covariate on time to each event.