PARAMETRIC LIKELIHOODS FOR MULTIPLE NONFATAL COMPETING RISKS AND DEATH

Clinical trials of fatal diseases often focus on one or more non-fatal events, in addition to survival, both to characterize morbidity and t o improve survival estimates. Three statistical complications are that the time to each non-fatal event and subsequent residual survival may be either positively or negatively associated, the times to death wit h or without an antecedent event often have very different distributio ns, and death may censor some of the non-fatal event times. Consequent ly, the overall survival time distribution is a mixture of the distrib utions corresponding to the possible antecedent non-fatal events. Thes e conditions violate the usual assumptions underlying many statistical methods for analysing multivariate time-to-event data. In this paper, we consider a general parametric model for multiple non-fatal competi ng risks and death. The model accounts for positive or negative associ ation between the time of each non-fatal event and subsequent survival while accommodating covariates and the usual administrative censoring . Each event time distribution is specified marginally by a three-para meter generalized odds rate model, and the time of each non-fatal even t and subsequent residual survival are combined under a bivariate gene ralized von Morgenstern distribution. The approach is illustrated by a pplication to two data sets from clinical trials in colon cancer and a cute leukaemia. (C) 1998 John Wiley & Sons, Ltd.