Survival analysis example based on an event history model from a clinical trial in cardiology

Background: The main purpose of this paper is to present a survival analysi s example based on an Event history model in a competitive risks framework. Our example is derived from a clinical trial designed for comparing surviv al of a beta-blocker therapy group and a placebo group. Methods: Competitive risks under study were non lethal cardiovascular event s, permanent treatment withdrawals and deaths. The event history model was assumed to be semi-Markovian. The Cox proportional hazards model was used f or modeling effects of treatments and regression variables on the transitio n rates. Results: After taking into account all covariables influencing survival, mo rtality in patients who had one non lethal cardiovascular event after rando mization but remained treated was shown to be reduced in the beta-blocker t herapy group (RR=0.41; CI 95%=[0.17-0.98], p=0.04). Conclusions: Our analysis indicates that beta-blocker therapy should not ne cessarily be stopped in patients who experience a non lethal cardiovascular event under treatment. This finding require confirmation in a separate set ting.