|
Abstract:
|
Many clinical studies (e.g., cardiovascular outcome trials) investigate the effect of an intervention on multiple event-time outcomes. The most common approach is a so-called "composite" analysis using the time to the first component event. Other approaches include the win ratio/difference for ordered outcomes and the application of the Wei-Lachin test. However, the influence of the marginal and joint distributions of the component events on the operating characteristics of these methods has not been explored. Herein these operating characteristics are investigated under both the joint null hypothesis of equal marginal hazards and the alternative hypothesis of specific marginal differences using a bivariate exponential model with a shared frailty, under which these properties can be expressed in closed form. We show that the composite analysis and the win ratio/difference can provide biased tests of the joint null hypothesis of equal marginal hazards in the setting where the correlation of event times differs between groups, while the operating characteristics of the Wei-Lachin test are unaffected. The methods are illustrated using data from the DCCT/EDIC study.
|
