Discrete Time-to-Event and Rank-Based Methods with Application to Composite Endpoint for Assessing Evidence of Disease Activity

Abstract

This chapter describes discrete time-to-event and rank-based statistical analyses methods for estimation of and inference about treatment effects on a composite measure of evidence of disease activity. The composite measure is a function of endpoints with known event times and another endpoint whose exact time of occurrence is unknown, but the endpoint is only known to have occurred within an interval. Unlike crude incidence rate approach that ignores subjects’ differential follow-up times and makes unverifiable assumptions about evidence of disease activity status of censored subjects, discrete time-to-event approaches allow incorporation of subjects’ differential follow-up times and appropriate handling of censoring. The rank-based method captures severity of disease activity that may be eluded by the collapsed binary composite outcome. Moreover, the rank-based method, besides providing crude estimates of proportion of subjects with no evidence of disease activity, also addresses possible clouding effect of any of the component endpoint that occurs with higher frequency.