| Frequently in clinical trials, the endpoint of interest is the time to a composite outcome. A composite outcome consists of one or more types of primary event constituents and the time to the first event in this set is of interest. Although one can assume that event times will be reported by the time of the final analysis, one cannot necessarily assume that all events will have been recorded to the database at the time of interim monitoring. Events may be unreported, and thus, using the Kaplan-Meier estimate of the time to the first event will overestimate survival. Several methods exist to compensate for delayed event ascertainment and these may be applied to estimate survival time to a composite outcome if events are reported in order. However, if events are reported in a non- monotone fashion (out of order), a later event may be reported at the time of interim monitoring but the first event (the composite outcome of interest) may be unreported; hence, the earliest reported event time is not necessarily the true composite outcome time and these methods yield a biased estimate of survival. Additionally, many trials require primary endpoint adjudication, which may be incomplete at the time of interim monitoring; thus, the primary status of events may be unknown, additionally biasing the estimate. We propose an estimator that correctly estimates the time to the composite endpoint under these conditions and will illustrate using data from the Reduction of Events by Darbepoetin Alfa in Heart Failure (RED-HF) Trial. We will demonstrate the benefit of correct survival estimation on trial planning. |
