|
Abstract:
|
Surrogate endpoints are attractive for use in clinical trials when a true endpoint is costly to obtain, difficult to measure, or requires a long period of follow-up to observe. Several methods exist for measuring the trial-level surrogacy of a potential time-to-event endpoint for a "true" time-to-event endpoint within a meta-analytic framework, but little has been done to investigate the potential effects of trial characteristics on the surrogacy measures. We perform a simulation study to measure the impact of several important factors, including number of trials, trial size, and censoring on a variety of trial-level surrogacy measures. Furthermore, we demonstrate the consequences of randomly splitting a small number of large trials into a greater number of smaller trials when performing a trial-level surrogacy analysis. Simulated data are generated to resemble overall survival (OS) and time-to-recurrence (TTR) events from a collection of Phase III adjuvant therapy clinical trials in colon cancer, based on the Adjuvant Colon Cancer Endpoints (ACCENT) data set.
|
