[PAST EVENT] Mathematics Colloquium and EXTREEMS-QED Lecture: Xue, Xiaoqiang (Quintiles)
Access & Features
- Open to the public
Title: Optimal Design of experiments with the observation censoring driven by random enrollment of study subjects
In clinical studies with time-to-event (e.g. death) as primary endpoint, beyond the uncertainties associated with observed endpoints, the operation process such as enrollment could cause significant amount of uncertainty. Subject arrivals at clinical trials are often staggered for lots of reasons, given the observed time to event, the exposure intervals or censoring windows are random. Unlike the traditional optimal design setting, the amount of information that can be gained during the execution of a planned study is random and becomes know only after completion of the study. I’ll discuss that for moderately large sample size the maximization of the average information is a sound strategy both theoretically and computationally. I will demonstrate through Poisson enrollment model for sake of simplicity purpose. The approach is based on the concept of elemental Fisher information matrix that allows the derivation of the very general results that work for a large family of survival models and convex optimality criteria.