Abstract:
|
This topic is motivated by the recent published white paper in NEJM: The Drug-Dosing Conundrum in Oncology – When Less is More, by Mirat Shah, et. al. together with FDA Guidance on Expansion Cohorts (Aug, 2018). Two statistical relevant items may deserve further discussions: 1. also include PK, PD, efficacy data in Bayesian methods for dose optimization; 2. randomization for the FIH trials. Although there are various Bayesian methods available for MTD dose finding, such as BLRM, BOIN, mTPI2, TiTE methods, etc. the limitation is due to mostly based on DLTs. However, since PK, PD, efficacy data also play key roles, should we develop more methods to re-enforce the robustness of the statistical modeling for dose selection. As for the randomization in FIH expansion cohorts, naturally there are quite a lot of challenges. Major part of it is due to the exploratory nature of shooting fast moving targets. How to work around it? Moreover, the sample sizes are usually small for FIH trials. Obviously, multiple dose-cohorts comparison in a randomized setting posts another challenge for dose optimization. Better statistical methods may deserve some further discussions as well.
|