New Statistical Designs for Drug-Combination Trials and Basket Trials

Professor Yin, Guosheng   (Principal Investigator (PI))

36

2018-07-01

445645

New Statistical Designs for Drug-Combination Trials and Basket Trials

Basket trial, Bayesian adaptive design, Dose finding, Phase I clinical trial, Phase II clinical trial

Probability & Statistics,Clinical Trials

Physical Sciences (P)

17307318

General Research Fund (GRF)

2018

Completed

1 We will modify the nonparametric overdose control design by introducing two new basis models for trial early stopping due to under- or over-dosing and developing a new sampling scheme for ordered uniform random variables. We further adopt the bootstrap aggregating ensemble method to boost the estimation of the order in the two-dimensional drug combination space. 2 We propose a new dose-finding method for cytostatic agents in phase I/II trials by constructing curve-free models for both toxicity and efficacy and introducing correlations through latent variables. The dose finding procedure is cast in a Bayesian model selection framework. 3 We will investigate basket trials by developing a two-stage trial design with futility stopping and homogeneity testing at the interim stage for one drug but with multiple biomarkers. In addition, we will explore the case with multiple drugs and multiple biomarkers.