Improving Clinical Trials Through Enrichment and Historical Controls

Abstract

Abstract In this dissertation, Bayesian adaptive design used to identify subgroup treatment effect is firstly explored. We investigate three Bayesian adaptive models for subgroup treatment effect identification: pairwise independent, hierarchical, and cluster hierarchical achieved via Dirichlet Process (DP). The impact of interim analysis and longitudinal data modeling on the personalized medicine study design is also explored. Interim analysis is considered since they can accelerate personalized medicine studies in cases where early stopping rules for success or futility are met. We apply integrated two-component prediction method (ITP) for longitudinal data simulation, and simple linear regression for longitudinal data imputation to optimize the study design. The designs’ performance in terms of power for the subgroup treatment effects and overall treatment effect, sample size, and study duration are investigated via simulation. We found that the hierarchical model with interim analysis and longitudinal modelling is an optimal approach to identifying subgroup treatment effects, and the cluster hierarchical model with interim analysis and longitudinal imputation is an excellent alternative approach in cases where sufficient information is not available for specifying the related priors. We then investigate several Bayesian designs incorporating historical control borrowing: power prior via overlapping area, commensurate prior, and some other methods. The impact of historical data type and different types of the threshold used in Bayesian decision rule are also explored. The designs’ performance in terms of power as a function of treatment effect, sample size, and posterior summary are investigated via simulation. It was found that it is a good consideration to apply the power prior adaptive design with power parameter determination via overlapping area of posterior distribution under certain values of true response rates of concurrent control, historical control, and treatment effect. Study design with commensurate prior is an admissible choice as well, however, appropriate priors need to be specified. Lastly, we use logistic regression and classification and regression tree (CART) models to identify the risk factors of early preterm birth (ePTB) from maternal perspective based on birth data from Center for Disease Control (CDC) and National Center for Health Statistics (NCHS)’ 2014 Natality public file. It revealed that the subgroup with a preterm birth history and a race designation as Black had the highest risk for ePTB. Those findings can provide valuable information for a future enrichment trial design. Moreover, both models can be applied to identify risk factors for other studies.