Precision Medicine Designs for Cancer Clinical Trials

This chapter reviews Bayesian adaptive designs for precision oncology clinical trials. We focus on two aspects of precision medicine: how to more accurately capture the multi-dimensional nature of clinical outcomes, and how to more accurately account for patient’s heterogeneity. We use the Bayesian optimal interval phase I-II (BOIN12) design as an example to illustrate the first aspect. The BOIN12 design uses the utility function to explicitly account for the risk-benefit tradeoff of the treatment and identify the optimal biological dose (OBD), and thus more accurately reflects the multiple clinical effects of the treatment underlying clinical decisions in practice. We then use a novel Bayesian adaptive design to illustrate how to determine subgroup OBD by accounting for patient’s genomic profile and heterogeneity. The design first uses a dimension-reduction technique to reduce the dimension of genomic biomarkers, and then incorporates this condensed biomarker information into the clinical outcome model and adaptive decision making. In the presence of patient heterogeneity, another effective approach to achieving precision medicine is to use the enrichment strategy. We describe a novel Bayesian adaptive enrichment design that simultaneously achieves two primary goals of precision medicine: to identify the sensitive subgroup and to test if the treatment is effective in that subgroup. The design uses the interim data to continuously and adaptively identify the subgroup of patients who are expected to benefit from the treatment, and actively enriches such subgroup by adaptively modifying patient enrollment criteria during the trial. The design has substantially higher power than the conventional designs to detect the treatment effect and identify the target subpopulation.