Jun Vivien Yin, PhD

Title: Model-based Phase I Designs with Incorporation of Personalized Dosing Using Toxicity Scores from Multiple Treatment CyclesAbstract: Dose-finding designs based on binary dose-limiting toxicity (DLT) endpoints are substantially inefficient for targeted and cancer immunotherapy agents by failing to incorporate efficacy signals, the type and severity of adverse events, and information from late treatment cycles. We developed a novel phase I repeated measures design (RMD) that uses a quasi-continuous total toxicity profile from multiple treatment cycles during the dose-finding process. In studies where preliminary efficacy is evaluated, an adaptive, multi-stage design to identify the most efficacious dose with acceptable toxicity was also developed. Furthermore, we extended our work to incorporate a dose algorithm that allows patients in subsequent treatment cycles to be treated with an individualized dose that is tailored to their specific tolerance and the cumulative adverse effects of the drug. These designs were calibrated with respect to specific operating characteristics. We conducted extensive simulations to assess the performance of the proposed design with comparison to previous published work, as well as published an R package on the Comprehensive R Archive Network (CRAN), to facilitate the implementation of these designs in practice.