Optimal Dose Allocation Considering Toxicity and Efficacy Jointly in Phase I/II Dose Finding Studies

Novel anticancer therapies are challenging the standards of drug development in Phase I dose-finding studies. Agents with specific biologic targets, unknown dose-efficacy trends, and limited toxicity motivate innovative designs to identify optimal biological dose (OBD), the dose jointly defined by toxicity and efficacy probabilities. In this thesis, a novel design to identify OBD is developed by combining optimal design theory with the continual reassessment method (CRM). The optimal design theory is implemented based on the continuation ratio model and straightforward OBD selection criteria. To better fit practical needs, the new design is progressed to identify adjusted-OBD with modified criteria. The possible candidates of OBD and adjusted-OBD are proposed. The C-optimal designs or variance functions of all candidates of both OBD and adjusted-OBD are demonstrated. Our simulation studies show that, under a wide range of dose-outcome scenarios compared to Isotonic, L-logistic, Logistic, EffTox and TriCRM, the proposed design has high probabilities in correctly recommending OBD when the assumed model is true. Under extensive settings of adjusted-OBD identification, our design keeps its outstanding performance in correctly recommending adjusted-OBD compared with TriCRM.