Considering dose reductions and dose interruptions in oncology dose -finding trial design

Within the past decade, the FDA has issued many post-marketing requirements
or commitments to study alternate doses in Molecularly Targeted Agents and Immunotherapies in reaction to a large percentage of Dose Reductions and Dose Interruptions, caused by chronic low-grade toxicities. Almost all of these trials used the Maximum Tolerated Dose (MTD) based on Dose Limiting Toxicity (DLT) information only. While extensions of traditional cytotoxic dose-finding designs address long-term toxicities as well as chronic low-grade toxicities, none has explicitly and appropriately estimated the probability of Dose Reduction and Dose Interruption. This dissertation introduces a novel modeling procedure to model Dose Reductions (and/or Dose Interruptions) and DLTs simultaneously, adopting a semi-competing risks framework for the first time in dose-finding. Simulations are presented to evaluate operating characteristics and design options, and to compare the proposed method to others likely to be used in the same setting. Important quantities and some theoretical results are derived. Additionally, Dose Reductions and DLTs affect the estimation of efficacy, which is necessary to evaluate in the current setting. No previous methods have addressed Dose Reductions in the context of early phase efficacy. A new efficacy estimand is defined for a tumor burden outcome to appropriately handle Dose Reductions and DLTs, based on Multiple Imputation. An illustration and discussion compare the proposed estimand with current practice. Finally, dose selection options are discussed using toxicity and efficacy information.