620 – Dose-Finding for Monotherapy and Combination Therapy in Oncology and Other Complex Studies
Design of Phase I/II Drug Combination Cancer Trials Using CRM and Adaptive Randomization
Matthieu Clertant
Lancaster University
Mourad Tighiouart
Cedars-Sinai Medical Center
We propose a design for a phase I/II cancer clinical trial in the context of drug combinations. The goal is to determine therapeutically efficacious dose combinations that are safe. In the first phase, univariate continual reassessment method (CRM) is used to allocate patients to dose combinations and only the toxicity outcomes are used to update the dose-toxicity model. In the second phase, toxicity and efficacy outcomes are used to update the joint model sequentially. At each step of the algorithm, an admissible set of safe doses is estimated using the posterior distribution of the model parameters and adaptive randomization is used to allocate patients to doses in that set. Since more than one dose combination can be recommended at the end of the trial, we propose a new estimator to assess the reliability of the recommended doses. We study the performance of the method by deriving the operating characteristics with two versions of the joint toxicity and efficacy model using scenarios under model misspecification.