Optimal Bayesian Dose-Finding in Two Treatment Cycles based on the Joint Utility of Efficacy and Toxicity
Yuan Ji, University of Chicago
Keywords: Bayesian Statistics; Clinical Trial, Dose-finding; Dynamic Treatment Regime; Utility
A method is described for adaptively optimizing dose in each of two cycles of therapy based on efficacy and toxicity in a phase I-II clinical trial. The proposed method relies on a Bayesian multivariate hierarchical latent variable dose-outcome model. Each patient's dose in each treatment cycle is chosen by adaptively optimizing the posterior expectation of an objective function defined in terms of numerical utilities of the possible (toxicity, efficacy) outcome values in each cycle. At the end of the trial, the method identifies an optimal two-stage regime consisting of an optimal cycle 1 dose, and an optimal function of the patient’s cycle 1 dose and (efficacy, toxicity) outcomes that either chooses a cycle 2 dose or says to not treat the patient in cycle 2. This is very different from simply choosing an “optimal” dose for each cycle. Because interim decisions are based on posterior quantities, the method is adaptive both within and between patients. Comparisons of the method via computer simulations to 2-cycle extensions of the continual reassessment method and 3+3 algorithm show that the optimal two-cycle regime is superior, in some cases by a very wide margin.
Important Dates & Deadlines
- October 9 - 11, 2013