A Comparative Study of the Estimation of the Maximum Tolerated Dose

• Activity Number: 556
• Type: Topic Contributed
• Date/Time: Wednesday, August 1, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Sujit Kumar Ghosh*+ and Seung Jun Shin
• Companies: North Carolina State University and North Carolina State University
• Keywords: Bernstein polynomial ; dose-response model ; toxicity probability function ; semiparametric estimation

Abstract

Dose-response analysis has been a very popular topic within statistical estimation especially in biomedical areas. In addition to estimating the entire dose-response relationship, assessing the risk itself at a certain dose level has been of much interest in some applications, such as in estimating the maximum tolerated dose (MTD) in Phase I clinical trials. Although a variety of parametric dose-response models have been proposed and successfully used in practice, many of the exiting models may not be appropriate in terms of capturing the underlying shape of the dose-response curve. As the estimation of the MTD depends crucially on the assumed dose-response curve, both parametric and nonparametric methods are explored and new flexible model is proposed. In addition, both frequentist and Bayesian MTD estimation methods are compared. It is shown via empirical studies that the proposed MTD estimation methods outperform the popular log-logit model under the violation of logistic assumptions and are competitive and robust against parametric model misspecifications.