Drug Approval and Labeling Based on the Bayesian Approach

• Activity Number: 396 - Drug Approval and Labeling Based on Bayesian Approach
• Type: Invited
• Date/Time: Wednesday, August 5, 2020 : 1:00 PM to 2:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Deborah Ashby* and Frank Bretz* and Maria Ciarleglio* and Brian Hobbs* and Ying Yuan* and Jeni Zhou*
• Companies: the School of Public Health, Imperial College Londo and Novartis Pharma AG and Yale School of Public Health and University of Texas at Austin and The University of Texas MD Anderson Cancer Center and Amgen
• Keywords: Bayesian approach; drug approval standard; subjectivity; selection bias ; multiplicity control; heterogeneity

Abstract

Bayesian designs are expected to play important roles in satisfying the substantial evidence standard that is needed for approval or licensure for medical products and supporting broad labels for which trials will evaluate multiple disease indications for “agnostic” effects, where subpopulation analysis becomes central to the study design and should be considered in the evaluation of trial operating characteristics. Experts will discuss the following key controversial aspects of Bayesian designs and provide potential solutions so that they will have been thoroughly considered in the drug development. A. Maintaining the same drug approval standard, regardless of methods or theories of inferences B. Understanding prior specification with respect to effective sample size, implications for bias, and subjectivity in relation to existing evidence, especially when one expects results to be sensitive to choices of the priors C. Preventing overuse of Bayesian inference in drug approval and labeling, in terms of being theoretically immune to selection bias or not requiring multiplicity control D. Dealing with heterogeneity across different sources of data or different subpopulation