SESSION D – The Annual Deming Conference on Applied Statistics

TITLE: Bayesian Methods for Sample Size Determinations
INSTRUCTORS: Sujit K. Ghosh, NCSU and Fei Wang, Boehringer Ingelheim
MODERATOR: Naitee Ting

Abstract:

Sample size determination is one of most widely used methods in bio-pharmaceutical applications. The primary goal is to determine minimal sample size that would achieve a desired sampling precision for statistical estimation and/or maintain a pre-specified Type-I and Type-II error rates for hypotheses testing. However, Bayesian methods are not usually designed to achieve such desired precision or control error rates despite the need for such desired characteristics from regulatory perspectives. On the other hand, the success of classical sample size determination methods crucially depends on finding a pivotal quantity which becomes increasingly difficult for general composite null hypothesis (e.g., bio-equivalence and non-inferiority tests) involving nuisance parameters. Modern clinical trial design features require simulations to show operating characteristics. Thus, a unified methodological framework is needed that not only provide theoretical guarantees for controlling desired level of errors but is also broadly applicable for composite null hypothesis. This tutorial will focus on presenting recent Bayesian methodologies for sample size calculation primarily for hypotheses testing framework.

In summary, the tutorial presents (i) a general Bayesian/Classical framework for sample size determination for estimation and testing hypotheses (ii) theoretical and numerical illustrations of controlling two types of error rates for hypotheses testing; and (iii) applications of the methodologies for a few popular clinical trials and some recent developments on adaptive methods. Selected software demos (R packages: BAEssd, BDP2, ph2bayes, gsbDesign, BayesPPD) will also be illustrated with working examples and use cases and supporting literature will also be provided.

Instructors’ Biography:

Fei Wang is a Senior Principal Data Scientist at Boehringer Ingelheim Pharmaceuticals, US. Her research interests are Bayesian design and modeling; Simulation-based Bayesian sample size determination; Translational medicine and biomarkers. She has more than 10 years’ experience in biopharmaceutical industry and 5 years’ experience in academic research and teaching. She worked in various therapeutic areas and supported submission projects. Before working in pharmaceutical industry, she was an Assistant Professor in the School of Public Health at Boston University for 5 years. She was in the Department of Health Policy and Management and worked in collaboration with the Center for Health Quality, Outcomes, & Economic Research. Her research interests are Bayesian hierarchical modeling in health policy, medical tests, and sample size determination. She did her postdoctoral work at Brown University after graduating from the Department of Statistics at the University of Connecticut.

Professor Sujit Kumar Ghosh earned a Ph.D. in Statistics from the University of Connecticut in 1996 and is currently a tenured faculty member at the rank of Full Professor in the Department of Statistics at North Carolina State University (NCSU) in Raleigh, NC, USA. He has over 22 years of experience in conducting, applying, evaluating and documenting statistical analysis of biomedical and environmental data. Prof. Ghosh is actively involved in teaching, supervising and mentoring graduate students at the doctoral and master levels. He has supervised over 35 doctoral graduate students and 5 post-doctoral fellows and he has served as a member on numerous other doctoral and master level committees.