Title: Estimands, Missing Data Handling and Bayesian Methods for Clinical Trials
Instructors: Frank Liu, Merck & Co. Inc and Fang Chen, SAS®
Moderator: Ivan S. F. Chan
Missing data often occur in longitudinal and survival clinical trials due to early discontinuation. The problem is closely associated with the estimand framework and assumptions about missing data and censoring. How to handle the missing data is critical in analysis of clinical trials. Methods such as likelihood-based approaches, multiple imputation, and Bayesian approaches are often considered. The course will provide overview of estimand frameworks, missing data methods for longitudinal and survival trials, and Bayesian approaches for handling missing data as well as for borrowing historical data, and illustrate how to implement these analyses methods using SAS.
• Review of estimands, concept of intercurrent events vs missing data
• Missing data methods for longitudinal trials o maximum likelihood methods: MMRM, cLDA o multiple imputation o generalized estimation equation approaches: GEE, wGEE o Repeated binary endpoints
• Introduction of Bayesian Method o Bayesian software o PROC MCMC o PROC BGLIMM o Examples
• Sensitivity analysis methods o General methods for missing not at random o More recent approaches
• Control-based imputation
• Control-based mean imputation
• Baseline-based imputation
• Sensitivity analysis methods (cont.) o More recent approaches
• Delta-adjustment methods and tipping point analysis
• Dropout as poor outcome o Bayesian sensitivity analysis
• Missing data methods for survival trials o Censoring issue in survival trials o Sensitivity analysis methods
• Control-based imputation
• Delta-adjustment methods
• Tipping point analysis o Bayesian sensitivity analysis
• Information borrowing with Bayesian
Dr. G. Frank Liu is a distinguished scientist at Merck & Co., Inc. and a Fellow of the American Statistical Association. For more than 26 years at Merck, Frank has worked on various therapeutical areas and conducted research in longitudinal trials, missing data, noninferiority trials, and Bayesian methods; and served as a technical consultation and leading the development of many methodological guidance documents. Before joining Merck, he received his PhD in statistics from UCLA and completed a post-doc in Biostatistics at Johns Hopkins University.
Dr. Fang Chen is Director of Analytical Software Development at SAS Institute Inc. and a Fellow of the American Statistical Association. He manages the development of statistical software for SAS/STAT®, SAS/QC®, and analytical components that drive SAS® Visual Statistics software. Also among his responsibilities is the development of Bayesian analysis software and the MCMC procedure. Before joining SAS, he received his PhD in statistics from Carnegie Mellon University.