Practical Comparison of Sample Size Re-Estimation and Group Sequential Designs: Case Studies

• Activity Number: 339
• Type: Topic Contributed
• Date/Time: Tuesday, August 6, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): William Prucka*+
• Companies: Eli Lilly and Company
• Keywords: Adaptive design ; Group sequential design ; Sample size re-estimation

Abstract

The FDA issued a draft guidance on adaptive designs (AD) in 2010. The draft guidance categorized adaptive methods into 2 groups; well understood designs (e.g. group sequential design (GSD)), versus less well understood designs (e.g. unblinded sample size re-estimation (SSR)). While this categorization is based primarily on regulatory experience with each design type, and secondarily on differences in statistical methodology, there remains a preference for GSD over SSR. This talk focuses on the practical comparison of GSD and SSR through case studies; demonstrating that in real world applications SSR is often preferable. While both GSD and SSR have similar goals, to right-size a trial, they have different philosophies and statistical implications. Literature comparisons of GSD and SSR have been made in idealized conditions, ignoring practical aspects such as enrollment rate, impact on secondary endpoints, and other considerations. Using a combination of analytic methods and trial simulation, this talk compares SSR and GSD on efficiency metrics that consider these factors. Two case studies are provided, one where blinded SSR was chosen, and one where SSR was preferred over GSD.