Online Program

Return to main conference page
Tuesday, September 24
Tue, Sep 24, 1:15 PM - 2:30 PM
Thurgood Marshall West
Optimizing Study Designs for Bioequivalence and Biosimilar Studies

A Two-Stage Adaptive Clinical Endpoint Bioequivalence (BE) Study with Sample Size Re-Estimation (300970)

*Wanjie Sun, FDA 

Keywords: Adaptive Design, Bioequivalence, Sample size re-estimation

A clinical endpoint bioequivalence (BE) study is often used to establish bioequivalence (BE) between a locally acting generic drug (T) and reference drug (R), where a pharmacokinetic (PK) study is uninformative. Typical study design is a double-blind, randomized three-arm (T, R and placebo: P) parallel clinical trial. BE between T and R can be established if two superiority tests (T vs. P, R vs. P) and one equivalence test (T vs. R) all pass. During study design, accurate estimate of the nuisance parameter (e.g. variance of the outcome variable) is essential in determining an adequate sample size to attain sufficient power. However, due to variations in the study design and procedures between NDA and Abbreviated NDA studies, as well as the high variability of clinical endpoints, the nuisance parameter is often under or over-estimated, resulting in unnecessary extra costs with a too large sample size or insufficient power with a too small sample size. Recently, Potvin et al (2017) proposed a two-stage adaptive design which allows sample size re-estimation based on the estimated variance from the first stage data. Potvin et al, however, only evaluated power and type 1 error of an equivalence test (T vs R), and did not assess family-wise type 1 error or power by incorporating the two superiority tests (T vs P, R vs P). In this work, we propose a two-stage clinical endpoint BE study with four sample size re-estimation approaches, which evaluate both individual and family-wise type 1 error and power. The four sample size re-estimation approaches are: 1) lumped variance with a naïve t test; 2) lumped variance with a stratified t test; 3) Gould’s variance estimate with a naïve t test; 4) pooled variance with a naïve t test. Unrestricted design (i.e., re-estimated sample size can increase or decrease from the original sample size) is adopted. Normal distribution is assumed with homogeneous or heterogeneous variances among treatment groups. Simulation studies are conducted