Inappropriate Covariance Structure May Lead to Smaller Spurious Variance Estimation of Drug Effect in TQT Crossover Study

• Activity Number: 246
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Li Fan*+ and Xiaodong Li
• Companies: Merck and Merck
• Keywords: Variance-covariance ; Crossover ; TQT

Abstract

Providing an evaluation of a new drug's effect on QT prolongation has become a routine requirement by regulatory agency. The recommended QTc study design is a crossover in 3 or 4 treatment including placebo, active control and supratherapeutic (therapeutic) doses of the test drug. Followed by ICH E14 guidance, a non-inferior hypothesis test of a new drug versus placebo is conducted. A linear mixed effect model for repeated measurement analysis is regarded as an appropriate approach with considering the correlation of measurements from the same subject. But variance-covariance structure selection in the above analysis is a concern. Due to potential effect of variance-covariance structure selection on the variance estimation of drug effects and power calculation, we reconfirm the "dual" structure, which means stronger correlation between two measurements under same treatment and weaker correlation between two measurements under different treatments, and explore the effects of the "single" structure with ignoring different magnitude of correlation between two measurements under same/different treatment.