Notably More Powerful Analyses of Thorough QT Crossover Trials

• Activity Number: 67
• Type: Topic Contributed
• Date/Time: Sunday, August 4, 2013 : 4:00 PM to 5:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Devan Mehrotra*+ and Li Fan and Xiaodong Li
• Companies: Merck and Merck Research Laboratories and Bristol Myers Squibb
• Keywords: Baseline ; Covariate ; Covariance Structure ; Crossover ; Mixed Model ; Repeated Measures

Abstract

A thorough QT trial is often conducted using a crossover design; in each treatment period, baseline and post-baseline QT and heart rate are determined for each subject. In the analysis, after a "correction" for potential heart rate changes, at each post-baseline time point, a confidence interval (CI) for the mean difference in the corrected QT values is calculated for placebo vs. moxifloxacin (active control) and placebo vs. new drug. The study is a success if the CIs indicate that (i) moxifloxacin is associated with a QT prolongation at one or more time points (assay sensitivity), and (ii) the new drug is not associated with a QT prolongation at any time point. We show that the probability of success can vary substantially depending on how the CIs are calculated. Specifically, we caution that a "full" crossover analysis spanning all time points with change from baseline as the dependent variable is notably less efficient than a by-time-point "partial" crossover analysis with the post-baseline response as the dependent variable and the sum of the corresponding period-specific baseline values as a covariate. An example and simulation results are used to reinforce the key message.