Use of the Average Baseline Versus the Time-Matched Baseline in Parallel Group Thorough QT/QTc Studies

Abstract

Using historical studies, we compared the impact of using the average baseline or time-matched baseline on diurnal effect correction, treatment effect estimation, and analysis of variance/covariance (ANOVA/ANCOVA) efficiency in a parallel thorough QT/QTc (TQT) study. Under a multivariate normal distribution assumption, we derived conditions for achieving unbiasness and better efficiency when using the average baseline, and confirmed these conditions using historical TQT studies. Furthermore, simulations were conducted under the randomized trial with and without observed imbalanced baseline settings. We conclude that the analyses using average baseline yield better efficiency and unbiased or less biased results under our TQT study conditions.

Key Words:

• ANCOVA
• ANOVA
• Bias
• Diurnal effect correction
• Efficiency

ACKNOWLEDGMENTS

The authors thank the Sanofi-Aventis ECG working group, Lynn Wei, Bernard Sebastien, Catherine Ortemann-Renon, Denis Bonnet, Professor Gary Koch, and the reviewer for their helpful comments.

Notes

Note: All studies had about 20 to 30 subjects in each treatment group.

Note: All studies had about 20 to 30 subjects in the placebo group.

Note. Each cell is based on 5000 simulations.

Note. Negative baseline imbalance: positive control baseline < placebo baseline. Positive baseline imbalance: positive control baseline > placebo baseline. Each cell is based on 500 simulations.

Note. LTMMD: largest time-matched mean difference. Stderr = mean of the treatment difference standard error from the model. For the balanced baseline scenario, each cell is based on 50,000 bootstrap samples; for imbalanced baseline scenarios, each cell is based on 5000 bootstrap samples.