Model averaging inconcentration-QT analyses.

Area of Science:

• Pharmacometrics
• Statistical modeling
• Clinical trial analysis

Background:

• Concentration-QT analyses are crucial in thorough QTc studies to assess drug-induced cardiac repolarization changes.
• Traditional methods, like linear models, can be sensitive to model misspecification.
• Robust statistical approaches are needed for reliable assessment of drug effects on QT interval.

Purpose of the Study:

• To evaluate the performance of a frequentist model averaging (FMA) approach for concentration-QT analyses.
• To compare the FMA approach with the classical model selection (MS) approach in thorough QTc studies.
• To assess the robustness of FMA against type I error control and model misspecification.

Main Methods:

• Simulations were conducted using three candidate model families: linear, exponential, and Emax.
• Frequentist model averaging was applied to combine estimates from these candidate models.
• Type I error rates and treatment effect estimates were evaluated across various simulated scenarios.

Main Results:

• The model averaging approach demonstrated robust control of type I error across most simulated scenarios.
• FMA showed less sensitivity to model misspecification compared to the widely used linear model.
• Performance differences between FMA and MS were minimal, but FMA offers advantages due to MS deficiencies.

Conclusions:

• Frequentist model averaging is a recommended approach for concentration-QT analyses in thorough QTc studies.
• The FMA approach provides reliable treatment effect estimates and superior type I error control.
• Systematic consideration of model averaging or model selection is advised for concentration-QT analyses.