Model-based bioequivalence approach for sparse pharmacokinetic bioequivalence studies: Model selection or model...

Area of Science:

• Pharmacokinetics
• Drug Development
• Statistical Modeling

Background:

• Bioequivalence (BE) studies assess drug product equivalence.
• Conventional methods like non-compartmental analysis (NCA) are challenging for sparse data.
• Model-based (MB) approaches are recommended but risk type I error due to model misspecification.

Purpose of the Study:

• To compare model selection (MS) and model averaging (MA) for model-based BE (MBBE) studies with sparse PK sampling.
• To evaluate the performance of MB-TOST using MS and MA against using a single specified model.

Main Methods:

• Simulated pharmacokinetic (PK) data from a two-way crossover BE study design.
• Applied model-based two one-sided test (MB-TOST) using candidate models, MS, and MA.
• Evaluated type I error rates and statistical power under null and alternative hypotheses.

Main Results:

• Model selection and model averaging controlled type I error rates at or below 0.05.
• These approaches achieved similar or higher statistical power compared to using the true PK model.
• Performance was robust even when the true model was not in the candidate pool.

Conclusions:

• Model selection prior to MB-TOST is proposed for BE studies with sparse PK data.
• Model averaging is recommended when candidate models exhibit similar Akaike information criterion values.
• These methods enhance the reliability of BE assessments for complex drug formulations.