Inference of bioequivalence for log-normal distributed data with unspecified variances

Area of Science:

• Pharmacokinetics and Drug Development
• Statistical Methods in Pharmaceutical Research
• Bioequivalence Studies

Background:

• Bioequivalence (BE) ensures drug product equivalence based on pharmacokinetic (PK) parameters.
• Current methods like TOST and Bayesian analysis have limitations in assessing the impact of unspecified variances on BE.
• Log-normal distribution is often assumed for PK parameters, leading to analysis on logarithmically transformed data.

Purpose of the Study:

• To propose a novel likelihood approach for bioequivalence assessment that explicitly incorporates unspecified variances.
• To compare the proposed method with existing Two One-Sided Tests (TOST) and Bayesian methods.
• To enhance understanding of how PK parameter variability influences BE determination.

Main Methods:

• Developed a likelihood approach partitioning the likelihood function into variance (F-statistic) and parameter difference (t-statistic) components.
• Retained unspecified variances within the statistical model.
• Applied the method to published real-life pharmacokinetic data.

Main Results:

• The proposed likelihood method yielded results consistent with TOST and comparable to Bayesian methods.
• The approach successfully identified ranges of variances influencing bioequivalence determination.
• Demonstrated that the method provides insights into the effect of unspecified variances on BE, surpassing TOST and Bayesian approaches.

Conclusions:

• The proposed likelihood method offers a more comprehensive assessment of bioequivalence by accounting for unspecified variances.
• This approach improves the understanding of how pharmacokinetic parameter variability affects BE outcomes.
• The method facilitates more achievable and informative bioequivalence determinations, especially in the presence of significant variability.