A new approach for outliers in a bioavailability/bioequivalence study

Area of Science:

• Pharmacokinetics
• Statistical Modeling
• Bioavailability and Bioequivalence Studies

Background:

• Outliers significantly impact bioavailability/bioequivalence study conclusions.
• Current outlier detection methods (ANOVA-based on log-AUC) are disconnected from pharmacokinetic principles and ignore the correlated nature of concentration data.
• This can lead to inaccurate AUC and variance estimates.

Purpose of the Study:

• To propose a novel residual analysis method for outlier detection in bioavailability/bioequivalence studies.
• To develop an approach that accounts for the correlation structure of observed concentrations.
• To align outlier detection with population pharmacokinetic (PK) concepts.

Main Methods:

• Utilized a functional linear model to predict concentrations, incorporating the correlation structure.
• Proposed a residual analysis based on predicted concentrations.
• Focused distributional assumptions on observed raw concentrations rather than summarized AUC parameters.
• Accounted for the repeated measurements inherent in concentration-time curves.

Main Results:

• The proposed method effectively detects outliers by considering the inherent correlation in concentration data.
• This approach provides a more accurate variance estimate compared to traditional methods.
• Demonstrated the method's utility using a real-world dataset.

Conclusions:

• The functional linear model-based residual analysis offers a more robust and PK-informed approach to outlier detection in bioavailability/bioequivalence studies.
• This method improves the accuracy of variance estimation by leveraging the full concentration-time profile.
• The approach is consistent with population PK principles and enhances the reliability of study conclusions.