(T1430-10-69) Using Simulation to Address Challenges in Bioequivalence Data Analysis for Drug Products Containing Endogenous Compounds

Purpose: The objective of bioequivalence (BE) studies is to assess the effect of formulation differences between the test and reference products on BE. As a result, baseline-corrected (BLC) concentrations of endogenous substances, which would generally reflect the amounts of compound released from a formulation and absorbed by each subject, are typically used for the BE determination when the drug product being studied contains an endogenous compound. However, since endogenous compounds are identical to the drug substance that is being administered, it can be difficult to determine the amount of drug substance that is released from the dosage form and subsequently absorbed by each subject. The purpose of this research was to identify challenges in the BE data analysis of drug products containing endogenous compounds and to use simulation to investigate different approaches to address these challenges.

Methods: We surveyed all Abbreviated New Drug Applications (ANDAs) for drug products containing endogenous compounds whose BE assessments are electronically available to identify challenges in BE data analysis (n =19 drugs). Subsequently, we simulated 10,000 datasets (Dataset A) of two-way crossover BE studies using a two-compartment pharmacokinetic model for each combination of bioavailability (F), absorption rate constant (ka), clearance (Cl), rate constant from the central compartment to the peripheral compartment (k₁₂), rate constant from the peripheral compartment to the central compartment (k₂₁), and volume of distribution (V_d) values. Each dataset A comprised 12 subjects receiving both test (T) and reference (R) treatments at different treatment periods with true T/R of 0.80 for Type I error (TIE) simulation or 0.95 for study power simulation. Eight baseline scenarios, encompassing high (40% of mean non-BLC Cmax) or low (5% of mean non-BLC Cmax) baselines, stable or circadian rhythmic baselines, and high (30%) or low (5%) variability, were simulated and added to Dataset A to generate Dataset B. For each baseline scenario, the baseline concentrations at sampling time used for baseline correction (e.g., -1, -0.5, and 0 hours, or time-matched baseline) were simulated and subtracted from Dataset B to obtain Dataset C, which was the BLC dataset. Statistical analysis was performed on each simulated dataset to compare TIE and study power between methods employed to address the challenges in BE data analysis of drug products containing endogenous compounds.

Results: Three specific data analysis challenges were identified from BE assessments of drug products containing endogenous compounds. The first challenge pertained to handling data where BLC pre-dose concentrations exceed 5% of their respective Cmax. The common practice is to exclude the affected subjects (‘5% Cmax Rule’) from the BE statistical analysis, irrespective of whether the drug product is endogenous. Our survey revealed a correlation between high baseline variability / high baseline concentrations of endogenous compounds and a high percentage of subjects with BLC pre-dose concentrations greater than 5% of their respective Cmax. Additionally, due to high baseline variability and high baseline concentrations, applying the ‘5% Cmax Rule’ to certain drug products is impractical as there was a large proportion of subjects (e.g., >50%) with BLC pre-dose concentrations greater than 5% of their respective Cmax. Simulation results demonstrated comparable TIE between the method of including all subjects regardless BLC pre-dose concentrations and the method of excluding the affected subjects. The second challenge concerned the methods to set BLC pre-dose concentrations, specifically, setting all negative BLC pre-dose concentrations to 0 or changing all BLC pre-dose concentrations to 0. Simulation results indicated similar TIE and study power between these two methods. The third challenge identified was related to the choice of baseline correction methods, namely, time-averaged method (e.g., average the concentrations from -1, -0.5, and 0 hours) vs. time-matched method. The commonly recommended baseline correction method is the time-averaged method, but it was also seen that some applicants used time-matched method. Simulation results showed no advantage of the time-matched method over the time-averaged method in terms of TIE and study power.

Conclusion: BE data analysis for certain drug products containing endogenous compounds poses challenges mainly due to high baseline variability and high baseline concentrations. The findings from our simulations provide insights into potential methods to address these challenges in the future.

Acknowledgements: This abstract reflects the views of the authors and should not be construed to represent FDA’s views or policies.