Evaluation of the Sensitivity of Concluding Bioequivalence Using Stochastic Simulation and Estimation

Purpose: Single dose (SD) bioequivalence (BE) studies can fail statistical criteria by maximum concentration (Cmax). Population pharmacokinetic (PoP PK) model based simulations are used in such instances to predict steady-state (SS) concentrations to evaluate BE criteria on simulated data. We evaluated the sensitivity of such PoP PK model based method to find true formulation differences in a simulation study. Methods: Stochastic simulation and estimation (SSE) method was applied. The structural model was a one compartment model with 1st order absorption. An exponential error (15% CV) for between subject variability and proportional error (15% CV) model for residual variability was assumed. PK studies (n=200) having 1000 virtual subjects dosed with test and reference formulations were simulated with NONMEM® software (v. 7.3.0) for SD and SS. The rate of absorption (Ka, hr-1) parameter varied between test and reference formulations to simulate passing and failing PK profiles for BE criteria. Different Ka combinations for test and reference formulations include 0.1/1, 0.2/1, and 0.5/1.5. Simulated PK profiles were analyzed using PKNCA (v. 0.8.4) package in R (v 1.0.143) software to conduct a non-compartmental analysis and iterated 200 times. The resulting Cmax and AUC were assessed using the Welch 2-sample t-test function to conclude BE. A 90% confidence interval of AUC and Cmax falling between 0.8-1.25 was set a priori as BE criteria. Lastly, the sensitivity of the PoP PK model based simulation method for concluding SD and SS based BE conclusion was calculated from 200 independent simulated studies. Results: Bioequivalent formulations of SS and SD that were designed by same absorption rate parameter met BE criteria set a priori for Cmax and AUC. Varying Ka parameters between test and reference formulations, with Ka combinations of 0.1/1 and 0.2/1, only failed BE criteria for Cmax on SD. The Ka parameter combination 0.5/1.5 did not result in failed Cmax on SD and is not considered as true failed criteria by design. Of the 200 simulated datasets in the fail BE criteria by design on SD, all of them passed the a priori BE criteria for Cmax for SS simulations. Conclusion: The PoP PK model based simulation method may not be sensitive to study BE of Cmax at SS after initial failure of two SD drug formulations. True differences in the formulation Ka cannot be detected in SS simulations. The method is still applicable for concluding the clinical relevance SS Cmax.