A comparison of classical bioequivalence analysis techniques with simulated annealing algorithm optimised sampling times and population pharmacokinetic modelling

Abstract

Bioequivalence (BE) studies are conducted to demonstrate that two drug for- mulations produce similar bioavalabilities or therapeutic e ect and safety when used. During this study, drug concentrations are obtained several times over a given period of time and a concentration vs. time graph is constructed. For a generic drug to be approved, this BE studies are conducted and the generic drug must be demonstrated to be therapeutically equivalent to the innovator drug. This study utilised a standard 2 2 crossover design to randomly assign sub- jects to each of the two sequences. Statistical methods such as con dence in- tervals, Schuirmann's two one-sided and Wilcoxon-Mann-Whitney tests were used to assess average bioequivalence (ABE). However, there are concerns that the use of ABE alone is not appropriate for drugs with high intra-subject and inter-subject variabilities. Under such a circumstance, population bioequiv- alence (PBE) and individual bioequivalence (IBE) are proposed. This study employs the PBE approach but not the IBE as it is not possible to perform IBE on the available data. Results indicated that the generic drug is average bioequivalent to the innovator drug although Cmax was outside the regulatory range set by the Food and Drug Administration (FDA). Most biological data are modelled using nonlinear xed e ect models. Popula- tion pharmacokinetic (PK) modelling has been used in clinical pharmacology to identify the sources of PK variability in the target population. This study was conducted to determine the characteristics of the PK parameters of the orally administered antibiotic given to pigs using a population approach. A population PK model was developed using a nonlinear mixed e ects model (NLMEM) with a one-compartment model using di erent residuals. For the NLMEM, the stochastic approximation expectation maximisation (SAEM) al- gorithm was implemented in MONOLIX. The models were used to estimate the population PK parameters and diagnostic plots obtained for model eval- uation. The results showed that the combined residual error model tted the data better than the constant error model. In addition, this study sought to nd optimal sampling times which will min- imise the number of blood samples required for pharmacokinetic study. The optimal sampling times were generated from a one compartment model and implemented in MATLAB. The parameters used in the optimisation were es- timated from the population PK model. These sampling times were generated using the simulated annealing (SA) algorithm. Keywords: Bioavailability, Bioequivalence, generic drugs, reference drugs, average bioequivalence, population bioequivalence, individual bioequivalence, pharmacokinetics, pharmacodynamics, therapeutic window, con dence inter- val, nonlinear xed e ects model, two one-sided tests, maximum likelihood, population parameters, optimal sampling design, concentration-time curve, model evaluation, Visual predictive checks (VPC), Normalised prediction dis- tribution errors (NPDE), Stochastic approximation expectation maximisation (SAEM) algorithm.