Sabrina Boutaghou (1,2), Sofia Vasilakaki (3), Vangelis Karalis (1)
1. Department of Pharmacy, School of Health Sciences, National and Kapodistrian University of Athens, Greece // 2. Department of Pharmacy, Aix-Marseille University, Marseille, France // 3. Department of Chemistry, National and Kapodistrian University of Athens, Greece
Objectives: Bioequivalence (BE) studies compare the concentration-time (C-t) profiles of two pharmaceutical products (test (T) and reference (R)) of the same active ingredient. The C-t profiles are not compared directly but by pharmacokinetic parameters, such as area under the curve (AUC), maximum observed plasma concentration (Cmax) and time (Tmax) at which Cmax is observed. Because pharmacokinetic parameters are calculated from the C-t data, the reliability of the calculations, and thus the results of a BE study, can be affected by the sampling scheme of the study. Presumably, an inadequate sampling design may lead to study inaccuracies and uncertain results, while a very dense sampling design may increase the workload and cost.
The aim of the present study was to evaluate the influence of different sampling schemes on the outcome of bioequivalence studies. In this analysis, two drugs (paracetamol and irbesartan) with different pharmacokinetic properties were studied and several sampling schemes were investigated using Monte Carlo simulations of 2×2 crossover BE studies.
Methods: Joint in vitro – in vivo simulations (IVIVS) were used to generate plasma concentration data [1]. Briefly, first-order kinetics were used to describe the dissolution kinetics, while the in vivo part was generated using literature information. For paracetamol, the absorption rate constant (Ka) was 0.1158min-1, the apparent volume of distribution was 70 l, and the apparent clearance (Cl) was 0.289 l/min. For irbesartan, the relevant values were Ka=0.00507 min-1, Cl= 0.225 l/min, inter-compartmental clearance 0.295 l/min and volume of distribution of the central and peripheral compartments equal to 13.8 l and 85.8 l, respectively. The use of IVIVS allowed the study of different release rates for the T and R products and thus the generation of different C-t profiles. Four levels of difference in drug release rate between T and R were examined: identical (i.e., 0% difference), 20%, 40%, and an extreme scenario of 60% difference. Variability between subjects (BSV) and within subjects (WSV) was added. A series of ten sampling schemes were applied that varied in terms of duration and density of sampling points at specific time points (namely, around Tmax or at later time points or homogenously sparse designs). These sampling schemes were set appropriately taking into account the pharmacokinetic properties of paracetamol and irbesartan. Based on the sampling scheme, the selected C-t data were used to calculate AUC and Cmax using the typical non-compartmental approach. Three sample sizes (N=12, 24 and 36) were considered and the generated subjects were divided accordingly in a 2×2 clinical design [1]. The procedure described above was repeated 5,000 times with Monte Carlo and the success or failure of each trial was recorded. After all repetitions were performed, the percentage probability of BE acceptance was determined.
Results: With identical release rates, the simulated C-t profiles were similar; minor differences resulted only from the applied BSV and WSV. In this case, all sampling schemes yielded 100% BE for both paracetamol and irbesartan. For paracetamol, a 20% difference in release rate resulted in statistical power values of 100% for Cmax and AUC for the scenarios in which sampling was moderate to sparse or in which few points were sampled in the falling part of the C-t curve. Scenarios with sparse sampling around Tmax had no effect on AUC power estimates, while they resulted in lower power for Cmax, but none was below 93%. For a 40% difference in rate, the probability of declaring BE was 93% (Cmax) and 100% (AUC) for the less sparse scenarios, but it reduced to 34% and 64% for sparse scenarios close to Tmax. In the case of the extreme scenario with 60% difference in release rate, the statistical significance for AUC remained at high level (range 87%-100%), while it was low for Cmax (32%, 48%, etc.). A similar trend was observed for irbesartan, but the differences were less pronounced compared to the paracetamol case. For irbesartan, the potency values were almost 10% higher compared to that of paracetamol. This suggests that irbesartan, which has slower absorption, is less affected by changes in sampling schedule. All the above results are for a sample size of 24. Higher values were taken when N=36 and lower for N=12.
Conclusions: It can be concluded that the demonstration of BE is not sensitive to differences in the duration and density of sampling schedules. All applied scenarios of sampling schemes were able to show bioequivalence when the pharmacokinetic parameters of the two products were identical. AUC was affected to a lesser extent than Cmax and only when large differences in the C-t profiles of the T and R products were present. Bioequivalence for Cmax was sensitive to the sparse schedules only when withdrawal sites were close to Tmax in association with large differences in drug release (greater than 20%). When the two drug products differ significantly in absorption, dense and sparse sampling schemes result in nearly similar BE acceptances.
References:
[1] Vlachou M, Karalis V. An In Vitro-In Vivo Simulation Approach for the Prediction of Bioequivalence. Materials (Basel). 2021 Jan 24;14(3):555. doi: 10.3390/ma14030555
Reference: PAGE 29 (2021) Abstr 9840 [www.page-meeting.org/?abstract=9840]
Poster: Methodology - Other topics