KILELA Songela Reagan 1,2, Helene Haguet Haguet 1, Lisa Hanquet 1, Happy Phanio Djokoto 1, Grace Shalom Govere 1, Adrien Olama 1, Camille Massaux 1, Lisa Wellin 1, Jean-Michel Dogné 1, Flora MUSUAMBA 1,2
1 University of Namur (Namur, Belgium), 2 University of Lubumbashi (Lubumbashi, Congo (DRC))
INTRODUCTION
Generic drugs play a crucial role in contemporary healthcare systems [1]. Their marketing authorization requires the demonstration of bioequivalence (BE) [2]. This regulatory requirement aims to establish that two formulations of the same active ingredient have similar exposure profiles, thus ensuring equivalent therapeutic efficacy and safety [3, 4].
Currently, BE assessment relies almost exclusively on non-compartmental analysis (NCA), which is recognized and required by regulatory agencies [5]. NCA offers several advantages: ease of implementation, transparency, and a direct link to standard BE criteria such as the area under the curve (AUC) and maximum plasma concentration (Cmax)[6].
However, despite its status as the gold standard, NCA also raises several scientific and practical challenges. It requires a sufficiently large number of samples and subjects to obtain statistically significant results, and it does not account for individual variability in the estimation of pharmacokinetic parameters. These requirements significantly increase the costs associated with studies while imposing logistical constraints [7].
Faced with these constraints, pharmacokinetic modeling and simulation (M&S) offer powerful tools that complement traditional experimental approaches. They allow for the integration of prior knowledge to predict drug behavior under various clinical and experimental conditions. These approaches can contribute to reducing the number of clinical trials required, optimizing protocols, and understanding the sources of inter-individual variability [8]
METHODS
The pharmacokinetic model used in this study was previously published in a journal of the American Society for Microbiology and describes the population pharmacokinetics (PopPK) of artemether, lumefantrine, and their metabolites following oral administration of artemether–lumefantrine in children with uncomplicated malaria in Papua New Guinea [9].
The validated model was used to simulate BE trials under multiple design scenarios. Each scenario was evaluated using 200 simulated replicate trials to ensure robust estimation of statistical power and Type I error. Simulations were initially conducted from a reference design including 60 subjects and 19 sampling time points per subject. In the first step, the number of sampling time points was progressively reduced to assess the impact of sampling density independently of sample size. In a second step, after identifying an optimal reduced sampling scheme, the population size was gradually decreased (from 60 to 6 subjects, in steps of 6) to evaluate design robustness.
BE was assessed using both a model-based (PopPK) approach and a traditional NCA approach. Statistical power and Type I error were compared across all simulated scenarios.
RESULTS
Model reproduction showed good agreement between the estimated parameters and those reported in the reference study.
For BE assessment, simulations were performed from a reference design including 60 subjects and 19 sampling time points. Reducing the number of samples showed that statistical power remained high down to 8 time points with both methods. Below this threshold, power declined markedly (to ~76–78% at 6–7 points), with a concomitant increase in Type I error. Thus, 8 sampling points were selected as the optimal compromise between power and error control.
With the 8-point design fixed, the sample size was progressively reduced from 60 to 6 subjects. Power remained ≥96% down to 12 subjects with the compartmental approach, while NCA maintained 80% power at 12 subjects, with Type I error well controlled (≤1.5%). At 6 subjects, power dropped substantially for both methods. Overall, a design combining 8 sampling points and at least 12 subjects preserved adequate statistical power and Type I error control compared with the initial 60-subject, 19-sample design.
CONCLUSION
This study demonstrates that an M&S approach represents a robust tool for optimizing BE study designs. The simulations showed that substantial reductions in both the number of sampling time points and the sample size can be achieved without compromising statistical power or control of Type I error. A design combining 8 sampling time points with 12 subjects emerged as the optimal balance between protocol simplification, operational feasibility, and maintenance of high statistical performance.
These findings highlight the value of M&S approaches to support the scientific justification of study design choices and to promote more efficient and quantitatively driven bioequivalence strategies.
References:
[1] M. Cassier, C. Baxerres, M. Cassier, and C. Baxerres, “Rationaliser les marchés de médicaments dans les Suds : réinventer et développer l ’ usage des médicaments essentiels,” HAL open science, vol. halshs-038, p. 16, 2022, [Online]. Available: https://shs.hal.science/halshs-03879605v1
[2] C. Hugnet, “Limites scientifiques de l’autorisation de mise sur le marché des médicaments génériques,” Bull. Acad. Vet. Fr., vol. 167, no. 2, pp. 153–156, 2014, doi: 10.4267/2042/53812.
[3] P. Baumann and J. Kahn, “Les médicaments génériques : quels sont les problèmes et d’où viennent-ils ?,” pp. 879–884, 2003.
[4] E. Masson and S. A. S. Tous, “Le médicament générique et la relation de soin . Sociologie d ’ un quiproquo Generic drugs and the care relationship : The sociology of a quid pro quo,” vol. 51, pp. 46–63, 2009.
[5] K. Möllenhoff et al., “Efficient model-based Bioequivalence Testing,” Feb. 2020, [Online]. Available: http://arxiv.org/abs/2002.09316
[6] B. M. Davit et al., “Comparing generic and innovator drugs: A review of 12 years of bioequivalence data from the United States Food and Drug Administration,” Annals of Pharmacotherapy, vol. 43, no. 10, pp. 1583–1597, 2009.
[7] P. Colucci, J. Turgeon, and M. P. Ducharme, “How critical is the duration of the sampling scheme for the determination of half-life, characterization of exposure and assessment of bioequivalence?,” Journal of Pharmacy and Pharmaceutical Sciences, vol. 14, no. 2, pp. 217–226, 2011.
[8] D. R. Mould and R. N. Upton, “Basic concepts in population modeling, simulation, and model-based drug development,” CPT Pharmacometrics Syst. Pharmacol., vol. 1, no. 1, pp. 1–14, 2012, doi: 10.1038/psp.2012.4.
[9] S. Salman et al., “Population pharmacokinetics of artemether, lumefantrine, and their respective metabolites in papua new guinean children with uncomplicated malaria,” Antimicrob. Agents Chemother., vol. 55, no. 11, pp. 5306–5313, 2011.
Reference: PAGE 34 (2026) Abstr 12030 [www.page-meeting.org/?abstract=12030]
Poster: Drug/Disease Modelling - Other Topics