D.E. (Esther) Lubberts (1,2), J.V. (Jeroen) Koomen (1,2), D.J. (Douglas) Eleveld (1), P.J. (Pieter) Colin (1)
(1) Department of Anesthesiology, University Medical Center Groningen, University of Groningen, Groningen, the Netherlands, (2) Medicines Evaluation Board (CBG-MEB), Utrecht, The Netherlands
Introduction Long-acting injectable (LAI) products are formulations intended for prolonged drug release that offer several benefits including improved patient outcomes. For the approval of generic LAI, general EMA guidance requests both a single-dose (SD) and multiple-dose (MD) study, with the MD having a possibility of a waiver if a low risk of accumulation is shown [1], [2]. Complex release characteristics hamper generic development including lengthy in vivo pharmacokinetic (PK) bioequivalence (BE) studies due to the need for long wash-out periods because of long apparent half-lives and feasibility issues accompanying safety concerns in healthy volunteers. As a consequence, a limited number of generics have been approved [3]. Allegedly model-based approaches might offer a novel direction in generic product development and regulatory approval [4], [5].
Objective This study investigates the potential to develop population pharmacokinetic (popPK) models from SD LAI data and the reliability of model-based extrapolation to MD situations.
Methods SD and MD PK blood concentration data were collected from the electronic common technical documents (eCTD) submitted in support of registration applications at the Medicines Evaluation Board (MEB), the Netherlands for 5 different LAI products. To secure the confidentiality of the included LAI, data was pseudo-anonymized before modeling analysis.
With the collected SD data, a range of models was tested per LAI using non-linear mixed effect modeling (NONMEM®) [6]. To handle samples reported as below the quantification limit (BLQ) the M3 method was applied and the Laplacian approximation method was used for parameter estimation [6], [7]. After selecting the best structural absorption model fit through fixed effects parameters (THETA), the random effects parameters (ETA) most reflective of the inter-individual variability (IIV) were selected [8]. Model diagnostics included goodness-of-fit (GOF) plots (e.g., OBS/IPRED and NPDE/PRED), visual predictive checks (VPCs), Akaike information criterion (AIC), and condition numbers ≤1000 [9]. The model where all of these model diagnostics aligned best was selected as the final model for MD simulations.
The MD simulations at (presumable) steady-state were performed in RxODE [10]. To account for uncertainty in the parameter estimates derived from the SD data, simulations were repeated 1000 times [9]. Hereafter, the simulated MD data were compared to the observed MD data by using VPCs.
Results Currently, the popPK models describing 3 out of the 5 LAI products after SD administration have been developed. This data showed that LAI PK is best described using (double, in parallel) first-order absorption, with one or two compartments, and a combined additive and proportional error model. In all models, IIV on clearance improved the model fit, which was expected considering the visual high variability in Cmax and elimination phase between the individual concentration-time curves. Other IIV parameters varied between the different models.
All SD models were built with treatment-naive healthy subjects data, whereas the observed MD data originates from studies with treatment-induction periods in patients, e.g. oral administration to acclimate the patient with the active substance. Hence, the similarity of predicted and observed MD was poor in describing the treatment onset. Nevertheless, once the predicted and observed MD reached their treatment (new) equilibrium again at the presumable steady-state level, we found that the predictive performances are satisfactory for all products so far based on VPCs. Since the bioequivalence assessment of MD is based on steady-state PK metrics, the poor predictive performance at treatment onset is considered to be irrelevant.
Conclusions For 3 out of 5 LAI, we found that the SD data seems to extrapolate well to the steady-state MD scenario. This pharmacometrics exercise will be extended to the remaining 2 LAI. In addition, the ratio of observed versus predicted for PK metrics frequently used to demonstrate bioequivalence (AUC(0-τ), Cmax,ss and Cτ,ss) will be calculated with a 0.8-1.25 acceptance range for similarity [1], [2], [5]. Further work will focus on quantifying the sensitivity of our approach for detecting scenarios that would violate the assumptions of the SD to MD extrapolation. Conceivably, this will guide the implementation of modeling and simulation as support for generic LAI marketing authorization applications.
References:
[1] Committee for Medicinal Products for Human Use (CHMP) and European Medicines Agency (EMA), “Guideline on the Investigation of Bioequivalence,” 2010. Accessed: Aug. 14, 2023. [Online]. Available: http://www.ema.europa.eu
[2] European Medicines Agency (EMA) and Committee for Medicinal Products for Human Use (CHMP), “Guideline on the pharmacokinetic and clinical evaluation of modified release dosage forms,” 2014, Accessed: Aug. 04, 2023. [Online]. Available: www.ema.europa.eu/contact
[3] M. N. O’Brien, W. Jiang, Y. Wang, and D. M. Loffredo, “Challenges and opportunities in the development of complex generic long-acting injectable drug products,” Journal of Controlled Release, vol. 336, pp. 144–158, Aug. 2021, doi: 10.1016/J.JCONREL.2021.06.017.
[4] P. Gajjar et al., “Determining bioequivalence possibilities of long acting injectables through population PK modelling,” European Journal of Pharmaceutical Sciences, vol. 179, p. 106296, Dec. 2022, doi: 10.1016/J.EJPS.2022.106296.
[5] L. Zhao, M. J. Kim, L. Zhang, and R. Lionberger, “Generating Model Integrated Evidence for Generic Drug Development and Assessment,” Clin Pharmacol Ther, vol. 105, no. 2, pp. 338–349, Feb. 2019, doi: 10.1002/CPT.1282.
[6] R. J. Bauer, “NONMEM Tutorial Part II: Estimation Methods and Advanced Examples,” CPT Pharmacometrics Syst Pharmacol, vol. 8, no. 8, pp. 538–556, Aug. 2019, doi: 10.1002/PSP4.12422.
[7] M. Bergstrand and M. O. Karlsson, “Handling Data Below the Limit of Quantification in Mixed Effect Models,” AAPS J, vol. 11, no. 2, p. 371, 2009, doi: 10.1208/S12248-009-9112-5.
[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. 9, p. e6, 2012, doi: 10.1038/PSP.2012.4.
[9] D. R. Mould and R. N. Upton, “Basic Concepts in Population Modeling, Simulation, and Model-Based Drug Development—Part 2: Introduction to Pharmacokinetic Modeling Methods,” CPT Pharmacometrics Syst Pharmacol, vol. 2, no. 4, p. e38, Apr. 2013, doi: 10.1038/PSP.2013.14.
[10] W. Wang, K. M. Hallow, and D. A. James, “A Tutorial on RxODE: Simulating Differential Equation Pharmacometric Models in R,” CPT Pharmacometrics Syst Pharmacol, vol. 5, no. 1, pp. 3–10, Jan. 2016, doi: 10.1002/PSP4.12052.
Reference: PAGE 32 (2024) Abstr 11136 [www.page-meeting.org/?abstract=11136]
Poster: Drug/Disease Modelling - Other Topics