(M1030-02-07) Establishment of Preclinical Mechanistic In Vitro-In Vivo Correlations for Long-Acting Injectable Suspensions

Purpose: Long acting injectables (LAIs) are parenteral formulations administered through either subcutaneous (SC) or intramuscular (IM) route. These formulations are designed to provide sustained drug release over an extended period ranging from weeks to several months. For LAIs that are formulated as crystalline suspensions, the extended release is obtained through the formation of in-situ drug depots at the injection site from which poorly soluble drugs slowly dissolve and are subsequently absorbed into the systemic circulation¹. Depo-SubQ Provera 104®, a SC suspension of medroxyprogesterone acetate (MPA), came off patent in 2020, hence the generic development of this product is anticipated. Nevertheless, the development of complex generic LAIs is challenging since the relationship between solubility, particle size, and in vivo release from a parenteral suspension has not yet been established systematically. In this scenario, the use of modeling and simulation may provide a unique opportunity to understand the in vivo release mechanisms of LAI suspensions and drug disposition. Establishing mechanistic in vitro-in vivo correlations (IVIVCs) is a valuable approach to further drug product development of LAIs. Hence, the purpose of this study was to use different modeling approaches to mechanistically understand the in vivo performance of MPA LAIs in preclinical species and to establish an IVIVC.

Methods: A mechanistic physiologically-based pharmacokinetic (PBPK) model for MPA was built using GastroPlus® 9.8.2. All tissues were modeled as perfusion-limited tissues and the tissue/plasma partition coefficients were calculated using the default Lukacova method. Physicochemical and biopharmaceutical properties were obtained from the literature² or were predicted from the structure using ADMET Predictor® 10.3. The change in depot volume over time due to tissue inflammation, to reflect the 1-3-fold increase as per literature report³, was taken into account in the PBPK model. The in vitro (formulation characterization⁴ and dissolution studies) and in vivo studies in rabbits were obtained from literature and from a collaboration with an academic research laboratory. Two approaches were taken to predict the in vivo dissolution. Firstly, the drug’s solubility and particle size distribution were investigated. In this approach, the difference between experimental data and the effective in vivo particle size distribution was assessed. The same scaling between in vitro and in vivo particle size distribution was used across several formulations. The second approach was a mechanistic IVIVC in which time scaling through Levy plots was applied. In this approach, the IVIVC was established with formulations F1, F2 and F3 (internal validation) and the reference product, Depo-SubQ Provera 104®, was used as an external validation.

Results: The developed PBPK model for MPA adequately described the plasma pharmacokinetic (PK) profile of MPA following intravenous (IV) administration in rabbit (R²>0.77). The established model and disposition parameters, such as clearance and volume of distribution, were then used to predict the plasma PK profile of MPA following SC administration in rabbits. The experimental particle size distribution resulted in overprediction of Cmax and AUC (data not shown), whereas the scaled particle size distribution resulted in more adequate prediction (Figure 1). The Cmax and AUC prediction errors (PE) were within 33.7% and 12.4%, respectively. The average absolute %PE for Cmax was 21.2% (±13.5) and for AUC 6.2% (±5.5). The inclusion of inflammation parameters in the model resulted in better Cmax prediction with a lower average absolute PE (11% ±5.2) while maintaining AUC average absolute PE < 10% and PE < 15% for each formulation. As shown in Figure 2, the IVIVC using the time scaling approach resulted in an overall successful prediction of plasma concentrations for rabbit. The Cmax and AUC %PEs were within 12.3% and 17.8%, respectively. The average absolute PE for Cmax was 9.1% (±8.6) and 9.2% (±10.7) for AUC. The %PEs for each formulation are given in Table 1.

Conclusion: Utilizing the same scaling between in vitro and in vivo particle size distribution across all the formulations and varying depot volume over time (due to inflammation) resulted in reasonable prediction of MPA PK profiles. The larger effective in vivo particle size can be related to in vivo particle aggregation. This shows the utility of in silico tools for mechanistic understanding of in vivo product performance. The mechanistic IVIVC approach also resulted in reasonable predictions of MPA profiles as demonstrated by average absolute %PE < 10% for both Cmax and AUC. Establishing IVIVC for LAIs is challenging given the need for time scaling (seen by the higher AUC %PE for RLD), hence further studies are needed to refine this approach.

References: 1. Shah, J. C. & Hong, J. Model for Long Acting Injectables (Depot Formulation) Based on Pharmacokinetics and Physical Chemical Properties. The AAPS Journal 24, 44 (2022).
2. Kozutsumi, D. et al. Pharmacokinetics of 9α-fluoromedroxyprogesterone acetate in rats: comparison with medroxyprogesterone acetate. Biopharmaceutics & Drug Disposition 20, 277–284 (1999).
3. Jucker, B. M. et al. Multiparametric magnetic resonance imaging to characterize cabotegravir long-acting formulation depot kinetics in healthy adult volunteers. British Journal of Clinical Pharmacology (2021) doi:10.1111/bcp.14977.
4. Bao, Q., Zou, Y., Wang, Y., Choi, S. & Burgess, D. J. Impact of Formulation Parameters on In Vitro Release from Long-Acting Injectable Suspensions. AAPS Journal 23, (2021).

Acknowledgements: Disclaimer: This research was funded through the FDA Office of Generic Drugs; contract # 75F40121C00133 and HHSF223201710135C. The views expressed here do not reflect official policies of the US FDA or the Department of Health and Human Services, nor does any mention of trade names imply endorsement by the US Government.


Table 1. Comparison of Cmax and AUC prediction errors after SC administration of the three test formulations and RLD.


Figure 1. Observed (squares) and predicted (line) plasma concentration vs time profile after SC administration of MPA in rabbits. Prediction based on particle size distribution using the same scaling across all formulations. (A) No inflammation considered, and (B) Inflammation parameters included. RLD: Reference Listed Drug


Figure 2. Observed (squares) and predicted (line) plasma concentration vs time profile after SC administration of MPA in rabbits. Prediction based on mechanistic IVIVC using time scaling approach. RLD: Reference Listed Drug