Wangui Mbuguiro: No relevant disclosure to display
Objectives: Healthy human controls in clinical trials are exposed to the risks of an investigational drug without benefiting. This study uses internal data from AstraZeneca (AZ) to evaluate the utility of modeling and simulation (M&S) in creating a virtual control population for organ impairment studies. Using virtual controls instead of healthy humans in renal and hepatic impairment studies could reduce the risks, costs, and time associated with these studies.
Methods: 17 AZ compounds with 30 renal and hepatic impairment studies conducted over the last 20 years (including normal/healthy controls) were identified as potential case studies for this analysis. Approximately 13 AZ compounds with 17 organ impairment studies were deemed permissible for this analysis due to legal/ethical restrictions in consideration of data privacy and language in each study’s informed consent. Population pharmacokinetic (PopPK) and physiologically based pharmacokinetic (PBPK) modeling were applied to simulate virtual controls in these studies, allowing a comparison of predicted PK parameters (e.g., AUC, Cmax) to observed PK from the control group. Simulations were performed using both early (developed using only Phase I data) and late (submitted in the NDA/published) models to examine the impact of model development timing and data-use on predictions.
Results: To date, late popPK model simulations have been completed for 7 studies involving 6 compounds. AUC and Cmax were accurately predicted ( < 30% error) for 7/7 and 5/7 late popPK simulations, respectively. The highest errors (~50%) were associated with one compound and could be explained by previously identified limitations in the model, underscoring the importance of assessing predictive accuracy of fit-for-purpose models prior to using virtual controls. For this compound, the early popPK model was able to predict Cmax with less error (34%). Late PBPK model simulations have been completed for 5 studies from 3 compounds, predicting Cmax and AUC with low error ( < 27%).
Conclusions: Using previous popPK and PBPK models, PK parameters in healthy subjects in impairment studies were predicted within 30% for nearly all case studies. Although others have used either popPK [1] or PBPK [2] modeling to simulate virtual controls in organ impairment studies, our approach uniquely applies both modeling techniques for a more robust analysis. Furthermore, this approach is distinctive in addressing the impact of model development timing and data-use on predictive ability, including both early and late models in the analysis. This study supports the evolution of regulatory frameworks that embrace innovative methods, integrating M&S approaches to support drug safety evaluations while reducing participant risks and expediting development.
Citations: [1] Younis, I. R., Wang, F., & Othman, A. A. (2024). Feasibility of Using Population Pharmacokinetics‐Based Virtual Control Groups in Organ Impairment Studies. The Journal of Clinical Pharmacology.
[2] Heimbach, T., Chen, Y., Chen, J., Dixit, V., Parrott, N., Peters, S. A., ... & Hall, S. (2021). Physiologically‐based pharmacokinetic modeling in renal and hepatic impairment populations: a pharmaceutical industry perspective. Clinical Pharmacology & Therapeutics, 110(2), 297-310.
Keywords: pharmacokinetic modeling, organ impairment, virtual controls
