(S-071) Assessment of Virtual Control Groups for Organ Impairment Studies Using Population Pharmacokinetic Simulation: Twin Matching versus Group Matching

Emily Bozenhardt – J&J Innovative Medicine; Lingjue Li – J&J Innovative Medicine; Terra Acri – J&J Innovative Medicine; Damayanthi Devineni – J&J Innovative Medicine

Objectives: The objectives of this analysis were to assess the use of population pharmacokinetic (PK) simulation for virtual controls in organ impairment (OI) studies, and to compare twin matching with group matching approaches for virtual controls within this context. Multiple examples (19 cohort/molecular entity combinations from OI studies, including parent drugs and metabolites) are explored using integrated population PK models based on early and late phase clinical data.

Methods: Virtual control groups were simulated using population PK based on previously developed final or near-final models using R and mrgsolve, with a similar approach as previously published1,2,3. For each approach (twin matching or group matching), N=1000 replicates of n=8 virtual controls were either drawn from the dataset, filtered by the range of covariate values in the control group in the clinical study report for the corresponding study (group matching), or had values set to exactly the values of the covariates in the OI patients (twin matching). For the twin matching, virtual controls had variables of specific interest [e.g. creatinine clearance (CRCL) for renally impaired participants] set to the median in the population PK model, while all other demographics such as age, body weight, and gender were set to the exact values of the matching OI patient. Only covariates that were included in the population PK model were included in the simulation. Calculation of derived PK parameters (e.g. Cmax, AUC) was done in R based on individual PK profiles simulated with the final population PK model. Direct calculation of Cmax and AUC geometric mean ratios (GMRs) was performed based on observed geometric means of PK parameters in mild, moderate, and severe cohorts and geometric means of simulated PK parameters in the virtual control group for each of 1000 replicates.

Results: The medians and 90% prediction intervals (based on the 5th to 95th percentile of GMRs from 1000 replicates) based on twin matching and group matching were presented in a forest plot and compared with the observed GMR and 90% CI from the OI study. In many cases, the observed GMR from the clinical study report fell within the 90% prediction interval for the GMR using virtual controls based on both approaches. Twin matching and group matching gave similar results, with twin matching sometimes resulting in a ratio closer to 1 than the observed ratio due to perfect matching of other characteristics besides those related to OI, depending on which characteristics were covariates in the population PK model.

Conclusions: Utilizing population PK simulation to generate a virtual matching healthy control group in OI studies holds some potential although requires some major considerations. Further research is warranted to assess whether there are scenarios that may lead to dose adjustment recommendations that differ between those based on virtual population PK simulation versus those based on observed data from historic studies to refine such a model-based approach. Lastly, utility of any such approach will also require a priori buy-in and alignment from all stakeholders including health authorities.

Citations: [1]: Prybylski, J, Wang, Y, et al. Simulating Healthy Participant Pharmacokinetics for Renal and Hepatic Impairment Studies: Retrospective Assessment of the Approach. AAPS J. 2024;26:65

[2]: Purohit V, Huh Y, Wojciechowski J, et al. Leveraging prior healthy participant pharmacokinetic data to evaluate the impact of renal and hepatic impairment on ritlecitinib pharmacokinetics. AAPS J. 2023;25:32
[3]: Younis, I.R., F. Wang, and A.A. Othman, Feasibility of Using Population Pharmacokinetics-Based Virtual Control Groups in Organ Impairment Studies. J Clin Pharmacol, 2024.

Keywords: organ impairment, virtual controls, population PK simulation