145: An Evaluation of POPPK & PBPK Model-based Virtual Controls to Replace Healthy Subject Controls in Organ Impairment PK Studies

Statement of Purpose: A dedicated organ impairment (OI) study typically enrolls healthy subjects with normal renal and hepatic function as the control group, matched for age, body weight and gender to the enrolled subjects with varying degrees (mild, moderate, and severe) of impaired renal or hepatic function. This healthy control group serves as a reference and enables a comparison of pharmacokinetic (PK) data between subjects with impaired organ function and normal organ function to assess the impact of organ dysfunction on drug exposure and PK. However, enrolling healthy subjects in dedicated OI studies can be challenging because of ethical concerns, safety considerations, and narrow eligibility criteria due to matching. The objective of this work is to explore the feasibility of replacing the healthy control group in OI studies with virtual populations generated by population PK (POPPK) and physiologically-based PK (PBPK) models.

Description of Methods & Materials: Five Genentech molecules (i.e., Molecule 1, 2, 3, 4 and 5) were identified for which dedicated OI studies were conducted. The PK, demographic (e.g., body weight), and other key covariates data (e.g., plasma protein concentration) were extracted from the healthy control in these OI studies. Previously established POPPK models and PBPK models of each molecule were used together with the individual covariate information (e.g., body weight) to simulate concentration-time profiles for each subject in the healthy control group. To evaluate the feasibility of POPPK and PBPK model-based replacement, the individual PK simulation result was compared to individual observed data to derive the predicted/observed ratio for each subject and the geometric mean of the ratio for the entire healthy control group.

Data & Results: Using the POPPK models, the geometric mean of simulated AUC/observed AUC ratio was 2.42, 1.22, 1.83, 0.801 and 0.844, and the geometric mean of simulated C_max/observed C_max ratio was 1.11, 1.77, 1.18, 0.81 and 0.84, for Molecule 1, 2, 3, 4 and 5, respectively. Overall, the POPPK models successfully predicted the healthy group PK of Molecule 2, 3, 4 and 5 within 2-fold and Molecule 4 and 5 within 1.25-fold, but overpredicted the AUC of Molecule 1. The results also showed a tendency to overpredict the C_max of Molecule 2 and AUC of Molecule 3 in the healthy control group.

Interpretation, Conclusion or Significance: Molecules 1, 2 and 3 are orally administered oncology compounds for which the POPPK models were built using PK data solely from cancer patients without any data from healthy subjects, whereas the POPPK model for Molecules 4 and 5 included PK data from healthy subjects. This suggests that PK data from healthy subjects are preferably incorporated in the POPPK models for better prediction. In addition, we hope to use our ongoing PBPK model evaluation to physiologically bridge the PK discrepancy between cancer patients and healthy subjects.

Disclosures: