150: Real-World Application of Physiologically Based Pharmacokinetic Modeling (PBPK) in Support of Decision-Making During Drug Discovery; Guidance & Recommendations for the Utility of PBPK Impact on Candidate Selection & Human Pharmacokinetic Prediction Through Case Examples

Statement of Purpose: Guidance on the typical steps involved in developing and applying a PBPK model during drug discovery to assist in decision making, illustrated through the description of three case examples, where PBPK has been successfully applied to aid in candidate selection, human PK projection and prediction of drug interaction liability for parent and metabolite.

Description of Methods & Materials: The application of PBPK during drug discovery has been illustrated through three case examples. In case example 1, a PBPK model was developed to identify the most promising candidates, from available compounds in a chemical series, for further studies in human subjects. The goal of this discovery program was to develop a compound with decreased clearance, relative to an approved drug, to improve on the current dosing regimen. In case example 2, PBPK modelling was used to predict human PK following oral absorption. In addition, the effect of CYP3A4 auto-inactivation and drug-drug interaction (DDI) liability as a perpetrator was investigated. In case example 3, a PBPK model was used for first in human (FIH) dose prediction and prospective DDI predictions both for parent (as victim and perpetrator) and active metabolite (as perpetrator).

Data & Results: In case example 1, the results of PBPK simulations were summarized in a heat map, to aid with candidate selection. The heatmap was created based on the lead compound properties, by plotting free potency as a function of CL_int,u (Figure 1). The PBPK model was applied using a series of different CL_int estimates and dose levels. The shaded region on the heatmap indicates the region of optimal free potency as a function of CL_int,u which is predicted to result in a lower efficacious dose. The black arrow indicates improvements required for non-optimal candidates. As a result of this PBPK project, an initial candidate was selected for GLP-toxicity studies. In case examples 2 and 3, human PK was predicted over a range of doses after model verification using preclinical data. Several scenarios, based on different methods of clearance prediction, were simulated to account for uncertainty in the prediction of human CL. Both tablet and solution formulations were simulated to account for uncertainty in the prediction of absorption. A human efficacious dose was predicted based on the target exposure for efficacy in human. Simulated human exposure was used to investigate DDI liability as victim or perpetrator.

Interpretation, Conclusion or Significance: Although PBPK modelling is most effectively performed over different stage gates (discovery, early and late development) in a learn and confirm cycle, there is potential for impact on decision making at the earliest stages of the drug development process. Beginning in discovery, PBPK supports lead molecule identification and optimization. PBPK can guide decisions regarding optimal laboratory objectives for a chemical series. As development progresses toward candidate selection, PBPK delivers early PK, pharmacodynamic (PD) and most importantly, a mechanistic FIH PK prediction.

Disclosures: