Background: Methods for applying physiologically based pharmacokinetic (PBPK) modeling to predict first-in patient (FIP) PK have been developed, shown to be relatively accurate, and grown in acceptance. However, often the literature on FIP predictions is limited to the assessment of how well the PBPK model predicts PK and does not include the potential utility of these models.
Objectives: The examples shown here are intended to provide an update on the accuracy of the method for prospective predictions. Also, examples of the utility of the models for FIP protocol development and early-stage development are presented.
Methods: Prospective PBPK models are described for 3 Pfizer compounds, BRAF inhibitors PF-07799933 and PF-07284890, and SHP2 inhibitor PF-07284892, based on standard PBPK modeling methods [1,2] and performed in GastroPlus. The models used the mechanistic ACAT model for absorption. The prospective simulations were compared to observed clinical data with a focus on AUC and Cmax but also including an assessment of predictions of dose-limited absorption. An assessment of accuracy of model predictions was performed. Additionally, the predictive accuracy of the PBPK ACAT absorption model used here was compared to another typical approach performed retrospectively, i.e., using a first-order absorption model.
Results: These prospective predictions provided reasonable, useful estimates of human PK before the first observed human PK was collected. Using the mechanistic absorption model provided a valuable tool to improve predictions and better understand properties driving PK, particularly for the compounds with low solubility. Even before human data were available, these bottom-up PBPK models were useful for FIP protocol development. Once comparison to observed clinical data was possible, the models were useful for a range of applications, including understanding data (eg, whether saturable absorption is expected and could that explain observed data) and predicting PK for a different regimen (eg, twice daily administration when the starting dose was once daily) or condition (eg, whether food may cause a difference in exposures).
Conclusions: PBPK models can provide accurate prospective predictions of human PK before any clinical data have been generated, providing a very early human PK model before sufficient clinical PK data for population PK modeling become available. Once clinical PK analyses have been generated, the model can be assessed and further refined as needed. The model can be a useful tool for a variety of applications at a very early stage in clinical development.
Citations: [1] Jones H, et al. Clin Pharmacokinet 2006; 45(5):511-42, PMID: 16640456. [2] Miller N, et al. Clin Pharmacokinet 2019; 58: 727-746, PMID: 30729397.
Keywords: FIH PK prediction, human PK projection, dose-limited absorption, FIH study design
