A novel physiologically based pharmacokinetics (PBPK) approach for human PK prediction

Accurate prediction of pharmacokinetics (PK) in human is essential to assess the potential of drug candidates. One of the major methods for human PK prediction is the physiologically based pharmacokinetics (PBPK) model that is a mechanistic approach using drug specific parameters and realistic organ compartmental structure. However, general “bottom-up” PBPK models have been difficult to get acceptable accuracy in human PK prediction, especially for highly lipophilic and bound drugs. For example, it has been reported that the mechanistic approach proposed by Rodgers et al could overpredict the volume of distribution at steady-state (Vdss) for the drugs with log n-octanol-water partition coefficient (logPow) more than 3 [1, 2]. In our previous study, the PBPK models also underpredicted the human plasma clearance (CLtot) for highly bound drugs probably due to miscalculation of unbound fraction in liver (fu,liver).

The purpose of this study is to develop the “bottom-up” PBPK model with high predictability. The in-house PBPK model consists of 9-compartments which are represented as independent compartments to influence the Vdss value. The liver and kidney were considered to be only the elimination sites, incorporating the metabolic clearance and the passive glomerular filtration of unbound drugs, respectively. Poulin et al [3] reported the “albumin-mediated uptake” theory, assuming the unbound drugs in plasma are transported to liver not only by passive diffusion but also ionic interaction between albumin-drug complex and hepatic cell surface. We applied this theory and calculated the fu,liver by multiplying the fu,plasma value with liver-to-plasma ratio of albumin concentration. The tissue:blood partition coefficients (Kp) values in human were predicted by considering the correlation of the volume of distribution for the unbound drugs between rat and human after the Kp values in rat were corrected by the ratio between the mechanistic approach proposed by Rodgers and experimental rat Vdss values. Using the in-house PBPK model, the plasma concentrations after a single intravenous administration to humans were simulated for 12 drugs. These model compounds are mainly eliminated by hepatic metabolism and have wide ranges of lipophilicity (logPow: 0.3 to 5.1) and fu,plasma (0.003 to 0.91). The prediction accuracy was compared with those in general “bottom-up” PBPK model and empirical scaling methods.

The in-house PBPK model predicted CLtot, Vdss and the plasma concentration at the last measurable time point (Clast) within 2-fold of observed values for 92%, 75% and 83% of those drugs, respectively. These predictabilities were better than those for general PBPK model or empirical scaling methods.

In conclusion, we developed the in-house PBPK model that enables accurate prediction of human PK parameters and the model could accelerate the lead optimization and successful nomination of the candidate compounds in the drug discovery process.

1. Rodgers T, et al. Pharm Res. 2007; 24 (5): 918-33.

2. Poulin P, et al. J Pharm Sci. 2012; 101 (6): 2250-61.

3. Poulin P, et al. J Pharm Sci. 2013 Sep; 102(9): 3239-51.