(M1430-05-32) Prediction of Human Volume of Distribution Using Animal PBPK Modeling and Simulation with Octanol-Water Partition Coefficient or Tissue-to-Plasma Partition Coefficient Scaling

Purpose: Although many methods have been validated for predicting human volume of distribution at steady state (Vd_ss), given that many compounds have multiphasic eliminations, prediction of Vd_ss alone is no longer sufficient. To address such issues, it has been previously reported that interspecies scaling of the logarithm of the octanol-water partition coefficient (logPo/w), optimized in a physiologically based pharmacokinetic (PBPK) model using animal data, can accurately predict the human volume of distribution of compounds with multiphasic eliminations¹. The concept is to optimize logPo/w based on in vivo PK data due to inaccuracies such as conversion from logPo/w to the logarithm of the olive oil-water and inter-experimental errors in measured values. The purpose of this study was to evaluate tissue-to-plasma partition coefficient (Kp) scalar tool and Method 3 implemented in the Simcyp simulator² as novel prediction approach for human volume of distribution. In addition, although the previous report required in vivo PK data from all animal species (rat, dog, and monkey)¹, the prediction accuracy was evaluated using in vivo PK data from single (rat, dog, or monkey) or two animal species (rat and dog or rat and monkey) to efficiently predict with less data. Furthermore, a total of 24 compounds with a wide range of physicochemical properties for which plasma concentration-time profiles were available after intravenous administration, in addition to the 10 compounds previously reported¹, were evaluated for prediction accuracy to better understand the success and limitations of logPo/w and Kp scaling.

Methods: Data collection: The digitized plasma concentration-time profiles and f_u in rat, dog, monkey, and human were obtained from the literature^1,3. The logPo/w and pKa were estimated in silico using ACD/Percepta V14. Total clearance was calculated by two-compartment model analysis using Phoenix WinNonlin V8.3.5. All blood/plasma ratios for each compound were set to 1 due to insufficient data. Animal PBPK modeling: After physicochemical and pharmacokinetic parameters were entered into Simcyp Animal V22, an animal PBPK model was constructed, and logPo/w or Kp scalar was optimized using the cross-species parameter estimate (CSPE) tool. The optimized logPo/w or Kp scalar was determined, respectively for 24 combinations of six animal species selection (rat, dog and monkey; rat and dog; rat and monkey; rat, dog, or monkey) and four Kp prediction methods (Method 1, 2, 3, or best-fit method). The best-fit method was defined as the Kp prediction method with the smallest objective function value (OFV) among Method 1, 2, and 3. Simulation and evaluation of prediction accuracy: The optimized logPo/w or Kp scalar obtained in animal PBPK modeling was input into Simcyp human V22 to simulate the plasma concentration-time profiles. Due to the multiphasic elimination of the predicted plasma concentrations and the difficulty in accurately estimating PK parameters, the prediction accuracy of human volume of distribution was evaluated using root mean squared error (RMSE) as an indicator.

Results: The individual RMSE values for neutral compounds ranged from 0.072 to 0.229, suggesting high prediction accuracy for all approaches. For acidic compounds, the combination of logPo/w optimization using in vivo PK data from monkeys and Method 3 showed the lowest RMSE value of 0.243, indicating high prediction accuracy. In the process of building the monkey PBPK model for acidic compounds, Method 3 was often selected as the best-fit method with the lowest OFV. These findings suggested that Method 3 likely describes the plasma concentration-time profiles of acidic compounds the best of three Kp prediction methods in both monkey and human. For basic compounds, the combination of logPo/w optimization using in vivo PK data from all animal species and the best-fit method showed the lowest RMSE value of 0.328, indicating higher prediction accuracy. Furthermore, the combination of logPo/w optimization using in vivo PK data from rat and monkey and the best-fit method resulted in an RMSE value of 0.338, suggesting that dog PK data are not necessarily required for basic compounds. The individual RMSE values for zwitterionic compounds ranged from 0.435 to 0.642, and no approach was found to provide high prediction accuracy. The RMSE values of Kp scaling tended to be comparable or relatively higher than those of logPo/w scaling, except for zwitterionic compounds.

Conclusion: It was found useful to select the appropriate animal species and Kp prediction methods depending on the physicochemical properties of the compounds except for zwitterionic compounds. This strategic approach is expected to realize drug discovery that is more efficient and promotes reduction of animal experiments. Although Kp scaling showed slightly lower prediction accuracy than logPo/w scaling, it may be an alternative method when logPo/w optimization is difficult. In the current situation where many studies have focused on Vdss and few on compounds with multiphasic eliminations, the results of this study provide important insight into accurately predicting the human volume of distribution or plasma concentration-time profiles of compounds with such characteristics.

References: 1. Shimizu, H., et al. Prediction of Human Distribution Volumes of Compounds in Various Elimination Phases Using Physiologically Based Pharmacokinetic Modeling and Experimental Pharmacokinetics in Animals. Drug Metab Dispos. 47, 114-123 (2019).
2. Fisher, C., Siméon, S., Jamei, M., Gardner, I. & Bois, Y.F. VIVD: Virtual in vitro distribution model for the mechanistic prediction of intracellular concentrations of chemicals in in vitro toxicity assays. Toxicology in Vitro. 58, 42-50 (2019).
3. Yahata, M., Ishii, Y., Nakagawa, T., Watanabe, T. & Miyawaki, I. Applicability of the Oie-Tozer model to predict three types of distribution volume (Vd) in humans: Vd in central compartment, Vd at steady state, and Vd at beta phase. Biopharm Drug Dispos. 41, 151-165 (2020).

Acknowledgements: Helen Musther, Frederic Bois, and Anthonia Onasanwo are employees of Certara UK Ltd. All other authors declared no competing interests for this work.


Figure 1　Individual plots and mean RMSE values for compounds classified by ionization

Abbreviations: RMSE, root mean squared error; N, Neutral; A, Acid; B, Basic; Z, Zwitterion; r, rat; d, dog; mk, monkey.


Table 1　Results of selecting the best-fit method with the lowest OFV among the three Kp prediction methods in the process of animal PBPK modeling

Abbreviations: OFV, objective function value; Kp, tissue-to-plasma partition coefficient; PBPK, physiologically-based pharmacokinetic; logPo/w, the logarithm of the octanol-water partition coefficient; M1, Method 1; M2, Method 2; M3, Method 3.


Figure 2　Direct comparison of RMSE for logPo/w scaling and Kp scaling approaches

Abbreviations: RMSE, root mean squared error; logPo/w, the logarithm of the octanol-water partition coefficient, Kp, tissue-to-plasma partition coefficient; r, rat; d, dog; mk, monkey.