Anneke Himstedt1, Hermann Rapp1, Peter Stopfer2, Ralf Lotz3, Stefan Scheuerer3, Thomas Arnhold2, Achim Sauer1, Jens Borghardt1
1Global Research DMPK, Global Drug Discovery Sciences, Boehringer Ingelheim Pharma GmbH & Co. KG, 2Clinical Pharmacology, Translational Medicine and Clinical Pharmacology, Boehringer Ingelheim Pharma GmbH & Co. KG, 3Nonclinical Pharmacokinetics, Global Nonclinical Safety and DMPK, Boehringer Ingelheim Pharma GmbH & Co. KG
Introduction: The prediction of human pharmacokinetics (PK) is a key task during drug discovery and drug development, with most published evaluations focusing on single PK parameters, like the mean residence time (MRT) or the half-life, and key points of the PK profile, such as the area under the curve (AUC), and the maximum concentration (Cmax). However, few evaluations exist for the trough concentration (Cmin, which is crucial for estimating the human efficacious dose in many cases) or the shape of the PK profile. Objectives: This study aims to evaluate the quality of human PK predictions by systematically incorporating in vitro absorption, distribution, metabolism, and excretion data (ADME), combined with in vivo PK data, and expert discussions, introducing an approach to assess the entire PK profile rather than focusing solely on single parameters. Methods: This study utilized a set of 40 compounds from the Boehringer Ingelheim pipeline, comprising data from preclinical profiling and single rising dose PK data in healthy human volunteers[1]. A combination of different methodologies for clearance (CL) and volume of distribution at steady state (Vss), including in vitro/in vivo correlation (IVIVC), mean IVIV ratio correction[2], and allometry[3], was employed in a mechanism-based approach where the choice of prediction method was tailored to each compound’s specific characteristics. Empirical PK modeling approaches were applied to predict absorption rates and intercompartmental parameters. The full set of human PK parameters allowed for the simulation of PK profiles, which were then compared to full derived plasma concentration-time profiles based on available clinical data. Due to incomplete human PK data after intravenous administration across compounds, only CL and Vss divided by bioavailability (F) could be compared to observed parameters. The quality of single parameter predictions, as well as the predictions of Cmax and the concentration 24 hours after dosing (C24h, representing trough concentrations), was assessed using the average absolute fold error (AAFE) and median fold error (MFE). The geometric mean of prediction errors in concentration at each simulation step (every six minutes for 24 hours) was calculated to evaluate the quality of the PK profile prediction (profile GMFE). Results: The predictions for CL/F, Vss/F, MRT, half-life, and Cmax showed high quality with most predictions within a 2-fold error range. The Cmin predictions were associated with higher prediction errors, resulting in an MFE and AAFE of 2.45 and 33.4, respectively. Profile predictions showed a median and average profile GMFE of 1.84 and 2.85, respectively, landing between the quality of predictions for Cmax and C24h. Conclusions: Tailoring the prediction method to the available data for each compound individually resulted in better outcomes than applying a uniform method across all compounds and performed well when benchmarked against other published prediction frameworks based on comparable in vitro/in vivo data (incl. PBPK[4], Wajima superposition[5]), and clearly superior to machine learning approaches[6,7], which were trained on available clinical or preclinical data bases. The use of mechanism-based prediction methods enabled the accurate prediction of human PK, including the full plasma concentration–time profile, within a two-fold error margin for most compounds analyzed. This approach underscores the importance of comprehensive evaluation beyond single PK parameters for the accurate prediction of human pharmacokinetics.
[1] Himstedt A et al. Beyond CL and VSS: A comprehensive approach to human pharmacokinetic predictions. Drug Discov Today. 2024;29:104238. [2] Naritomi Y, Terashita S, Kimura S, Suzuki A, Kagayama A, Sugiyama Y. Prediction of human hepatic clearance from in vivo animal experiments and in vitro metabolic studies with liver microsomes from animals and humans. Drug Metab Dispos: Biol fate Chem. 2001;29:1316–24. [3] Mahmood I, Balian JD. Interspecies scaling: Predicting clearance of drugs in humans. Three different approaches. Xenobiotica. 1996;26:887–95. [4] Mao J et al. Shared learning from a physiologically based pharmacokinetic modeling strategy for human pharmacokinetics prediction through retrospective analysis of Genentech compounds. Biopharm Drug Dispos. 2023; [5] Vuppugalla R et al. PhRMA CPCDC initiative on predictive models of human pharmacokinetics, part 4: prediction of plasma concentration-time profiles in human from in vivo preclinical data by using the Wajima approach. J Pharm Sci. 2011;100:4111–26. [6] Parrott N, Manevski N, Olivares-Morales A. Can We Predict Clinical Pharmacokinetics of Highly Lipophilic Compounds by Integration of Machine Learning or In Vitro Data into Physiologically Based Models? A Feasibility Study Based on 12 Development Compounds. Mol Pharmaceut. 2022;19:3858–68. [7] Gruber A, Führer F, Menz S, Diedam H, Göller AH, Schneckener S. Prediction of human pharmacokinetics from chemical structure: combining mechanistic modeling with machine learning. J Pharm Sci. 2023
Reference: PAGE 33 (2025) Abstr 11470 [www.page-meeting.org/?abstract=11470]
Poster: Methodology - Other topics