A New Guidance for the Prediction of Hepatic Clearance in the Early Drug Discovery and Development from the in Vitro-to-in Vivo Extrapolation Method and an Approach for Exploring Whether an Albumin-Mediated Hepatic Uptake Phenomenon Could be Present Under in Vivo Conditions

First-in-Human Predictions of Hepatic Clearance for Drugs With the Well-Stirred Model: Comparative Assessment Between Models of Fraction Unbound Based Either on the Free Drug Hypothesis, Albumin-Facilitated Hepatic Uptake or Dynamic Binding Kinetics

The well-stirred model (WSM) is commonly used to predict the hepatic clearance in vivo (CLH) of drugs. The necessary intrinsic clearance of the unbound drug (CLint-in vitro-unbound) is generated in the in vitro assays in the presence of microsomes or hepatocytes but in the absence of plasma proteins. The value of CLint-in vitro-unbound can be extrapolated with the fraction unbound determined in vitro in plasma (fup) only if the fraction unbound in vivo in liver is the same. However, this approach resulted to a systematic underprediction bias of CLH. With the goal of reducing this bias, two new models of fraction unbound were published in this journal. These models estimate the binding kinetics of the rates of association and dissociation of the drug-protein complex and propose that more dissociation in the liver compared to plasma will increase the fraction unbound available for the metabolism. Consequently, these two models generated higher values of fraction unbound, implying a lower underprediction bias of CLH with the WSM. The first model was developed by Poulin et al. and is referring to the value of fup that is adjusted (fu-adjusted) to quantify the effect of a full dissociation of the drug-protein complex at the hepatocyte membrane in accordance with the theory of the albumin-facilitated hepatic uptake. A second model was developed by Yan et al. who presented a dynamic fraction unbound (fu-dynamic) measuring the real dissociation kinetics of the drug-protein complex with a new in vitro assay in the presence and absence of a recombinant liver enzyme in plasma. Therefore, the objective of this study was to make the first comparative assessment between these two models. The results indicate that, in general, the WSM combined with the values of fu-adjusted was the most accurate approach for predicting CLH. The WSM combined with the values of fu-dynamic has underperformed particularly with the acidic and neutral drugs binding to the albumin and presenting a low metabolic turnover in vitro. Therefore, the new in vitro assay for fu-dynamic gave an underprediction bias of CLH for these drug properties. However, the values of fu-adjusted are significantly higher than those values of fu-dynamic, and, this resulted to no underprediction bias, which is reinforcing the theory of the ALB-facilitated hepatic uptake. For the other neutral and acidic drugs, the models of fu-dynamic and fu-adjusted are in closer agreement. Finally, for the basic drugs, the models of fu-adjusted and fu-dynamic as well as a third model only considering a pH gradient effect on fup are almost accurately equivalent.

Using the Dynamic Well-Stirred Model to Extrapolate Hepatic Clearance of Organic Anion-Transporting Polypeptide Transporter Substrates without Assuming Albumin-Mediated Hepatic Drug Uptake

Extrapolating in vivo hepatic clearance from in vitro uptake data is a known challenge, especially for organic anion-transporting polypeptide transporter (OATP) substrates, and the well-stirred model (WSM) commonly yields systematic underpredictions for those anionic drugs, hypothetically due to "albumin-mediated hepatic drug uptake". In the present study, we demonstrate that the WSM incorporating the dynamic free fraction (f D), a measure of drug protein binding affinity, performs reasonably well in predicting hepatic clearance of OATP substrates. For a selection of anionic drugs, including atorvastatin, fluvastatin, pravastatin, rosuvastatin, pitavastatin, cerivastatin, and repaglinide, this dynamic well-stirred model (dWSM) correctly predicts hepatic plasma clearance within 2-fold error for six out of seven OATP substrates examined. The geometric mean of clearance ratios between the predicted and the observed values falls in the range of 1.21-1.38. As expected, the WSM with unbound fraction (f u) systematically underpredicts hepatic clearance with greater than 2-fold error for five out of seven drugs, and the geometric mean of clearance ratios between the predicted and the observed values is in the range of 0.20-0.29. The results suggest that, despite its simplicity, the dWSM operates well for transporter-mediated uptake clearance, and that clearance under-prediction of OATP substrates may not necessarily be associated with the chemical class of the anionic drugs, nor is it a result of albumin-mediated hepatic drug uptake as currently hypothesized. Instead, the superior prediction power of the dWSM confirms the utility of the dynamic free fraction in clearance prediction and the importance of drug plasma binding kinetics in hepatic uptake clearance. SIGNIFICANCE STATEMENT: The traditional well-stirred model (WSM) consistently underpredicts organin anion-transporting polypeptide transporter (OATP)-mediated hepatic uptake clearance, hypothetically due to the albumin-mediated hepatic drug uptake. In this manuscript, we apply the dynamic WSM to extrapolate hepatic clearance of the OATP substrates, and our results show significant improvements in clearance prediction without assuming albumin-mediated hepatic drug uptake.

Therapeutic potential of orally applied KB-R7943 in streptozotocin-induced neuropathy in rats

Both peripheral neuropathy and depression can be viewed as neurodegeneration's consequences of diabetes, at least in part coexisting with or resulting from sodium-calcium dysbalance. This study aims to assess the therapeutic potential of the orally applied reverse-mode inhibitor of the sodium-calcium exchanger (NCX) KB-R7943 in the streptozotocin (STZ) diabetes model in rats. A pilot pharmacokinetic (PK) study with high-performance liquid chromatography with high-resolution tandem mass spectrometric detection revealed higher drug exposure (AUC), lower volume of distribution (Vd) and clearance (Cl), and faster decline of the plasma concentration (ƛ) in rats with diabetes vs. controls. Brain and heart accumulation and urinary excretion of the unmetabolized KB-R7943 at least 24 h were also demonstrated in all rats. However, heart and hippocampus KB-R7943 penetration (AUCtissue/AUCplasma) was higher in controls vs. diabetic rats. The development of thermal, mechanical, and chemical-induced allodynia was assessed with the Cold plate test (CPT), Randall-Stiletto (R-S) test, and 0.5% formalin test (FT). Amitriptyline 10 mg/kg, KB-R7943 5 mg/kg, or 10 mg/kg p.o once daily was applied from the 28th to the 49th day. The body weight, coat status, CPT, R-S, and FT were evaluated on days (-5), 0, and 42. On day 41, a forced swim test and 24-h spontaneous physical activities were assessed. The chronic treatment effects were calculated as % of the maximum. A dose-depended amelioration of neuropathic and depression-like effects was demonstrated. The oral application of KB-R7943 for potentially treating neurodegenerative consequences of diabetes merits further studies. The brain, heart, and kidneys are essential contributors to the PKs of this drug, and their safety involvement needs to be further characterized.

Acetylcholinesterase Inhibition in Rats and Humans Following Acute Fenitrothion Exposure Predicted by Physiologically Based Kinetic Modeling-Facilitated Quantitative In Vitro to In Vivo Extrapolation

Worldwide use of organophosphate pesticides as agricultural chemicals aims to maintain a stable food supply, while their toxicity remains a major public health concern. A common mechanism of acute neurotoxicity following organophosphate pesticide exposure is the inhibition of acetylcholinesterase (AChE). To support Next Generation Risk Assessment for public health upon acute neurotoxicity induced by organophosphate pesticides, physiologically based kinetic (PBK) modeling-facilitated quantitative in vitro to in vivo extrapolation (QIVIVE) approach was employed in this study, with fenitrothion (FNT) as an exemplary organophosphate pesticide. Rat and human PBK models were parametrized with data derived from in silico predictions and in vitro incubations. Then, PBK model-based QIVIVE was performed to convert species-specific concentration-dependent AChE inhibition obtained from in vitro blood assays to corresponding in vivo dose-response curves, from which points of departure (PODs) were derived. The obtained values for rats and humans were comparable with reported no-observed-adverse-effect levels (NOAELs). Humans were found to be more susceptible than rats toward erythrocyte AChE inhibition induced by acute FNT exposure due to interspecies differences in toxicokinetics and toxicodynamics. The described approach adequately predicts toxicokinetics and acute toxicity of FNT, providing a proof-of-principle for applying this approach in a 3R-based chemical risk assessment paradigm.

ADME characterization and PBK model development of 3 highly protein-bound UV filters through topical application

Estimating human exposure in the safety assessment of chemicals is crucial. Physiologically based kinetic (PBK) models which combine information on exposure, physiology, and chemical properties, describing the absorption, distribution, metabolism, and excretion (ADME) processes of a chemical, can be used to calculate internal exposure metrics such as maximum concentration and area under the concentration-time curve in plasma or tissues of a test chemical in next-generation risk assessment. This article demonstrates the development of PBK models for 3 UV filters, specifically octyl methoxycinnamate, octocrylene, and 4-methylbenzylidene camphor. The models were parameterized entirely based on data obtained from in vitro and/or in silico methods in a bottom-up modeling approach and then validated based on human dermal pharmacokinetic (PK) data. The 3 UV filters are "difficult to test" in in vitro test systems due to high lipophilicity, high binding affinity for proteins, and nonspecific binding, for example, toward plastic. This research work presents critical considerations in ADME data generation, interpretation, and parameterization to assure valid PBK model development to increase confidence in using PBK modeling to help make safety decisions in the absence of human PK data. The developed PBK models of the 3 chemicals successfully simulated the plasma concentration profiles of clinical PK data following dermal application, indicating the reliability of the ADME data generated and the parameters determined. The study also provides insights and lessons learned for characterizing ADME and developing PBK models for highly lipophilic and protein-bound chemicals in the future.

Drug Design and Success of Prospective Mouse In Vitro-In Vivo Extrapolation (IVIVE) for Predictions of Plasma Clearance (CLp) from Hepatocyte Intrinsic Clearance (CLint)

Hepatocyte intrinsic clearance (CL_int) and methods of in vitro-in vivo extrapolation (IVIVE) are often used to predict plasma clearance (CL_p) in drug discovery. While the prediction success of this approach is dependent on the chemotype, specific molecular properties and drug design features that govern these outcomes are poorly understood. To address this challenge, we investigated the success of prospective mouse CL_p IVIVE across 2142 chemically diverse compounds. Dilution scaling, which assumes that the free fraction in hepatocyte incubations (f_u,inc) is governed by binding to the 10% of serum in the incubation medium, was used as our default CL_p IVIVE approach. Results show that predictions of CL_p are better for smaller (molecular weight (MW) < 500 Da), less polar (total polar surface area (TPSA) < 100 Ų, hydrogen bond donor (HBD) ≤1, hydrogen bond acceptor (HBA) ≤ 6), lipophilic (log D > 3), and neutral compounds, with low HBD count playing the key role. If compounds are classified according to their chemical space, predictions were good for compounds resembling central nervous system (CNS) drugs [average absolute fold error (AAFE) of 2.05, average fold error (AFE) of 0.90], moderate for classical druglike compounds (according to Lipinski, Veber, and Ghose guidelines; AAFE of 2.55; AFE of 0.68), and poor for nonclassical "beyond the rule of 5" compounds (AAFE of 3.31; AFE of 0.41). From the perspective of measured druglike properties, predictions of CL_p were better for compounds with moderate-to-high hepatocyte CL_int (>10 μL/min/10⁶ cells), high passive cellular permeability (P_app > 100 nm/s), and moderate observed CL_p (5-50 mL/min/kg). Influences of plasma protein binding (f_u,p) and P-glycoprotein (Pgp) apical efflux ratio (AP-ER) were less pronounced. If the extended clearance classification system (ECCS) is applied, predictions were good for class 2 (P_app > 50 nm/s; neutral or basic; AAFE of 2.35; AFE of 0.70) and acceptable for class 1A compounds (AAFE of 2.98; AFE of 0.70). Classes 1B, 3 A/B, and 4 showed poor outcomes (AAFE > 3.80; AFE < 0.60). Functional groups trending toward weaker CL_p IVIVE were esters, carbamates, sulfonamides, carboxylic acids, ketones, primary and secondary amines, primary alcohols, oxetanes, and compounds liable to aldehyde oxidase metabolism, likely due to multifactorial reasons. Multivariate analysis showed that multiple properties are relevant, combining together to define the overall success of CL_p IVIVE. Our results indicate that the current practice of prospective CL_p IVIVE is suitable only for CNS-like compounds and well-behaved classical druglike space (e.g., high permeability or ECCS class 2) without challenging functional groups. Unfortunately, based on existing mouse data, prospective CL_p IVIVE for complex and nonclassical chemotypes is poor and hardly better than random guessing. This is likely due to complexities such as extrahepatic metabolism and transporter-mediated disposition which are poorly captured by this methodology. With small-molecule drug discovery increasingly evolving toward nonclassical and complex chemotypes, existing CL_p IVIVE methodology will require improvement. While empirical correction factors may bridge the gap in the near future, improved and new in vitro assays, data integration models, and machine learning (ML) methods are increasingly needed to address this challenge and reduce the number of nonclinical pharmacokinetic (PK) studies.

Plasma Protein-Mediated Uptake and Contradictions to the Free Drug Hypothesis: A Critical Review

According to the free drug hypothesis (FDH), only free, unbound drug is available to interact with biological targets. This hypothesis is the fundamental principle that continues to explain the vast majority of all pharmacokinetic and pharmacodynamic processes. Under the FDH, the free drug concentration at the target site is considered the driver of pharmacodynamic activity and pharmacokinetic processes. However, deviations from the FDH are observed in hepatic uptake and clearance predictions, where observed unbound intrinsic hepatic clearance (CL_int,u) is larger than expected. Such deviations are commonly observed when plasma proteins are present and form the basis of the so-called plasma protein-mediated uptake effect (PMUE). This review will discuss the basis of plasma protein binding as it pertains to hepatic clearance based on the FDH, as well as several hypotheses that may explain the underlying mechanisms of PMUE. Notably, some, but not all, potential mechanisms remained aligned with the FDH. Finally, we will outline possible experimental strategies to elucidate PMUE mechanisms. Understanding the mechanisms of PMUE and its potential contribution to clearance underprediction is vital to improving the drug development process.

Computational Predictions of Nonclinical Pharmacokinetics at the Drug Design Stage

Although computational predictions of pharmacokinetics (PK) are desirable at the drug design stage, existing approaches are often limited by prediction accuracy and human interpretability. Using a discovery data set of mouse and rat PK studies at Roche (9,685 unique compounds), we performed a proof-of-concept study to predict key PK properties from chemical structure alone, including plasma clearance (CLp), volume of distribution at steady-state (Vss), and oral bioavailability (F). Ten machine learning (ML) models were evaluated, including Single-Task, Multitask, and transfer learning approaches (i.e., pretraining with in vitro data). In addition to prediction accuracy, we emphasized human interpretability of outcomes, especially the quantification of uncertainty, applicability domains, and explanations of predictions in terms of molecular features. Results show that intravenous (IV) PK properties (CLp and Vss) can be predicted with good precision (average absolute fold error, AAFE of 1.96-2.84 depending on data split) and low bias (average fold error, AFE of 0.98-1.36), with AutoGluon, Gaussian Process Regressor (GP), and ChemProp displaying the best performance. Driven by higher complexity of oral PK studies, predictions of F were more challenging, with the best AAFE values of 2.35-2.60 and higher overprediction bias (AFE of 1.45-1.62). Multi-Task approaches and pretraining of ChemProp neural networks with in vitro data showed similar precision to Single-Task models but helped reduce the bias and increase correlations between observations and predictions. A combination of GP-computed prediction variance, molecular clustering, and dimensionality-reduction provided valuable quantitative insights into prediction uncertainty and applicability domains. SHAPley Additive exPlanations (SHAPs) highlighted molecular features contributing to prediction outcomes of Vss, providing explanations that could aid drug design. Combined results show that computational predictions of PK are feasible at the drug design stage, with several ML technologies converging to successfully leverage historical PK data sets. Further studies are needed to unlock the full potential of this approach, especially with respect to data set sizes and quality, transfer learning between in vitro and in vivo data sets, model-independent quantification of uncertainty, and explainability of predictions.

In Vitro-In Vivo Extrapolation by Physiologically Based Kinetic Modeling: Experience With Three Case Studies and Lessons Learned

Physiologically based kinetic (PBK) modeling has been increasingly used since the beginning of the 21st century to support dose selection to be used in preclinical and clinical safety studies in the pharmaceutical sector. For chemical safety assessment, the use of PBK has also found interest, however, to a smaller extent, although an internationally agreed document was published already in 2010 (IPCS/WHO), but at that time, PBK modeling was based mostly on in vivo data as the example in the IPCS/WHO document indicates. Recently, the OECD has published a guidance document which set standards on how to characterize, validate, and report PBK models for regulatory purposes. In the past few years, we gained experience on using in vitro data for performing quantitative in vitro–in vivo extrapolation (QIVIVE), in which biokinetic data play a crucial role to obtain a realistic estimation of human exposure. In addition, pharmaco-/toxicodynamic aspects have been introduced into the approach. Here, three examples with different drugs/chemicals are described, in which different approaches have been applied. The lessons we learned from the exercise are as follows: 1) in vitro conditions should be considered and compared to the in vivo situation, particularly for protein binding; 2) in vitro inhibition of metabolizing enzymes by the formed metabolites should be taken into consideration; and 3) it is important to extrapolate from the in vitro measured intracellular concentration and not from the nominal concentration to the tissue/organ concentration to come up with an appropriate QIVIVE for the relevant adverse effects.

Evidence of the Need for Modified Well-stirred Model in In Vitro to In Vivo Extrapolation

In vitro to in vivo extrapolation (IVIVE), an approach for hepatic clearance (CL_H) prediction used worldwide, remains controversial due to systematic underprediction. Among the various probable factors, the original assumption of the hepatic mathematical model (i.e., the well-stirred model, WSM) may become problematic, leading to the underestimation of drug CL_H. Having a similar prerequisite that the well-stirred conditions are homogenous with perfectly mixed reactants, but using a different driving concentration, the modified well-stirred model (MWSM) stands apart from the WSM. However, we believe that both models should coexist so that the entire well-stirred scenario can be completely illustrated. Consequently, we collected published data from the literature and employed a logistic regression method to differentiate the optimal timing of use between WSM and MWSM in drug CL_H prediction. Generally, variances adopted in the regression, including partition coefficient (logP), fraction unbound (fu), volumes of distribution at steady-state (Vss), and mean residence time (MRT), corresponded to our assumption when protein-facilitated uptake was considered. Furthermore, a new empirical approach was introduced to allow practical use of the MWSM. The results showed that this model could provide a more precise prediction compared to previous empirical approaches. Therefore, these preliminary results not only delineated a more detailed structure and mechanism of MWSM but also highlighted its necessity and potential.

Application of empirical scalars to enable early prediction of human hepatic clearance using IVIVE in drug discovery: an evaluation of 173 drugs

The utilization of in vitro data to predict drug pharmacokinetics (PK) in vivo has been a consistent practice in early drug discovery for decades. However, its success is hampered by mispredictions attributed to uncharacterized biological phenomena/experimental artifacts. Predicted drug clearance (CL) from experimental data (i.e. hepatocyte intrinsic clearance: CLint, fraction unbound in plasma: fu,p) is often systematically underpredicted using the well-stirred model (WSM). The objective of this study was to evaluate using empirical scalars in the WSM to correct for CL mispredictions. Drugs (N=28) were used to generate numerical scalars on CLint (α), and fu,p (β) to minimize the error (AAFE) for CL predictions. These scalars were validated using an additional dataset (N=28 drugs) and applied to a non-redundant AstraZeneca (AZ) dataset available in the literature (N=117 drugs) for a total of 173 compounds. CL predictions using the WSM were improved for most compounds using an α value of 3.66 (~64%<2-fold) compared to no scaling (~46%<2-fold). Similarly, using a β value of 0.55 or combination of α and β scalars (values of 1.74 and 0.66, respectively) resulted in a similar improvement in predictions (~64%<2-fold and ~65%<2-fold, respectively). For highly bound compounds (fu,p{less than or equal to}0.01), AAFE was substantially reduced across all scaling methods. Using the β scalar alone or a combination of α and β appeared optimal; and produce larger magnitude corrections for highly-bound compounds. Some drugs are still disproportionally mispredicted, however the improvements in prediction error and simplicity of applying these scalars suggests its utility for early-stage CL predictions. Significance Statement In early drug discovery, prediction of human clearance using in vitro experimental data plays an essential role in triaging compounds prior to in vivo studies. These predictions have been systematically underestimated. Here we introduce empirical scalars calibrated on the extent of plasma protein binding that appear to improve clearance prediction across multiple datasets. This approach can be used in early phases of drug discovery prior to the availability of pre-clinical data for early quantitative predictions of human clearance.