Transfer of drugs into milk and the clearance of drugs in neonates are critical determinants of the exposure of infants to drugs in breast milk. Models predicting both parameters have been proposed. The objective of this dissertation is to test two models predicting milk to serum ratio and an ontogeny clearance model predicting clearance in the neonate. Predicted milk to serum ratio (M/S) values were generated according to the Atkinson and Begg model. The model did not adequately predict M/S when comparing the predicted values to observed values in the literature. The Fleishaker model was also tested. The model was able to predict whether the drugs appeared in milk by passive diffusion only or whether active transport processes were involved. This model, together with appropriate animal models, is useful in understanding the mechanism of drug transfer into milk. An ontogeny model that predicts clearance was proposed earlier by our laboratory. In order to test the model prediction and assumptions of constant microsomal protein and constant Km for an enzyme-substrate system with age, the male rat was used as an animal model. The ontogeny of Cyp3a1, Cyp3a2, Mdr1a and Mdr1b mRNA was examined in the male rat liver and intestine. The ontogeny pattern of Cyp3a2 mRNA, protein and in vitro Cyp3a activity were found to be similar in male rat liver. The microsomal protein content was found to vary with age in the liver. Km was found to be constant with age for the midazolam 4-hydroxylation by male rat liver microsomes. Scaling factors that extrapolate adult clearance to infant clearance were calculated from in vitro data. The model did not predict the in vivo oral clearance of midazolam for day 7 and 21 age groups from the 112 day age group (adult). The assumption that intestinal availability in the rat pups and adults was equal to unity might not be true resulting in overprediction of rat pup clearance when compared to the adult. Intestinal first pass effect for midazolam in adult rats might be significant. More experiments are needed to further test the model adequacy in clearance prediction.

Abstract Risk assessment of chemicals needs reliable scientific information and one source of information is the characterization of the metabolic fate and toxicokinetics of a chemical. Metabolism is often the most important factor contributing to toxicokinetics. Cytochrome P450 (CYP) enzymes are a superfamily of microsomal proteins playing a pivotal role in xenobiotic metabolism. In the present study, pesticides were used as representative xenobiotics since exposure to pesticides is a global challenge to risk assessment. Human and animal in vitro hepatic models were applied with the advantage of novel analytical techniques (LC/TOF-MS and LC/MS-MS) to elucidate the in vitro metabolism and interaction of selected pesticides. The results of these studies demonstrate that CYP enzymes catalyze the bioactivation of profenofos, diuron and carbosulfan into their more toxic metabolites desthiopropylprofenofos, N-demethyldiuron and carbofuran, respectively. The suspected carcinogenic metabolite of metalaxyl, 2,6-dimethylaniline, was not detected. CYP3A4 and CYP2C19 activities may be important in determining the toxicity arising from exposure to profenofos and carbosulfan. Individuals with high CYP1A2 and CYP2C19 activities might be more susceptible to diuron toxicity. Qualitative results of in vitro metabolism were generally in agreement with the results obtained from the published in vivo data, at least for the active chemical moiety and major metabolites. Considerable differences in the quantities of the metabolites produced within the species, as well as in the ratios of the metabolites among the species, were observed. These findings illustrate that in vitro screening of qualitative and quantitative differences are needed to provide a firm basis for interspecies and in vitro-in vivo extrapolations. Based on our findings, in vitro-in vivo extrapolation based on the elucidation of the in vitro metabolic pattern of pesticides in human and animal hepatic models could be a good model for understanding and extending the results of pesticides metabolism studies to human health risk assessment.

The impact of the intestine in determining the oral bioavailability of drugs has been extensively studied. Its large surface area, metabolic content and positioning at the first site of exposure for orally ingested xenobiotics means its contribution can be significant for certain drugs. However, prediction of the exact metabolic component of the intestine is limited, in part due to limitations in validation of in vitro tools as well as in vitro-in vivo extrapolation scaling factors. Microsomes are a well established in vitro tool for extrapolating hepatic metabolism, however standardised methodologies for preparation in the intestine are limited, in light of complexities in preparation (e.g. presence of multiple non-metabolic cells, proteases and mucus). Therefore, the aims of this study were to establish an optimised method of intestinal microsome preparation via elution in the proximal rat intestine, and to determine microsomal scaling factors by correcting for protein losses during preparation. In addition, to assess species in another preclinical species (dog) and human as well as assessing and regional differences in scaling factors and metabolism. Following optimisation of a reproducible intestinal microsome preparation method in the rat, the importance of heparin in limiting mucosal contamination was established. These microsomes were characterised for total cytochrome P450 (CYP) content, and CYP and uridine 5′-diphosphate glucuronosyltransferase (UGT) activities using maker probes of testosterone and 4-nitrophenol. Loss corrected microsomal scaling factors between two pools of n=9 rats was 9.6±3.5 (recovery 33%). A broad range of compounds (n=25) in terms of metabolic activity and physicochemical properties were screened in rat intestinal microsomes. The prediction accuracy relative to in house generated or literature in vivo estimates of the fraction escaping intestinal metabolism (FG) through in vitro-in vivo extrapolation of observed metabolism and the derived scaling factors and either Caco-2 permeability of physicochemical permeability estimates utilising the Qgut model. In the dog, regional differences in intestinal scaling factors and metabolic activities were explored, as well as relationships between the proximal intestine and liver in matched donors. Positive correlations in both hepatic activity and microsomal scalars were observed. Robust scaling factors were established using the 3 microsomal markers. A total of 24 compounds were screened for hepatic and intestinal metabolism in order to make in vivo estimates of FG, the fraction escaping hepatic metabolism (FH) and oral bioavailability (F). Estimates based on Caco-2 and physicochemical based scaling, as well as utilising a commercial PBPK software platform (ADAM model, Simcyp® v12) were broadly similar with generally reduced prediction accuracy in proximal physicochemical based Qgut scaling, and improved predictions using Caco-2 Qgut or PBPK approaches. Worse predictions were observed for compounds with high protein binding, transporter substrates and/or CYP3A inhibitors. Regional metabolism demonstrated peak metabolism in the proximal intestine, before declining distally. Human intestinal microsomes were prepared for jejunum and ileum tissue. Although samples were limited, regional differences in metabolic activities and scaling factors were also assessed, using correction markers and activity in 23 compounds. In all, 20 compounds overlapped between all three species. Comparison in Fa.FG between rat and human CYP3A substrates showed a modest relationship, however relationships between species and human were generally poor given the observed differing metabolic contributions of testosterone and 4-NP metabolite formation between species limited the observed relationships between species. However, within species, good estimates of oral bioavailability were observed. This is the largest know interspecies comparison of intestinal metabolism and scaling factors with microsomes prepared within the same lab.

Conventional ‘well-stirred’ extrapolation methodology using intrinsic metabolic clearance data from rat liver microsomes poorly predicts in vivo clearance for approximately half of drug discovery compounds. The aim of this present study was to gain a more detailed understanding of the hepatobiliary disposition pathways which influence drug clearance. A set of 77 new chemical entities (NCEs), demonstrating a range of physicochemical properties and in vitro-in vivo clearance correlations (IVIVC), were employed to explore relationships between hepatobiliary disposition pathways in rat and physicochemical, structural and molecular properties of the NCEs. Primary rat hepatocytes with >80% cell viability were successfully isolated from male Han Wistar rats and used to establish in vitro models of drug uptake and biliary efflux. Preliminary studies with cultured primary rat hepatocytes indicated that uptake of d8-taurocholic acid and pitavastatin was time, concentration and temperature dependent. Initial studies with sandwich cultured primary rat hepatocytes demonstrated that cellular accumulation and biliary efflux of [3H]-Taurocholic acid was time and concentration dependent. These in vitro rat hepatocyte models were then used to investigate drug uptake and biliary efflux for all NCEs. In general, NCEs with high (passive) permeability showed better IVIVC and a lower incidence of active uptake and biliary efflux compared to NCEs with lower permeability, suggesting permeability is a key property influencing hepatobiliary drug disposition in rat. Preliminary in silico models analysing structural and molecular descriptors of substrates of active transport in rat hepatocytes were developed and indicated modest potential to highlight clearance pathways beyond hepatic metabolism but further follow up work with larger, more diverse compound sets is warranted to gain confidence in these models. Extended clearance models were investigated to estimate the effect of hepatic transporters on clearance and to predict the overall hepatic clearance of the NCEs. None of these models resulted in a 1 to 1 correlation but in general, improvements in clearance predictions were made when drug transport processes were accounted for. In vivo excretion studies using bile duct cannulated rats demonstrated that NCEs with high permeability and good IVIVC were not directly eliminated in bile or urine as unchanged drug, whereas NCEs with lower permeability and poor IVIVC (> 3-fold under predicted) were all directly eliminated unchanged indicating key drivers of clearance beyond metabolism. In conclusion these investigations confirmed a role for hepatic transporters in clearance but the complex nature of active transport mechanisms and a lack of robust in vitro tools create challenges in the quantitative prediction of hepatobiliary clearance. However, one of the key findings from this research, which is highly applicable in early drug discovery, was to identify the existence of disposition permeability relationships. These can be anticipated by observing physicochemical parameters of NCEs in conjunction with conventional IVIVC, since NCEs that are not highly permeable, possess some hydrophobic characteristics, and which are poor substrates of cytochrome P450 enzymes are more likely to be good substrates of transporters and be directly eliminated in bile and/or urine. The present study focused on exploring hepatobiliary disposition pathways using rat as the investigative species. Whilst there is no guarantee that pathways relevant to rat will be similar to other preclinical species or even humans, an early diagnosis of dominant clearance pathways can guide a more efficient use of the ADME-PK toolbox.

The kidneys have a significant role in drug elimination through both metabolic and excretory routes. Despite a recent paradigm shift towards systems pharmacology approaches, prediction of renal drug disposition using 'bottom-up' and mechanistic modelling approaches remains underdeveloped. Lack of 'gold-standard' in vitro assays and corresponding in vitro-in vivo extrapolation (IVIVE) approaches for prediction of renal metabolic (CLR,met) and excretory (CLR) clearances contribute to this. A comprehensive literature analysis of quantitative physiological data to inform renal IVIVE scaling factors and systems parameters relevant for physiologically based pharmacokinetic (PBPK) kidney models was initially performed to identify existing knowledge gaps. Following this, microsomal protein content in dog kidney cortex (MPPGK) and liver (MPPGL) were measured in 17 samples from the same animal. Mean dog MPPGK (44.0 mg/ g kidney) and MPPGL (63.6 mg/ g liver) obtained using glucose-6-phosphatase activity as the microsomal protein marker where systematically higher than when CYP content was used as the marker (33.9 mg/ g kidney and 41.1 mg/ g liver respectively). Dog MPPGK was lower than MPPGL, with no direct correlation between the organs. In addition to dog, MPPGK and cytosolic protein per gram kidney (CPPGK) were obtained from 31 human samples, which represent the largest dataset currently available. Mean human MPPGK (25.7 mg/ g kidney) and CPPGK (52.7 mg/ g kidney), were measured using glucose-6-phosphatase and glutathione-S-transferase activities as recovery markers, respectively. Activity of prepared kidney microsomes was assessed using mycophenolic acid glucuronidation as a marker. Novel scaling factor of 25.7 mg/ g kidney was applied for IVIVE of mycophenolic acid microsomal glucuronidation data, resulting in a 2-fold increase in scaled intrinsic clearance compared with data scaled by the commonly used literature MPPGK value (12.8 mg/ g kidney). In addition to the microsomal scaling factor, several elements of a modified stereology method were developed for quantifying human proximal tubule cellularity. The methods included implementation of a systematic uniform random sampling protocol and investigation of tinctorial and immunohistochemistry based staining approaches that could be used identify and count proximal tubule cells in histology sections. A range of mechanistic models for prediction of CLR via either tubular reabsorption or active secretion were developed. A novel 5-compartment model for prediction of tubular reabsorption and CLR from Caco-2 apparent permeability data was developed. This model accounted for relevant physiological complexities of the kidney, such as regional differences in tubular filtrate flow rates and tubular surface area, including consideration of the impact of microvilli. The model predicted the CLR of 45 drugs with overall good accuracy (geometric mean fold error of 1.96), although a systematic under-prediction was noted for basic drugs. The novel 5-compartment model represents an important addition to the IVIVE toolbox for physiologically-based prediction of renal tubular reabsorption and CLR and can be implemented in the more complex mechanistic kidney models, as shown in the case of prediction of urine flow dependent CLR of theophylline and caffeine. Final part of the Thesis focused on the refinement of digoxin PBPK kidney model and its ability to predict effect of aging and renal impairment on digoxin CLR. The analysis has identified that reducing either the proximal tubule cellularity or OATP4C1 abundance parameters in the mechanistic model recovers well observed reduced tubular secretion and CLR of digoxin in renal impairment populations whereas no effect of modification of P-gp abundance was observed. Conversely, reducing the proximal tubule cellularity, OATP4C1 abundance or P-gp abundance parameters in the model resulted in negligible change, decreased or increased accumulation of digoxin in proximal tubule cells, respectively. In conclusion, the current study provides to date the most comprehensive kidney microsomal and cytosolic metabolic scaling factors, together with revised database on renal physiological data necessary for quantitative prediction of renal drug disposition. Mechanistic modelling work shown here has highlighted a need for physiological data from different population groups to inform kidney model parameters, in order to improve the scope and utility of such models within the systems pharmacology paradigm.

The elimination of drugs from the body is in many cases performed by the liver. Much could be gained if an accurate prediction of this process could be made early in the development of new drugs. However, for the elimination to occur, the drug molecule needs first to get inside the liver cell. Disposition is the expression used to encapsulate both elimination and distribution. This thesis presents novel approaches and models based on simple in vitro systems for the investigation of processes involved in the hepatic drug disposition. An approach to the estimation of enzyme kinetics based on substrate depletion data from cell fractions was thoroughly evaluated through experiments and simulations. The results that it provided were confirmed to be accurate and robust. In addition, a new experimental setup suitable for a screening environment, i.e., for a reduced number of samples, was generated through optimal experimental design. The optimization suggested that sampling at late time points over a wide range of concentration was the most advantageous. A model, based on data from primary hepatocytes in suspension, for the investigation of cellular disposition of metabolized drugs was developed. Information on the relative importance of metabolism and membrane protein related distribution was obtained by analysis of changes in the kinetics by specific inhibition of the various processes. The model was evaluated by comparing the results to those obtained from an in vivo study analyzed with an especially constructed mechanistic PBPK model. These investigations showed that the suggested model produced good predictions of the relative importance of metabolism and carrier mediated membrane transport for hepatic disposition. In conclusion, new approaches for the investigation of processes involved in hepatic disposition were developed. These methods were shown to be robust and increased the output of information from already commonly implemented in vitro systems.

The liver is the major detoxifying organ, clearing the blood from drugs and other xenobiotics. The extent of hepatic clearance (CL) determines drug exposure and hence, the efficacy and toxicity associated with exposure. Drug-drug interactions (DDIs) that alter the hepatic CL may cause more or less severe outcomes, such as adverse drug reactions. Accurate predictions of drug CL and DDI risk from in vitro data are therefore crucial in drug development. Liver CL depends on several factors including the activities of transporters involved in the hepatic uptake and efflux. The work in this thesis aimed at developing new in vitro and in silico methods to predict hepatic transporter-mediated CL and DDIs in vivo. Particular emphasis was placed on interactions involving the hepatic uptake transporters OATP1B1, OATP1B3, and OATP2B1. These transporters regulate the plasma concentration-time profiles of many drugs including statins. Inhibition of OATP-mediated transport by 225 structurally diverse drugs was investigated in vitro. Several novel inhibitors were identified. The data was used to develop in silico models that could predict OATP inhibitors from molecular structure. Models were developed for static and dynamic predictions of in vivo transporter-mediated drug CL and DDIs. These models rely on a combination of in vitro studies of transport function and mass spectrometry-based quantification of protein expression in the in vitro models and liver tissue. By providing estimations of transporter contributions to the overall hepatic uptake/efflux, the method is expected to improve predictions of transporter-mediated DDIs. Furthermore, proteins of importance for hepatic CL were quantified in liver tissue and isolated hepatocytes. The isolation of hepatocytes from liver tissue was found to be associated with oxidative stress and degradation of transporters and other proteins expressed in the plasma membrane. This has implications for the use of primary hepatocytes as an in vitro model of the liver. Nevertheless, by taking the altered transporter abundance into account using the method developed herein, transport function in hepatocyte experiments can be scaled to the in vivo situation. The concept of protein expression-dependent in vitro-in vivo extrapolations was illustrated using atorvastatin and pitavastatin as model drugs.

La prédiction des paramètres pharmacocinétiques/toxicocinétiques (PK/TK), tels que la clairance intrinsèque (CLint) et la clairance hépatique (CLh) des médicaments, demeure un défi majeur en modélisation quantitative. Selon « l’hypothèse du médicament libre », seul le médicament libre peut traverser la membrane plasmique et la CLh de ce médicament est calculée en fonction de sa fraction libre (fup). Néanmoins, la captation hépatique facilitée par l’albumine (ALB) représente clairement une violation de « l’hypothèse du médicament libre ». Cette captation hépatique se base sur la possibilité que le complexe ALB-médicament puisse assurer un apport supplémentaire en médicament aux hépatocytes. Ainsi, cela pourrait expliquer en grande partie les sous-prédictions observées de CLh. Par ailleurs, certains médicaments peuvent se lier fortement à plusieurs protéines plasmatiques telles que l’ALB et l’alpha-1-glycoprotéine acide (AGP). Ainsi, la forte liaison d’un même médicament à l’ALB, à l’AGP, ou aux deux, pourrait avoir des répercussions bien distinctes sur la prédiction de ces paramètres PK/TK. Cependant, aucune étude n’a été faite pour simuler la différence entre leurs effets. L’objectif principal de cette thèse est donc d’évaluer (avec plus d’exactitude et de précision), pour une série de médicaments, en condition in vivo (ou in situ), ces répercussions en présence des deux protéines plasmatiques, conjointement ou séparément. En outre, il est indispensable de vérifier si une approche générique en modélisation peut être appliquée. Cette thèse est répartie en trois objectifs spécifiques. Le premier est de proposer un arbre décisionnel pour faciliter la sélection des approches prédictives appropriées de CLhin vivo pour des médicaments ayant des caractéristiques différentes. Le second est d’évaluer les répercussions de fortes liaisons aux deux protéines plasmatiques ALB et AGP sur la CLh de deux xénobiotiques choisis (perampanel (PER) et fluoxétine (FLU)) ; ces médicaments ont de fortes affinités pour les deux protéines et un métabolisme exclusif (ou prédominant) dans le foie. Et, le dernier est de développer et valider un nouveau modèle prédictif de CLh pour les xénobiotiques ayant le potentiel de se lier fortement dans le plasma, à l’ALB ainsi qu’à l’AGP. Dans un premier temps, des données in vitro rapportées chez l’humain ont été colligées pour 19 médicaments (substrats des transporteurs OAT2 et OATP1B1), et ont été ensuite utilisées dans six modèles d’extrapolation in vitro-in vivo (IVIVE) pour prédire lesdits paramètres. Après une comparaison statistique, les résultats ont montré que l’approche 2 (c’est-à-dire « fup-adjusted model ») qui se base sur la captation hépatique facilitée par l’ALB, avait la meilleure performance prédictive. Cependant, l’approche 5 (c’est-à-dire « Extended Clearance Model ») qui se base sur le transport facilité, en était une très pertinente à appliquer pour les substrats de transporteurs membranaires. Lesdits substrats seraient potentiellement moins affectés par l’ALB. Ainsi, un arbre décisionnel a été proposé pour choisir rapidement et judicieusement la meilleure approche IVIVE servant à prédire la CLhin vivo pour chaque xénobiotique en présence de l’ALB. Dans un deuxième temps, les médicaments PER et FLU ont été sélectionnés à partir d’une collecte de données (N= 1907 médicaments) en fonction de certains critères (avoir un métabolisme exclusif ou prédominant dans le foie, pas de transport facilité par les transporteurs membranaires, une haute affinité pour les deux protéines ALB et AGP, et un ratio de liaison à l’AGP sur celle à l’ALB proche de l’unité). Cette sélection a été réalisée pour faire des expériences sur des foies isolés et perfusés de rats (IPRL), en présence et en absence des protéines ALB et AGP (c’est-à-dire quatre scénarios IPRL). Les résultats IPRL ont démontré que PER est faiblement à moyennement métabolisé (extraction hépatique= 0,2-0,7), tandis que FLU est fortement métabolisé (extraction hépatique= 0,8-0,99). Le modèle Michaelis-Menten a été ajusté aux cinétiques métaboliques, et différents paramètres Vmax, Km et Km, u ont été obtenus de ce modèle. À de faibles concentrations libres pour les deux médicaments (c’est-à-dire à des concentrations thérapeutiques) et en présence des protéines plasmatiques, les valeurs de CLint non liée ont augmenté pour PER (avec l’ALB et le mélange des deux protéines (MIX)) et FLU (avec l’ALB, l’AGP et le MIX) par rapport à celles obtenues du scénario sans protéine (sauf pour PER avec AGP, lesdites valeurs ont diminué). Par ailleurs, les calculs des ratios CLint (SANS versus AVEC protéine) ont permis d’indiquer l’occurrence d’une facilitation de la captation hépatique de médicaments par l’ALB ou l’AGP. Ces ratios ont aussi permis de vérifier si la cinétique métabolique pour PER et FLU suivait soit « l’hypothèse du médicament libre » soit celle de « la captation hépatique facilitée par les protéines plasmatiques ». Dans un dernier temps, une nouvelle approche prédictive de CLh (approche WO-to-MIX) est développée en se basant sur une nouvelle notion de liaison fractionnelle et en intégrant dans le « fup-adjusted model » de nouveaux paramètres tels que la fraction liée à l’ALB (fB-ALB) et celle liée à l’AGP (fB-AGP) à partir du scénario MIX. Ce modèle est basé sur la captation facilitée par l’ALB. Contrairement à l’approche WO-to-MIX, le « well-stirred model » (ou modèle conventionnel) est basé sur l’hypothèse du médicament libre. Ensuite, les paramètres Vmax et Km obtenus in situ pour PER et FLU lors des expériences IPRL sans protéines, ont été utilisés en combinaison avec le paramètre intrant de la fraction libre ajustée (fup-adjusted) pour le « fup-adjusted model » ou avec la fraction libre (fup) pour le « well-stirred model ». Une comparaison des performances prédictives globales des deux modèles a été faite. Les performances prédictives du nouveau modèle étaient prometteuses, en particulier pour FLU qui montrait le plus haut degré de captation hépatique médiée par l’ALB, par rapport au modèle conventionnel. L’approche WO-to-MIX est une première validation d’un nouveau modèle d’extrapolation proposé pour les médicaments comme FLU qui se lient à l’ALB et à l’AGP. Néanmoins, le modèle conventionnel reste utile à utiliser pour les médicaments comme PER. L’exactitude de prédiction était inférieure pour ce dernier médicament probablement parce que la captation hépatique par l’ALB ne semble pas être maximale, et, par conséquent, l’utilisation de fup-adjusted a surestimé la CLhin vivo. Par conséquent, plus de travail est nécessaire en particulier pour PER. Cette thèse démontre qu’une seule approche générique pour prédire la CLh n’existe pas. Néanmoins, le choix d’une approche IVIVE ayant une performance prédictive satisfaisante est maintenant possible. Les résultats de cette thèse contribuent à : 1) mieux comprendre les répercussions sur les paramètres PK/TK de la forte liaison des médicaments à l’ALB et à l’AGP ; 2) choisir la meilleure approche prédictive de CLh sur la base de l’affinité du xénobiotique (médicament ou contaminant) pour chacune des protéines plasmatiques et des mécanismes impliqués dans le foie ; et 3) prédire la CLh avec précision et exactitude des xénobiotiques qui se lient aux deux protéines plasmatiques. Ces approches IVIVE pour la CLh pourront assurément être intégrées dans des modèles PK/TK à base physiologique pour les xénobiotiques afin d’améliorer la prédiction de leur pharmacocinétique et d’accélérer le processus de développement de médicaments. / The prediction of pharmacokinetic/toxicokinetic (PK/TK) parameters such as intrinsic clearance (CLint) and hepatic clearance (CLh) for highly bound drugs is a major challenge in quantitative modeling. According to the ‘free drug hypothesis’, only the free drug can pass through the plasma membrane and the CLh of this drug is calculated according to its free fraction (fup). Nevertheless, the hepatic uptake facilitated by albumin (ALB) is a violation of the ‘free drug hypothesis’. This facilitated hepatic uptake is based on the possibility that the ALB-drug complex may provide additional drug intake to the hepatocytes. Thus, this could largely explain the underpredictions of CLh. In addition, some drugs can bind extensively in plasma, and to several plasma proteins such as ALB and alpha-1-glycoprotein acid (AGP). Thus, the high binding of the same drug to either ALB or AGP, or to both, could have distinct impacts on the prediction of these PK/TK parameters. However, no study has yet explored how to simulate the difference between these impacts. The main objective of this thesis is therefore to evaluate (with accuracy and precision) for a series of drugs, in the in vivo (or in situ) condition, these impacts in the presence of the two plasma proteins, jointly or separately. Also, it is important to verify if a generic model can be applied. This thesis is divided into three specific objectives. The first is to propose a decision tree to facilitate the selection of appropriate predictive approaches of CLhin vivo for drugs with different characteristics. The second is to assess the impacts of extensive binding to the two plasma proteins ALB and AGP on the CLh of two selected xenobiotics (perampanel (PER) and fluoxetine (FLU)); these drugs have strong affinities to both proteins and an exclusive (or predominant) metabolism in the liver. And the last objective is to develop and validate a new predictive model of CLh for xenobiotics with the potential to bind extensively to ALB as well as to AGP. Firstly, in vitro data obtained in humans were collected for 19 drugs (i.e. substrates of OAT2 and OATP1B1 transporters) and were then used in six in vitro-to-in vivo (IVIVE) extrapolation models to predict these PK/TK parameters. After a statistical comparison, the results showed that the approach 2 (i.e. ‘fup-adjusted model’) that is based on the ALB-facilitated hepatic uptake, had the best predictive performance. However, the approach 5 (i.e. ‘Extended Clearance Model’) that is based on the membrane transporter-mediated uptake, was very relevant to apply for the substrates of membrane transporters. These substrates would potentially be less affected by ALB. Thus, a decision tree has been proposed to quickly and judiciously select the best IVIVE approach to predict CLhin vivo for each xenobiotic in the presence of ALB. Secondly, the PER and FLU drugs were selected from a data collection of 1907 drugs depending on certain criteria (exclusive or predominant metabolism in the liver, no transport facilitated by membrane transporters, high affinity for the two proteins ALB and AGP, and having a binding ratio between AGP and ALB close to the unity). This selection was made to conduct experiments using the isolated and perfused rat liver (IPRL) apparatus, in the presence, and in the absence of the ALB and AGP proteins (i.e. four IPRL scenarios). The IPRL results showed that PER is low to moderately metabolized (hepatic extraction= 0.2-0.7), while FLU is highly metabolized (hepatic extraction= 0.8-0.99). The Michaelis-Menten model was fitted to the obtained metabolic kinetics, and different parameters Vmax, Km and Km, u were obtained from the model. At low free concentrations for both drugs (i.e. therapeutic concentrations) and in the presence of plasma proteins, the values of unbound CLint increased for PER (with ALB and the mixture of the two proteins (MIX)) and FLU (with ALB, AGP and MIX); when compared to those obtained from the protein-free scenario (except for PER with AGP, the unbound CLint values decreased). In addition, the calculations of CLint ratios (WITHOUT versus WITH protein) indicated the occurrence of a hepatic uptake facilitated by ALB or AGP. These ratios also helped in verifying whether the metabolic kinetics for PER and FLU followed either ‘the free drug hypothesis’ or that of ‘plasma protein-facilitated hepatic uptake’. Finally, a new predictive approach of CLh (WO-to-MIX approach) was developed based on a new notion of fractional binding and incorporating new parameters such as the ALB bound fraction (fB-ALB) and the AGP bound fraction (fB-AGP) from the MIX scenario into the ‘fup-adjusted model’. This model is based on the ‘ALB-facilitated hepatic uptake’. Unlike the WO-to-MIX approach, the ‘well-stirred model’ is based on ‘the free drug hypothesis’. Then, the Vmax and Km parameters that were obtained in situ for PER and FLU from the protein-free IPRL experiments, were used in combination with the fup-adjusted input parameter for the ‘fup-adjusted model’ or with the free fraction (fup) for the ‘well-stirred model’. A comparison of the two models’ overall predictive performances was made. The predictive performances of the new model were promising for FLU, which showed the highest degree of ‘ALB-mediated hepatic uptake’, compared to the conventional model. This WO-to-MIX approach is a first validation of a novel extrapolation model suggested for drugs such as FLU that bind to both ALB and AGP. The well-stirred model remains however a useful tool to predict the clearance for drugs such as PER. The prediction accuracy was lower for the latter drug probably because the ALB-mediated hepatic uptake does not seem to be maximal, and, hence, the use of fup-adjusted has overestimated its CLhin vivo. Therefore, more work is needed particularly for PER. This thesis shows that a generic approach to predict the CLh in vivo does not exist. Nevertheless, the choice of an IVIVE approach with satisfactory predictive performances is now possible. The results of this thesis contribute to: 1) better understand the impacts on the PK/TK parameters of extensive drug binding to ALB and AGP; 2) choose the best predictive approach to CLh based on the affinity of xenobiotic (drug or contaminant) to each of the plasma proteins and the mechanisms involved in the liver; and 3) predict accurately and with precision the output CLh of xenobiotics that bind to the two plasma proteins. These IVIVE approaches for CLh can certainly be integrated into physiologically based PK/TK models for xenobiotics to improve the prediction of their pharmacokinetics and to accelerate the drug development process.