1521-009X/43/4/510–522$25.00 http://dx.doi.org/10.1124/dmd.114.062596DRUG METABOLISM AND DISPOSITION Drug Metab Dispos 43:510–522, April 2015Copyright ª 2015 by The American Society for Pharmacology and Experimental Therapeutics

Physiologically Based Pharmacokinetic Modeling for SequentialMetabolism: Effect of CYP2C19 Genetic Polymorphism on

Clopidogrel and Clopidogrel Active Metabolite Pharmacokinetics

Nassim Djebli, David Fabre, Xavier Boulenc, Gérard Fabre, Eric Sultan, and Fabrice Hurbin

Sanofi R&D, Drug Disposition, Disposition Safety and Animal Research, Montpellier, France

Received December 9, 2014; accepted January 21, 2015

ABSTRACT

Clopidogrel is a prodrug that needs to be converted to its activemetabolite (clopi-H4) in two sequential cytochrome P450 (P450)-dependent steps. In the present study, a dynamic physiologicallybased pharmacokinetic (PBPK) model was developed in Simcyp forclopidogrel and clopi-H4 using a specific sequential metabolite modulein four populations with phenotypically different CYP2C19 activity(poor, intermediate, extensive, and ultrarapid metabolizers) receivinga loading dose of 300 mg followed by a maintenance dose of 75 mg.This model was validated using several approaches. First, a compar-ison of predicted-to-observed area under the curve (AUC)0–24 obtainedfrom a randomized crossover study conducted in four balancedCYP2C19-phenotypemetabolizer groupswas performed using a visualpredictive check method. Second, the interindividual and intertrialvariability (on the basis of AUC0–24 comparisons) between the predicted

trials and the observed trial of individuals, for each phenotypicgroup, were compared. Finally, a further validation, on the basis ofdrug-drug–interaction prediction, was performed by comparingobserved values of clopidogrel and clopi-H4 with or withoutdronedarone (moderate CYP3A4 inhibitor) coadministration usinga previously developed and validated physiologically based PBPKdronedarone model. The PBPK model was well validated for bothclopidogrel and its active metabolite clopi-H4, in each CYP2C19-phenotypic group, whatever the treatment period (300-mg loadingdose and 75-mg last maintenance dose). This is the first studyproposing a full dynamic PBPK model able to accurately predictsimultaneously the pharmacokinetics of the parent drug and of itsprimary and secondary metabolites in populations with geneti-cally different activity for a metabolizing enzyme.

Introduction

The antiplatelet agent clopidogrel is a prodrug that is metabolizedby two main metabolic pathways: an esterase-dependent pathwayleading to hydrolysis into an inactive carboxylic acid derivative(85–92% of circulating metabolites) and a cytochrome P450 (P450)-dependent pathway leading to its active metabolite (clopi-H4) (Linset al., 1999; Kazui et al., 2010; Tuffal et al., 2011; Dansette et al.,2012). Clopi-H4 is formed in a two-step oxidative process (Fig. 1)mediated by CYP1A2, CYP2B6, CYP2C19, and CYP3A4 (Kazuiet al., 2010). Clopi-H4 leads to inhibition of adenosine diphosphate–induced aggregation by irreversible binding of the platelet P2Y12 receptor(Bhatt and Topol, 2003).Polymorphisms of CYP2C19 affect both the pharmacodynamic and

pharmacokinetic profiles of clopi-H4, and it has been determined thatthis isoform is one of the major determinants of interindividual variabilityin clopidogrel pharmacodynamic and pharmacokinetic responsiveness(Hulot et al., 2006; Kim et al., 2008; Umemura et al., 2008; Mega et al.,2009). CYP2C19 contribution to the formation of clopi-H4 was confirmedin a randomized crossover study conducted in four balanced CYP2C19-phenotyped metabolizer groups (poor, intermediate, extensive, andultrarapid metabolizers) (Simon et al., 2011). The authors of this study

also performed a meta-analysis on data from 396 healthy subjects andconfirmed that CYP2C19 is the most important polymorphic P450involved in clopi-H4 formation and antiplatelet response, whereasCYP1A2, CYP2C9, CYP2D6, and CYP3A5 played no significantroles. The in vivo impact of CYP3A4 on clopi-H4 pharmacokineticvariability appears to be minimal as observed after coadministrationwith CYP3A4 inhibitors such as ketoconazole and dronedarone (Faridet al., 2007; Sanofi, 2014).We have previously reported a static model (Boulenc et al., 2012),

which can be generalized for more metabolic steps, to estimate the netcontribution of a given polymorphic enzyme to secondary metaboliteformation (or its total inhibition). We also used a dynamic model inthe Simcyp software to compare predictions with the two types ofmodels. The limitation, as was stated in the publication, was that it wasa preliminary physiologically based pharmacokinetic (PBPK) modeland that it was not validated, strictly speaking, with a formal comparisonbetween observed and predicted exposure parameters. The aim of theinvestigation was to use the same metabolized fraction values in thedynamic and static models for comparison of exposure ratios only be-tween the different CYP2C19-phenotyped populations. The PBPKmodels are models consisting of a physiologically realistic compart-mental structure into which input parameters from different sources(e.g., in vitro and in vivo experiments and in silico predictions) can becombined to predict plasma and tissue concentration-time profiles.dx.doi.org/10.1124/dmd.114.062596.

ABBREVIATIONS: AUC, area under the plasma concentration-versus-time curve; BID, twice a day; clopi-H4, active metabolite isomer of clopidogrel(H4); Cmax, maximum plasma concentration; DDI, drug-drug interaction; EM, extensive metabolizer; IM, intermediate metabolizer; MBI, mechanism-based inhibitor; MIIS, secondary metabolite of the substrate in the specific module; P450, cytochrome P450; PBPK, physiologically basedpharmacokinetic; PM, poor metabolizer; SAC, single-adjusting compartment; UM, ultrarapid metabolizer; Vmax, maximum velocity of the metabolizingenzyme; VPC, visual predictive check.

510

at ASPE

T Journals on A

pril 18, 2020dm

d.aspetjournals.orgD

ownloaded from

PBPK models have gained widespread use as a mechanistic and realisticmodeling approach in critical areas of clinical pharmacology, includingpediatrics (Edginton et al., 2006; Khalil and Läer, 2011; Barrett et al.,2012; Leong et al., 2012), pharmacogenetics (Djebli et al., 2009; Yeoet al., 2013), formulation effect (Jamei et al., 2009; Lukacova et al.,2009), organ impairment (Thompson et al., 2009; Johnson et al., 2010),and drug-drug interaction (DDI) (Rostami-Hodjegan et al. 2004; Djebliet al., 2009; Rowland-Yeo et al., 2010; Boulenc and Barberan, 2011,2012; Vieira et al., 2012). PBPK tools that incorporate interindividualvariability of intrinsic factors, such as Simcyp, can improve evaluationof pharmacokinetic interindividual variability and consequentlyanticipate DDI impact and better determine optimal formulation,dosing regimen, and sampling schemes in the general population aswell as in special populations (e.g., renal impaired patients, differentethnic groups, etc.).In the present study, a dynamic PBPK model was developed and

validated for clopidogrel and for its active metabolite clopi-H4, usingthe specific sequential metabolite module, in the four CYP2C19 phenotypemetabolizers groups (poor, intermediate, extensive, and ultrarapidmetabolizers).

Materials and Methods

Physiologically Based Pharmacokinetics Model Building

Simcyp algorithms (version 10.20 SE; Simcyp Limited, Sheffield, UK, aCERTARA company) were used to predict clopidogrel and clopi-H4 exposuresin CYP2C19 PM (poor metabolizers), intermediate metabolizers (IM ), extensivemetabolizers (EM), and ultra-rapid metabolizers (UM).

In addition, a specific module was implemented, through collaborationbetween Simcyp Limited and Sanofi, and used for the present analysis to enablemodeling of a compound with a secondary metabolite. This module is availableas free add-on package for all Simcyp users.

Assumptions of the Secondary Metabolite Module. The clopidogrel PBPKmodel involved the development of a module that incorporated a secondarymetabolite formed sequentially from a primary metabolite. The following assumptionswere made:

The secondary metabolite is only formed from a primary metabolite of thesubstrate.

The secondary metabolite is available for metabolism and inhibitioninstantaneously.

The substrate is given orally or intravenously and can be administered asa single dose or multiple doses.

As for the primary metabolite, the gut transporters kinetic parameterscould not be applied for the secondary metabolite.

The distribution of the secondary metabolite was described by a minimalPBPK model. As a result, transporter kinetic models (e.g., hepatictransporters), if any, could not be applied.

Mutual interactions (competitive inhibition, mechanism-based inhibition,and induction) between the secondary metabolite and other compounds(substrate, the primary metabolite of the substrate, inhibitors, and theprimary metabolite of the inhibitor) were considered, as well asautoinhibition, via mechanism-based inhibition and autoinduction.

Implementation of the Secondary Metabolite Module. It was assumedthat the formed secondary metabolite was instantaneously available for furtherelimination (metabolism and excretion) and interactions. MIIS was used torepresent the secondary metabolite of the substrate.

The fraction of secondary metabolite escaping first-pass metabolism in thegut, Fg–MIIS, could be calculated in the same way as for the primary metabolite:

Fg2MIIS ¼ Qvilli

Qvilli þ fugut2MIISCLintG2MIISð1Þ

where Qvilli was the villi blood flow, fugut–MIIS, and CLintG–MIIS were thesecondary metabolite unbound fraction in the gut and the total gut intrinsicclearance, respectively. The formation rate of the secondary metabolite in thegut was described by:

AMIIS ¼ AP   +M

n¼1

fugut2PCLintG2P2 n

Qgut þ fugut2PCLintG2Pð2Þ

where AP was the formation rate of the primary metabolite in the gut, fugut–P isthe unbound fraction of the primary metabolite in the gut, Qgut was the gutblood flow, CLintG–P was the total gut clearance of the primary metabolite, andCLintG–P–n was the nth metabolic pathway of the primary metabolite to form thesecondary metabolite. The intrinsic clearances were corrected for nonspecificbinding and if Vmax/Km values were provided, CLintG–P–n was computed as:

CL int G2P2 n¼VmaxG2P2 n

KmP2 n þ fugut2PPpvð3Þ

where VmaxG–P–n and KMP–n were the gut metabolism kinetic parameters of thenth pathway, fugut–P was the fraction unbound in the gut, and Ppv was theprimary metabolite concentration in portal vein.

The secondary metabolite portal vein concentration was determined using:

dMIISpvdt

¼ 1Vpv

�QpvðMIISsys 2MIISpvÞ þ Fg2MIISAMIISPOF

� ð4Þ

where POF was 0 when the parent drug was given by intravenous route, and 1when the parent drug was given by oral route. Also, MIISsys and MIISpv werethe secondary metabolite systemic and portal vein plasma concentrations andFg–MIIS is the secondary metabolite fraction escaping gut metabolism, and Qpv,the portal vein blood flow.

The secondary metabolite liver was defined as:

dMIISLivdt

¼ 1VLiv

QpvMIISpv þ QhaMIISsys þ UptakeP   fub2PPLiv +M

n¼1CLint uH2P2 n

2CLintH2MIIS  UptakeMIIS fub2MIISMIISliv 2QhMIISLiv

264

375

ð5Þ

where Qpv and Qha were the portal vein and hepatic artery blood flows,UptakeP, and UptakeMIIS were the active uptake into hepatocyte for the primaryand secondary metabolites, fub–P and fub–MIIS were the unbound fraction of drugin blood of the primary and secondary metabolites, PLiv was the primarymetabolite concentration in the liver, and MIISliv was the liver concentration ofthe secondary metabolite.

Fig. 1. Biotransformation pathway of clopidogrel leading to its pharmacologically active metabolite (H4) via 2-oxo-clopidogrel.

PBPK Modeling for Clopidogrel and Its Active Metabolite 511

at ASPE

T Journals on A

pril 18, 2020dm

d.aspetjournals.orgD

ownloaded from

TABLE 1

Physicochemical and in vitro ADME parameters used in Simcyp for clopidogrel, 2-oxo-clopidogrel, and active metabolite (clopi-H4)

Parameters Value Implemented in Simcyp Source Data

Clopidogrel

Physicochemical

MW (g/mol) 321.8 Internal dataLog Po:w 3.89

Compound type Monoprotic acidPka 4.55

Hematocrit (%) 45.0 Simcyp library

Absorption

Absorption model/input type First order —

fa; Ka (h21) 0.5; 0.5 Internal dataPeff, man (10

24 cm/s) 0.466 Predicted Pcaco2 = 0.399�1026 cm/sFormulation Solution —

fuGut 0.02 Set equal to fup

Distribution

Distribution model Full PBPK model —

Vss (l/kg) Predicted, 0.217 Prediction method Rodgers et al. (2005a, b); Rodgers andRowland (2006, 2007)

B/P ratio Predicted; 0.72 Prediction method Uchimura et al. (2010)fup 0.02 Internal data

Metabolism Clearance type Enzyme kineticsIn vitro metabolic system Human recombinant P450 isoforms Kazui et al. (2010)

rhCYP1A2 Vmax (pmol/minper pmol)

2.27

KM (mM) 1.58fumic 0.015

rhCYP2B6 Vmax (pmol/minper pmol)

7.66

KM (mM) 2.08fumic 0.015

rhCYP2C19 Vmax (pmol/minper pmol)

7.52

KM (mM) 1.12fumic 0.015 N.B.: fumic obtained using the prediction toolbox and refined by

sensitivity analysisAdditional systemic clearance (l/h) 600 Representing about 90% of clopidogrel clearance (esterase-

dependent pathway)

2-Oxo-clopidogrel (primary metabolite)

Physicochemical

MW (g/mol) 337.8 Internal dataLog Po:w 2.96

Compound type Monoprotic acidPka 3.41

Hematocrit (%) 45.0 Simcyp library

Distribution

Distribution model Minimal PBPK model —

Vss (l/kg) 0.100 Sensitivity analysisB/P ratio Predicted; 1.00 Prediction method Uchimura et al. (2010)

fup Predicted; 0.0310 Prediction method Lobell and Sivarajah (2003)

Metabolism

Clearance type Enzyme kineticsIn vitro metabolic system Human recombinant P450 isoforms Kazui et al. (2010)

rhCYP2B6 Vmax (pmol/minper pmol)

2.48

KM (mM) 1.62fumic 0.180

rhCYP2C9 Vmax (pmol/minper pmol)

0.855

KM (mM) 18.1fumic 0.180

rhCYP2C19 Vmax (pmol/minper pmol)

9.06

KM (mM) 12.1fumic 0.180

rhCYP3A4 Vmax (pmol/minper pmol)

3.63

KM (mM) 27.8fumic 0.180 N.B.: fumic obtained using the prediction toolbox

and refined by sensitivity analysisAdditional clearance HLM Clint (ml/min

per mg)50 Representing about 50% of the total clearance

(esterase-dependent pathway)fumic 0.180

Active uptake into hepatocyte 2 Sensitivity analysis

Clopi-H4 (secondary metabolite = active metabolite)

PhysicochemicalMW (g/mol) 355.8 Internal dataLog Po:w 3.60

(continued )

512 Djebli et al.

at ASPE

T Journals on A

pril 18, 2020dm

d.aspetjournals.orgD

ownloaded from

The secondary metabolite systemic compartment was defined as:

dMIISsysdt

¼ 1Vd2MIIS

�Qh

�MIISLiv 2MIISsys

�2CLr2MIIS

BPMIISMIISsys

�ð6Þ

where Vd–MIIS was the secondary metabolite volume of distribution atsteady-state, Qh was the hepatic blood flow, CLr–MIIS was the secondarymetabolite renal clearance and BPMIIS was the secondary metabolite blood-to-plasma ratio, MIISLiv was the liver concentration of the secondary

metabolite, and MIISsys was the secondary metabolite systemic vein plasmaconcentration;

Depending on the extent of sequential metabolism, a certain amount of thesecondarily formed metabolite will go to systemic circulation.

Input Data

Simcyp model was set up using clopidogrel and its metabolites (i.e., 2-oxo-clopidogrel, the primary P450-dependent metabolite, and clopi-H4, the

TABLE 1—Continued

Parameters Value Implemented in Simcyp Source Data

Compound type Diprotic acidPka 1; Pka 2 3.20; 5.10Hematocrit (%) 45.0 Simcyp library

Distribution

Distribution model Minimal PBPK model —

Vss (l/kg) Predicted; 0.230 Prediction method Rodgers et al. (2005a, b); Rodgers andRowland (2006, 2007)

B/P ratio Predicted; 0.820 Prediction method Uchimura et al. (2010)fup 0.018 Prediction method Lobell and Sivarajah (2003)

ClearanceClearance type In vivo clearance Representing the direct irreversible covalent binding to platelets

CLpo (l/h) 500

B/P, blood-to-plasma ratio; Clint, intrinsic clearance; CLpo, oral clearance; fa, fraction absorbed; fumic, unbound fraction in microsomes; fup, unbound fraction in plasma; KM, Michaelis-Mentencoefficient; Peff, effective permeability; Po:w, octanol/water partition coefficient; Vss, steady state.

TABLE 2

Physicochemical and in vitro ADME parameters used in Simcyp for dronedarone

Parameters Value Implemented in Simcyp Source Data

Physicochemical

MW (g/mol) 557 Analytical dossierLog Po:w 7.80

Compound type Monoprotic basePka 9.30

Hematocrit (%) 45.0 Simcyp library

Absorption

Absorption model/input type ADAM model —

fa; Ka (h21) Predicted; 0.898; 0.816 Predicted using ADAM modelPeff, (10

24 cm/s) 1.98 Predicted Pcaco2 = 5.30 �1026 cm/s

Formulation Solid; Controlled-Released —

Dissolution-timeprofile

Time (h): 0, 0.083, 0.167, 0.25, 0.33, 0.42,0.5, 0.75, 1 and 1.5

Dissolution (%): 0, 6.6, 12.8, 28.5, 38.9, 47.7,55.2, 75.9, 92.2 and 100

Analytical dossier

Solubility– pHprofile

pH: 3, 4, 5, 6 and 7 Solubility (mg/ml): 1.6, 1.6, 1.5, 0.1 and 0.05 Analytical dossier

fuGut 1.00 —

Distribution

Distribution model Minimal PBPK model —

Vss (l/kg) 10 Analytical dossier; PopPkanalysis

B:P ratio 1.00 Analytical dossierfup 0.003

Metabolism

Clearance type Enzyme kineticsIn vitro metabolic system Recombinant Analytical dossier

rhCYP3A4 Vmax

(pmol/min per pmol)13.7

KM (mM) 4.2fumic 0.0011

rhCYP3A5 Vmax

(pmol/min per pmol)4.87

KM (mM) 3.10fumic 0.0011

Additional liverclearance

Clint(l/min per mg)

40

fumic 0.0011 N.B.: fumic obtained using theprediction toolbox and refined

by sensitivity analysis

Interaction

CYP2B6 (comp.inhibition)

Ki (mM) 12.0 Analytical dossierfumic 0.0011

CYP2D6 (comp.inhibition)

Ki (mM) 5.0fumic 0.0011

CYP3A4 (MBI) Ki (mM) 2.44Kinact (h

21) 9.16fumic 0.0011

fa, fraction absorbed; fup, unbound fraction in plasma; fumic, unbound fraction in microsomes; Ka, first-order rate constant; Ki and Kinact, mechanism-based inactivation parameters; Peff, effectivepermeability in human; Vss, steady state.

PBPK Modeling for Clopidogrel and Its Active Metabolite 513

at ASPE

T Journals on A

pril 18, 2020dm

d.aspetjournals.orgD

ownloaded from

secondary metabolite), with the physicochemical, absorption, distribution, andclearance parameters described in Table 1.

Physicochemical Parameters. As physicochemical input parameters, themolecular weight, the chemical nature, the Pka, and the LogP values were usedfor clopidogrel, 2-oxo-clopidogrel, and clopi-H4.

Absorption. The absorption process was described for clopidogrel only. Thefirst-order absorption model was selected with a fraction absorbed (fa), a first-order rate constant (Ka), and an effective permeability (Peff) in human wereused as input parameters. The formulation was considered as a solution.

Distribution. Two PBPK distribution models were available in Simcyp: theminimal and the full PBPK models. The minimal PBPK model can be describedas a “lumped” model that has only three compartments, when there is no single-adjusting compartment (i.e., peripheral compartment), predicting the systemic,portal vein, and liver concentrations. The full PBPK distribution model proposeda number of time-based differential equations to simulate the concentrations invarious organ compartments: blood (plasma), adipose, bone, brain, gut, heart,kidney, liver, lung, muscle, skin, and spleen. The interindividual variability oftissue volume was estimated taking account of age, sex, weight, and height. Thedistribution is assumed to be perfusion-limited, using the full PBPK model, unlessthe membrane transporters are taken into account, whereby permeability-limiteddistribution is handled in the liver, kidney, and brain. For the current analysis, thefull PBPK distribution model was selected for clopidogrel, and the minimalPBPK model was selected for 2-oxo-clopidogrel and for clopi-H4. The volumesof distribution at steady-state (Vss) were 0.217, 0.10, and 0.23 l/kg forclopidogrel, 2-oxo-clopiogrel, and clopi-H4, respectively. These values werepredicted using the model proposed by Rodgers and Rowland (Rodgers et al.,2005a, b; Rodgers and Rowland 2006, 2007), with the exception of 2-oxo-clopidogrel, for which the sensitivity analysis model was used to refine this valueon the basis of its impact on the observed clopidogrel and clopi-H4 exposures[maximum plasma concentration (Cmax) and area under the curve (AUC)]. Theblood-to-plasma ratio was predicted using the model proposed by Uchimura et al.(2010): 0.72, 1.00, and 0.82 for clopidogrel, 2-oxo-clopidogrel, and clopi-H4,

respectively. The unbound fraction in plasma (fup) was set to 0.02 for clopidogrelas stated in the analytical dossier and to 0.031 and 0.018 for 2-oxo-clopidogreland clopi-H4, respectively, using the model proposed by Lobell and Sivarajah(2003).

Elimination. For clopidogrel metabolism, enzyme kinetics informationusing human recombinant P450 isoforms was selected. The maximum velocityof the metabolizing enzyme (Vmax), and Michaelis-Menten coefficient value(KM) from Kazui et al. (2010) were used. The unbound fraction in humanhepatic microsomes (fumic) of 0.015 was predicted using the QSAR modelpublished by Gao et al. (2008), in a first step, and refined using the sensitivityanalysis module on the basis of observed clopidogrel and clopi-H4 exposuresin a second step. Moreover, an additional systemic clearance of 600 l/h wasconsidered, representing the esterase-mediated clearance using the retrogrademodel (about 90% of clopidogrel total clearance).

The enzyme kinetic information (Vmax and KM) from Kazui et al. (2010)using human recombinant P450 isoforms was also used for 2-oxo-clopidogrel.Moreover, an additional clearance of 50 ml/min per milligram was consideredfor 2-oxo-clopidogrel, representing the esterase-mediated clearance (about 50%of the total 2-oxo-clopidogrel clearance). An active uptake into hepatocytes of2 was set for 2-oxo-clopidogrel using the sensitivity analysis module.

Regarding clopi-H4, an in vivo clearance of 500 l/h was programmed intoSimcyp. This value represented the immediate direct irreversible binding of thisactive metabolite to platelets.

Dronedarone PBPK Model. The dronedarone Simcyp model has beenpreviously developed and validated for a large range of doses (200–1600 mgBID). The input parameters are detailed in Table 2.

This model accurately predicted the pharmacokinetics of dronedarone andcorrectly took into account the nonlinearity of dronedarone pharmacokinetics(Fig. 2). This nonlinearity resulted from the moderate mechanism-based inhibitionof CYP3A4, which is the main isoform involved in dronedarone clearance itself.The main purpose of the dronedarone model was its application as a guide fordose selection in pediatrics. This model was also used to evaluate the feasibility of

Fig. 2. Comparison of observed versus predicted dronedarone AUC0–12 and Cmax at steady state after 400-mg BID administration (A and B) and 200–1600-mg BIDadministration (C and D).

514 Djebli et al.

at ASPE

T Journals on A

pril 18, 2020dm

d.aspetjournals.orgD

ownloaded from

a sustained-release formulation for a 800-mg once-daily administration instead ofthe marketed 400 mg BID.

Simcyp Simulations

The simulations were performed using four virtual populations (10 trials of10 individuals each) of 100 healthy volunteers aged between 20 and 50 witha male/female ratio of 50/50, in fasted condition, representing PM-, IM-, EM-,and UM-CYP2C19 individuals. The number of virtual subjects (10 trials of 10subject in each trial) was selected on the basis of the subject number in study 1(10 subjects in each CYP2C19-phenotyped group) to optimize the relevance ofthe comparison with observed values at the model validation step.

The difference between these four CYP2C19-phenotyped groups resultedmainly from mean liver CYP2C19 abundance. For EM individuals, the meanliver abundance of CYP2C19 was set in the Simcyp library to 14 pmole/mg ofprotein, whereas it was defined as 0 pmole/mg for PM individuals. RegardingIM and UM individuals, the liver abundances of CYP2C19 were set to 10 and18 pmole/mg, respectively, after repeated simulations using a conditionalsensitivity analysis with mean value for IM individuals ranging between 0 and14 pmol/mg (between PM and EM individuals) and for UM individuals higherthan 14 pmol/mg (higher than EM individuals).

A customized design similar to that of study 1 was used, with a 5-day treatmentduration: a loading dose of 300 mg clopidogrel at day 1 and a maintenance dose of75 mg once daily clopidogrel from day 2 to day 5.

The simulations could be performed in Simcyp using either the “PK/PDparameters” or the “PK/PD profiles” options. The PK/PD parameters option

was the only way to estimate the relative enzyme contribution (static modeling).This option excludes time- and, in some cases, concentration-dependentphenomena. The PK/PD profiles options provided time- and concentration-dependent predictions. The inter- and intramoieties interactions (metabolite,inhibitor, inducer, effect of the substrate on its own metabolism) were alsotaken into account, as well as the organ parameters (e.g., changes in enzymessynthesis or degradation rates following administration of an inducer and/ora mechanism-based inhibitor).

In the present analysis, the PK/PD parameters option was used to estimatethe relative enzyme contribution for both clopidogrel and 2-oxo-clopidogrelmetabolism, and the PK/PD profiles option was used to predict the PK profilesof the compounds in the different CYP2C19-phenotyped groups.

Clinical Trials

Two clinical studies were used for model validation purposes in the presentanalysis. The first one aimed at validating the contribution of CYP2C19 inclopidogrel metabolism and clopi-H4 formation. The second study aimed atvalidating the model in terms of CYP3A4-based DDI.

The first clinical study (study 1) was conducted to compare clopidogrel andclopi-H4 in four CYP2C19-defined metabolizer groups. This single-center,randomized, placebo-controlled, two-treatment, two-period crossover study infour CYP2C19-defined metabolizer groups (PM, IM, EM, and UM) wasconducted to determine whether CYP2C19 polymorphisms affected thepharmacokinetics of clopidogrel and clopi-H4 after clopidogrel oral adminis-tration of 300-mg loading dose followed by 75 mg for 4 days or 600-mg

Fig. 3. Mean predicted contribution of P450 isoforms andesterase (additional clearance) to clopidogrel (A) and to 2-oxo-clopidogrel (B) metabolism in CYP2C19-extensive metabolizers.

TABLE 3

Observed-versus-predicted ratio estimates (90% CI) of Cmax and AUC0–24 for clopidogrel and clopi-H4 without and withdronedarone coadministration (400 mg BID) at day 1 (300-mg loading dose) and at day 5 (75-mg maintenance dose)

ParameterClopidogrel Secondary metabolite (clopi-H4)

Observed (n = 63) Simcyp (n = 100) Observed (n = 63) Simcyp (n = 100)

Ratio estimate (90% CI) at day 1 (300-mg clopidogrel loading dose)Cmax 0.89 (0.80–0.99) 1.01 (1.01–1.01) 0.93 (0.84–1.04) 0.72 (0.71–0.76)AUC0–24 1.00 (0.94–1.07) 1.00 (1.00–1.00) 1.09 (0.65–1.20) 0.73 (0.72–0.76)

Ratio estimate (90% CI) at day 5 (75-mg clopidogrel maintenance dose)Cmax 0.96 (0.85–1.08) 1.00 (1.00–1.00) 0.81 (0.73–0.89) 0.72 (0.71–0.75)AUC0–24 1.03 (0.96–1.11) 1.00 (1.00–1.00) 1.05 (0.67–1.22) 0.72 (0.71–0.76)

CI, confidence interval.

PBPK Modeling for Clopidogrel and Its Active Metabolite 515

at ASPE

T Journals on A

pril 18, 2020dm

d.aspetjournals.orgD

ownloaded from

loading dose followed by 150 mg for 4 days (Simon et al., 2011). The numberof subjects was 40 (10 per group).

The second clinical study (study 2) was conducted to evaluate the impactof dronedarone, a moderate CYP3A4 inhibitor, on the pharmacokinetics ofclopidogrel and clopi-H4 pharmacokinetics. This was a randomized, single-center, double-blind, placebo-controlled, repeated-dose, two-treatment two-period, two-sequence crossover pharmacokinetic interaction study conducted inFrance. The two treatment periods, separated by a 7-day washout period, wereas follows: repeated doses of 400 mg BID dronedarone or placebo (for 14 days)on repeated doses of clopidogrel (300-mg loading dose followed by 75-mg

maintenance dose for 4 days) started the 10th day after dronedarone initiation.Dronedarone was administered for 9 days before clopidogrel initiation to achievesteady-state pharmacokinetic conditions. Regardless of the sequence, a washoutduration between of the two periods was at least 7 days to ensure that plateletaggregation returned to baseline during the $17 days between the last clopidogreladministration (day 14) in the first treatment period and the loading dose of clopidogrel(day 10) in the second treatment period. Healthy male subjects 18–65 years of age wereeligible for enrollment if they provided informed consent; had a body weight of 50–95 kg, body mass index of 18-28 kg/m2; and no contraindication to clopidogrel anddronedarone. Only CYP2C19 EM individuals were considered in this analysis.

TABLE 4

Observed-versus-predicted mean (SD) of Cmax and AUC0–24 of clopi-H4 without and with dronedarone coadministration(400 mg BID) at day 1 (300-mg loading dose) and at day 5 (75-mg last maintenance dose)

Data are presented as mean (S.D.).

ParameterWithout dronedarone With dronedarone

Obs. (n = 63) Sim. (n = 100) Obs. (n = 63) Sim. (n = 100)

Day 1 (300-mg clopidogrel loading dose)Cmax (ng/ml) 17.0 (10.5) 16.8 (10.6) 17.6 (14.2) 12.1 (6.82)AUC0–24 (ng × h21 × ml21) 37.9 (19.2) 59.5 (37.4) 43.1 (27.9) 43.2 (26.5)

Day 5 (last 75-mg clopidogrel last maintenance dose)Cmax (ng/ml) 6.63 (4.08) 6.56 (4.43) 5.21 (2.89) 4.66 (2.95)AUC0–24 (ng.h

21 × ml21) 12.8 (5.12) 20.4 (13.5) 14.2 (8.11) 14.7 (9.67)

Fig. 4. Predicted and observed median AUC0–24 and error bars for clopidogrel with the 300-mg loading dose (day 1) in CYP2C19-poor, -intermediate, -extensive, and-ultrarapid metabolizers.

516 Djebli et al.

at ASPE

T Journals on A

pril 18, 2020dm

d.aspetjournals.orgD

ownloaded from

Clopidogrel and clopi-H4 analyses: plasma samples for pharmacokineticassessment of unchanged clopidogrel and clopi-H4 were collected after loadingdose and last maintenance dose of each of the two periods at T0 (time ofclopidogrel administration) and at time points (in hours) T0.25, T0.5, T1, T1.5,T2, T3, T4, T6, T10, T16, and T24 under conditions already described forstudy 1 (Simon et al., 2011). Clopidogrel and clopi-H4 plasma concentrationswere assayed by Sanofi (Bridgewater, NJ; Malvern, PA; and Montpellier, France),using validated liquid chromatography-tandem mass spectrometry with lower limitsof quantification of 5 pg/ml and 0.5 ng/mL, respectively (Tuffal et al., 2011).

Cmax and area under the plasma concentration-versus-time curve from T0 toT24 using the trapezoidal rule (AUC0–24) were calculated using noncompart-mental techniques using PKDMS Version 2.0, incorporating WinNonlinProfessional Version 5.2.1 (Pharsight, Mountain View, CA).

Validation of the PBPK Model

Comparison of Predicted-to-Observed Clopidogrel and Clopi-H4 AUC0–24

Values. The predicted AUC0–24 values of clopidogrel and clopi-H4, using thePBPK model, at loading dose (day 1) and maintenance dose (day 5) were comparedwith those observed in study 1. This comparison was performed for eachCYP2C19-phenotyped group.

As previously mentioned, for each simulation (i.e., each CYP2C19-phenotypedgroup), 10 trials of 10 virtual individuals in each trial were generated. The medianAUC0–24 value and error bar of the group of “real” subjects was presented togetherwith the median and error bar of each virtual trial. This representation allowedboth the predicted interindividual variability and intertrial variability to be well

evaluated and confirmed that the group of “real” patients behaved as one of thevirtual trials.

Visual Predictive Check. A visual predictive check (VPC) was adapted toPBPK and performed to allow a graphical qualification of the clopidogrel/2-oxo-clopidogrel/clopi-H4 Simcyp model. This evaluation method of the model’spredictive performance by comparing the predictions to clinical data were describedas the reference at the 2012 FDA Pediatric Advisory Committee (US Food andDrug Administration, 2012). Briefly, to graphically validate the model’s pre-dictability, the 50th (median), 5th and 95th percentiles of predicted concentration-time profiles (obtained from the Simcyp simulations of 100 generated virtualindividuals for each CYP2C19-phenotyped group and for each compound) werepresented with the observed data obtained in study 1.

Validation Based on DDI prediction. The last validation was performed onthe basis of DDI predictions using a previously developed and validateddronedarone Simcyp model, given the fact that dronedarone is a P450-dependent substrate and is a CYP3A4 mechanism–based inhibitor (MBI) (USFood and Drug Administration, 2009). This validation was performed by thecomparison of the Simcyp predictions to observed values (from clinical study2), with or without dronedarone coadministration, on clopidogrel and on theactive metabolite clopi-H4 plasma pharmacokinetics.

Results

Predicted Contribution in Clopidogrel and 2-Oxo-ClopidogrelMetabolism. The Simcyp simulation using the “PK/PD parameters”option with 100 virtual CYP2C19-EM individuals resulted in mean

Fig. 5. Predicted and observed median AUC0–24 and error bars for clopidogrel with the 75-mg maintenance dose at day 5 in CYP2C19-poor, -intermediate, -extensive, and-ultrarapid metabolizers.

PBPK Modeling for Clopidogrel and Its Active Metabolite 517

at ASPE

T Journals on A

pril 18, 2020dm

d.aspetjournals.orgD

ownloaded from

clopidogrel metabolism contributions for CYP1A2, CYP2B6, CYP2C19and additional systemic clearance as presented in Fig. 3A, and in mean2-oxo-clopidogrel metabolism contributions for CYP2B6, CYP2C9,CYP2C19, and CYP3A4 and additional clearance as presented inFig. 3B.The mean contribution of non-P450-mediated systemic clearance,

i.e., mainly esterase-dependent hydrolysis to the overall clopidogrelclearance was about 90.2%. Specifically regarding P450-mediatedoxidative clearance of clopidogrel and representing about 10% ofoverall clearance, predicted relative mean contributions were of 29.0%for CYP1A2, 22.4% for CYP2B6, and 48.6% for CYP2C19.For 2-oxo-clopidogrel metabolism, mean contribution of non-P450-

mediated clearance to the overall 2-oxo-clopidogrel clearance was about50.7%. Specifically regarding P450-mediated oxidative clearance of2-oxo-clopidogrel, Simcyp predicted relative mean contributions of 39.0%for CYP2B6, 6.47% for CYP2C9, 21.1% for CYP2C19, and 33.5% forCYP3A4. These predictions are consistent with those published by Kazuiet al. (2010).Comparison of Predicted-to-Observed Clopidogrel and Clopi-H4

AUC0–24 Values. The predicted AUC0–24 values of clopidogrel andclopi-H4, using the PBPK model, at loading dose (day 1) and main-tenance dose (day 5) were compared with those observed in study 1. Thiscomparison was performed for each CYP2C19-phenotyped group. The

median AUC0–24 value and error bar of the group of observed subjectswas presented together with the median and error bar of each virtualtrial. In addition, the corresponding global median and 5th and 95thpercentiles for each dose in each CYP2C19-phenotyped group obtainedfrom the 100 virtual subjects taken together were presented. In additionto allowing a comparison of predicted to observed median AUC0–24

values, this representation provided an accurate evaluation of both pre-dicted interindividual and intertrial variability. This is also a way toconfirm that the group of observed patients (n = 10) behaved as one ofthe virtual trials (n = 10 each).Figures 4 and 5 present the results for clopidogrel 300-mg loading

(day 1) and 75-mg maintenance doses (day 5), respectively. Figures 6and 7 are the results for clopi-H4 at these doses.These figures showed the good predictive performance of the Simcyp

model for both clopidogrel and clopi-H4, whatever the treatment period(for both loading and maintenance doses) and whatever the CYP2C19-phenotyped group.Visual Predictive Check. The VPC was adapted to PBPK and

performed to allow a graphical qualification of the Simcyp model fromday 1 (300-mg loading dose) to day 5 of treatment (75-mg maintenancedose), for clopidogrel and for clopi-H4 in each CYP2C19-phenotypedgroup. The results are presented in Fig. 8 for clopidogrel and Fig. 9 forclopi-H4.

Fig. 6. Predicted and observed median AUC0–24 and error bars for clopi-H4 with the 300-mg loading dose (day 1) in CYP2C19-poor, -intermediate, -extensive, and-ultrarapid metabolizers.

518 Djebli et al.

at ASPE

T Journals on A

pril 18, 2020dm

d.aspetjournals.orgD

ownloaded from

These results, obtained from the 100 virtual individuals for eachCYP2C19-phenotyped group, for both clopidogrel and clopi-H4, confirmedthe accuracy of the predictions and the good CYP2C19 contribution in bothmetabolic steps (clopidogrel to 2-oxo-clopidogrel and 2-oxo clopidogrel toclopi-H4). In addition, these figures confirmed the accurate estimation of theinterindividual variability for both compounds.Model Qualification on the Basis of DDI Prediction. This model

qualification on the basis of DDI prediction uses a previously developedand validated dronedarone Simcyp model (see Table 2 and Fig. 2).As previously mentioned, study 2 was conducted to evaluate the

impact of dronedarone coadministration (400-mg BID repeated dosesfrom day 9 to day 5) on the pharmacokinetics of clopidogrel and clopi-H4pharmacokinetics (300-mg clopidogrel loading dose at day 1 followed by75 mg from day 2 to day 5), since dronedarone is a moderate CYP3A4MBI. The Simcyp simulations were performed on the basis of the samedesign and similar population (healthy CYP2C19-EM individuals) study2 to make the comparison the most reliable. The observed andpredicted ratio estimates and 90% confidence intervals are presentedin Table 3 and the observed and predicted clopi-H4 exposures arepresented in Table 4.Regarding clopidogrel, this comparison confirmed the absence of

any DDI on clopidogrel when coadministered with dronedarone. Theobserved ratio estimates ranged between 0.89 and 1.03 and predictedvalues ranged between 1.00 and 1.01.

For the active metabolite, i.e., clopi-H4, the predicted ratio estimates(90% confidence interval) were slightly underestimated: observed Cmax ratioranged between 0.81 (0.73–0.89) at day 5 and 0.93 (0.84–1.04) at day 1,and the predicted Cmax ratio was about 0.72 (0.71–0.76) at day 5 and 0.72(0.71–0.75) at day 1; the observed AUC0–24 ratio ranged between 1.05(0.67–1.22) at day 5 and 1.09 (0.65–1.20) at day 1, and the predicted valuewas about 0.72 (0.71–0.76) at day 5 and 0.73 (0.72–0.76) at day 1.When looking at the clopi-H4 exposures (see Table 4), the predicted

Cmax values were very similar to the observations when clopidogrelwas administered without dronedarone comedication and slightlyunderestimated when coadministered with dronedarone (for both day 1and day 5). On the contrary, the predicted AUC0–24 seemed to beslightly overestimated when clopidogrel was administered withoutdronedarone and very close to observations when coadministered withdronedarone (also observed for at both day 1 and day 5). Overall, thepredicted interindividual variability on clopi-H4 exposures (56.4–66.2%)was close to the observations (40.0–80.7%).

Discussion

This is the first study proposing a full dynamic PBPK model able toaccurately predict simultaneously the pharmacokinetics of the parentdrug, its primary and secondary metabolites, in populations withgenetically different activity for a metabolizing enzyme. This PBPK

Fig. 7. Predicted and observed median AUC0–24 and error bars for clopi-H4 with the 75-mg maintenance dose at day 5 in CYP2C19-poor, -intermediate, -extensive, and-ultrarapid metabolizers.

PBPK Modeling for Clopidogrel and Its Active Metabolite 519

at ASPE

T Journals on A

pril 18, 2020dm

d.aspetjournals.orgD

ownloaded from

model was developed for clopidogrel, its primary metabolite 2-oxo-clopidogrel, and its secondary active metabolite clopi-H4. The modelqualification was performed on the parent drug clopidogrel and onclopi-H4 whatever the treatment period (loading or maintenancedoses) for each CYP2C19-phenotyped group (PM, IM, EM, and UM).The model was not validated for 2-oxo-clopidogrel because of the

absence of plasma concentrations for this metabolite. The 2-oxo-clopidogrelplasma concentrations were not assayed because of the weak stability andof the fleeting property of this metabolite.The PBPK model was built using an approach integrating all of the

available physicochemical and in vitro information of the three com-pounds, gathered via the enzyme kinetic parameters that govern thetwo metabolic steps.The observed data from two clinical studies were used for model

qualification: 1) the first study with well-balanced genetic polymorphic

populations (CYP2C19-PMs, -IMs, -EMs, and -UMs), on the basis ofthe important CYP2C19 involvement in both metabolic steps, and 2) thesecond study with or without dronedarone coadministration for DDIprediction purpose, given that dronedarone is a moderate CYP3A4 MBIand that CYP3A4 is involved in the second step of clopidogrel metabolism,i.e., 2-oxo-clopidogrel to clopi-H4. Three qualification methods were usedfor this PBPK model: the comparison of observed-to-predicted AUC0–24

coupled with an estimation of the variability, the VPC method, on thebasis of a visual inspection of the predictive performance of the model,and the last method on the basis of DDI prediction.The first qualification method is presented in Figs. 4 and 5 for

clopidogrel and in Figs. 6 and 7 for clopi-H4 on the basis of a comparisonof the observed median AUC0–24 values (and the corresponding variability)in each CYP2C19-phenotyped group (with n = 10 for each group) to thepredicted values of 10 virtual trials (10 individuals in each trial). This

Fig. 8. Visual predictive check of clopidogrel in CYP2C19-poor, -intermediate, -extensive, and -ultrarapid metabolizers. Observed concentrations (blue dots) and median ofpredictions (red line) and the ranges of 5th and 95th percentiles of predictions (pink area).

520 Djebli et al.

at ASPE

T Journals on A

pril 18, 2020dm

d.aspetjournals.orgD

ownloaded from

representation allowed, in addition to evaluating the predictive performanceof the model regarding median values, estimation of the interindividual andintergroup variability. This method allowed validation of the model forboth clopidogrel and clopi-H4 in each CYP2C19-phenotyped group,whatever the treatment period, and clearly showed that the clinicalstudy, in terms of AUC0–24, behaved as one of the virtual trials.The second qualification method was developed on the basis of the

adaptation of VPC method to PBPK. Classically, the VPC method isperformed to validate a model developed using the population approach(population pharmacokinetics and pharmacodynamics), where hundredsof simulations were launched once the final model was built, and at theend the observations were visually compared with the statistics of thepredictions (Holford 2005; Karlsson and Holford, 2008; Post et al.,2008). This method is uses as a basis the presentation of the observedconcentrations on a “time-versus-concentrations” plot on which were

superimposed the 5th, 50th (median), and 95th percentiles of the pre-dictions obtained from 100 virtual individuals. A plot was presented foreach compound and for each CYP2C19-phenotyped group (Fig. 8 forclopidogrel and Fig. 9 for clopi-H4). In the present study, this methodconfirmed the good predictive performance of the PBPK model.The complexity of clopidogrel pharmacokinetics, linked to the

metabolic cascade with many metabolic enzymes involved (differentP450s and esterases), was a strong incentive to perform a modelqualification on the basis of DDI prediction and on the comparison ofratio estimates of Cmax and AUC0–24 of clopidogrel and clopi-H4, withand without dronedarone coadministration. A PBPK model previouslydeveloped and validated accurately predicted the pharmacokinetics ofdronedarone and correctly accounted for the nonlinearity of dronedaronepharmacokinetics resulting from the moderate mechanism-based in-hibition of CYP3A4, which is itself the main isoform involved in

Fig. 9. Visual predictive check of clopi-H4 in CYP2C19-poor, -intermediate, -extensive, and -ultrarapid metabolizers. Observed concentrations (blue dots) and median ofpredictions (red line) and the ranges of 5th and 95th percentiles of predictions (pink area).

PBPK Modeling for Clopidogrel and Its Active Metabolite 521

at ASPE

T Journals on A

pril 18, 2020dm

d.aspetjournals.orgD

ownloaded from

dronedarone clearance. Regarding clopidogrel, this comparison con-firmed the absence of any DDI interaction on Cmax and AUC0–24 ofclopidogrel when coadministered with dronedarone. On the other hand,for the active metabolite clopi-H4, the predicted ratio estimates wereunexpectedly slightly underestimated. There are some hypotheses thatcould explain this underestimation. The first one would be an over-estimation of the impact of dronedarone CYP3A4 inhibition via inaccuratemechanism-based inactivation input parameters (KI/Kinact) and/or CYP3A4turnover (Kdeg) values in the Simcyp population library. This hypothesiswas evaluated since the Kdeg value used in this library (0.0077 hour

21) wasdiscussed (Rowland-Yeo et al., 2011), suggesting that the use of a highervalue (0.0193 hour21) resulted in less bias and greater precision in thepredictions. At the end, this hypothesis was improbable since thedronedarone model was well validated using a large range of doses (from200 to 1600 mg BID) and accurately predicted the CYP3A4 saturationattributable to mechanism-based inhibition. The second hypothesis wouldbe an overestimation of CYP3A4 contribution to 2-oxo-clopidogrelmetabolism. This idea is debatable given that the clopidogrel model waswell validated for the four CYP2C19-phenotyped groups, suggesting thatthe contribution of CYP2C19 was well documented and consistent with theobserved values. Given that P450 isoforms other than CYP2C19 andCYP3A4 were involved in 2-oxo-clopidogrel metabolism, a clinical studyto validate this hypothesis would be of interest.This work is the first study accurately describing the pharmacoki-

netics of a drug and its sequential metabolite using the PBPK approachin different phenotypic groups. This can be considered as the first step inbuilding up a PBPK-PD model able to predict the therapeutic effect indifferent subpopulations and/or different clinical conditions (Chettyet al., 2014).

Authorship ContributionsParticipated in research design: Djebli, Boulenc, D. Fabre, G. Fabre, Sultan,

Hurbin.Performed data analysis: Djebli.Wrote or contributed to the writing of the manuscript: Djebli, Boulenc, Hurbin.

References

Barrett JS, Della Casa Alberighi O, Läer S, and Meibohm B (2012) Physiologically basedpharmacokinetic (PBPK) modeling in children. Clin Pharmacol Ther 92:40–49.

Bhatt DL and Topol EJ (2003) Scientific and therapeutic advances in antiplatelet therapy. Nat RevDrug Discov 2:15–28 Review.

Boulenc X and Barberan O (2011) Metabolic-based drug-drug interactions prediction, recent approachesfor risk assessment along drug development. Drug Metabol Drug Interact 26:147–168.

Boulenc X, Djebli N, Shi J, Perrin L, Brian W, Van Horn R, and Hurbin F (2012) Effects ofomeprazole and genetic polymorphism of CYP2C19 on the clopidogrel active metabolite. DrugMetab Dispos 40:187–197.

Chetty M, Rose RH, Abduljalil K, Patel N, Lu G, Cain T, Jamei M, and Rostami-Hodjegan A(2014) Applications of linking PBPK and PD models to predict the impact of genotypicvariability, formulation differences, differences in target binding capacity and target site drugconcentrations on drug responses and variability. Front Pharmacol 5:258.

Dansette PM, Rosi J, Bertho G, and Mansuy D (2012) Cytochromes P450 catalyze both steps ofthe major pathway of clopidogrel bioactivation, whereas paraoxonase catalyzes the formationof a minor thiol metabolite isomer. Chem Res Toxicol 25:348–356.

Djebli N, Rauch C, Fabre D, and Boulenc X (2009) Quantification of the pharmacokineticinhibition of a CYP3A4 substrate with low hepatic first pass effect and long terminal half-life;importance of study design and inhibition mechanism: a simulation study. Poster, 16th NorthAmerican Regional ISSX Meeting, Baltimore. (http://issx.confex.com/issx/16na/webprogram/Paper17095.html).

Edginton AN, Schmitt W, and Willmann S (2006) Development and evaluation of a genericphysiologically based pharmacokinetic model for children. Clin Pharmacokinet 45:1013–1034.

Farid NA, Payne CD, Small DS, Winters KJ, Ernest CS, 2nd, Brandt JT, Darstein C, Jakubowski JA,and Salazar DE (2007) Cytochrome P450 3A inhibition by ketoconazole affects prasugrel andclopidogrel pharmacokinetics and pharmacodynamics differently. Clin Pharmacol Ther 81:735–741.

Gao H, Yao L, Mathieu HW, Zhang Y, Maurer TS, Troutman MD, Scott DO, Ruggeri RB,and Lin J (2008) In silico modeling of nonspecific binding to human liver microsomes. DrugMetab Dispos 36:2130–2135.

Holford N (2005) The Visual Predictive Check – Superiority to Standard Diagnostic (Rorschach)Plots. PAGE 14 Absr 738 [http://www.page-meeting.org/default.asp?abstract=738]

Hulot JS, Bura A, Villard E, Azizi M, Remones V, Goyenvalle C, Aiach M, Lechat P,and Gaussem P (2006) Cytochrome P450 2C19 loss-of-function polymorphism is a majordeterminant of clopidogrel responsiveness in healthy subjects. Blood 108:2244–2247.

Jamei M, Turner D, Yang J, Neuhoff S, Polak S, Rostami-Hodjegan A, and Tucker G (2009)Population-based mechanistic prediction of oral drug absorption. AAPS J 11:225–237.

Johnson TN, Boussery K, Rowland-Yeo K, Tucker GT, and Rostami-Hodjegan A (2010) A semi-mechanistic model to predict the effects of liver cirrhosis on drug clearance. Clin Pharma-cokinet 49:189–206.

Karlsson MP, Holford N (2008) A tutorial on visual predictive checks. PAGE 17 Abstr 1434[http://www.page-meeting.org/default.asp?abstract=1434]

Kazui M, Nishiya Y, Ishizuka T, Hagihara K, Farid NA, Okazaki O, Ikeda T, and Kurihara A(2010) Identification of the human cytochrome P450 enzymes involved in the two oxidativesteps in the bioactivation of clopidogrel to its pharmacologically active metabolite. Drug MetabDispos 38:92–99.

Khalil F and Läer S (2011) Physiologically based pharmacokinetic modeling: methodology,applications, and limitations with a focus on its role in pediatric drug development. J BiomedBiotechnol 2011:907461.

Kim KA, Park PW, Hong SJ, and Park JY (2008) The effect of CYP2C19 polymorphism on thepharmacokinetics and pharmacodynamics of clopidogrel: a possible mechanism for clopidogrelresistance. Clin Pharmacol Ther 84:236–242.

Leong R, Vieira ML, Zhao P, Mulugeta Y, Lee CS, Huang SM, and Burckart GJ (2012) Reg-ulatory experience with physiologically based pharmacokinetic modeling for pediatric drugtrials. Clin Pharmacol Ther 91:926–931.

Lins R, Broekhuysen J, Necciari J, and Deroubaix X (1999) Pharmacokinetic profile of 14C-labeled clopidogrel. Semin Thromb Hemost 25 (Suppl 2):29–33.

Lobell M and Sivarajah V (2003) In silico prediction of aqueous solubility, human plasma proteinbinding and volume of distribution of compounds from calculated pKa and AlogP98 values.Mol Divers 7:69–87.

Lukacova V, Woltosz WS, and Bolger MB (2009) Prediction of modified release pharmacoki-netics and pharmacodynamics from in vitro, immediate release, and intravenous data. AAPS J11:323–334.

Mega JL, Close SL, Wiviott SD, Shen L, Hockett RD, Brandt JT, Walker JR, Antman EM,Macias W, and Braunwald E, et al. (2009) Cytochrome p-450 polymorphisms and response toclopidogrel. N Engl J Med 360:354–362.

Post TM, Freijer JI, Ploeger BA, and Danhof M (2008) Extensions to the visual predictive checkto facilitate model performance evaluation. J Pharmacokinet Pharmacodyn 35:185–202.

Rodgers T and Rowland M (2006) Physiologically based pharmacokinetic modelling 2: pre-dicting the tissue distribution of acids, very weak bases, neutrals and zwitterions. J Pharm Sci95:1238–1257.

Rodgers T and Rowland M (2007) Mechanistic approaches to volume of distribution predictions:understanding the processes. Pharm Res 24:918–933.

Rodgers T, Leahy D, and Rowland M (2005a) Physiologically based pharmacokinetic modeling1: predicting the tissue distribution of moderate-to-strong bases. J Pharm Sci 94:1259–1276.

Rodgers T, Leahy D, and Rowland M (2005b) Tissue distribution of basic drugs: accounting forenantiomeric, compound and regional differences amongst beta-blocking drugs in rat. J PharmSci 94:1237–1248.

Rostami-Hodjegan A, Amin AM, Spencer EP, Lennard MS, Tucker GT, and Flanagan RJ (2004)Influence of dose, cigarette smoking, age, sex, and metabolic activity on plasma clozapineconcentrations: a predictive model and nomograms to aid clozapine dose adjustment and toassess compliance in individual patients. J Clin Psychopharmacol 24:70–78.

Rowland Yeo K, Jamei M, Yang J, Tucker GT, and Rostami-Hodjegan A (2010) Physiologicallybased mechanistic modelling to predict complex drug-drug interactions involving simultaneouscompetitive and time-dependent enzyme inhibition by parent compound and its metabolite inboth liver and gut - the effect of diltiazem on the time-course of exposure to triazolam. Eur JPharm Sci 39:298–309.

Rowland Yeo K, Walsky RL, Jamei M, Rostami-Hodjegan A, and Tucker GT (2011) Predictionof time-dependent CYP3A4 drug-drug interactions by physiologically based pharmacokineticmodelling: impact of inactivation parameters and enzyme turnover. Eur J Pharm Sci 43:160–173.

Sanofi (2014) Summary of Product Characteristics for Multaq®. http://www.ema.europa.eu/docs/en_GB/document_library/EPAR_-_Product_Information/human/001043/WC500044534.pdf.

Simon T, Bhatt DL, Bergougnan L, Farenc C, Pearson K, Perrin L, Vicaut E, Lacreta F, Hurbin F,and Dubar M (2011) Genetic polymorphisms and the impact of a higher clopidogrel doseregimen on active metabolite exposure and antiplatelet response in healthy subjects. ClinPharmacol Ther 90:287–295.

Thompson CM, Johns DO, Sonawane B, Barton HA, Hattis D, Tardif R, and Krishnan K (2009)Database for physiologically based pharmacokinetic (PBPK) modeling: physiological data forhealthy and health-impaired elderly. J Toxicol Environ Health B Crit Rev 12:1–24.

Tuffal G, Roy S, Lavisse M, Brasseur D, Schofield J, Delesque Touchard N, Savi P, Bremond N,Rouchon M-C, and Hurbin F, et al. (2011) An improved method for specific and quantitativedetermination of the clopidogrel active metabolite isomers in human plasma. Thromb Haemost105:696–705.

Uchimura T, Kato M, Saito T, and Kinoshita H (2010) Prediction of human blood-to-plasma drugconcentration ratio. Biopharm Drug Dispos 31:286–297.

Umemura K, Furuta T, and Kondo K (2008) The common gene variants of CYP2C19 affectpharmacokinetics and pharmacodynamics in an active metabolite of clopidogrel in healthysubjects. J Thromb Haemost 6:1439–1441.

US Food and Drug Administration (2009) MULTAQ® (DRONEDARONE). Briefing Document.Advisory committee meeting of the cardiovascular and renal drugs division of the US Food andDrug Administration. www.fda.gov/downloads/AdvisoryCommittees/.../UCM134981.pdf

US Food and Drug Administration (2012) Summary Minutes of the Advisory Committee for Phar-maceutical Science and Clinical Pharmacology (http://www.fda.gov/downloads/AdvisoryCommittees/CommitteesMeetingMaterials/Drugs/AdvisoryCommitteeforPharmaceuticalScienceandClinicalPharmacology/UCM306989.pdf).

Vieira ML, Zhao P, Berglund EG, Reynolds KS, Zhang L, Lesko LJ, and Huang SM (2012)Predicting drug interaction potential with a physiologically based pharmacokinetic model: a casestudy of telithromycin, a time-dependent CYP3A inhibitor. Clin Pharmacol Ther 91:700–708.

Yeo KR, Kenny JR, and Rostami-Hodjegan A (2013) Application of in vitro-in vivo extrapolation(IVIVE) and physiologically based pharmacokinetic (PBPK) modelling to investigate the impact ofthe CYP2C8 polymorphism on rosiglitazone exposure. Eur J Clin Pharmacol 69:1311–1320

Address correspondence to: Dr. Nassim Djebli, Drug Disposition, DispositionSafety and Animal Research, Sanofi Recherche et Développement, 371 rue duProfesseur Blayac, Montpellier, France. E-mail: nassim.djebli@sanofi.com

522 Djebli et al.

at ASPE

T Journals on A

pril 18, 2020dm

d.aspetjournals.orgD

ownloaded from