1521-009X/44/7/934–943$25.00 http://dx.doi.org/10.1124/dmd.115.068031DRUG METABOLISM AND DISPOSITION Drug Metab Dispos 44:934–943, July 2016Copyright ª 2016 by The American Society for Pharmacology and Experimental Therapeutics

Special Section on Pediatric Drug Disposition andPharmacokinetics—Minireview

Physiologically Based Pharmacokinetic Modeling in PediatricOncology Drug Development

Nathalie Rioux1 and Nigel J. Waters2

Epizyme Inc., Cambridge, Massachusetts

Received October 22, 2015; March 1, 2016

ABSTRACT

Childhood cancer represents more than 100 rare and ultra-rarediseases, with an estimated 12,400 new cases diagnosed eachyear in the United States. As such, this much smaller patientpopulation has led to pediatric oncology drug development laggingbehind that for adult cancers. Developing drugs for pediatricmalignancies also brings with it a number of unique trial designconsiderations, including flexible enrollment approaches, age-appropriate formulation, acceptable sampling schedules, andbalancing the need for age-stratified dosing regimens, given thesmaller patient populations. The regulatory landscape for pediatricpharmacotherapy has evolved with U.S. Food and Drug Adminis-tration (FDA) legislation such as the 2012 FDA Safety and In-novation Act. In parallel, regulatory authorities have recommendedthe application of physiologically based pharmacokinetic (PBPK)modeling, for example, in the recently issued FDA Strategic Plan

for Accelerating the Development of Therapies for Pediatric RareDiseases. PBPK modeling provides a quantitative and systems-based framework that allows the effects of intrinsic and extrinsicfactors on drug exposure to be modeled in a mechanistic fashion.The application of PBPK modeling in drug development forpediatric cancers is relatively nascent, with several retrospectiveanalyses of cytotoxic therapies, and latterly for targeted agentssuch as obatoclax and imatinib. More recently, we have employedPBPK modeling in a prospective manner to inform the firstpediatric trials of pinometostat and tazemetostat in geneticallydefined populations (mixed lineage leukemia–rearranged andintegrase interactor-1–deficient sarcomas, respectively). In thisreview, we evaluate the application of PBPK modeling in pediatriccancer drug development and discuss the important challengesthat lie ahead in this field.

Introduction

Childhood cancer represents a collection of more than 100 rare andultra-rare diseases, with an estimated 12,400 new cases diagnosed eachyear in the United States in patients aged,21 years (Ries et al., 1999).This much smaller patient population has, in large part, led to cancerdrug development for pediatric patients being an afterthought, withapproval for pediatric indications following the development andapproval of the agent in adult cancers (Kearns et al., 2003). The rateof development has also been slow; from 1948 to 2003, there were 120new cancer drug approvals, only 30 of which were used in pediatrics(Adamson et al., 2014). Developing drugs for pediatric malignanciesalso brings with it a number of unique challenges in terms of trial

design, potential for disparity in genetics and pathophysiology relativeto the adult disease, and what is considered acceptable clinical benefit.Trial design considerations include age-stratified dosing regimens, age-appropriate formulations (Breitkreutz and Boos, 2007), amenablesampling schemes (given the limitations on the number and volumeof blood draws), as well as the necessity for flexible enrollmentapproaches in phase I, such as the rolling six design (Skolnik et al.,2008). The specific oncogenic pathways in adult carcinomas may notbe active in the childhood malignancy, although common cellularpathways have emerged (Rossig et al., 2011). The benchmark forclinical benefit can also be quite different. Increasing overall survival bya few months may be acceptable in adults but the focus in pediatriconcology is cure, with cure rates in most cases of .70% and 5-yearevent free survival of approximately 80% (Adamson et al., 2014).However, novel drug approaches are needed for children with difficult-to-treat malignancies such as stage IV neuroblastoma, sarcomas, braintumors, and relapsed leukemia (Horton and Berg, 2011). As a direct

1Current affiliation: H3 Biomedicine, Cambridge, Massachusetts.2Current affiliation: Syros Pharmaceuticals, Cambridge, Massachusetts.dx.doi.org/10.1124/dmd.115.068031.

ABBREVIATIONS: AAG, a-1-acid glycoprotein; ADME, absorption, distribution, metabolism, and excretion; AUC, area under the curve; AUCss,area under the curve at steady state; BSA, body surface area; CL/F, oral clearance; Cmax, peak serum concentration; CLint, intrinsic clearance; DDI,drug–drug interaction; DME, drug-metabolizing enzyme; DOT1L, disruptor of telomeric silencing 1-like; EPZ-5676, 9-[5-deoxy-5-[[cis-3-[2-[6-(1,1-dimethylethyl)-1H-benzimidazol-2-yl]ethyl]cyclobutyl](1-methylethyl)amino]-b-D-ribofuranosyl]-9H-purin-6-amine; EPZ-6438, N-[(1,2-dihydro-4,6-dimethyl-2-oxo-3-pyridinyl)methyl]-5-[ethyl(tetrahydro-2H-pyran-4-yl)amino]-4-methyl-49-(4-morpholinylmethyl)-[1,19-biphenyl]-3-carboxamide; FDA, Foodand Drug Administration; GST, glutathione-S-transferase; HMT, histone methyltransferase; INI, integrase interactor; MLL, mixed lineage leukemia; MLL-r,mixed lineage leukemia–rearranged; P450, cytochrome P450; PBPK, physiologically based pharmacokinetics; P-gp, P-glycoprotein; PK, pharmacokinetics;t1/2, terminal half-life; UGT, uridine diphosphoglucuronosyl transferase.

934

at ASPE

T Journals on A

ugust 29, 2020dm

d.aspetjournals.orgD

ownloaded from

consequence of the unmet need in pediatric pharmacotherapy ingeneral, the regulatory landscape has evolved with the introduction ofU.S. Food and Drug Administration (FDA) legislation including theBest Pharmaceuticals for Children Act in 2002 and the PediatricResearch Equity Act in 2003 providing incentives for evaluating anddeveloping drugs for children, which were both made permanent underthe FDA Safety and Innovation Act in 2012.In developing and optimizing pediatric pharmacotherapy, an appre-

ciation of the comparative physiology and biochemistry between adultsand children is paramount, being particularly acute in prenatal, infant,and toddler populations. As the long-standing adage states, “childrenare not small adults”; thus, simple dose reductions may not be suffi-cient and the nature of disease can also be divergent. As our knowledgeof normal growth and development has increased, so too has ourunderstanding of how these growth changes can greatly affectpharmacokinetics (PK) and ultimately response to therapy. Drugs inpediatric oncology are typically administered by either intravenous ororal routes; as such, there are a number of physical, chemical, andbiologic barriers that must be overcome to reach the target site of action.The gastrointestinal absorption of drugs can be markedly affected bythe developmental changes in absorptive surface area, intraluminal pH,splanchnic blood flow, gastric emptying time, and intestinal motility(Batchelor et al., 2014). The maturation profile in biliary function alsoleads to age-related differences in duodenal bile salt concentrations,which can affect drug dissolution and solubility at the site of absorption(Poley et al., 1964). The ontogeny in intestinal expression of drug-metabolizing enzymes and transporters can also be a contributing factorfor oral bioavailability. This is illustrated with the alkylating antineo-plastic agent, busulfan, in which biopsies of distal duodenum indicatedthat glutathione-S-transferase (GST) activity decreased from infancythrough adolescence and manifested as a reduced oral clearance ofbusulfan, a GST substrate (Gibbs et al., 1999). Drug distribution canalso show marked age-related changes as a result of physiologic andbiochemical ontogeny. Infants and children have proportionally higherextracellular water and total body water than adults, in addition to theobvious differences in cardiac output, tissue perfusion, and organ size(Kearns et al., 2003). The major plasma proteins, albumin and a-1-acidglycoprotein (AAG), also show developmental changes, which canaffect the distribution of highly protein bound drugs (Ehrnebo et al.,1971). The delayed maturation of drug-metabolizing enzymes (DMEs)and potential for toxicity was typified with the case of newborns treat-ed with doses of the antimicrobial chloramphenicol, extrapolatedfrom those shown to be safe and effective in adults, which led tocardiovascular collapse associated with Gray syndrome in neonates.The mechanistic basis for these drug-induced sequelae was later shownto be an immature uridine diphosphoglucuronosyl transferase (UGT)system resulting in impaired metabolic clearance. The maturationprofile of DMEs, the most studied in this regard being the cytochromeP450 (P450) superfamily (Hines, 2009), can follow one of threepossible trajectories: 1) the enzyme exhibits high expression during thefirst trimester and either remains high or decreases during gestationbefore a steep decline in expression level postnatally, within a few days orup to 2 years after birth (e.g., CYP3A7); 2) the enzyme shows relativelyconstant expression throughout gestation and postnatally (e.g.,CYP2C19); and 3) the enzyme may be not expressed or present at lowlevels in the fetus, followed by a marked increase in expression level,depending on the isoform in question, between the second and thirdtrimesters to the first years of life (e.g., CYP3A4). Similarly, isoforms ofthe other DME families can be categorized in this manner, such as thesulfotransferases, flavin monooxygenases, and UGTs. The prevalence ofenzymes falling into the third class underpins the consistent observation inclinical studies of drugs metabolized by the liver showing an age-

dependent increase in plasma clearance in children aged,10 years, suchas dextromethorphan (Blake et al., 2007), caffeine (Pons et al., 1988), andmidazolam (Altamimi et al., 2015). It should also be noted that excretorymechanisms also show an age-based relationship in terms of renalfunction, glomerular filtration rate, and renal blood flow that can allaffect the elimination of renally excreted drugs. More detailed reviews ofthe physiologic and biochemical ontogeny influencing the absorption,distribution, metabolism, and excretion (ADME) of drugs were reportedpreviously (Kearns et al., 2003; Hines, 2009).It is clear that ontogenesis plays a key role in the metabolism and

disposition of drugs in children; as a result, simplified dosingalgorithms based on body size alone are usually insufficient. Traditionalmethods such as allometry have been used to scale pediatric doses,wherein absolute dose (or clearance) is related to the child-to-adultbody weight ratio raised to the power of 3/4 (eq. 1) (Björkman, 2006;Mahmood, 2006). The basis for the 3/4 power law has been discussedelsewhere (West et al., 1999). Since this does not account for maturationprocesses and has a tendency to overpredict clearance in youngerchildren, its use has been limited to children aged.1 to 2 years. Furtherrefinement with the inclusion of terms for maturation or organ functionhas shown utility within predefined limits (eq. 2) but still lacks theholistic and mechanistic facets that are possible with physiologicallybased pharmacokinetic (PBPK) modeling.

Dosechild ¼ Doseadult � ðBWchild=BWadultÞ0:75 ð1ÞCLchild ¼ CLadult � ðBWchild=70Þ0:75 �MF� OF ð2Þ

In eqs. 1 and 2, BW indicates body weight, CL is clearance, MF ismaturation function, and OF is organ function.PBPK modeling provides a quantitative and systems-based framework

with each compartment representing a physiologic volume interconnectedby flow rates that are anatomically representative of the circulatory system(Fig. 1). Mass balance equations describe the transfer of drug from thearterial blood into tissues and from tissues into venous blood. As such,simulations are dynamic in nature, with model fitting and predictionfocused on time-concentration data. Importantly, PBPK models are morecomprehensive than empirical PK models, incorporating both system-specific parameters as well as drug-specific parameters. Although it is nota new approach, PBPK modeling applied in pharmaceutical research anddevelopment has gained increasing attention in recent years with theavailability of robust in vitro and in silico data, advancements in in vitro toin vivo extrapolation, and the substantial investment that has beenmade inthe curated databases of system-dependent and probe drug-dependentparameters that are the foundation of the commercially available softwarepackages. This technological progress has enhanced the number andbreadth of robust applications in the area of clinical pharmacology,including first-in-human studies, special populations, drug interactions,and biopharmaceutics/formulations (Jones et al., 2015; Sager et al., 2015).Considerable experience has now accumulated with the development ofPBPKmodels to describe drug concentration-time profiles in adults (Joneset al., 2009; Kostewicz et al., 2014), and, more recently, in cancer drugdevelopment (Block, 2015) and special populations such as pediatrics(Khalil and Läer, 2011; Barrett et al., 2012; Maharaj and Edginton, 2014).There have been multiple applications of PBPK in pediatric drugdevelopment, many of which are in support of the first pediatric trial,including starting dose selection, prediction of exposures across the agecontinuum, optimization of blood sampling strategy, prediction of targetorgan exposure [safety and pharmacodynamics (PD)], and evaluation ofdrug interaction potential.The momentum in the PBPK modeling arena has evolved to a

regulatory science. As mentioned earlier, the health authorities, FDA,European Medicines Agency, and the Ministry of Health, Labor and

PBPK Modeling in Pediatric Oncology Drug Development 935

at ASPE

T Journals on A

ugust 29, 2020dm

d.aspetjournals.orgD

ownloaded from

Welfare of Japan (European Medicines Agency, 2014; Zhao et al.,2012; Shepard et al., 2015; Wagner et al., 2015) have all highlightedPBPK modeling in guidances covering drug interactions and hepaticimpairment. Furthermore, PBPK modeling and simulation has recentlybeen used to directly support labeling statements (e.g., Imbruvica;related to the use of moderate CYP3A inhibitors with ibrutinib;Imbruvica product label; Janssen, Titusville, NJ). Between 2008 and2012, the FDA received 33 Investigational New Drug/New DrugApplication submissions containing PBPK modeling approaches (in-cluding six pediatric submissions, although none in oncology). Inparallel with the increasing number of submissions, the agency hasincreasingly used de novo (i.e., FDA initiated) PBPK modeling in itsreviews to help characterize PK in a variety of complex clinicalscenarios (Huang et al., 2013). In 2013, the cumulative number ofsubmissions containing PBPK modeling increased to 84, with 22% ofthem related to pediatric applications (Zhao, 2014). Furthermore, theFDA recently issued a “Strategic Plan for Accelerating the Develop-ment of Therapies for Pediatric Rare Diseases,” which specificallyhighlights the role of modeling and simulation approaches such asPBPKmodeling to inform the design and conduct of PK/PD studies andother clinical trials for investigational drugs in pediatric rare diseasepopulations (http://www.fda.gov). Given this regulatory focus, thePBPK community is redoubling efforts to formulate industry-wide bestpractices and standardize the level of rigor and verification that isneeded in PBPK model submissions (Sager et al., 2015).The construction of PBPK models has typically used the various in

vitro and physicochemical data available in early preclinical drugdevelopment in what is often termed a “bottom-up” approach, in anattempt to recapitulate the concentration-time data (Fig. 1). Alterna-tively, there may be reasons to consider a “top-down” approach inwhich the underlying drug-specific parameterization of the PBPKmodel is accomplished by optimizing the fit of the time-concentrationdata in question. In practice, there is usually an approach that landssomewhere between the two, termed “middle out,” in which some of thedrug-specific parameters based on in vitro or in silico data may not scaleappropriately or there are missing data and incomplete informationregarding some aspects of drug disposition, and an element of top-downmodel optimization is necessary. Furthermore, mechanistic understand-ing can be gained through parameter sensitivity analysis and asking“what if” questions of the optimized model.

Case Studies of PBPK Modeling in Pediatric Oncology

The application of PBPK modeling in pediatric oncology drugdevelopment is still quite nascent and has largely been focused onretrospective analyses in validating the approach. A summary ofreported applications is shown in Table 1. The first published modelin this area of drug research dates back to 1982, with this seminal PBPKmodel being built with a limited number of compartments from theBischoff–Dedrick multiorgan model. The observed parent cisplatin andtotal platinum serum concentrations in 14 pediatric patients wereadequately recapitulated in the model and helped to better understandchanges in the renal clearance of total platinum (Evans et al., 1982).Similar reports based on a modified Bischoff–Dedrick model followedand were expanded to incorporate effusion spaces as a compartmentof drug distribution and clearance (Li and Gwilt, 2002). Simulatedmethotrexate plasma concentrations in adults without effusions wereadequately modeled prior to providing congruent simulated andmeasured plasma and effusion methotrexate concentrations in onepediatric patient with a malignant pleural effusion after intravenousadministration. This early work has been expanded on with PBPKmodeling applications reported for a number of drugs used in thepediatric oncology setting, including busulfan, docetaxel, etoposide,coadministered 6-mercaptopurine and methotrexate, and, more recently,imatinib, obatoclax, pinometostat (EPZ-5676; 9-[5-deoxy-5-[[cis-3-[2-[6-(1,1-dimethylethyl)-1H-benzimidazol-2-yl]ethyl]cyclobutyl](1-methylethyl)amino]-b-D-ribofuranosyl]-9H-purin-6-amine), and tazemetostat(EPZ-6438; N-[(1,2-dihydro-4,6-dimethyl-2-oxo-3-pyridinyl)methyl]-5-[ethyl(tetrahydro-2H-pyran-4-yl)amino]-4-methyl-49-(4-morpholinyl-methyl)-[1,19-biphenyl]-3-carboxamide).Unlike most pediatric applications wherein adult PK data are

leveraged in the PBPK model building process, the cisplatin model wasinitially developed with PK data in dogs and adjusted for humanpediatric physiology (Evans et al., 1982). As summarized in Table 1,the more common approach to pediatric PBPK modeling is to scale avalidated adult PBPK model to children by incorporating ontogeny andmaturation processes, thereby negating the issue of potential speciesdifferences. Another exception to the adult-to-children model workflowwas presented by Barrett et al. (2011). After establishing quantitativepharmacology relationships between in vitro activity data and a pre-clinical diseased mouse model, obatoclax exposures were recapitulatedin the mouse prior to scaling to a human infant, with the goal of

Fig. 1. (A) A typical workflow scheme for the development of a PBPK model in support of pediatric dose selection early in clinical development. (B) Representative drug-specific data inputs used in “bottom-up” PBPK model building. (C) Generalized scheme for a PBPK model. IV, intravenous; PO, oral.

936 Rioux and Waters

at ASPE

T Journals on A

ugust 29, 2020dm

d.aspetjournals.orgD

ownloaded from

evaluating the potential to achieve targeted exposure in this population(Barrett et al., 2011). This approach allowed tissue distribution to bemodeled and scaled, such that the PBPK model based on obatoclaxphysicochemical properties compared well with observed obatoclaxlevels in the plasma, spleen, liver, and kidney of leukemia-bearing micefor two single intravenous dose levels; by contrast, exposure valueswere overpredicted in the brain. After scaling to human infants, PBPKsimulation of obatoclax administration to 1-year-old patients suggestedthat adequate drug exposure to target organs should be achievable atclinically relevant doses. To the best of our knowledge, this alternateapproach for obatoclax has not been verified with clinical data from apediatric population. By contrast, and in line with a recently proposedpediatric PBPK model development workflow (Maharaj and Edginton,2014), Thai et al. (2015) first developed a model for docetaxel from invitro ADME and intravenous PK data frommore than 500 patients withcancer after either single or multiple dosing, to demonstrate howmodeling can be applied to optimize dose and sampling times for apediatric PK bridging study. Docetaxel is highly protein bound, has ahigh volume of distribution, and is highly metabolized. In the full-bodyPBPK model developed for adult patients, all 16 compartments werewell stirred or perfusion limited, except for the liver, where organicanion transporting polypeptide 1B1 and1B3 activities were considered,and, for the muscle, where apparent passive diffusion was used. Thetissue/plasma partition was determined using the volume of distributionas a primary predictor, in addition to physicochemical descriptors.Hepatic (CYP3A), biliary, and renal clearances were fit in a top-downapproach. The model was customized for oncology patients byconsidering demographic and pathophysiological differences fromhealthy subjects. After scaling to a pediatric population, and asdetermined by visual inspection, the model adequately predicted

docetaxel plasma concentration-time profiles in neonates to 18-year-old patients by accounting for age-dependent physiologic differencesand CYP3A ontogeny, with predicted clearance and volume of dis-tribution within 1.5-fold of observed data (Thai et al., 2015).Although P450 ontogeny functions are implemented in commercial

PBPK software, scaling of other metabolic pathways is less common. Inthat regard, the modeling exercise described by Diestelhorst et al.(2014) for the DNA-alkylating agent busulfan provides a proof ofconcept for GST substrates. In brief, an adult PBPK model was refinedby implementing GST-A1 in 11 organs, using the PK-Sim integratedenzyme expression database, and adding irreversible DNA binding andplasma protein binding processes. Age-dependent enzyme activity andmaturation factors were considered and the adult-to-child scalingindicated lower clearance values for children relative to adults.Intravenous administration of busulfan was simulated in pediatricpatients, with a mean percentage error for all patients of 3.9%; 3 of 23children demonstrated a mean percentage error of greater than 630%,showing an adequate predictive performance of this retrospectivemodel (Diestelhorst et al., 2014). In another study, both CYP3A4 andUGT1A1 ontogeny information was included to model etoposideexposure in children (Kersting et al., 2012). Etoposide is highly boundto plasma proteins, metabolized by CYP3A4 and UGT1A1, andeliminated in urine and bile. An adult PBPK model was developedusing data from nine women with primary breast cancer receiving thedrug intravenously, as part of a polychemotherapy regimen before stemcell transplantation. The physicochemical parameters of etoposide,along with the unbound plasma fraction at high and low doses in adults,kinetics for in vitro metabolic enzymes [CYP3A4 and UGT1A1], andactive biliary and renal transporters [P-glycoprotein (P-gp), multidrugresistance-associated protein 2, and a hypothetical influx transporter]

TABLE 1

Summary of PBPK modeling applications in pediatric oncology drug development

DrugRoute of

AdministrationObjective(s) Software

DataUsed inModelBuild

PerformedPrior

to FTPaReference

Busulfan i.v. Predict plasma concentration-time curves PK-Sim (Bayer TechnologyServices, Germany)

Adult No Diestelhorst et al., 2014

Cisplatin i.v. Describe the PBPK model development;simulate cisplatin disposition in humans

Not reported Dog No Evans et al., 1982

Docetaxel i.v. Demonstrate how PBPK modeling can be appliedto optimize dose and sampling times for apediatric PK bridging study in oncology

Simcyp (Certara, UK) Adult No Thai et al., 2015

Etoposide i.v. Evaluate the ability of PBPK to predict thesystemic drug exposure in children andthe effect of comedications

PK-Sim Adult Kersting et al., 2012

Imatinib p.o. Predict plasma concentration-time profilesin plasma and tissue in pediatric subjects,and assess the effect of pediatric growthprocesses using a PBPK model

Not reported Adult No European MedicinesAgency, 2013

Evaluate factors influencing imatinibexposure in pediatric patients

6-Mercaptopurine i.v. and p.o. Improve dose individualization and dosageregimen optimization; model DDIinteraction with methotrexate

MATLAB (Mathworks,Natick, MA)

Adult No Ogungbenro et al., 2014b

Methotrexate i.v. and p.o. Evaluate the role of modeling and simulationin the development of drugs for rarediseases

MATLAB Adult No Ogungbenro et al., 2014a

Methotrexate i.v. Evaluate the effect of malignant effusionson the PK of methotrexate

Berkeley Madonna (Universityof California–Berkeley,Berkeley, CA)

Adult No Li and Gwilt, 2002

Obatoclax i.v. Evaluate the potential to achieve targetexposures in infants

PK-Sim Mouse Yes Barrett et al., 2011

Pinometostat i.v. First-time-in-pediatrics starting dose selection Simcyp Adult Yes Waters et al., 2014Tazemetostat p.o. First-time-in-pediatrics starting dose selection GastroPlus (Simulations Plus,

Lancaster, CA)Adult Yes Rioux et al., 2015

aFirst time in pediatrics.

PBPK Modeling in Pediatric Oncology Drug Development 937

at ASPE

T Journals on A

ugust 29, 2020dm

d.aspetjournals.orgD

ownloaded from

were incorporated into the adult model. The tissue/plasma partitioncoefficients were generated using the Rodgers and Rowland model. Inadults, the simulated plasma concentration-time profiles of protein-bound and free etoposide were in good agreement with observed data,with mean relative deviation of 1.12 and 1.36 for low and high doses ofetoposide, respectively. The model was scaled to children, incorporat-ing ontogeny for both CYP3A4 and UGT1A1, glomerular filtration,and tubular excretion mediated by P-gp and adequately simulated theexposure seen in 18 children with various diagnoses and normal renaland liver function. Of interest, the effects of the coadministration ofcyclosporine A on the metabolism and excretion of etoposide wasdetermined in five patients, suggesting that the PBPK model could beuseful for performing hypothesis testing on the effect of concomitantmedications, a relatively poorly explored application of PBPK inpediatric drug development but extremely relevant given the poly-pharmacy in this patient population. A recent model developed for6-mercaptopurine, a purine antimetabolite and prodrug used incombination with methotrexate as therapy for childhood acute lym-phoblastic leukemia, is another example of drug–drug interaction (DDI)risk assessment by PBPKmodeling in pediatric oncology (Ogungbenroet al., 2014a,b). 6-Mercaptopurine undergoes very extensive intestinaland hepatic metabolism after oral dosing, owing to the activity ofxanthine oxidase leading to low and highly variable bioavailability.Dose adjustment during treatment is still based on toxicity rather thanroutine therapeutic drug monitoring and this work was an attemptto improve dose individualization and dosage regimen optimizationthrough modeling and simulation, ultimately to achieve a betteroutcome in patients with childhood acute lymphoblastic leukemia.The PBPK model, based on the assumption of the same eliminationpathways in adults and children with age-dependent parametersadjusted for pediatric scaling, adequately predicted plasma and redblood cell concentrations both in terms of population mean andvariability versus observed data after intravenous or oral administrationin children aged 3–18 years. Further model refinements incorporatedparameters such as net secretion clearance, biliary transit time, and redblood cell distribution and binding and enabled the oral absorption ofmethotrexate to be well described with recapitulation of the nonlinearrelationship between the fraction absorbed and the methotrexate dose.The inhibition of 6-mercaptopurine first-pass metabolism by metho-trexate (via xanthine oxidase inhibition in the gut and liver) predicted aninteraction but to a greater extent than that reported clinically. Thepredicted percentage increase in the area under the curve (AUC) andpeak serum concentration (Cmax) was approximately 65% and 50% in5- to 18-year-olds, respectively, whereas the observed increase in AUCand Cmax has been reported as 31% and 26%, respectively.In the context of a Pediatric Investigation Plan, the European

Medicines Agency (2013) requested PBPK simulations as part of thesubmission package for the use of imatinib (Gleevec; Novartis, Basel,Switzerland) in the treatment of pediatric patients with diagnosedPhiladelphia chromosome–positive acute lymphoblastic leukemiacombined with chemotherapy. PBPK modeling was used to predictarea under the curve at steady state (AUCss) and concentration-timeprofiles in pediatric subjects, to evaluate factors influencing imatinibexposure in pediatric patients. An adult PBPK model was validated andthen transformed using growth and maturation processes for pediatricscaling. The model evaluations compared the predicted versus observedimatinib AUCss and concentration-time profile from 67 pediatricpatients. The simulations showed that 29 of 31 actual AUCss valuesnormalized to 340 mg/m2 in pediatric patients fell within the 0.5 and99.5 percentiles of the model-projected range scaled from adultmeasurements and the predicted plasma concentration-time profileswere generally in good agreement for the pediatric cohort (n = 67),

except for subjects aged #2 years, for which the exposure appeared tobe overpredicted. Nonetheless, the prediction was 1.5-fold of the adultvalue at 1 year of age. Of interest, the model was used to betterunderstand the differences in prediction of children and adults, whichseems to be the mixed effect of changing distribution volume and bloodflow, in addition to clearance maturation with age. Overall, in con-junction with a population PK analysis, PBPK modeling suggested thatthe proposed posology of 340 mg/m2 in pediatric patients from 1 to 18years of age was appropriate for imatinib.

Pediatric PBPK Modeling In-House at Epizyme

Building on many of the retrospective analyses described above andthe recent regulatory precedent, we have applied PBPK modeling ina prospective manner to enable the first trials in pediatric patients fortwo first-in-class epigenetic therapies: pinometostat in mixed lineageleukemia–rearranged (MLL-r) and tazemetostat in integrase interactor(INI)-1–deficient solid tumors. The preemptive utilization of thisapproach early in development has informed some important aspectsof phase I pediatric trial design, including starting dose regimen andblood sampling strategy. With model verification and optimization onavailability of PK data in pediatric patients, the PBPK modelingworkflow can be further applied to address additional clinical pharma-cology questions such as drug interactions and PK/PD analysis, throughthe course of development of these investigational therapies in pediatricpopulations.Relapsed/refractory MLL-r acute leukemia in children has a poor

prognosis, with a reported 5-year overall survival for children aged,1year with recurrentMLL-r leukemia being 4%–20%. Themixed lineageleukemia (MLL) gene, on chromosome 11q23, codes for a histonemethyltransferase (HMT) that is responsible for methylation of H3K4, amodification associated with active transcription. Translocationsof MLL result in the loss of the SET or catalytic domain of the proteinwith the most common translocation partners, AF4, AF9, and ENL,recruiting another HMT, the disruptor of telomeric silencing 1-like(DOT1L). The aberrant recruitment of DOT1L to MLL fusion targetgenes results in ectopic H3K79 methylation and increased expressionof genes including HOXA9 and MEIS1, which are involved in leuke-mogenesis of MLL-r leukemias. Pinometostat (EPZ-5676) is a smallmolecule inhibitor of DOT1L with subnanomolar affinity and. 37,000-fold selectively against other HMTs. Preclinically, pinometostatselectively inhibits intracellular histone H3K79 methylation anddownstream target gene expression and demonstrated complete tumorregressions in an MLL-r leukemia xenograft model (Daigle et al.,2013). Because of the unmet need of MLL-r in children, we employed aprospective PBPK modeling approach leveraging pinometostat pre-clinical data (Basavapathruni et al., 2014) and early clinical data fromthe first-in-human phase I open label study in adult patients withrelapsed/refractory leukemia (Stein et al., 2014), to guide dose selectionand trial design for a companion pediatric study (Waters et al., 2014;Shukla et al., 2015). A PBPK model describing the concentration-timeprofile of pinometostat after continuous intravenous administration inadult patients at dose levels of 24–90 mg/m2 per day was built usingSimcyp (Simcyp Ltd., Sheffield, UK). The quantitative contribution ofP450 enzymes to the metabolic clearance of pinometostat was based onintrinsic clearance (CLint) data derived from in vitro recombinant P450systems using a relative activity factor approach. Pinometostat is asubstrate for CYP3A4 and CYP2C19 in human and these enzymes arebelieved to account for 90% and 10% of the metabolic intrinsicclearance respectively. The metabolic intrinsic clearance was scaledusing the well stirred model to generate a total hepatic in vivo CLint,which was then allocated to the P450 isoforms involved in the

938 Rioux and Waters

at ASPE

T Journals on A

ugust 29, 2020dm

d.aspetjournals.orgD

ownloaded from

metabolism of pinometostat and thus allowed isoform-specific ontog-eny functions to be used in the translation to the pediatric setting. Therenal clearance in human was low (0.1 l/h) and was also incorporatedinto the model. Pinometostat was shown to preferentially bind to AAGin vitro (manuscript in preparation) and this allowed both the extent ofbinding and identity of plasma proteins involved to be included in themodel, together with the maturation function for AAG. Early in modeloptimization, perfusion-limited kinetics were not able to recapitulatethe plasma steady-state profile due to a marked overprediction of totalclearance (several fold above the median observed clearance of 5.5 l/h).Using permeability-limited kinetics and a low steady-state volume ofdistribution (0.08 l/kg, derived from a separate compartmental analysis)was consistent with the short terminal half-life (t1/2) observed post-infusion and was able to fit the data well in terms of CL, steady-stateplasma concentration, and the initial phase after the start of infusion. Asensitivity analysis of the passive diffusion clearance indicated that avalue of 0.0075 mL/min/million hepatocytes gave the best model fit.This was plausible and consistent with the physicochemical propertiesand preclinical ADME data which showed low permeability in MDCKcell monolayers and a greater than 20-fold higher liver microsomalscaled CLint compared with hepatocyte scaled CLint (Basavapathruni

et al., 2014). The adult model simulations and observed data are shownin Fig. 2A, with the Css predicted within 2-fold of the observed valuesand CL predicted within 20% of observed across the dose range 24–90 mg/m2 per day. Having built and qualified the PBPK model forpinometostat using adult PK data, the next step was to prospectivelyestimate exposures across the pediatric age range from1month to 18 years.The median clearance projected in pediatric patients ranged from0.5 l/h in 1- to 3-month-olds to 4.2 l/h in 6- to 18-year-olds, and whennormalized for body surface area (BSA) ranged from 1.8 l/h per squaremeter to 3.2 l/h per square meter, respectively. A modest 1.7-folddifference in BSA-normalized clearance between infant and adolescentsuggested a dampening of the effect of P450 ontogeny on the clearanceof pinometostat, as a direct result of age-independent, permeability-limited hepatic uptake incorporated into the model. This observation isin stark contrast to the clearance of prototypical CYP3A substrates inchildren which show an age-dependent relationship, concordant withthe maturation of CYP3A expression during the first 2 years postnatal(Björkman, 2005; Hines, 2009). The model assumption with potentialto affect predictive accuracy was clearly the permeability-limitedmetabolic clearance being independent of age. This was consideredreasonable since the basis for the rate-limiting passive diffusion

Fig. 2. (A) Representative plot showing the model fit for interim adult PK data at 90 mg/m2 per day. Data at 24 mg/m2 per day (n = 5) and 36 mg/m2 per day (n = 3)showed similar fits. The black line represents the mean, the dotted line is the 95% confidence interval, and gray lines represent mean results from individual simulationtrials (20 trials of four subjects). The majority of data fall within the 95% confidence interval of the model. (B) Representative plots of the simulated time-concentrationprofile for pinometostat administered at 70 mg/m2 per day in age groups ranging from 6 months to 18 years (solid gray lines), with observed plasma concentrations inpediatric patients receiving an infusion of 70 mg/m2 per day (open circles). Medians are shown as solid black symbols. (C) Box and whisker plot of the plasma Css ofpinometostat administered at 45, 70, and 90 mg/m2 per day in pediatric patients (median age range, 0.33, 5.5, and 4 years, respectively). The dotted line corresponds tothe observed median plasma Css in adults at 90 mg/m2 per day, and the shaded area represents the predicted Css range in pediatric patients aged .1 year at the dose levelof 90 mg/m2 per day.

PBPK Modeling in Pediatric Oncology Drug Development 939

at ASPE

T Journals on A

ugust 29, 2020dm

d.aspetjournals.orgD

ownloaded from

clearance of pinometostat was physicochemical in nature, as opposed totransporter-mediated active efflux, for example, which may show agedependent expression or activity. Simulations of the predicted steady-state systemic exposure of pinometostat in pediatric virtual populationswere used to support starting dose selection. At a fixed BSA-normalizeddose, adjustments of 55%–65% of the adult dose were projected ininfants up to 2 years old to achieve equivalent exposures to adults; inchildren aged .6 years, the dose was predicted to be equivalent to theadult dose. For the practical purposes of trial conduct in this rare patientpopulation, the starting dose level and age stratification was furthersimplified in a conservative manner to derive a pediatric starting dose of80% and 50% of the highest adult dose (90 mg/m2 per day) in .12-month-olds and ,12-month-olds, respectively. PK data recentlyobtained from this study (Shukla et al., 2015) have enabled modelverification and this is summarized in Fig. 2, B and C. The observed Css

at 70 mg/m2 per day (n = 6; age range, 1.25–15 years) was within 0.75-to 2.33-fold of the predicted Css across this age range, showing that thedata were consistent with the postulated effect of dampened P450ontogeny on the clearance of pinometostat. In this cohort, clearanceranged from 1.8 to 3.7 l/h per square meter, showing good concordancewith the PBPK model CL projection of 2.4–3.2 l/h per square meter insubjects older than 1 year. The limited enrollment of patients less than 1year of age in the relapsed/refractory setting precluded a more thoroughassessment of P450 ontogeny on pinometostat clearance. Notwith-standing, the observed Css at 90 mg/m2/d (n = 5; age range, 1–15 years)showed a similar level of agreement with model predictions and wascomparable with the plasma exposure observed in adult at this doselevel (Fig. 2C; Shukla et al., 2015).Tazemetostat (EPZ-6438) is a selective, orally active, small molecule

inhibitor of the histone-lysine methyltransferase Enhancer of ZesteHomolog 2 (EZH2), which has been implicated in the pathogenesis of avariety of malignancies, including B-cell non-Hodgkin lymphoma andintegrase interactor (INI)-1–deficient tumors (Knutson et al., 2013,2014). INI-1–negative malignant rhabdoid tumors, for instance, occurmainly in young children. Other INI-1–negative tumors includeepithelioid sarcomas, epithelioid malignant peripheral nerve sheathtumors, extraskeletal myxoid chondrosarcoma, renal medullary carci-noma, and myoepithelial carcinomas, characterized by their rarity andunmet medical need. INI-1–deficient tumors include several tumors forwhich there is no established standard of care, with a median survival of0.3 years after recurrence in children. The primary objective of thePBPK modeling approach was to simulate tazemetostat exposures inadults prior to scaling to a pediatric population and to predict startingdoses in children that would provide systemic exposures in the safe andefficacious range observed in adults. Tazemetostat PK data frompatients enrolled in the dose escalation cohorts of the first-in-humanphase I clinical study across the oral dose range of 100 mg (suspension)and 100, 200, 400, 800, and 1600 mg (tablet) twice daily, together withphysicochemical and in vitro data including plasma protein binding,blood partitioning, metabolic stability, and P450 phenotyping, wereused to simulate adult exposures by PBPK modeling using GastroPlus(version 8.5; Simulation Plus, Inc). In these adult patients, tazemetostatwas rapidly absorbed with a maximum plasma concentration observedapproximately 1 to 2 hours postdose. Plasma concentrations declinedin a monoexponential manner with a mean t1/2 of approximately 3–6 hours. The tazemetostat AUC was slightly higher than dose pro-portional after a single dose, reasonably dose proportional at steadystate, and comparable between the tablet and suspension formulations.After multiple dosing (day 15), the median time to reach maximumplasma concentrations (tmax) and t1/2 remained unchanged across thedose range, with a modest decrease in systemic exposure on day 15 andno further reduction onward (Ribrag et al., 2015). In the PBPK model,

partition coefficients were set using the Lucakova (Rodgers–RowlandSingle) equation for perfusion-limited tissues. Effective permeabilitywas fitted to the clinical PK data in a top-down approach, refining theinitial Cmax estimated from an in vitro permeability assay. Usingpermeability data from porcine kidney epithelial cells resulted in a 1.6-fold underestimation ofCmax at 800 mg. This difference likely reflects aconcentration-dependent effect since the in vitro assay was performedat low micromolar concentrations, whereas the local gastrointestinalconcentration could be as high as molar range (800 mg in 250ml). Totalclearance consisted of hepatic, intestinal, and renal components. Renalclearance was set as passive renal filtration (i.e., product of free fractionand glomerular filtration rate), which was in line with the preliminaryestimate of renal clearance in the phase I study. Metabolic clearances inthe gut and liver were based on system parameters for the expressionlevels of CYP3A4 in each gut compartment and the average expressionof CYP3A4 in liver. Human liver microsome kinetics (Km and Vmax)were used as initial inputs to model tazemetostat hepatic clearancein humans, prior to refinement based on observed clinical data, and itwas assumed that the fraction of the drug metabolized by CYP3A4was close to unity. The refinement of Km and Vmax values was notsubstantial, being within 20% of the experimentally determined valuesin human liver microsomes. Although the tazemetostat accumulationratio (AUCday15/AUCday1) was suggestive of a modest induction ofmetabolism, omitting to include CYP3A4 induction was not detrimen-tal to the simulation of the steady-state (day 15) exposures (Fig. 3A).Because accumulation ratios remained modest throughout the dosingrange (accumulation ratio of 0.8 to 0.5), and since the simulatedconcentration-time profile was very similar to the observed, CYP3A4induction kinetics were not added to this model. This is a limitation ofthe model, which likely accounts for the nonoptimal capture of the timedependence of tazemetostat PK, particularly at higher doses (althoughpredicted day 1 AUC and Cmax were within 2-fold of observed values atdoses of#800 mg). Furthermore, the scaling of CYP3A autoinductionto a pediatric population (i.e., ontogeny of nuclear hormone receptorssuch as the pregnane X receptor) is not sufficiently understood to allowquantitative modeling. This would have made it difficult to account forautoinduction in the simulations of the pediatric population even if ithad been included in the adult model. Similarly, because no transporterkinetic data were available at the time of this study and sufficient P-gpontogeny data are lacking, the effect of potential efflux transport ontazemetostat PK was not considered in this model. Given these lim-itations, the potential effects of autoinduction should be considered inthe interpretation of the model results. Based on visual inspection, themodel fit of the adult exposures (n = 24) adequately described the time-concentration profiles of tazemetostat and resulted in prediction ofmean AUCss and oral clearance (CL/F) within 630% of the observedresults across the dose range. In addition, mean Cmax,ss values werepredicted within 2-fold of the observed values (Fig. 3B; Rioux et al.,2015). The resultant adult model was then used to simulate tazemetostatsteady-state exposures in discrete pediatric age ranges (6 months to 1year, .1–2 years, .2–6 years, .6–12 years, and .12–18 years) aftertwice-daily administration of an oral suspension. In addition to pediatricphysiology, the simulations accounted for ontogeny in hematocrit,plasma protein levels, and P450 expression (liver and gut). AlthoughCYP3A4 ontogeny wasmodeled, the propensity for CYP3A7-mediatedmetabolism of tazemetostat was unknown and so no interaction withthis major fetal isoform was accounted for in the pediatric model.Nevertheless, the effect of potential metabolism by CYP3A7 isexpected to be limited, since projection of pediatric doses was limitedto patients aged $6 months (Hines, 2009). Using this exposure-basedanalysis, pediatric doses that afforded the target AUC (80% of adultAUCss at 800 mg or 390 mg/m2 twice daily) were identified. On a

940 Rioux and Waters

at ASPE

T Journals on A

ugust 29, 2020dm

d.aspetjournals.orgD

ownloaded from

BSA-normalized basis, the projected doses were comparable across theage range (1–18 years), from 270 to 305 mg/m2 twice daily, with aslightly lower projected dose of 230 mg/m2 twice daily, for the groupaged 6 months to 1 year. This lower dose was attributable to multiplefactors, including CYP3A4 ontogeny (60%–80% of adult expression inthe liver for the group aged 6 months to 1 year), change in free fraction,volume of distribution, and blood flow. By contrast, the tazemetostatfraction absorbed was not projected to change significantly from adultsto children for doses of#300 mg/m2 twice daily. Because the projecteddoses by age were comparable, population simulations were performedto determine the corresponding exposures for each age range at twofixed doses (240 and 300 mg/m2 twice daily; Fig. 3B), since using anage-independent dose was expected to facilitate trial design in this rarepatient population. In brief, at 300 mg/m2 twice daily, mean steady-stateAUC0–t values ranged from approximately 80% to 100% of thatobserved in adults at 800 mg twice daily. By contrast, the mean steady-state Cmax was projected to range from 110% to 190% of that observedat 800 mg twice daily in adults, but within the safe and efficacious

exposure range defined in adults at doses up to 1600 mg twice daily. Alower dose of 240 mg/m2 twice daily was projected to maintain AUCss

values# 80% of the adult target exposure across the age range of 1–18years.These studies demonstrate the prospective application of PBPK early

in clinical development to support clinical trial design, and they haveprovided dosing recommendations for the ongoing trials of pinometo-stat and tazemetostat in pediatric patients.

Current Challenges and Limitations in Pediatric PBPK Modeling

Although multiple ontogenic physiologic processes are fairly wellcharacterized in PBPK models, knowledge gaps remain because somegrowth and maturation trajectories are still only partially or poorlycharacterized. Among these, the Pediatric Biopharmaceutics Classifi-cation System Working Group determined that additional research wasrequired to fully understand age-based changes in gastrointestinalfluid composition, motility, absorptive surface area, and pH ranges

Fig. 3. (A) Representative plots showing the model fit for tazemetostat interim adult PK data at steady state (day 15, n = 24) showed a good fit for each dose group. Symbolsrepresent observed data from individuals (6 S.D., n = 3 or 6 per dose) and the solid line represents the mean profiles predicted from the PBPK model in GastroPlus. Dottedgray lines represent the 90% confidence interval. (B) Predicted mean total steady-state plasma concentration-time profiles of tazemetostat administered as a 300-mg/m2 twice-daily or a 240-mg/m2 twice-daily oral dose across the age ranges, with mean measured total steady-state plasma concentration-time profile of tazemetostat administered as a800-mg (390 mg/m2) or 1600-mg (780 mg/m2) twice-daily oral dose in adults (n = 6 per dose). BID, twice daily.

PBPK Modeling in Pediatric Oncology Drug Development 941

at ASPE

T Journals on A

ugust 29, 2020dm

d.aspetjournals.orgD

ownloaded from

encountered along the gastrointestinal tract, all of which are criticalfactors needed to better understand age-dependent drug absorption(Abdel-Rahman et al., 2012).For clearance, incomplete knowledge of the developmental patterns

for some P450 isoforms, UGTs, and other conjugative enzymes remainsa challenge. Partial coverage of the pediatric age spectrum, sparse datain extrahepatic tissues, and an undetermined fraction metabolized byCYP3A7 (a major isoform in neonates) are some of the factors that canmake scaling adult clearance to children a complex endeavor (Kearnset al., 2003; Abdel-Rahman et al., 2012). In addition, the parameter-ization of P450 ontogeny functions used in PBPKmodels may be basedon data derived either in vitro (expression and/or activity) or in vivo(activity only), and recent studies have highlighted the potential formarked differences between the two methods (Salem et al., 2014;Upreti and Wahlstrom, 2016). In parallel, understanding of thepharmacokinetic effect of the transporter-specific expression level indiscrete age strata is still limited and remains a challenge whenperforming PBPK modeling for pediatric applications. Screening ofthe mRNA expression of uptake and efflux transporters in the humanliver reveals that only a small number of isoforms are detectable up to1 month after birth; in general, although expression increased with age,the timing at which adult mRNA levels were reached did vary (Klaassenand Aleksunes, 2010; Abdel-Rahman et al., 2012; Mooij et al., 2014).Although more data are becoming available on transporter expressionlevels, studies on transporter activity in pediatric populations are stillscarce. As data highlighting the importance of drug transporters in adultmedicine continue to emerge, this critical knowledge gap in thepediatric population becomes even more evident, as raised by theNational Institutes of Health Pediatric Transporter Working Group intheir recent white paper (Brouwer et al., 2015).Although PBPK modeling is increasingly used for prediction of DDI

potential in adult populations, there are a number of gaps in the robustprediction of DDI potential in pediatrics. From the perspective of avictim drug interaction, the magnitude of a change in metabolicclearance depends on the fraction metabolized by the inhibited orinduced pathway(s), and age-related changes owing to ontogeny mayaffect the predictive accuracy in the quantitative contribution ofindividual clearance mechanisms (Salem et al., 2013). Furthermore,nuclear hormone receptor mRNA expression (e.g., the pregnane Xreceptor and the constitutive androstane receptor, known to regulateexpression of genes involved in drug disposition) demonstrate consid-erable interindividual variability in human fetal and pediatric livers(Vyhlidal et al., 2006), with the effect on P450 induction not yet fullyunderstood. This confounds modeling of complex drug interactionssuch as simultaneous inhibition and induction of metabolic pathwaysand/or transporters by a drug (and/or its metabolites), presentingsignificant challenges to the prediction of clinical DDIs in adults(Varma et al., 2015) and even more so in pediatric populations.Ultimately, the ability to model the effect of disease state on PK

would further enhance the PBPKmodeling approach by factoring in thephysiologic sequelae of cancer in children relative to the current data-bases for healthy pediatric subjects; this is not currently a surmountabletask, given the pleiotropic nature of malignant neoplastic disease.Furthermore, the extension of pediatric PBPK modeling to include apharmacodynamics component remains an area of much-neededresearch.The evaluation of the predictive performance of published PBPK

models is an area that is in need of further scrutiny in moving toward amore standardized and systematic reporting of PBPK modeling andsimulation efforts, and this issue has been raised by a number of groupsfrom academia, industry, and regulatory authorities (Maharaj andEdginton, 2014; Zhao, 2014; Sager et al., 2015).

Summary and Conclusions

PBPK modeling and simulation in pediatric oncology is still quitenascent, limited to a few well characterized drugs and often performedas part of a retrospective analysis validating the methodology in thisarea of drug research. Nevertheless, these studies demonstrated theability of PBPK modeling to simulate PK in pediatric patients, refineclinical trial design, inform on covariates influencing drug exposurelevels in children, and evaluate the potential for clinically relevantDDIs. The case studies presented for pinometostat and tazemetostatprovide further validation of their prospective utility a priori to the first-in-pediatrics clinical investigation, providing guidance on starting doseregimen and judicious selection of blood sampling times. Modeling andsimulation endeavors in this area of drug research are particularlychallenging since the patient population is rare. First-in-pediatric studypopulations tend to be relatively small, which impedes the statisticalrigor usually associated with evaluating age as a covariate in later-stagetrials. However, as detailed herein, PBPK modeling applied with a fit-for-purpose tenet can enable informed decisions on the starting dose inphase I pediatric oncology trials, in an effort to ensure that the agecontinuum is treated with a starting dose at or close to a safe andefficacious one. The future direction of PBPK modeling in the fieldof pediatric oncology drug development is likely to see expandedapplication in early clinical development with the assessment of DDIpotential, influence of disease covariates and providing mechanisticinsight into observed variability in the target patient population.

Acknowledgments

The authors thank Epizyme colleagues, clinical investigators, and their teamsand, most importantly, the patients and families who participated in the studiesreported herein.

Authorship ContributionsWrote or contributed to the writing of the manuscript: Rioux, Waters

References

Abdel-Rahman SM, Amidon GL, Kaul A, Lukacova V, Vinks AA, and Knipp GT; Members ofthe BCS Task Force (2012) Summary of the National Institute of Child Health and HumanDevelopment–Best Pharmaceuticals for Children Act Pediatric Formulation InitiativesWorkshop–Pediatric Biopharmaceutics Classification System Working Group. Clin Ther 34:S11–S24.

Adamson PC, Houghton PJ, Perilongo G, and Pritchard-Jones K (2014) Drug discovery inpaediatric oncology: roadblocks to progress. Nat Rev Clin Oncol 11:732–739.

Altamimi MI, Sammons H, and Choonara I (2015) Inter-individual variation in midazolamclearance in children. Arch Dis Child 100:95–100.

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

Barrett JS, Zhang AY, Urtishak KA, Danet-Desnoyers G, Carroll AJ, Hunger SP, and Felix CA(2011) Predicting clinical dose-exposure and exposure-response relationships of pan-antiapoptotic BCL-2 family inhibitor obatoclax in MLL rearranged infant leukemias frompreclinical disease models and adult experience (Abstract). Blood 118:2580.

Basavapathruni A, Olhava EJ, Daigle SR, Therkelsen CA, Jin L, Boriack-Sjodin PA, Allain CJ,Klaus CR, Raimondi A, and Scott MP, et al. (2014) Nonclinical pharmacokinetics and me-tabolism of EPZ-5676, a novel DOT1L histone methyltransferase inhibitor. Biopharm DrugDispos 35:237–252.

Batchelor HK, Fotaki N, and Klein S (2014) Paediatric oral biopharmaceutics: key considerationsand current challenges. Adv Drug Deliv Rev 73:102–126.

Björkman S (2005) Prediction of drug disposition in infants and children by means of physio-logically based pharmacokinetic (PBPK) modelling: theophylline and midazolam as modeldrugs. Br J Clin Pharmacol 59:691–704.

Björkman S (2006) Prediction of cytochrome p450-mediated hepatic drug clearance in neonates,infants and children: how accurate are available scaling methods? Clin Pharmacokinet 45:1–11.

Blake MJ, Gaedigk A, Pearce RE, Bomgaars LR, Christensen ML, Stowe C, James LP, WilsonJT, Kearns GL, and Leeder JS (2007) Ontogeny of dextromethorphan O- and N-demethylationin the first year of life. Clin Pharmacol Ther 81:510–516.

Block M (2015) Physiologically based pharmacokinetic and pharmacodynamic modeling incancer drug development: status, potential and gaps. Expert Opin Drug Metab Toxicol 11:743–756.

Breitkreutz J and Boos J (2007) Paediatric and geriatric drug delivery. Expert Opin Drug Deliv 4:37–45.

Brouwer KLR, Aleksunes LM, Brandys B, Giacoia GP, Knipp G, Lukacova V, Meibohm B,Nigam SK, Rieder M, and de Wildt SN; Pediatric Transporter Working Group (2015) Human

942 Rioux and Waters

at ASPE

T Journals on A

ugust 29, 2020dm

d.aspetjournals.orgD

ownloaded from

ontogeny of drug transporters: review and recommendations of the pediatric transporterworking group. Clin Pharmacol Ther 98:266–287.

Daigle SR, Olhava EJ, Therkelsen CA, Basavapathruni A, Jin L, Boriack-Sjodin PA, Allain CJ,Klaus CR, Raimondi A, and Scott MP, et al. (2013) Potent inhibition of DOT1L as treatment ofMLL-fusion leukemia. Blood 122:1017–1025.

Diestelhorst C, Boos J, McCune JS, Russell J, Kangarloo SB, and Hempel G (2014) Predictiveperformance of a physiologically based pharmacokinetic model of busulfan in children. PediatrHematol Oncol 31:731–742.

Ehrnebo M, Agurell S, Jalling B, and Boréus LO (1971) Age differences in drug binding byplasma proteins: studies on human foetuses, neonates and adults. Eur J Clin Pharmacol 3:189–193.

European Medicines Agency (2013) EMA/CHMP/161314/2013: Assessment Report for Glivec(Imatinib) (May 30, 2013), European Medicines Agency, London.

European Medicines Agency (2014) EMA/CHMP/211243/2014: Concept paper on Qualificationand Reporting of Physiologically-Based Pharmacokinetic (PBPK) Modelling and Analyses(June 26, 2014), European Medicines Agency, London.

Evans WE, Crom WR, Tsiatis A, Green AA, Hayes FA, and Pratt CB (1982) Pharmacokineticmodeling of cisplatin disposition in children and adolescents with cancer. Cancer ChemotherPharmacol 10:22–26.

Gibbs JP, Liacouras CA, Baldassano RN, and Slattery JT (1999) Up-regulation of glutathioneS-transferase activity in enterocytes of young children. Drug Metab Dispos 27:1466–1469.

Hines RN (2009) The ontogeny of drug metabolizing enzymes/pediatric exclusivity, in Handbookof Drug Metabolism (Pearson PG and Wienkers LC eds) 2nd ed, pp 227–239, InformaHealthcare, New York.

Horton TM and Berg SL (2011) Educational paper. The development of new therapies forpediatric oncology. Eur J Pediatr 170:555–559.

Huang SM, Abernethy DR, Wang Y, Zhao P, and Zineh I (2013) The utility of modeling andsimulation in drug development and regulatory review. J Pharm Sci 102:2912–2923.

Jones HM, Chen Y, Gibson C, Heimbach T, Parrott N, Peters SA, Snoeys J, Upreti VV, Zheng M,and Hall SD (2015) Physiologically based pharmacokinetic modeling in drug discovery anddevelopment: a pharmaceutical industry perspective. Clin Pharmacol Ther 97:247–262.

Jones HM, Gardner IB, and Watson KJ (2009) Modelling and PBPK simulation in drug dis-covery. AAPS J 11:155–166.

Kearns GL, Abdel-Rahman SM, Alander SW, Blowey DL, Leeder JS, and Kauffman RE (2003)Developmental pharmacology–drug disposition, action, and therapy in infants and children.N Engl J Med 349:1157–1167.

Kersting G, Willmann S, Würthwein G, Lippert J, Boos J, and Hempel G (2012) Physiologicallybased pharmacokinetic modelling of high- and low-dose etoposide: from adults to children.Cancer Chemother Pharmacol 69:397–405.

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.

Klaassen CD and Aleksunes LM (2010) Xenobiotic, bile acid, and cholesterol transporters:function and regulation. Pharmacol Rev 62:1–96.

Knutson SK, Kawano S, Minoshima Y, Warholic NM, Huang KC, Xiao Y, Kadowaki T, Uesugi M,Kuznetsov G, and Kumar N, et al. (2014) Selective inhibition of EZH2 by EPZ-6438 leads to potentantitumor activity in EZH2-mutant non-Hodgkin lymphoma. Mol Cancer Ther 13:842–854.

Knutson SK, Warholic NM, Wigle TJ, Klaus CR, Allain CJ, Raimondi A, Porter Scott M,Chesworth R, Moyer MP, and Copeland RA, et al. (2013) Durable tumor regression in ge-netically altered malignant rhabdoid tumors by inhibition of methyltransferase EZH2. ProcNatl Acad Sci USA 110:7922–7927.

Kostewicz ES, Aarons L, Bergstrand M, Bolger MB, Galetin A, Hatley O, Jamei M, Lloyd R,Pepin X, and Rostami-Hodjegan A, et al. (2014) PBPK models for the prediction of in vivoperformance of oral dosage forms. Eur J Pharm Sci 57:300–321.

Li J and Gwilt P (2002) The effect of malignant effusions on methotrexate disposition. CancerChemother Pharmacol 50:373–382.

Maharaj AR and Edginton AN (2014) Physiologically based pharmacokinetic modeling andsimulation in pediatric drug development. CPT Pharmacometrics Syst Pharmacol 3:e150.

Mahmood I (2006) Prediction of drug clearance in children from adults: a comparison of severalallometric methods. Br J Clin Pharmacol 61:545–557.

Mooij MG, Schwarz UI, de Koning BA, Leeder JS, Gaedigk R, Samsom JN, Spaans E, vanGoudoever JB, Tibboel D, and Kim RB, et al. (2014) Ontogeny of human hepatic and intestinaltransporter gene expression during childhood: age matters. Drug Metab Dispos 42:1268–1274.

Ogungbenro K and Aarons L; CRESim & Epi-CRESim Project Groups (2014a) Physiologicallybased pharmacokinetic modelling of methotrexate and 6-mercaptopurine in adults and children.Part 1: methotrexate. J Pharmacokinet Pharmacodyn 41:159–171.

Ogungbenro K and Aarons L; CRESim & Epi-CRESim Project Groups (2014b) Physiologicallybased pharmacokinetic modelling of methotrexate and 6-mercaptopurine in adults and children.Part 2: 6-mercaptopurine and its interaction with methotrexate. J Pharmacokinet Pharmacodyn41:173–185.

Poley JR, Dower JC, Owen CA, Jr, and Stickler GB (1964) Bile acids in infants and children.J Lab Clin Med 63:838–846.

Pons G, Carrier O, Richard MO, Rey E, d’Athis P, Moran C, Badoual J, and Olive G (1988)Developmental changes of caffeine elimination in infancy. Dev Pharmacol Ther 11:258–264.

Ribrag V, Soria JC, Thomson B, Reyderman L, Kraljevic S, Chen R, Salazar P, Kuznetsov G,Keilhack H, Blakemore SJ, et al. (2015) Phase 1 study of EPZ-6438 (E7438), an Enhancer ofZeste Homolog-2 (EZH2) inhibitor: dose determination and preliminary activity in non-Hodgkin lymphoma, in Proceedings of the International Conference on Malignant Lymphoma;2015 June 17–20; ICML, Bellinzona, Italy.

Ries LAG, Smith MA, Gurney JG, Linet M, Tamra T, Young JL, and Bunin GR, editors (1999)Cancer Incidence and Survival among Children and Adolescents: United States SEER Pro-gram 1975–1995, SEER Program Pub No. 99-4649, National Cancer Institute, Bethesda, MD.

Rioux N, Keilhack H, Thomson B, Chesworth R, Copeland RA, Hedrick E, Ho P, and Waters NJ(2015) Physiologically-based pharmacokinetic modeling to support clinical investigation of theEZH2 inhibitor, tazemetostat (EPZ-6438) in INI1-deficient pediatric tumors (Abstract). MolCancer Ther 14:A135.

Rossig C, Juergens H, and Berdel WE (2011) New targets and targeted drugs for the treatment ofcancer: an outlook to pediatric oncology. Pediatr Hematol Oncol 28:539–555.

Sager JE, Yu J, Ragueneau-Majlessi I, and Isoherranen N (2015) Physiologically based phar-macokinetic (PBPK) modeling and simulation approaches: a systematic review of publishedmodels, applications and model verification. Drug Metab Dispos 43:1823–1837.

Salem F, Johnson TN, Abduljalil K, Tucker GT, and Rostami-Hodjegan A (2014) A re-evaluationand validation of ontogeny functions for cytochrome P450 1A2 and 3A4 based on in vivo data.Clin Pharmacokinet 53:625–636.

Salem F, Johnson TN, Barter ZE, Leeder JS, and Rostami-Hodjegan A (2013) Age relatedchanges in fractional elimination pathways for drugs: assessing the impact of variable ontogenyon metabolic drug-drug interactions. J Clin Pharmacol 53:857–865.

Shepard T, Scott G, Cole S, Nordmark A, and Bouzom F (2015) Physiologically based models inregulatory submissions: Output from the ABPI/MHRA forum on physiologically based mod-eling and simulation. CPT Pharmacometrics Syst Pharmacol 4:221–225.

Shukla N, O’Brien M, Silverman L, Pauly M, Wetmore C, Loh M, Whitlock J, Brown P, DaigleS, Blakemore SJ, et al. (2015) Preliminary report of the phase 1 study of the DOT1L inhibitorpinometostat (EPZ-5676) in children with MLL-r acute leukemia: safety, exposure and evi-dence of target inhibition, in Proceedings of the 57th American Society of Hematology AnnualMeeting; 2015 December 5–8; Orlando, FL. American Society of Hematology, Washington,DC.

Skolnik JM, Barrett JS, Jayaraman B, Patel D, and Adamson PC (2008) Shortening the timelineof pediatric phase I trials: the rolling six design. J Clin Oncol 26:190–195.

Stein E, Garcia-Manero G, Rizzieri D, Savona MR, Tibes R, Altman J, Jongen-Lavrencic M,Döhner H, Armstrong S, and Pollock RM, et al. (2014) The DOT1L inhibitor EPZ-5676: safetyand activity in relapsed/refractory patients with MLL-rearranged leukemia (Abstract). Blood124:387.

Thai HT, Mazuir F, Cartot-Cotton S, and Veyrat-Follet C (2015) Optimizing pharmacokineticbridging studies in paediatric oncology using physiologically-based pharmacokinetic model-ling: application to docetaxel. Br J Clin Pharmacol 80:534–547.

Upreti VV and Wahlstrom JL (2016) Meta-analysis of hepatic cytochrome P450 ontogeny tounderwrite the prediction of pediatric pharmacokinetics using physiologically based pharma-cokinetic modeling. J Clin Pharmacol 56:266–283.

Varma MV, Pang KS, Isoherranen N, and Zhao P (2015) Dealing with the complex drug-druginteractions: towards mechanistic models. Biopharm Drug Dispos 36:71–92.

Vyhlidal CA, Gaedigk R, and Leeder JS (2006) Nuclear receptor expression in fetal and pediatricliver: correlation with CYP3A expression. Drug Metab Dispos 34:131–137.

Wagner C, Zhao P, Pan Y, Hsu V, Grillo J, Huang SM, and Sinha V (2015) Application ofphysiologically based pharmacokinetic (PBPK) modeling to support dose selection: report ofan FDA public workshop on PBPK. CPT Pharmacometrics Syst Pharmacol 4:226–230.

Waters NJ, Thomson B, Gardner I, Johnson TN, Olhava EJ, Pollock RM, Legler M, CopelandRA, and Hedrick E (2014) Pediatric dose determinations for the phase I study of the DOT1Linhibitor, EPZ-5676, in MLL-r acute leukemia: leveraging early clinical data in adults throughphysiologically-based pharmacokinetic modeling. Blood 124:3619.

West GB, Brown JH, and Enquist BJ (1999) The fourth dimension of life: fractal geometry andallometric scaling of organisms. Science 284:1677–1679.

Zhao P (2014) Application of PBPK modeling and simulations in pediatric drug development, inProceedings of the American Society for Clinical Pharmacology and Therapeutics AnnualMeeting; 2014 March 20; Atlanta, GA. American Society for Clinical Pharmacology andTherapeutics, Alexandria, VA.

Zhao P, Rowland M, and Huang SM (2012) Best practice in the use of physiologically basedpharmacokinetic modeling and simulation to address clinical pharmacology regulatory ques-tions. Clin Pharmacol Ther 92:17–20.

Address correspondence to: Dr. Nigel J. Waters, Syros Pharmaceuticals, 620Memorial Drive, Cambridge, MA 02139. E-mail: nwaters@syros.com

PBPK Modeling in Pediatric Oncology Drug Development 943

at ASPE

T Journals on A

ugust 29, 2020dm

d.aspetjournals.orgD

ownloaded from