doxorubicin pharmacokinetics: macromolecule binding, metabolism, and excretion in the context of a...

14
Doxorubicin Pharmacokinetics: Macromolecule Binding, Metabolism, and Excretion in the Context of a Physiologic Model DANIEL L. GUSTAFSON, 1 JEFFREY C. RASTATTER, 1 TINA COLOMBO, 2 MICHAEL E. LONG 1 1 Department of Pharmaceutical Sciences, School of Pharmacy, University of Colorado Health Sciences Center, 4200 East 9th Avenue, Denver, Colorado 80262 2 Laboratory of Cancer Chemotherapy, Mario Negri Institute for Pharmacological Research, Via Eritrea 62, 20157 Milan, Italy Received 19 June 2001; revised 25 September 2001; accepted 25 February 2002 ABSTRACT: The studies described herein were designed to determine whether doxo- rubicin (DOX) pharmacokinetics (PKs) could be described by a physiologically based PK model that incorporated macromolecule-specific binding and organ-specific metabolism and excretion. Model parameters were determined experimentally, or were gathered from the literature, in a species-specific manner, and were incorporated into a phy- siologically based description of DOX blood and tissue distribution for mice, dogs, and humans. The resulting model simulation data were compared with experimentally determined data using PK parameters calculated using compartmental or noncompart- mental analysis to assess the predictability of the models. The resulting physiologically based PK model that was developed could accurately predict blood and tissue PKs of DOX in mice. When this model was interspecies extrapolated to predict DOX levels in dogs and humans undergoing treatment for cancer, predictions in dog plasma or human serum were also consistent with the actual clinical data. This model has potential utility for predicting the magnitude of PK interactions of DOX with other drugs, and for pre- dicting changes in DOX PKs in any number of clinical situations. ß 2002 Wiley-Liss, Inc. and the American Pharmaceutical Association J Pharm Sci 91:1488–1501, 2002 Keywords: doxorubicin; pharmacokinetics; physiologically based modeling; simula- tion; PBPK INTRODUCTION Doxorubicin (DOX) is a naturally occurring an- thracycline that has a broad spectrum of activity for the treatment of cancer. The five anthracy- clines currently in clinical use worldwide are DOX, daunorubicin, idarubicin, epirubicin, and pirarubicin. Standard combination chemotherapy regimens for the treatment of solid tumors, lymphomas, and leukemias usually contain an anthracycline component. 1 The dose-limiting toxi- cities of the two commonly used anthracyclines, DOX and daunorubicin, are myelosuppression, mucositis, and cardiac toxicity. 2 The anthracyclines can react with numerous cellular components to induce a number of effects that are thought to have a role in the antineo- plastic and toxic effects of these compounds. An- thracyclines are capable of DNA intercalation and inhibition of RNA and DNA polymerases, 3 inter- action with topoisomerase II, 4 and alkylation of DNA. 5 DOX is also capable of generating reactive oxygen species through quinone redox cycling, 6,7 and perturbing cellular Ca 2þ homeostasis through both receptor-mediated 8,9 and redox-mediated 10 mechanisms. Other mechanisms of action have also been investigated, including inhibition of 1488 JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002 Correspondence to: Daniel L. Gustafson (Telephone: 303- 315-0755; Fax: 303-315-6281; E-mail: [email protected]) Journal of Pharmaceutical Sciences, Vol. 91, 1488–1501 (2002) ß 2002 Wiley-Liss, Inc. and the American Pharmaceutical Association

Upload: daniel-l-gustafson

Post on 15-Jun-2016

217 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

Doxorubicin Pharmacokinetics: Macromolecule Binding,Metabolism, and Excretion in the Context of aPhysiologic Model

DANIEL L. GUSTAFSON,1 JEFFREY C. RASTATTER,1 TINA COLOMBO,2 MICHAEL E. LONG1

1Department of Pharmaceutical Sciences, School of Pharmacy, University of Colorado Health Sciences Center,4200 East 9th Avenue, Denver, Colorado 80262

2Laboratory of Cancer Chemotherapy, Mario Negri Institute for Pharmacological Research, Via Eritrea 62, 20157 Milan, Italy

Received 19 June 2001; revised 25 September 2001; accepted 25 February 2002

ABSTRACT: The studies described herein were designed to determine whether doxo-rubicin (DOX) pharmacokinetics (PKs) could be described by a physiologically based PKmodel that incorporated macromolecule-specific binding and organ-specific metabolismand excretion. Model parameters were determined experimentally, or were gatheredfrom the literature, in a species-specific manner, and were incorporated into a phy-siologically based description of DOX blood and tissue distribution for mice, dogs, andhumans. The resulting model simulation data were compared with experimentallydetermined data using PK parameters calculated using compartmental or noncompart-mental analysis to assess the predictability of the models. The resulting physiologicallybased PK model that was developed could accurately predict blood and tissue PKs ofDOX in mice. When this model was interspecies extrapolated to predict DOX levels indogs and humans undergoing treatment for cancer, predictions in dog plasma or humanserumwere also consistent with the actual clinical data. This model has potential utilityfor predicting the magnitude of PK interactions of DOX with other drugs, and for pre-dicting changes in DOX PKs in any number of clinical situations. � 2002 Wiley-Liss, Inc.

and the American Pharmaceutical Association J Pharm Sci 91:1488–1501, 2002

Keywords: doxorubicin; pharmacokinetics; physiologically based modeling; simula-tion; PBPK

INTRODUCTION

Doxorubicin (DOX) is a naturally occurring an-thracycline that has a broad spectrum of activityfor the treatment of cancer. The five anthracy-clines currently in clinical use worldwide areDOX, daunorubicin, idarubicin, epirubicin, andpirarubicin. Standard combination chemotherapyregimens for the treatment of solid tumors,lymphomas, and leukemias usually contain ananthracycline component.1 The dose-limiting toxi-

cities of the two commonly used anthracyclines,DOX and daunorubicin, are myelosuppression,mucositis, and cardiac toxicity.2

The anthracyclines can react with numerouscellular components to induce a number of effectsthat are thought to have a role in the antineo-plastic and toxic effects of these compounds. An-thracyclines are capable of DNA intercalation andinhibition of RNA and DNA polymerases,3 inter-action with topoisomerase II,4 and alkylation ofDNA.5 DOX is also capable of generating reactiveoxygen species through quinone redox cycling,6,7

and perturbing cellular Ca2þ homeostasis throughboth receptor-mediated8,9 and redox-mediated10

mechanisms. Other mechanisms of action havealso been investigated, including inhibition of

1488 JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002

Correspondence to: Daniel L. Gustafson (Telephone: 303-315-0755; Fax: 303-315-6281;E-mail: [email protected])

Journal of Pharmaceutical Sciences, Vol. 91, 1488–1501 (2002)� 2002 Wiley-Liss, Inc. and the American Pharmaceutical Association

Page 2: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

thioredoxin reductase11 and interaction with com-ponents of the plasma membrane.12 The role ofeach of these processes in the causation of anti-tumor and side effects is still debated, although itis reasonable to assume that all have some role inthe pharmacology of these compounds.

The pharmacokinetics (PKs) of DOX has beenwell studied in many species. After intravenousdosing, DOX blood levels fall dramatically as thedrug distributes into tissues, followed by a slowelimination phase due to renal and biliary clear-ance and metabolism. DOXmetabolism occurs viareduction of a side chain carbonyl group by aldo-keto reductases13 yielding doxorubicinol, and byreductive cleavage of the sugar moiety to form the7-hydroxy aglycone.14 DOX partitioning fromblood to tissues has been shown to correlate withDNA concentration,15 and DOX is also known tobind to anionic lipids, particularly cardiolipin.16,17

To relate drug dosage to therapeutic and/ortoxic effects, physiologically based PK (PBPK)models are useful tools that allow for simulationand prediction of target tissue concentrations ofactive drug or metabolites. PBPK models are amathematical representation of a biological sys-tem constructed using known physiologic andbiochemical constants. The body is divided intocompartments that represent individual organsand tissue groupings, and the transport, clearance,andmetabolism of xenobiotics between these com-partments or within these compartments is de-scribed using mass balance ordinary differentialequations.18 Physiologically based modeling hasmany advantages which include: (a) utilization ofa large body of physiologic and physiochemicaldata; (b) extrapolation of results both across spe-cies and routes of administration; and (c) predic-tion of PKs and target tissue dosimetry over awiderange of doses.19

Previous PBPK models for DOX have beendeveloped,20,21 but these models have used tissuepartitioning as defined by plasma/tissue concen-tration ratios to describe DOX tissue distribution,and have not included specific terms for thevarious metabolic and elimination pathways thathave a role in DOX disposition. The model de-veloped with these studies uses experimentallydefined biochemical parameters for DOX metabo-lism and macromolecule binding, and places themin the context of a physiologic model representingorgans that are targets for toxicity or have a rolein DOX metabolism and excretion. The resultingmodel output describesDOXdistribution inmousetissues and calculatedDOXPKparameters.When

the model is scaled to canine or human physiologyand biochemical parameters, the resulting outputagain correlates with DOX serum PKs of patientsreceivingDOX therapy. These results suggest thatthe DOX PBPK model developed herein incorpo-rated the major factors that dictate DOX PK inmammalian species.

MATERIALS AND METHODS

Chemicals

DOX hydrochloride, daunorubicin hydrochloride,calf-thymus DNA, and Hoechst 33258 were pur-chased from Sigma Chemical Co. (St. Louis, MO).All other reagents were of analytical grade.

DOX PK Studies in Mice

Female, Balb/c mice (4–5 weeks old) were pur-chased from Harlan Sprague Dawley (Indiana-polis, IN) and allowed to acclimate for 7 days.Animals were housed (three per cage) in poly-carbonate cages and kept on a 12-h light/darkcycle. Food andwater were given ad libitum. Afteracclimation, mice were randomly assigned to thetime-point groups, with three mice per group. Allstudies were conducted in accordance with theNational Institutes of Health guidelines for thecare and use of laboratory animals, and animalswere housed in a facility accredited by theAmerican Association of Laboratory Animal Care.

DOX was administered by intravenous (iv) in-jection into the tail vein at a dose of 6 mg/kg in avolume of 50 mL. Animals were killed by cervicaldislocation after methoxyflurane anesthesia, andtissue and blood samples collected at 5, 10, 30, 60,120, and 480 min after injection. Heart, liver,kidney, small intestine, large intestine, skeletalmuscle, fat, spleen, and bone marrow were collec-ted, rinsed with phosphate buffered saline (pH7.6), and stored at�808C until assayed. Blood wascollected by cardiac puncture, placed in hepar-inized vials, and stored at �208C until analyzed.

DOX PKs in Dogs

Serum samples were obtained from client animalsbeing treated for lymphoma with DOX (30 mg/m2)every 3 weeks for five cycles at the VeterinaryTeaching Hospital at Colorado State University(Fort Collins, CO). All samples were obtained fromdogs whose owners had signed a consent agree-ment regarding the inclusion of the patient in aclinical PK study.

DOXORUBICIN PHARMACOKINETICS 1489

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002

Page 3: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

DOX High-Pressure Liquid Chromatography(HPLC) Analysis

An extraction protocol modified from De Jonget al.22 was used for blood, serum, and tissuesamples. Briefly, tissue was homogenized in waterto give a final concentration of 10 mg/mL. Analiquot of 100 mL homogenate, 50 mL whole blood,or 100 mL serum was transferred to a 1.5-mL poly-propylene microfuge tube; 50 mL 1 mM daunor-ubicin was added as an internal standard, and0.6 mL methanol was added. The samples werevortex mixed for 15 min, 0.25 mL 12 mM phos-phoric acid was added, and then centrifuged at10,000g for 8 min. The resulting supernatant wascollected, the volume adjusted to 1.0 mL, andanalyzed by HPLC with fluorescence detection aspreviously described.22

The HPLC system used was a Beckman GoldSystem (Beckman-Coulter, Fullerton, CA) con-sisting of a 126 pump module, 508 autosampler,and a Jasco FP-920 fluorescence detector (JascoCorp., Tokyo, Japan) with excitation and emissionwavelengths set at 480 and 580 nm, respectively.The mobile phase consisted of 15 mM NaH2PO4

(pH 4)/acetonitrile in a 2:1 volume-to-volumeratio at a flow rate of 1 mL/min. Separationwas done on an Alltima C18 (5 mm) 250� 4.6 mmcolumn (Alltech Associates Inc., Deerfield, IL)fitted with an Alltima C18 guard cartridge (AlltechAssociates).

DNA Content Analysis

Tissue preparation and DNA analysis weredone using a protocol modified from Downs andWilfinger.23 Mouse tissues were collected as de-scribed above from untreated animals. Dog tis-sues were collected upon necropsy from clientanimals at the Veterinary Teaching Hospital(Colorado State University) and stored at �808C.Human tissues, either surgical or autopsy speci-mens, were acquired through the CooperativeHuman Tissue Network Western Division (CaseWestern Reserve University, Cleveland, OH).Tissues (10 mg per 0.75 mL) were homogenizedin AT extraction solution (1 N NH4OH, 0.2%Triton X-100) and incubated at 378C for 10 min.An aliquot of 100 mL homogenate was then dilutedto 1 mL with Assay Buffer (100 mM NaCl, 10 mMEDTA, 10 mM Tris, pH 7.0) and centrifugedat 2500g for 30 min at room temperature. Thesupernatant was collected and stored on ice untilassayed.

The assay mixture consisted of 10 mMTris (pH 7.0), 100 mM NaCl, 10 mM EDTA, and100 ng/mL Hoechst 33258. Fifty microliters ofsample (blank, standards, and unknowns) wasadded to 2 mL of the assay mixture, and thefluorescence determined using a Hitachi F-2000fluorescence spectrophotometer (Hitachi Instru-ments Inc., San Jose, CA) with excitation andemission wavelengths set at 350 and 455 nm,respectively. Blanks, samples, and unknownswere all measured in triplicate, and measure-ments were considered to be valid if replicatesamples differed by less than 10%. The concentra-tion of DNA in unknown samples was calculatedusing a standard curve generated by plottingfluorescence units versus DNA concentration instandards made up using calf thymus DNA.MolarDNA concentration was calculated for each tissueusing a value of 330.9 formula weight for DNAbases and a density of 1 g/mL for all tissues.

DOX PBPK Model Development

A PBPK model for DOX was developed incorpor-ating DNA and cardiolipin binding, tissue-specificmetabolism, and biliary and urinary eliminationof DOX into a seven-compartment flow-limitedmodel. A schematic of this model is shown inFigure 1. Physiologic parameters (tissue volumesand blood flows) for mouse, dog, and human aretaken from Brown et al.24 and are presented inTable 1. Bone marrow blood flow parameters wereestimated from studies in rat25 for the mousemodel, and humans26 for the dog and humanmodels. Tissue DNA content was measured inmouse, dog, and human tissues and the resultsare presented in Table 2. Cardiolipin levels inmouse, dog, and human tissues were estimatedfrom measurements made in rat tissues.27 Theuse of the rat cardiolipin values to estimate thelevels in mouse, dog, and human tissues is sup-ported by previous studies that show similartissue levels of cardiolipin, when measured aspercent of total lipid phosphorous, in mouse,human, and other species as compared with therat.28,29

Affinity constants for DOX–DNA (KDNA) andDOX–cardiolipin (KCAL) binding were set at 200and 400 nM, respectively. These values are con-sistent with the in vitro binding characteristics ofDOX to DNA and cardiolipin.17 The concentrationof DNA and cardiolipin in the tissues was modi-fied by a factor that represents the number ofmolecules of DNA or cardiolipin that bind one

1490 GUSTAFSON ET AL.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002

Page 4: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

DOX molecule. From in vitro studies, it has beenapproximated that the binding site in DNA forDOX is made up of two base pairs (four bases)30;however, in the modeling process, this term was

derived as 500 bases. This discrepancy betweenthe in vitro number of bases constituting a DOXbinding site and the model-derived term can beeasily accounted for because of obvious differencesin ‘‘naked’’ DNA used in in vitro studies andprotein-coated dynamic DNA that is presentin vivo. From in vitro data with cardiolipin, amolar ratio of 2:1 (DOX/cardiolipin) has beenproposed,17 and the number used in the model forcardiolipin molecules per DOX binding site is two.

A generic tissue compartment mass balanceequation is represented by eq. 1:

dAT

dt¼ QTðCA � CVT

Þ ð1Þ

where AT is the amount of drug in the compart-ment, QT is blood flow to the compartment(Table 1), and CA and CVT

are the arterial andvenous blood concentrations of drug being deliv-ered to and exiting from the compartment. Thearterial blood concentration available to all tis-sues, with the exception of liver, is considered tobe the free drug concentration in the blood, andit is assumed that DOX is 70% bound to plasmaprotein.31 For liver, the model assumes that bothfree and bound drugs are available for tissueuptake.32 To account for binding of DOX to DNAand cardiolipin within the respective tissues, theconcentration of DOX leaving in the tissue venousblood flow is modified by eq. 2:

CVT¼ AT

VT þ TDNA � VT

KDNA þ CVT

þ TCAL � VT

KCAL þ CVT

ð2Þ

where VT is the volume of the tissue compartment(Table 1), TDNA, and TCAL are the DNA and car-diolipin binding capacity available (DNA or car-diolipin concentration divided by the number ofmolecules per DNA binding site), and KDNA andKCAL are the binding affinities of DOX to DNAand cardiolipin, respectively. This mathematicalrepresentation of saturable chemical-specific bind-ing to macromolecules in tissues is modified fromthose used to describe 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) binding and PKs using a phy-siologic model.33

Metabolic and Excretory Model Parameters

DOX metabolism to doxorubicinol and the 7-OH-aglycone occurs in hepatic and extrahepatictissues. Km and Vmax values for mice, dogs, andhumans were used as previously reported forDOX metabolism by aldo-keto reductases in liver

Table 1. Physiologic Parameters Used in PBPKModel Simulations

Mouse Dog Human

% of body weightOrgan volumea

Liver 5.5 3.3 2.6Heart 0.5 0.8 0.5Kidney 1.7 0.5 0.4Bone marrow 2.1 2.1 2.1Gut 4.2 3.7 1.7Slowly perfused 70.5 75.8 77.3Rapidly perfused 10.6 5.6 7.5Blood 4.9 8.2 7.9

% of cardiac outputOrgan blood flowLiver 2.0 4.6 4.6Heart 6.6 4.6 4.0Kidney 9.1 17.3 17.5Bone marrowb 1.0 3.0 3.0Gut 14.1 25.1 18.1Slowly perfused 40.0 38.0 34.3Rapidly perfused 27.2 7.4 18.5

aPhysiologic parameters were obtained from Brown et al.24bFemoral bone marrow blood flow in mice is estimated from

data in rats by Iversen et al.,25 and dog and human values areestimated from a study by Lahtinen et al.26

Figure 1. Schematic representation of a physiologi-cally based pharmacokinetic model for DOX. Solidarrows represent blood flow, and dotted or dashedarrows represent metabolic or excretory pathways thateliminate the parent drug.

DOXORUBICIN PHARMACOKINETICS 1491

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002

Page 5: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

and kidney.34 Glycosidase activity in liver, kidney,and heart is expressed as a first-order rate con-stant based on measured values in mouse, dog,and human.34 These first-order rate constantswere calculated by transforming published resultsin nanomoles per milligram of protein, to hour�1

by incorporating the time frame in which theexperiments were performed and using the inte-grated first-order rate expression (At¼A0 � e�kt)and solving for k. Values were converted fromper milligram of protein to per organ volume byusing a conversion of 120mg of protein per gram oftissue inkidney, 180mgof protein per gramof livertissue, and 15 mg of microsomal protein per gramof tissue for kidney, liver, and heart. These con-version valuesweredetermined experimentally byusing known tissue amounts and determining theprotein content. Parameters for fecal and urinaryexcretion of DOXwere optimized within themodelto best describe the tissue distribution data andtaking into account published fecal and urinaryelimination kinetics.35,36 Themetabolic and excre-tory parameters used are shown in Table 3.

Data Analysis

The predictive capability of the model was asses-sed by calculating the median absolute perfor-mance error (MAPE%), the median performanceerror (MPE%), and the root mean squared perfor-mance error (RMSPE%).37 The performance er-rors were calculated as the difference between themeasured values and the predicted values nor-malized to the predicted value as shown in eq. 338:

Performance Error ðPEÞ

¼ Cmeasured � Cpredicted

Cpredicted� 100% ð3Þ

The MAPE%, which is a measure of the accuracyof the prediction, was calculated as shown in eq. 4where n is the total number of samples for thattissue:

MAPE% ¼ medianðjPE1j; jPE2j; . . . jPEnjÞ ð4Þ

The MPE%, which is a measure of the bias of theprediction, was calculated as:

MPE% ¼ median ðPE1;PE2; . . .PEnÞ ð5Þ

Table 2. DNA and Cardiolipin Tissue Levels Used in PBPK Model Simulations

Organ Mouse Dog Human

DNA contenta

Liver 26.0� 4.0 14.6� 0.4 23.7� 2.3Heart 20.6� 4.0 3.2� 1.0 8.3� 4.0Kidney 50.2� 7.9 13.5� 2.7 16.2� 2.2Bone marrow 175.1� 36.8 23.7� 12.7 19.1� 13.7Gutb 60 8.1� 2.8 25.2� 2.3Slowly perfusedc 10 4.5 4.5Rapidly perfusedd 60 15 15

Cardiolipin contente Mouse, dog, and humanLiver 44.6Heart 43.8Kidney 52.3Bone marrow 25Gut 25Slowly perfused 15Rapidly perfused 30

aValues are expressed in mmol/L, and represent the mean�SD of at least three independentdeterminations.

bMouse value is based on a proportional average of DNAmeasurements made in large and smallintestine and stomach. Dog and human values are based on measurements made in necropsycollected tissues along the gastrointestinal tract.

cMouse value is based on a proportional average of DNAmeasurements made in skeletal muscleand adipose tissue. Dog and human values are based on DNA measurements made in dog skeletalmuscle.

dMouse value is based on DNA measurements made in spleen, kidney, and liver.eTissue levels of cardiolipin are expressed in mmol/L assuming a molecular weight of 1300.

Mouse, dog, and human tissue cardiolipin levels are estimated based on values obtained in rattissues.27

1492 GUSTAFSON ET AL.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002

Page 6: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

The RMSPE%, which is a measure of the accuracyof the prediction, was calculated as:

RMSPE% ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPni¼1

PE2i

n

vuuutð6Þ

Total Protein Determination

Total protein was determined using the BCAtotal protein assay reagent (Pierce Chemical,Rockford, IL) with bovine serum albumin as astandard.

Computer Simulation and Software

The simulation model was implemented in ad-vanced continuous simulation language (ACSL)Tox version 11.5.2 (Pharsight Corp., MountainView, CA) on a PC-based pentium computer.Compartmental modeling and calculation of PKparameters were done using WinNonlin version3.0 (Pharsight Corp.).

RESULTS

DOX PKs and Model Simulations in Mice

DOX blood and tissue levels were determinedin female Balb/C mice after a single iv dose of6 mg/kg; the resulting blood and tissue levels over8 h are shown in Figure 2. Tissue and serumlevels from CD2F1/Crl BR male mice treated with10 mg/kg DOX via a single iv dose are alsoshown.39 Data from the previously publishedstudy were kindly provided by Dr. MaurizioD’Incalci (Mario Negri Institute for Pharmacolo-gical Research, Milan, Italy). Model simulationsat both dose levels (6 and 10 mg/kg) are shownalong with the data points for each tissue. Theresulting model simulations are in agreementwith the actual data for blood and serum at bothdosing levels (Fig. 2A). Levels in all other tissuesare consistent with the predicted values, andthe simulations have the same tissue-specificcharacteristics as the data values. This is mostprominently shown in bone marrow data andsimulations (Fig. 2F), where both the data and the

Table 3. Metabolic and Excretory Parameters Used in PBPK Model Simulations

Metabolism

Mouse Dog Human

Km Vmax Km Vmax Km Vmax

Aldo-keto reductasea

Liver 97 410 118 4633 275 1804Kidney 172 1080 133 1685 539 3161

Glycosidaseb

Liver 9524 13,295 12,104Kidney 7531 1805 484Heart 883 2061 760

Excretory parameters Mouse, dog, and humanKm Vmax

Fecal eliminationc

Biliary transport 10 300Gut transport 0.2 80

Urinary eliminationd

Active secretion 10 100Fraction renal bloodflow cleared

0.1

aAldo-keto reductase activities are expressed as Km (mM) and Vmax (mmol h�1 kg tissue�1) and were calculated from in vitro databy Loveless et al.34

bGlycosidase activity is expressed as a first-order rate constant (h�1 kg tissue�1) as calculated from in vitro data by Lovelesset al.34

cBiliary transport and gut transport of DOX into the gut lumen is expressed as Km (mM) and Vmax (mmol h�1 kg tissue�1). Theseparameters were optimized in the model using available mouse tissue data.

dUrinary elimination of DOX is expressed as both glomerular filtration and active secretion from arterial blood flowing to thekidney. Glomerular filtration is estimated to be equal to 10% of renal blood flow, and active secretion is expressed as Km (mM) andVmax (mmol h�1 kg tissue�1).

DOXORUBICIN PHARMACOKINETICS 1493

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002

Page 7: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

simulation show a slow accumulation of drug inthe tissue over the time frame of the sampling.

Predictive performance of the model was anal-yzed in mouse tissues by calculating the MAPE%,MPE%, and RMSPE%, as described previously,and the results are shown in Table 4. The predic-tive accuracy, as measured by MAPE%, variedbetween 16.7 and 40.9%, with a majority of thetissues having MAPE% values of approximately20. The median percent variability in the actualtissue data sets for the 6 mg/kg dose varied from16.4 to 65.7%. Therefore, model predictions weregenerally within the variance of the data. Thepredictive simulations showed no consistent biastoward over or under prediction as shown by theMPE%. These values ranged from�22.3 to 12.8%,with four of the tissues showing a negative MPE%and two having positive MPE%.

Whole blood and serum PK parameters werecalculated using a two-compartment descrip-

tion and the results were compared with model-generated values taken at the same time points.This allowed for a more direct comparison of theactual and model-generated data; the results areshown in Table 5. PK parameters estimated fromactual time-course data were similar to para-meters estimated from data points generated bythe PBPK model at the sampling time points.

Tissue PK parameters, area under the curve(AUC), and terminal half-life, were also calcu-lated in tissues for both the measured and thesimulated data, and the results are shown inTable 6. From these data, the ratios (measured/simulated) for each parameter, at each dose, werecalculated to compare the measured and thesimulated values. These results show that forAUCs, the simulated data compared very well tothe measured data at both 6 and 10 mg/kg doses,with the largest discrepancy being 54%. Themeasured data used to calculate the tissue PKparameters had coefficients of variation (CV%) of42.8� 32.7 and 17.5� 13.0 for the 6 mg/kg and10 mg/kg determinations at each time-point,respectively. The terminal half-life ratios weresimilar in most cases, except for kidney at 6 mg/kgand heart at 10 mg/kg, where the measured andsimulated determinations were different by ap-proximately three-fold. Because this parameteris calculated from the slope of a putative linear

Table 4. Measures of Predictive Performance forPBPK Model Simulations in Mice and Canine andHuman Cancer Patients

Species Tissue MAPE%a MPE%b RMSPE%c

Mouse Blood/Plasma 16.7 �3.2 57.9Liver 20.8 11.6 45.9Kidney 22.3 �22.3 25.2Gut 40.9 �13.5 48.2Heart 21.0 �7.1 31.6Bone marrow 24.2 12.8 25.3

Canine Serum 11.2 0.7 25.2

Human Patient 1 17.2 �6.5 25.4Patient 2 51.8 49.7 98.9Patient 3 46.9 22.4 61.7Patient 4 30.1 �14.2 40.5Patient 5 21.7 �20.8 28.3

aMAPE% represents the median absolute performanceerror, which is a measure of accuracy and is calculated asdescribed in the text.

bMPE% represents the median performance error, which isa measure of bias and is calculated as described in the text.

cRMSPE% represents the root mean squared performanceerror, which is a measure of accuracy and is calculated asdescribed in the text.Figure 2. Tissue levels and PBPKmodel simulations

in mouse (A) whole blood or serum, (B) liver, (C) gut, (D)kidney, (E) heart, and (F) bone marrow after iv treat-ment with 6 or 10 mg/kg DOX. Filled symbols representthe 6 mg/kg dose and open symbols the 10 mg/kg dose.Solid lines are model simulations of a 6 mg/kg dose, anddashed lines are model simulations of a 10 mg/kg dose.The two filled symbol sets in gut (C) are from measure-ments made in the large (squares) and small (circles)intestine.

1494 GUSTAFSON ET AL.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002

Page 8: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

terminal elimination phase, discrepancies inindividual points can have a large impact on thisparameter. As shown in Figure 2D with the 1-hpoint, and 2E with the 12-h, this seems to be thecase with the calculation of terminal half-life inthese tissues at the respective doses.

DOX PKs and Model Simulations in Dogs

Time course samples were collected from clientdogs being treated with DOX (30 mg/m2) forlymphoma at the Veterinary Teaching Hospital atColorado State University, and plasma levels of

Table 5. DOXPharmacokinetic Parameters inMice Calculated FromWhole Blood (6 mg/kg) and Serum (10mg/kg)Data and PBPK Model Simulationsa

PK Parameters 6 mg/kgb 6 mg/kgc 10 mg/kgb 10 mg/kgc

AUC (nmol�h/L) 4243� 1394 3297� 226 8351� 1621 9116� 566T1/2a (h) 0.036� 0.005 0.118� 0.035 0.067� 0.045 0.22� 0.076T1/2b (h) 10.3� 4.0 10.3� 0.9 7.3� 1.8 8.4� 0.7Vss (L) 0.734� 0.073 0.987� 0.019 0.456� 0.033 0.486� 0.011CL (L/h) 0.052� 0.017 0.067� 0.005 0.044� 0.009 0.040� 0.003

aPK parameters were calculated using a two-compartment model. Values represent the parameter estimates� estimatedstandard error.

bPK parameters are calculated from measured DOX levels in mouse whole blood or serum after a single iv dose.cPK parameters are calculated from PBPK model-predicted DOX levels in mouse whole blood or serum after a single iv dose.

Table 6. Tissue Pharmacokinetic Parameters From Mouse DOX Pharmacokinetic Data and PBPK ModelSimulationsa

Sample AUCtb AUCinf

c t1/2d AUCt Ratio

e AUCinf Ratioe t1/2 Ratio

e

Liver6 mg/kg 74.3 112.7 5.96 mg/kg sim 65.0 95.7 4.7 1.14 1.18 1.2610 mg/kg 166.3 251.9 7.810 mg/kg sim 128 212.5 9.9 1.30 1.19 0.79

Kidney6 mg/kg 133.6 465.5 18.16 mg/kg sim 210.2 353.3 6.2 0.64 1.32 2.9210 mg/kg 387.1 676.3 9.210 mg/kg sim 398.2 739.8 11.5 0.97 0.91 0.80

Gut6 mg/kg 73.3 113.5 5.56 mg/kg sim 83.7 116.6 4.5 0.88 0.97 1.2210 mg/kg 229.2 405.1 9.810 mg/kg sim 169.4 262.7 9.3 1.35 1.54 1.05

Heart6 mg/kg 78.6 174.6 9.86 mg/kg sim 97.5 222.9 9.9 0.81 0.78 0.9910 mg/kg 220.5 255.5 4.610 mg/kg sim 199.5 412.4 13 1.11 0.62 0.35

Bone marrow6 mg/kg 71.8 — —6 mg/kg sim 58.8 — — 1.22 — —

aTissue pharmacokinetic parameters (AUCs and t1/2) were calculated using noncompartmental analysis with extravascular inputof drug.

bAUCt represents the calculated AUC from time 0 to time 8 or 12 h for the 6 mg/kg and 10 mg/kg data sets, respectively.cAUCinf represents the calculated AUC extrapolated to infinity.dt1/2 represents the estimated half-life in hours for DOX elimination from the tissue as estimated from linear regression of the

terminal elimination phase.eRatio of the value generated for that parameter from the measured versus the simulated data sets.

DOXORUBICIN PHARMACOKINETICS 1495

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002

Page 9: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

DOX were determined. The mouse PBPK modelused above was scaled to dog physiologic (Table 1)and metabolic parameters (Table 3), and DNAmeasurements were made in dog tissues for use inthe model (Table 2). The model was also modifiedby adding an iv infusion component that allowsfor the delivery of the dose, at a constant rate,over the course of an infusion time period. Modelsimulations were then performed, and the result-ing model output was compared with mean valuesfrom four animals (Fig. 3). The predictive perfor-mance of the simulation is shown in Table 4, witha MAPE% of 11.2% and a MPE% of 0.7%.

Plasma PK parameters were calculated fromthe individual plasma time-course data using athree-compartment description. The resulting PKparameters were compared with PK parametersthat had been estimated from simulated plasmalevels in which weight and dose were matched tothe client dogs from which the plasma sampleoriginated. Because many of the model para-meters are scaled to body weight, it is importantto take this source of variability into account. Theresulting comparison is shown in Table 7, with the

actual and simulated-data-generated parametersshowing good agreement. One potential source ofdifference in the measured and simulated data isthe fact that the end-of-infusion measurementis generally much greater in the simulated versusmeasured values. This is probably because anydelay in sampling between the actual end-of-infusion and the drawing of the blood samplewill lead to large changes in DOX concentrationbecause of the very steep slope of the initial dis-tribution phase for this drug. For this reason, tobetter compare our model output to the actualmeasured data, we included an evaluation of thesimulated data set using measured end-of-infu-sion values, and the resulting PK parameters arealso included in Table 7.

DOX PKs and Model Simulations in Humans

A human PBPK model was generated usinghuman parameters for organ size and blood flow(Table 1), tissue DNA content (Table 2), andmetabolism (Table 3). Excretory parameters andcardiolipin content values used were the sameas those used in the mouse and dog models, butwere scaled relative to the size of the organ. Theresulting model outputs were compared withhuman DOX PK data sets kindly provided byDr. Leonard Liebes (New York University Schoolof Medicine, New York, NY) from a previouslypublished study.40 Individual model simulationswere performed to match patient weight, totaldose, and length of infusion, and the results areshown in Figure 4. The accuracy of the predic-tions for each patient was determined from theMAPE%, and ranged from 17.2 to 51.8%. Therewas no trend toward over-prediction or under-prediction, because the MPE% ranged from �20.8to 49.7%, with three of the patients having nega-tive values and two of the patients having positivevalues.

PK parameters were estimated from actual andsimulated data for each patient using a three-compartment model, and the results are summar-ized in Table 8. The simulated data showedsignificant differences from the actual data whencomparing AUC, CL (clearance), and Vss (steady-state volume of distribution). However, thesedifferences were attributable to the magnitudeof the end-of-infusion point in the simulated data,as shown in Table 8, when the simulated end-of-infusion value is substituted with an actual value.From our simulations, it can be estimated that thelarge differences that we see in the simulated and

Figure 3. Plasma levels of DOX in dogs after a20-min iv infusion at a dose of 30 mg/m2. Pointsrepresent the mean� standard deviation of four ani-mals treated at this dose. The solid line represents thePBPK model-predicted values using dog physiologicand biochemical parameters. Themodel simulation wasdone using the mean body weight of the four animalswhose time-course data are shown.

1496 GUSTAFSON ET AL.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002

Page 10: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

the measured end-of-infusion values can occur ina span of minutes. Therefore, it seems reasonablethat our simulated values would be much higherthan the measured values because of a time lag insampling, or from sampling at a distant site. Theend-of-infusion corrected values are consistentwith the PK parameters that we calculated fromthe data set, as well as previously publishedliterature values on DOX PKs in which a three-compartment model was applied.41

DISCUSSION

PBPK models have been developed for a numberof chemotherapeutic agents42–45 includingDOX.20,21,46 Most of these models have beenflow-limited models that describe tissue-compart-ment distribution of drugs based on blood/tissuepartitioning and include metabolic and excretoryparameters that are optimized to fit the data sets.An exception to this is the PBPK methotrexatemodels42,47 that incorporated specific binding tomacromolecules and enterohepatic circulation inthe description of methotrexate PKs. The utility ofthese models has been limited to describing tissuelevels in model systems, with the predictive na-ture of these models being largely ignored.

The PBPK model described herein is unique inthat it relies on macromolecule-specific bindingto describe the tissue distribution of DOX, and in-corporates measured physiochemical parameters

Table 7. DOX Pharmacokinetic Parameters in Canine Patients and Canine PBPKModel Simulations

PK Parametersa Actual Datab Simulated Datac Simulated Datad

t1/2 a (h) 0.043� 0.013 0.035� 0.003 0.049� 0.015t1/2 b (h) 0.44� 0.20 0.24� 0.01 0.26� 0.04t1/2 g (h) 19.8� 22.3 25.1� 6.0 28.3� 11.4AUC (nmol�h/L)e 66.7� 66.1 92.6� 11.5 66.1� 10.4CL (L/h) 46.5� 28.5 20.2� 0.7 28.4� 3.0Vss (L) 496� 223 318� 79 803� 453

aPharmacokinetic parameters were derived using a three-compartment model with iv infusioninput and first-order elimination. All parameters are calculated in a model-dependent manner andrepresent the mean�SD from four patient data sets and model simulations.

bTime-course DOX serum data were generated from samples obtained from four client dogsbeing treated for lymphoma at a dose of 30 mg/m2 at the Veterinary Teaching Hospital at ColoradoState University.

cTime-course data for DOX were generated using our PBPK model with patient weight, dose,and infusion time set in the model to match that from the actual data.

dThe simulated end-of-infusion DOX plasma concentration was replaced by the actualdetermined value to compensate for real-time versus delayed sampling in the actual data sets.

eAUC values were normalized for dose (per mg/m2).

Figure 4. Serum levels of DOX in human patientsfollowing (A) 13-min, (B) 6-min, (C) 5-min, and (D)10-min iv infusions. Panel C contains simulations anddata from patients receiving doses of 30 (open symbols,dashed line) and 60 (closed symbol, solid line) mg/m2.All other data and simulations are from patientsreceiving a dose of 60 mg/m2. Points represent thedetermined value for the patient at that time point.Lines represent the PBPK model-predicted value usinghuman physiologic and biochemical parameters. Modelsimulations were done using the weight of the indivi-dual patients and using dosing that delivered the exactamount of drug the patient received over the infusionperiod.

DOXORUBICIN PHARMACOKINETICS 1497

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002

Page 11: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

into this description. The relationship betweenDOX tissue partitioning and DNA content hasbeen previously established.15 The binding ofDOX to cardiolipin has also been well estab-lished,17 and the relative and absolute affinities ofDOX for these macromolecules have been stud-ied.17,30 The use of this information, coupledwith the flow dynamics and tissue compartmentsof a PBPK model, enables prediction of thedistribution of DOX to mouse tissues. The addi-tion of organ-specific metabolism that is based onin vitro metabolic studies,34 and biliary and uri-nary elimination that are mechanistic in nature,produces a model based on fundamental physiol-ogy and physio- and biochemical properties ofDOX.

DOX is one of the most widely used agents inthe treatment of cancer, and studies that deter-mine potential interactions of this agent withother drugs that are given concurrently to cancerpatients are continuously being performed. Amodel of DOX disposition that includes termsthat describe process-specific excretion and meta-bolism is potentially very useful for predictingthe impact of various agents on DOX PKs. Forexample, this model could be used to predict andquantify some of the sequence-dependent PK ef-fects that have been seen when anthracyclineshave been used in combination with taxanes,48,49

or to predict the effects that drug vehicles mayhave on DOX disposition.50

This model may also be coupled to Monte Carlosimulation that will allow for the incorporation

of variability across the model parameters.51

Because of the limited number of patients thatare used in phase I trials of new drug combina-tions, the use of a model that can generate popu-lation-based data would be useful for identifyingsubsets of individuals that may be susceptibleto alterations in DOX PKs that are detrimental.This type of approach could be very useful in pre-dicting the variation in DOX PKs that could beexpected because of induction of P-glycoproteinby concurrent taxane treatment.52 Preliminarystudies using this model coupled to Monte Carlosimulation have shown that when parametervariability is incorporated into the model, themodel generates DOX PK variability that is verysimilar to that seen in the population (data notshown).

In summary, we have developed a model forDOX distribution in animals that incorporatesmacromolecule-specific binding, metabolism, andbiliary and urinary excretion in the context of aflow-based physiologic model. This model is uni-que from previous physiologically based modelsof DOX distribution in that it relies on knowninteractions of DOX with tissue macromoleculesrather than measured tissue to blood concentra-tion differences to determine DOX tissue uptakeand release. The resulting model was able to bescaled from mice to dogs to humans, and predictDOX PKs in each species. This model has utilityin the prediction of potential drug interactionsand predicting PK alterations in special patientpopulations.

Table 8. DOX Pharmacokinetic Parameters in Human Patients and Human PBPK Model Simulations

PK Parametersa Actual Datab Simulated Datac Simulated Datad Literature Datae

t1/2 a (h) 0.034� 0.014 0.035� 0.003 0.056� 0.027 —t1/2 b (h) 0.64� 0.46 0.32� 0.01 0.38� 0.12 —t1/2 g (h) 23.0� 5.6 26.0� 4.3 27.0� 6.4 28.5� 10.7AUC (nmol�h/L)f 63.4� 13.1 94.7� 3.7 66.4� 7.0 79.4� 29.8CL (L/h) 43.5� 6.9 28.4� 1.0 41.0� 5.5 44.5� 18.1Vss (L) 1091� 307 533� 85 1230� 601 1758� 850

aPharmacokinetic parameters were derived using a three-compartment model with iv infusion input and first-order elimination.All parameters were calculated in a model-dependent manner and represent the mean�SD from five patient data sets and modelsimulations.

bHuman time-course data for DOX were kindly provided by Dr. Leonard Liebes (New York University School of Medicine,New York, NY).

cTime-course data for DOXwere generated using our PBPKmodel with patient weight, dose, and infusion time set in themodel tomatch that from the actual human data.

dThe simulated end-of-infusion DOX plasma concentration was replaced by the actual determined value to compensate for real-time versus delayed sampling in the actual data sets.

eDOX pharmacokinetic parameters as determined after the first course of therapy in cancer patients from a previously publishedstudy.41

fAUC values were normalized for dose (per mg/m2).

1498 GUSTAFSON ET AL.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002

Page 12: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

NOMENCLATURE

Q Blood flow (L/h)A Amount of drug (moles)V Tissue volume (L)CA Arterial blood concentration of free DOX

(M)CV Venous blood concentration of total DOX

leaving tissues (M)CBL Arterial blood concentration of total DOX

(M)FB Fraction of DOX bound to plasma proteinsTDNA Tissue-specific DNA binding capacity for

DOX (M)TCAL Tissue-specific cardiolipin binding capa-

city for DOX (M)KDNA Binding affinity of DOX for DNA (M)KCAL Binding affinity of DOX for cardiolipin (M)AM Amount metabolized (moles)AE Amount excreted (moles)VMAX Maximum rate of activity (moles/hr/L

tissue)KM Michaelis’ constant (M)KMET First ordermetabolic rate constant (h�1 kg

tissue�1)FFILT Fraction renal blood flow filtered at the

glomerulus

Subscripts

T Generic tissue compartmentC Total cardiac outputL LiverK KidneyG GutB BloodAKR Aldo-keto reductaseAG AglyconePGP P-glycoprotein

ACKNOWLEDGMENTS

We are grateful for the help of Dr. Kim Selting,Dr. Greg Ogilvie, Dr. Wendy Pott, andMrs. LauraChubb (College of Veterinary Medicine and Bio-medical Sciences, Colorado State University, FortCollins, CO) for their help in performing thesestudies. We also thank Drs. Melvin E. Andersenand Russell S. Thomas for helpful discussions.

APPENDIX

The mathematical equations used to simulate themodel structure shown in Figure 1 are shown

below. The complete model code as written inACSL is available by contacting the correspond-ing author.

Tissue compartment mass balance

dAT

dt¼ QTðCA � CVT

Þ

CVT¼ AT

VT þ TDNA � VT

KDNA þ CVT

þ TCAL � VT

KCAL þ CVT

Blood compartment mass balance

dAB

dt¼

XQT � CVT

� ðQC �QLÞ � CA �QL � CBL

CA ¼ AB

VB� ð1� FBÞ

CBL ¼ AB

VB

DOX metabolism by aldo-keto reductases

dAMAKR

dt¼ VMAX�AKT � CVT

KM�AKT þ CVT

DOX metabolism to aglycone

dAMAG

dt¼ KMET�AG � CVT

� VT

Amount of DOX eliminated in urine

dAEU

dt¼ FFILT �QK � CA þ VMAX�PGP�K � CBL

KM�PGP�K þ CBL

Amount of DOX eliminated in feces

dAEF

dt¼ VMAX�PGP�L � CVL

KM�PGP�L þ CVLþ VMAX�PGP�G � CVG

KM�PGP�G þ CVG

REFERENCES

1. Robert J. 1998. Anthracyclines. In: Grochow LB,Ames MM, editors. A clinician’s guide to chemo-therapy pharmacokinetics and pharmacodynamics.Baltimore: Williams & Wilkins, pp 93–173.

2. Young RC, Ozols RF, Myers CE. 1981. The anthra-cycline neoplastic drugs. N Engl J Med 305:139–153.

3. Zunino F, Gambetta R, Di Marco A. 1975. The inhi-bition in vitro of DNA polymerase and RNA poly-merase by daunomycin and adriamycin. BiochemPharmacol 24:309–311.

4. Tewey KM, Chen GI, Nelson EM, Liu LF. 1984.Intercalative anti-tumor drugs interfere with thebreakage-reunion reaction of mammalian DNAtopoisomerase. J Biol Chem 259:9182–9187.

DOXORUBICIN PHARMACOKINETICS 1499

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002

Page 13: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

5. Taatjes DJ, Gaudiano G, Resing K, Koch TH. 1996.Alkylation of DNA by the anthracycline, antitumordrugs adriamycin and daunomycin. J Med Chem39:4135–4138.

6. Doroshow JH. 1985. Role of hydrogen peroxide andhydroxyl radical in the killing of Ehrlich tumorcells by anticancer quinones. Proc Natl Acad SciUSA 83:4514–4518.

7. Bachur NR, Gordon SL, Gee MV. 1978. A generalmechanism for microsomal activation of quinoneanticancer agents to free radicals. Cancer Res 38:1745–1750.

8. Pessah IN, Durie EL, Schiedt MJ, Zimanyi I. 1990.Anthraquinone-sensitized Caþ release channelfrom rat cardiac sarcoplasmic reticulum: Possiblereceptor-mediated mechanism of doxorubicin car-diomyopathy. Mol Pharmacol 37:503–514.

9. Oakes SG, Schlager JJ, Santone KS, Abraham RT,Powis G. 1990. Doxorubicin blocks the increase inintracellular Caþþ, part of a second messengersystem in N1E-115 murine neuroblastoma cells. JPharmacol Exp Ther 252:979–983.

10. Bielack SS, Erttmann R, Kempf-Bielack B,WinklerK. 1996. Impact of scheduling on toxicity andclinical efficacy of doxorubicin: what do we knowin the mid-nineties? Eur J Cancer 32A:1652–1660.

11. Mau B-L, Powis G. 1992. Inhibition of cellularthioredoxin reductase by diaziquone and doxorubi-cin: Relationship to the inhibition of cell prolifera-tion and decreased ribonucleotide reductaseactivity. Biochem Pharmacol 43:1621–1626.

12. Juranka PF, Zastawny RL, Ling V. 1989. P-glyco-protein: Multidrug-resistance and a superfamily ofmembrane-associated transport proteins. FASEB J3:2583–2592.

13. Ahmed NK, Felsted RL, Bachur NR. 1981. Daunor-ubicin reduction mediated by aldehyde and ketonereductases. Xenobiotica 11:131–136.

14. Pan S-S, Bachur NR. 1980. Xanthine oxidase cata-lyzed reductive cleavage of anthracycline antibio-tics and free radical formation. Mol Pharmacol 17:95–99.

15. Terasaki T, Iga T, Sugiyama Y, Hanano M. 1982.Experimental evidence of characteristic tissuedistribution of adriamycin: Tissue DNA concentra-tion as a determinant. J Pharm Pharmacol 34:597–600.

16. Nicolay K, Timmers RJM, Spoelstra E, van derNeut R, Fok JJ, Huigen YM, Verkleij AJ, de KruijffB. 1984. The interaction of adriamycin with cardio-lipin in model and rat liver mitochondrial mem-branes. Biochim Biophys Acta 778:359–371.

17. Goormaghtigh E, Chatelain P, Caspers J,Ruysschaert JM. 1980. Evidence of a specific com-plex between adriamycin and negatively chargedphospholipids. Biochim Biophys Acta 597:1–14.

18. Bischoff KB. 1975. Some fundamental considera-tions of the applications of pharmacokinetics to

cancer chemotherapy. Cancer Chemother Rep 59:777–793.

19. Yang RSH, Andersen ME. 1994. Pharmacokinetics.In: Hodgson E, Levi P, editors. Introduction to bio-chemical toxicology. Norwalk: Appleton and Lange,pp 49–73.

20. Chan KK, Cohen JL, Gross JF, Himmelstein KJ,Bateman JR, Tsu-Lee Y, Marlis AS. 1978. Predic-tion of adriamycin disposition in cancer patientsusing a physiologic pharmacokinetic model. CancerTreat Rep 62:1161–1171.

21. Harris PA, Gross JP. 1975. Preliminary pharma-cokinetic model for adriamycin (NSC-123127).Cancer Chemother Rep 59:819–825.

22. De Jong J, Guerand WS, Schoofs PR, Bast A, vander Vijgh WJF. 1991. Simple and sensitive quanti-fication of anthracyclines in mouse atrial tissueusing high-performance liquid chromatographyand fluorescence detection. J Chromatogr 570:209–216.

23. Downs TR, Wilfinger WW. 1983. Fluorometricquantification of DNA in cells and tissue. AnalBiochem 131:538–547.

24. Brown RP, Delp MD, Lindstedt SL, Rhomberg LR,Beliles RP. 1997. Physiological parameter valuesfor physiologically based pharmacokinetic models.Toxicol Ind Health 13:407–484.

25. Iversen PO, Thing-Mortensen B, Nicolaysen G,Benestad HB. 1993. Decreased blood flow to ratbone marrow, bone, spleen, and liver in acuteleukemia. Leuk Res 17:663–668.

26. Lahtinen R, Lahtinen T, Romppanen T. 1982. Boneand bone-marrow blood flow in chronic granulocyticleukemia and primary myelofibrosis. J Nucl Med23:218–224.

27. Courtade S, Marinetti GV, Stotz E. 1967. Thestructure and abundance of rat tissue cardiolipins.Biochim Biophys Acta 137:121–134.

28. Pearce PH, Kakulas BA. 1980. Skeletal musclelipids in normal and dystrophic mice. Aus J ExpBiol Med Sci 58:397–408.

29. Hostetler KY. 1982. Polyglycerophospholipids:Phosphatidylglycerol, diphosphatidylglycerol andbis(monoacylglycero)phosphate. In: HawthorneJN, Ansell GB, editors. Phospholipids. New York:Elsevier Biomedical Press, pp 215–261.

30. Mustonen P, Kinnunen PKJ. 1993. On the reversalby deoxyribonucleic acid of the binding of adria-mycin to cardiolipin-containing liposomes. J BiolChem 268:1074–1080.

31. Myers CE, Chabner BA. 1990. Anthracyclines. In:Chabner BA, Collins JM, editors. Cancer chemo-therapy: Principles and practice. Philadelphia: JBLippincott, pp 356–381.

32. Meijer DKF, van der Sluijs P. 1989. Covalent andnoncovalent protein binding of drugs: Implicationsfor hepatic clearance, storage, and cell specific drugdelivery. Pharm Res 6:105–118.

1500 GUSTAFSON ET AL.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002

Page 14: Doxorubicin pharmacokinetics: Macromolecule binding, metabolism, and excretion in the context of a physiologic model

33. Leung HW, Paustenbach DJ, Murray FJ, AndersenME. 1990. A physiological pharmacokinetic de-scription of the tissue distribution and enzyme-inducing properties of 2,3,7,8-tetrachlorodibenzo-p-dioxin in the rat. Toxicol Appl Pharmacol 103:399–410.

34. Loveless H, Arena E, Felsted RL, Bachur NR. 1978.Comparative mammalian metabolism of adriamy-cin and daunorubicin. Cancer Res 38:593–598.

35. Wilkinson PM, Israel M, Pegg WJ, Frei E III. 1979.Comparative metabolism and excretion of adria-mycin in man, monkey and rat. Cancer ChemotherPharmacol 2:121–125.

36. Krarup-Hansen A,WassermannK, Rasmussen SN,DalmarkM. 1988. Pharmacokinetics of doxorubicinin man with induced acid or alkaline urine. ActaOncol 27:25–30.

37. Sheiner LB, Beal SL. 1981. Some suggestions formeasuring predictive performance. J Pharmacoki-net Biopharm 9:503–512.

38. Gustafsson LL, Ebling WF, Osaki E, Harapat S,Stanski DR, Shafer SL. 1992. Plasma concentrationclamping in the rat using a computer-controlledinfusion pump. Pharm Res 9:800–807.

39. Colombo T, Zucchetti M, D’Incalci M. 1994. Cyclo-sporin A markedly changes the distribution ofdoxorubicin in mice and rats. J Pharmacol ExpTher 269:22–27.

40. Sparano JA, Speyer J, Gradishar WJ, Liebes L,Sridhara R, Mendoza S, Fry D, Egorin MJ. 1999.Phase I trial of escalating doses of paclitaxel plusdoxorubicin and dexrazoxane in patients withadvanced breast cancer. J Clin Oncol 17:880–886.

41. Jacquet J-M, Bressolle F, Galtier M, Bourrier M,Donadio D, Jourdan J, Rossi J-F. 1990. Doxorubicinand doxorubicinol: Intra- and inter-individualvariations of pharmacokinetic parameters. CancerChemother Pharmacol 27:219–225.

42. Bischoff KB, Dedrick RL, Zaharko DS, LongstrethJA. 1971.Methotrexate pharmacokinetics. J PharmSci 60:1128–1133.

43. Dedrick RL, Forrester DD, Cannon JN, El DareerSM, Mellett LB. 1973. Pharmacokinetics of 1-b-D-

arabinofuranosylcytosine (Ara-C) deamination inseveral species. Biochem Pharmacol 22:2405–2417.

44. Farris FF, Dedrick RL, King FG. 1988. Cisplatinpharmacokinetics: Applications of a physiologicalmodel. Toxicol Lett 43:117–137.

45. Lutz RL, Galbraith WM, Dedrick RL, Schrager R,Mellett LB. 1996. A model for the kinetics ofdistribution of actinomycin D in the beagle dog.J Pharmacol Exp Ther 200:469–478.

46. Gallo JM, Hung CT, Gupta PK, Perrier DG. 1989.Physiological pharmacokinetic model of adriamy-cin delivered via magnetic albuminmicrospheres inthe rat. J Pharmacokinet Biopharm 17:305–326.

47. Bischoff KB, Dedrick RL, Zaharko DS. 1970. Preli-minary model for methotrexate pharmacokinetics.J Pharm Sci 59:149–154.

48. Venturini M, Lunardi G, Del Mastro L, VannozziMO, Tolino G, Numico G, Viale M, Pastrone I,Angiolini C, Bertelli G, Straneo M, Rosso R,Esposito M. 2000. Sequence effect of epirubicinand paclitaxel treatment on pharmacokinetics andtoxicity. J Clin Oncol 18:2116–2125.

49. Esposito M, Venturini M, Vannozzi MO, Tolino G,Lunardi G, Garrone O, Angiolini C, Viale M,Bergaglio M, Del Mastro L, Rosso R. 1999. Com-parative effects of paclitaxel and docetaxel on themetabolism and pharmacokinetics of epirubicinin breast cancer patients. J Clin Oncol 17:1132–1140.

50. Webster LK, Cosson EJ, Stokes KH, Millward MJ.2001. Effect of the paclitaxel vehicle, CremophorEL, on the pharmacokinetics of doxorubicin anddoxorubicinol in mice. Br J Cancer 73:522–524.

51. Thomas RS, Lytle WE, Keefe TJ, Constan AA, YangRSH. 1996. Incorporating Monte Carlo simulationinto physiologically based pharmacokinetic modelsusing advanced continuous simulation language(ACSL): A computational method. Fundam ApplToxicol 31:19–28.

52. Synold TW, Dussault I, Forman BM. 2001. Theorphan nuclear receptor SXR coordinately regu-lates drug metabolism and efflux. Nat Med 7:584–590.

DOXORUBICIN PHARMACOKINETICS 1501

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 6, JUNE 2002