EPA/600/R-06/014A
March 2006
Use of Physiologically Based Pharmacokinetic
Models to Quantify the Impact of Human Age
and Interindividual Differences in Physiology
and Biochemistry Pertinent to Risk
Final Report
Cooperative Agreement
CR828047010
John C. Lipscomb, Ph.D., DABT
U.S. Environmental Protection Agency
Office of Research and Development
National Center for Environmental Assessment
26 W. Martin Luther King Drive, MS-190
Cincinnati, OH 45268
Tel 5137 569-7217
lipscomb.john@epa.gov
Gregory L. Kedderis, Ph.D.
Independent Consultant
1803 Jones Ferry Road
Chapel Hill, NC 27516
Tel 9197 942-3 921
GKedderis@msn.com
-------
NOTICE
The U.S. Environmental Protection Agency through its Office of Research and
Development funded and managed the research described here under cooperative agreement
CR828047010 to Gregory L. Kedderis, Ph.D. It has been subjected to the Agency's peer and
administrative review and has been approved for publication as an EPA document. Mention of
trade names or commercial products does not constitute endorsement or recommendation for use.
11
-------
TABLE OF CONTENTS
Page
TABLE OF CONTENTS iii
LIST OF TABLES vii
LIST OF FIGURES x
EXECUTIVE SUMMARY xii
1. INCORPORATING HUMAN INTERINDIVIDUAL BIOTRANSFORMATION
VARIANCE IN HEALTH RISK ASSESSMENT 1-1
ABSTRACT 1-1
INTRODUCTION 1-2
DEFINITION OF THE PROBLEM 1-7
OBJECTIVE 1-8
APPROACH 1-9
REQUIRED CONDITIONS 1-12
APPLICABILITY 1-12
CONCLUSIONS 1-13
REFERENCES 1-14
2. HOW DIFFERENCES IN ENZYME EXPRESSION CAN TRANSLATE
INTO PHARMACOKINETIC VARIANCE AND SUSCEPTIBILITY TO
TOXICITY 2-1
ABSTRACT 2-1
BACKGROUND 2-2
TOXICITY AND ASSESSMENT 2-2
METABOLISM AND PHARMACOKINETICS IN RISK ASSESSMENT 2-3
USING ENZYME KINETIC DATA TO ESTIMATE
PHARMACOKINETIC VARIANCE 2-6
IMP ACT OF STUDY DESIGN 2-10
SUMMARY AND CONCLUSIONS 2-11
REFERENCES 2-11
3. APPLICATION OF IN VITRO BIOTRANSFORMATION DATA AND
PHARMACOKINETIC MODELING TO RISK ASSESSMENT 3-1
ABSTRACT 3-1
TOXICOLOGY AND HUMAN HEALTH RISK ASSESSMENT 3-2
PHYSIOLOGICALLY BASED PHARMACOKINETIC MODELS 3-4
EXTRAPOLATION OF IN VITRO DATA TO HUMANS 3-5
INCORPORATION OF PHARMACOKINETICS INTO RISK
ASSESSMENTS 3-9
REFERENCES 3-13
in
-------
TABLE OF CONTENTS cont.
Page
4. THE IMPACT OF CYTOCHROME P450 2E1-DEPENDENT METABOLIC
VARIANCE ON A RISK-RELEVANT PHARMACOKINETIC OUTCOME
IN HUMANS 4-1
ABSTRACT 4-1
INTRODUCTION 4-2
METHODS 4-7
Human Samples and Quantification of CYP Proteins 4-8
Distribution of CYP2E1 to Human Hepatic MSP 4-8
Estimation of Proteins in Intact Liver 4-9
CYP2E1-Dependent Oxidation of TCE 4-10
Statistical Analysis 4-10
Combination of Data Sets 4-14
PBPK Model 4-15
RESULTS 4-16
Distribution of CYP2E1 to Human Hepatic MSP 4-16
Distribution of CYP2E1 to Intact Human Liver 4-16
In vitro Metabolic Rate Constant (Vmax) 4-18
Determining the Metabolic Capacity of Intact Tissue and
Extrapolation of Units 4-19
PBPK Model Predictions 4-20
DISCUSSION 4-20
SUMMARY AND CONCLUSIONS 4-28
REFERENCES 4-30
APPENDIX: MATLAB Code to Generate AxD 4-45
5. PHARMACOKINETIC ANALYSIS TO SUPPORT AN INHALATION RfC
FOR CHLOROFORM 5-1
BACKGROUND 5-1
Purpose 5-2
Objective 5-2
RISK ASSESSMENT APPLICATION 5-2
SCOPE AND LIMITATIONS 5-4
APPROACH 5-5
APPLICATION OF PHYSIOLOGICALLY BASED
PHARMACOKINETIC MODELING 5-11
IV
-------
TABLE OF CONTENTS cont.
Page
METHODS: PBPK MODEL STRUCTURE AND PARAMETERS 5-14
Model Structure 5-14
Partition Coefficient Derivation 5-14
Mice 5-15
Adult Rats 5-15
Adult Humans 5-17
Children 5-18
Parameter Values 5-18
METABOLIC VARIABILITY 5-24
Simulated Exposure Conditions 5-27
RESULTS 5-34
Derivation of the Rodent NO AEL Values 5-34
Model Response 5-41
Chloroform Model Verification 5-42
Comparison to Mouse and Rat Pharmacokinetic Data 5-45
Model Sensitivity 5-49
Comparisons to Human Pharmacokinetic Data 5-55
Choice of Values for Km 5-58
Human Equivalent Concentration 5-61
Human Variability 5-62
Deriving CM24 at Which to Determine Human Variability 5-62
Human Variability: Metabolic Capacity and Age 5-62
Human Variability: Hepatic Blood Flow 5-62
Human Variability: Blood:Air PC Value 5-65
Combined Variability in the Adult 5-65
Obesity in Adult Males 5-68
Summary of Results 5-68
DISCUSSION 5-69
CONCLUSIONS 5-80
REFERENCES 5-82
APPENDIX A: COMPUTER CODE FOR THE CHLOROFORM
PBPK MODEL 5-86
APPENDIX B: DERIVATION OF CHLOROFORM PARTITION
COEFFICIENTS 5-89
-------
TABLE OF CONTENTS cont.
Page
APPENDIX C: VARIABILITY OF HEPATIC BLOOD FLOW IN
ADULT HUMANS 5-102
APPENDIX D: EXTRAPOLATION OF IN VITRO DERIVED
CHLOROFORM METABOLIC RATE CONSTANTS AND
VARIABILITY FOR PBPK MODELING 5-113
VI
-------
LIST OF TABLES
No. Title Page
2-1 Effect of a 10-Fold Induction of Vmax on Hepatic Clearance over
a Four-Log Increase in Substrate Delivery 2-14
3-1 Variance in Human Hepatic CYP2E1-Dependent TCE Oxidative
Capacity and the Amount of TCE Oxidized 3-16
4-1 Identification of Data Sets and Parameters for Statistical Evaluation 4-3 5
4-2 Liver Enzyme Data 4-36
4-3 CYP2E1 Content and TCE Metabolic Activity Used to Produce
Data Set3, Describing Parameter D 4-38
4-4 Distributions of TCE Metabolism Rate Constant, Microsomal Protein
andCYP2El Content of Adult Human Liver 4-39
4-5 Effect of Human Hepatic CYP2E1 Activity Distribution on the
Bioactivation of TCE Following an Inhalation and Oral Exposure 4-39
5-1 Adult Rat Blood: Air and Tissue: Air Partition Coefficient Values 5-16
5-2 Adult Rat Tissue:Blood Partition Coefficient Values Derived from
Paired Tissues 5-16
5-3 Adult Human Blood: Air Partition Coefficient Values 5-17
5-4 Tissue:Blood Partition Coefficient Values for Adult Humans 5-17
5-5 Tissue: Air PC Values in PND 10 Rat Pups 5-19
5-6 Tissue:Blood PC Values Used for Children PBPK Model 5-19
5-7 Age and Species Dependent Model Parameter Values 5-20
5-8 Parameters and Selected Percentile Values for the Fraction of
Cardiac Output as Hepatic Blood Flow 5-21
5-9 Distribution of Chloroform Metabolic Rate Constants in Adults and
Children 5-26
5-10 Derivation of VmaxC Values for Inclusion in PBPK Modeling for Adults,
Children and Rats 5-28
vn
-------
LIST OF TABLES cont.
No. Title Page
5-11 Exposure Conditions, Relative Liver Weight and Hepatocyte Labeling
Index from Constan et al. (2002) 5-35
5-12 Exposure Conditions and Endpoints Examined in Rodent Inhalation
Bioassays 5-37
5-13 Endpoints and Response Levels Identified in Rodent Inhalation Bioassays 5-38
5-14 Chloroform Metabolized in Liver in Mice, Rats and General Adult
(Male) Humans 5-43
5-15 Results from Open Chamber Metabolism Studies in Rats and Mice 5-48
5-16 Results from Studies with Male Humans 5-59
5-17 Results from Studies with Female Humans 5-59
5-18 Derivation of the Human Equivalent Concentration from Liver
Effects Observed in Mice and Rats 5-61
5-19 The Impact of CYP2E1 -Dependent Metabolic Parameters on Chloroform
Metabolism Among Selected Segments of the Human Population 5-64
5-20 Effect of Variance in Hepatic Blood Flow on Chloroform
Metabolism Among Adult Humans 5-66
B-l Rat and Human Blood: Air and Tissue:Blood Partition Coefficient
Values as Reported in Corley etal. (1990) 5-90
B-2 Fractional Fat and Water Content of Adult Human Tissues 5-92
B-3 Blood: Air and Tissue: Air Partition Coefficient Values Derived
from Studies with Rat Tissues 5-93
B-4 Tissue:Blood Partition Coefficient Values Derived from
Studies with Rat Tissues 5-94
B-5 Blood: Air Partition Coefficient Values Derived from Studies
with Adult Human Blood 5-94
B-6 Tissue:Blood Partition Coefficient Values Derived by Combining Mean
Human B:A PC Values with Individual Rat T:A PC Values 5-95
Vlll
-------
LIST OF TABLES cont.
No. Title Page
B-7 Predictions of Human Blood: Air and Tissue:Blood
Partition Coefficient Values 5-95
B-8 Human Tissue: Air and Tissue:Blood Partition Coefficient
Values Based on Adjusted Predictions of Tissue: Air PC Values 5-96
B-9 Comparison of B:A and T:B PC Values for Mature Individuals 5-96
B-10 Fractional Fat and Water Content of Children's Tissues 5-98
B-ll Predicted T: A PC Values in Humans 5-98
B-12 Child-Specific Blood:Air and Tissue:Blood PC Values 5-98
B-13 Adult-Specific Blood:Air and Tissue:Blood PC Values 5-99
B-14 Blood:Air and Tissue:Blood PC Values in PND 10 Rat Pups 5-100
C-l Hepatic Blood Flow and Fraction of CO as HBF, Estimated
from Data of Ceasaretal. (1961) 5-105
C-2 Hepatic Blood Flow and Fraction of CO as HBF Estimated
from Data of Wiegandetal. (1960) 5-106
C-3 Hepatic Blood Flow and Fraction of CO as HBF, as Reported
by Feruglio et al. (1964), "Before" Condition 5-107
C-4 Hepatic Blood Flow and Fraction of CO as HBF, Derived from HBF and
Patient Characteristics Presented by Reemtsma et al. (1960) 5-108
C-5 Hepatic Blood Flow and CO Data which Served as the Basis
for Results Presented in lijima et al. (2001) 5-109
C-6 Hepatic Blood Flow and CO Data which Served as the Basis
for Results Presented in Sakkaetal. (2001) 5-110
C-7 Parameters and Selected Percentile Values for the Fraction of
Cardiac Output as Hepatic Blood Flow 5-111
D-l Information Used to Extrapolate in vitro Derived Vmax Value
to the Intact Liver 5-115
D-2 Donor Demographic Information for 10 Child Organ Donors 5-116
D-3 Determination of Cytochrome P450 Enzyme in Liver Tissue of
10 Child Organ Donors 5-117
D-4 Distribution of CYP2E1 to Adults and Children 5-118
IX
-------
LIST OF FIGURES
No. Title
1-1 Isolation of Microsomal Protein from the Intact Liver 1-16
1-2 Relationship Between Intact Liver, Microsomal Protein and
Some CYP Forms 1-17
1-3 Extrapolation and Incorporation of in vitro Derived Metabolic
Rates in PBPK Modeling 1-18
2-1 Pharmacokinetics and the Risk Assessment Paradigm 2-15
2-2 The Relationship Between Km Value and Substrate Concentration 2-16
2-3 The Relationship Between Substrate Delivery, Metabolic Capacity
and the Amount of Metabolite Formed 2-17
2-4 The Impact of Substrate Delivery and Metabolic Capacity on the
Formation of Metabolites in Liver 2-18
3-1 The in vitro/in vivo Parallelogram Approach 3-17
3-2 Use of a PBPK Model to Interpret the Risk Significance of Differences
in Metabolic Capacity 3-18
4-1 Application of PBPK Modeling to Link External Dose with
Concentration of Toxicant in Target Organs 4-40
4-2 Relationship Between Liver, Microsomal Protein and Some CYP Forms 4-41
4-3 Correlation Between mg MSP/gram and pmol CYP2El/mg MSP 4-42
4-4 Classification Tree Model for the Distribuiotn of A = pmol
CYP2El/gram Liver 4-43
4-5 Extrapolation and Incorporation of in vitro Derived Metabolic
Rates in PBPK Modeling 4-44
5-1 A Conceptual Presentation of the Application of TK Information and
the TK Approach to Human Health Risk Assessment 5-3
5-2 Subdivision of Uncertainty Factors into Toxicokinetic and
Toxicodynamic Components 5-8
-------
LIST OF FIGURES cont.
No. Title
5-3 Application of PBPK Modeling to Extrapolate Internal Dosimetry
Between Species 5-9
5-4 Five Compartment PBPK Model 5-10
5-5 Achieving Steady State in Liver Tissue 5-29
5-6 Relationship Between Exposure Concentration and CM24 in Rats,
Mice and Humans at Steady State 5-31
5-7 Derivation of the Human Equivalent Concentration 5-32
5-8 Relationship Between the Blood:air Partition Coefficient for CF and CM24 .... 5-51
5-9 Relationship Between Hepatic Blood Flow and CM24 for CF 5-52
5-10a Relationship Between the Maximal Rate of CF Metabolism and CM24 5-53
5-1 Ob Expanded Scale Showing the Relationship Between the Maximal
Rate of CF Metabolism and CM24 5-54
5-1 la Relationship Between the Michaelis Constant for CF Metabolism
andCM24 5-56
5-1 Ib Expanded Scale of the Relationship Between the Michaelis Constant
for CF Metabolism and CM24 5-57
5-12 Application of Toxicokinetic Analysis to Address Animal-to-Human and
Human Interindividual Differences for Chloroform Metabolism and Risk
Assessment 5-67
XI
-------
EXECUTIVE SUMMARY
Outbred species, like humans, are more diverse than the animal species routinely used for
testing and research. This diversity often complicates the extrapolation of animal findings to and
among humans; lack of understanding of the impact of that diversity on tissue dosimetry and
response confounds the level of certainty we place on default measures of human variability used
in risk assessment. The uncertainty factors governing animal to human extrapolation (UFA) and
human interindividual extrapolation (UFH) have recently begun to be addressed by information
which can inform their subsequent division into their respective toxicokinetics (TK) and
toxicodynamics (TD) components. This is explicit in the RfC methodology, where TK
information is used to develop the human equivalent concentration, effectively reducing the TK
component of UFA to 1. Specific and advanced risk assessments have used pharmacokinetic
models and information, often physiologically based pharmacokinetic (PBPK) models and
predictions, to inform the ascription of values for UFA and UFH.
This report is intended to communicate a framework developed for the extrapolation and
integration of in v/Yro-derived measures of chemical metabolism, including those that define
human interindividual variability. It presents instruction on the most useful measures of
chemical metabolism constants and guides the proper interpretation of differences in enzyme
content, so that risk assessors who summarize these types of data for inclusion in human health
risk assessment make optimal use of the available information. Likewise, the demonstration of
the approach and data requirements will guide the construction of laboratory research protocols
which result in the development of data optimally relevant to informing the true nature of human
variability. Finally, this approach is communicated through instruction and demonstration of the
xn
-------
approach and framework; it is useful to those who develop and evaluate PBPK models intended
to address human variability for risk assessment application.
Chapters 1-3 represent generally-applicable works already published, Chapter 4
represents work specific to cytochrome P450 2E1 (CYP2E1) using trichloroethylene as an
example and accepted for publication, and Chapter 5 represents the original presentation of
results describing variance in the CYP2E1-mediated metabolism of chloroform.
Most of the past measures of human variability conducted to evaluate susceptibility have
not been conducted under the proper guidance; faulty interpretations based on unstated and
untenable assumptions fuel inaccurate evaluations of variability, susceptibility and risk. While
banks of liver preparations have been available for more than a decade, they have been exploited
to ascertain measures of human interindividual variability, expressed per unit of isolated
subcellular fraction without regard to measuring variability at the level of the intact tissue. This
has resulted in inaccuracies in estimations of variability, and their well-intentioned inclusion in
evaluations of susceptibility. With regard to chemical-metabolizing enzymes, there are two
fundamental errors in this approach: (1) the variability of the enzyme is not expressed at the level
of the intact tissue, disregarding the impact of the isolation procedure, and (2) enzyme variability
is assumed to represent variability in chemical metabolism. This second error disregards the
dependency of metabolism on substrate (chemical) availability; it disregards the anatomic,
physiologic and biochemical constraints which govern transport of the chemical from air or
ingested material into the bloodstream, and its ultimate diffusion from blood into liver tissue.
This report contains tutorial materials, data quality objectives and guidance (Chapters
1-3) as well as case study investigations (Chapters 4 and 5) which address the development,
evaluation and integration of laboratory-derived data aimed at quantifying human interindividual
variability for application in human health risk assessment. The integrating framework is that of
xiii
-------
physiologically based pharmacokinetic modeling, a procedure popularized two decades ago.
This framework allows the application of appropriate constraints of the intact body on chemical
absorption, distribution, metabolism and elimination.
The case studies developed for two important water contaminant chemicals,
trichloroethylene and chloroform. These case studies comprise data which are chemical-specific,
enzyme-specific, and general in nature. Chemical-specific data include measured or predicted
measures of chemical partitioning into blood and solid tissues, and metabolic rate constants. In
each case study, cytochrome P450 2E1 (CYP2E1) is the enzyme responsible for metabolism; in
each case study metabolism is a fundamental requirement for toxicity - the metabolite is toxic.
Enzyme-specific data include those which characterize the distribution of CYP2E1 to the intact
liver, including age-dependent differences in humans. These data were developed from the
analysis of intact liver tissue samples obtained from human organ donors. Measures of enzyme
content of the liver, and age-dependent differences will be applicable to the many environmental
toxicants metabolized by CYP2E1. This investigation does not include an evaluation of genetic
polymorphisms, as no polymorphisms in the CYP2E1 gene have been linked to changes in the
expression or activity of this enzyme. Data generally applicable in PBPK modeling in humans,
regardless of chemical, are those which characterize the distribution of hepatic blood flow. This
analysis of hepatic blood flow variability is noteworthy in that hepatic blood flow often limits the
metabolism of solvents. This analysis is the first of its kind, and is based on the collection of
peer-reviewed data from the open literature, as well as individual data sets collected from
respective authors. While the case studies have addressed chemicals for which metabolism
represents a bioactivation step, the data on enzyme variability are equally applicable when
metabolism of the contaminant represents a detoxication step.
xiv
-------
These results demonstrate the successful application of this approach to guide the
collection, interpretation compilation and evaluation of risk-relevant pharmacokinetic data. Data
on human biochemical and physiologic variability have been incorporated into PBPK models for
adults and children which were designed to assess human interindividual differences in the
production of a PK outcome linked with risk. While both examples have involved chemicals
whose metabolism is limited by hepatic blood flow; both examples have produced data which
allow us to dispel certain events or co-exposures as modifiers of risk, from the standpoint that
they change the level of CYP2E1 expression. From these results, it is clear that interindividual
differences in the content of the CYP2E1 enzyme can now be quantified, and the data are useful
in investigations of additional environmental contaminants that are also metabolized by
CYP2E1, including benzene, bromobenzene, styrene, the trihalomethanes drinking water
disinfection byproducts and several low molecular weight halogenated solvents.
xv
-------
1. INCORPORATING HUMAN INTERINDIVIDUAL BIOTRANSFORMATION
VARIANCE IN HEALTH RISK ASSESSMENT
ABSTRACT
The protection of sensitive individuals within a population dictates that measures other
than central tendencies be employed to estimate risk. The refinement of human health risk
assessments for chemicals metabolized by the liver to reflect data on human variability can be
accomplished through 1) the characterization of enzyme expression in large banks of human
liver samples, 2) the employment of appropriate techniques for the quantification and
extrapolation of metabolic rates derived in vitro, and 3) the judicious application of
physiologically based pharmacokinetic (PBPK) modeling. While in vitro measurements of
specific biochemical reactions from multiple human samples can yield qualitatively valuable
data on human variance, such measures must be put into the perspective of the intact human to
yield the most valuable predictions of metabolic differences among humans. For quantitative
metabolism data to be the most valuable in risk assessment, they must be tied to human anatomy
and physiology, and the impact of their variance evaluated under real exposure scenarios. For
chemicals metabolized in the liver, the concentration of parent chemical in the liver represents
the substrate concentration in the Michaelis-Menten description of metabolism. Metabolic
constants derived in vitro may be extrapolated to the intact liver, when appropriate conditions are
met. Metabolic capacity (Vmax; the maximal rate of the reaction) can be scaled directly to the
concentration of enzyme (or enzyme fraction) contained in the liver. Several environmental,
genetic and lifestyle factors can influence the concentration of cytochrome P450 forms (CYP) in
the liver by affecting either 1) the extent to which the CYP forms are expressed in the
endoplasmic reticulum of the cell (isolated as the microsomal fraction from tissue homogenates),
1-1
-------
or 2) the expression of microsomal protein in intact liver tissue. Biochemically sound measures
of the hepatic distribution of xenobiotic metabolizing enzymes among humans, based on
expression of the enzymes within microsomal protein and the distribution of microsomal protein
among intact livers, can be combined with metabolic constants derived in vitro to generate
values consistent with those employed in PBPK models. When completed, the distribution (and
bounds) of Vmax values can be estimated and included in PBPK models. Exercising such models
under plausible exposure scenarios will demonstrate the extent to which human interindividual
enzyme variance can influence parameters (i.e., the detoxication of a toxic chemical through
metabolism) that may influence risk. In this article, we describe a methodology and conditions
which must exist for such an approach to be successful.
INTRODUCTION
The establishment of safe exposure limits for chemicals is a primary concern for
individuals exposed in their occupations, through the environment, or the through the
consumption of sustenance or medication. Several different federal agencies (e.g., National
Institute for Occupational Safety and Health, US Environmental Protection Agency, US Food
and Drug Administration) are charged with developing safe exposure guidelines for xenobiotics
(chemicals and drugs). Often, these xenobiotics are encountered through more than one
exposure scenario. An industrial chemical may become an environmental pollutant, and a
therapeutic agent (either human or animal) may find its way into the environment and the food
chain. Thus, the uniform recognition of some fundamental underlying concepts and their
consistent application to identify and protect sensitive individuals seem to be in order.
In the US EPA's methodology to ascribe Reference Dose (RfD) values to
environmentally-occurring contaminants, adequate information from studies with test animals
1-2
-------
are used to determine human doses deemed to be without increased probability of risk when
encountered continuously over a lifetime. Uncertainty factors (UF) are employed to adjust
Lowest-Observed-Adverse-Effect Level (LOAEL) or No-Observed-Adverse-Effect Level
(NOAEL) values determined in animal (or human) studies to doses (concentrations) deemed to
be without significant human health risk. Within the US EPA, specific attention has been
focused on the dissection of metabolic differences from the general UF in risk assessments by
dividing the UF used in the derivation of RfD values into their constituent pharmacokinetic (PK)
and pharmacodynamic (PD) components. This division follows similar earlier advances in the
establishment of safe exposure limits set for inhaled substances. Although the genetic similarity
among humans is remarkable (when compared to differences between humans and animals), the
degree of human interindividual variance in key biochemical machinery produces differences in
xenobiotic distribution and sensitivity to toxic insult among humans. Small differences in the
genetic code can result in lower or higher expression of certain genes, resulting in lower or
higher expression of coded enzymes or in the expression of enzymes whose function is
compromised compared to the normal expression. Genetic polymorphisms can exist in
unencoded genetic domains (introns) or in coded domains (exons); the latter are responsible for
the transcription of a compromised protein. Some polymorphisms are responsible for allelic
expression, where the number of alleles expressed is quantitatively related to the enzyme
content, and thus to enzyme activity. Other polymorphisms may exist that alter the three-
dimensional structure of an enzyme without altering the degree to which the enzyme is
expressed. These alterations may affect how an enzyme interacts with substrate molecules, thus
potentially altering substrate specificity, substrate affinity, and maximal activity.
1-3
-------
Factors beyond genetic polymorphisms may also be responsible for human
interindividual differences in chemical metabolism. Dietary factors, including types and
quantity of food, control the expression of some forms of cytochrome P450 (CYP). Lifestyle
factors such as cigarette smoking, stress, and alcohol consumption may alter the expression of
CYP forms. Health conditions such as diabetes and obesity may alter the expression of CYP
forms, and the expression of some members of other enzyme families (i.e, UDP glucuronyl
transferase, glutathione S-transferase) are themselves recognized as markers of disease
processes. Immortalized cell lines used for in vitro studies express different levels of some
enzymes compared to the corresponding normal cells. These differences in enzyme expression
and activity in vivo can alter the circulating levels and tissue distribution of xenobiotics and their
metabolites. Thus, differences in the expression of xenobiotic-metabolizing enzymes can have
an appreciable impact on risk relevant PK outcomes (i.e., rate of degradation of a toxic parent
compound, rate of formation of atoxic metabolite, tissue concentrations of atoxic metabolite,
etc.). Not all PK outcomes may quantitatively correlate to risk (i.e, circulating levels of a
compound whose bioactivated metabolite formed in target tissues produces toxicity). However,
studies on the variance in expression of hepatic (and extrahepatic) drug metabolizing enzymes
will be useful when the risk relevant PK outcome is dependent on their activity and/or
expression.
Factors other than the expression of an enzyme may also lead to changes in metabolic
parameters which can influence risk. Drug-drug interactions based on either enzyme inhibition
due to competition between two xenobiotics for a single enzyme or xenobiotic-dependent
induction are well recognized and serve as the basis for therapeutic contraindications.
Sometimes, the balance between the content of two or more enzymes determines toxicity. For
1-4
-------
example, bioactivated CYP-derived metabolites of acetaminophen are detoxicated by GST-
catalyzed conjugation. Increases in CYP activity or decreases in conjugating activity will
increase the level of acetaminophen toxicity.
The extent to which differences in enzyme expression and activity (two different
measures) predispose individuals to the toxic effects of xenobiotics is worthy of study, given the
increasing degree to which toxicity and PK information are becoming available. The
relationships between xenobiotics (environmental contaminants or therapeutic agents) and the
enzymes responsible for their metabolism, and between xenobiotic metabolites and toxicity are
critical factors in determining susceptibility. Individuals who allelically express low (or no)
levels of CYP2D6 are at increased risk for toxicity (adverse drug reaction) due to the
accumulation of some commonly employed therapeutic compounds. Likewise, individuals who
over-express an enzyme which bioactivates a non-toxic parent compound to bioactive (toxic)
metabolites may, to some degree, be protected from the toxic response.
The CYP enzymes are studied most often and with the highest degree of certainty in
vitro. The preparation of metabolically active tissue fractions for in vitro investigations
necessarily involves their removal from surrounding tissue, and often requires an artificial
increase in their relative concentration, usually produced via ultracentrifugation (Figure 1-1).
The variance in the content of CYP enzymes in the metabolically active subcellular fraction
isolated from human liver has been evaluated in several key publications (Iyer and Sinz, 1999;
Shimada et al., 1994; Snawder and Lipscomb, 2000; see Figure 1-2). In the absence of data
describing the microsomal protein (MSP) content of liver, however, these evaluations of
metabolic rates and enzyme expression in MSP isolated (literally, in isolation) from the intact
liver provide data which are of limited value in determining the expression of microsomally-
1-5
-------
contained enzymes in the intact liver or metabolic rates representative of the intact liver
sufficient for inclusion in PK models.
Most CYP enzymes follow saturable Michaelis-Menten kinetics:
v = Vmax * [S] / (Km + [S]) (1-1)
where v is the initial velocity (rate; d[S]/dt) of the reaction, Vmax is the maximal rate of the
reaction, [S] is the substrate concentration, and Km is the Michaelis constant. Equation (1-1)
indicates that the initial velocity of the reaction increases hyperbolically as a function of the
substrate concentration. The Vmax is a horizontal tangent to the zero-order (saturated) portion
of the curve, while the tangent to the initial linear (first-order) portion of the curve at low
substrate concentrations is the initial rate of the reaction, V/K (Vmax/Km). The Km is the
substrate concentration that gives one-half Vmax. Thus, at the low tissue concentrations attained
after occupational or environmental exposures to chemicals, the V/K (reflected by the Km) is
more important than the Vmax in describing the kinetics of the reaction. In situations such as
bolus exposures to chemicals or drugs, Vmax can be more important. For rapidly metabolized
chemicals, the slowest overall step in disposition is not biotransformation but rather delivery of
the substrate to the liver via hepatic blood flow. In many exposure scenarios involving rapidly
metabolized chemicals, differences in CYP activity due to genetics or induction do not result in
differences in metabolic activation because of the overall limitation of blood flow delivery of the
chemical to the liver (Kedderis, 1997).
PBPK modeling offers an opportunity to study the impact of differences in enzyme
expression on both risk relevant and other PK outcomes in humans. When PK models are
constructed to include metabolic rates (and rate constants) derived in vitro, several extrapolations
are necessary, not the least of which is the extrapolation of enzyme content. PBPK models
1-6
-------
include the apparent Vmax expressed as mg/hr/kg body mass, while typical in vitro studies
express Vmac in terms of nmoles product formed/minute/mg microsomal protein. Accurate
extrapolation requires initially that enzyme content be expressed per unit intact liver (i.e., pmoles
CYP2El/gram liver), and the extrapolation has usually included a numerical estimation of the
MSP content of liver (i.e, 50 mg MSP/gram liver). Measures of the MSP content of the intact
liver have been inferred or developed in several PBPK studies in which rates of metabolism
derived in vitro have been extrapolated to the in vivo setting (Lipscomb et al., 1998; Reitz et al.,
1996).
This manuscript identifies the types of data required, communicates an approach,
describes the limitations of the approach and proposes the applicability of the approach to
estimate the human interindividual PK variance of outcomes which are relevant to risk and
which may signify susceptibility to chemical injury.
DEFINITION OF THE PROBLEM
The wealth of information being generated from genomic and proteomic investigations
places more and more opportunities for refinement within the grasp of risk assessors. These
results allow investigators to more correctly identify the fraction of the population which differs
from the norm for enzyme expression, and to gather data on the degree to which enzyme content
and/or activity varies among the population under investigation for risk. For example, advances
in US EPA's inclusion of specific mechanistic, biochemical and PK modeling techniques in risk
assessments encourages the development of methods to incorporate emerging data that can be
used to quantify differences in enzyme expression. Because metabolism and PK are components
of chemical exposure that have been specifically linked with risk, we were compelled to develop
1-7
-------
and communicate methods to include these new data sets in risk assessments. Several key
concepts defining the problem are given below.
• Environmental human health protection guidelines are designed to limit exposure to
chemicals producing adverse effects in tissues, organs or organ systems.
• Newer risk assessment methods used to ascertain "safe" doses include specific
mechanistic and biochemical information.
• Attention is given to the absorption, distribution, metabolism and elimination of a
chemical.
• US EPA has been encouraged to develop risk assessment approaches that reduce
uncertainty in the extrapolation of results from animals to humans and within the human
species in order to protect sensitive humans.
• In vitro measurements of human enzyme activity toward a toxicant can yield qualitatively
and quantitatively valuable data on human variance.
• PBPK models can be used to investigate the extent to which human interindividual
enzyme variance can influence parameters (i.e., the detoxication of a toxic parent
chemical through metabolism) that may influence risk.
OBJECTIVE
In order to refine human health risk assessments for chemicals metabolized by the liver
to reflect data on human interindividual metabolic and PK variance, the objective of this report is
to communicate a method developed to 1) extrapolate measures of enzyme activity derived in
vitro that capture human interindividual variance, and 2) demonstrate the applicability of the data
and approach for risk assessment purposes. This report communicates a means by which new
information on the variance of enzyme expression may be linked with PBPK modeling to
estimate the degree of human interindividual variance with respect to a risk-relevant PK
outcome.
-------
APPROACH
Extrapolation of metabolic rates derived in vitro from the subcellular preparations (i.e.,
cytosolic protein or MSP) to the intact liver requires two sets of data. The first set is the liver (or
pertinent extrahepatic tissue) content of the enzyme (e.g., pmoles CYP2El/gram liver). This
may be empirically derived by quantitatively recovering liver homogenate protein (i.e., pmoles
CYP2El/mg homogenate protein) and assessing the homogenate protein's content of the
enzyme. This is seldom done. Another approach to determining the liver's content of the
enzyme involves the application of information presently available which describes the MSP's
content of enzymes. However, this approach requires the generation of an additional data set:
that describing the liver's content of MSP (i.e, mg MSP/gram liver). This measurement is rarely
reported in metabolism studies, regardless of the subcellular fraction investigated. In instances
where the individual enzyme responsible for metabolism has been identified and data on its
distribution (and variance) have been determined in the corresponding subcellular fraction (e.g.,
pmoles CYP2El/mg MSP), data can be included to quantify the variance of enzyme expression
within the subcellular fraction as well as the variance in the distribution of the subcellular
fraction to the intact liver. The liver content of an enzyme (Figure 1-3, panel C) can thus be
determined by combining separate data sets describing the MSP content of the liver (panel A)
and the CYP content of MSP (panel B). Because the objective of this initial step is to develop a
measurement of the amount of enzyme present per gram liver tissue, a direct measurement of this
parameter may also be obtained from the quantification of the enzyme directly in liver
homogenate protein, although this has not been demonstrated. When estimating variance of the
summed measurements, it is important to characterize and capture adequate measures of the
variance within each individual data set.
1-9
-------
The second required data set is that which describes the metabolic rate expressed per unit
pertinent enzyme (i.e., CYP2E1; Equation 1-2; Figure 1-3, panel D) or per unit subcellular
protein (i.e., mg MSP; see Equation 1-3). This measurement is typically available as a point
estimate, although it can be measured as a distribution when adequate data are available. Data
describing the metabolic rate may be available as a turnover number, the quantity of substrate
metabolized per unit time per unit of enzyme. This measurement may be developed from
purified enzymes or from genetically expressed enzymes. In exceptional cases, variance about
this measurement may be characterized and that variance also captured in the extrapolation
procedure. Ideally, the turnover number should be independent of preparation as long as each
enzyme is genetically identical. However, some human CYPs exhibit genetic variance that
affects their catalytic activity. Additionally, CYP and other xenobiotic metabolizing enzymes are
membrane bound and interact with other proteins and lipids in ways that affect their catalytic
activity. These lipids and other cellular constituents, which are present in human tissue
preparations, are not present in genetically expressed systems. The potential instability of some
purified enzymes must also be considered.
(product formed/min/pmol CYP) * (pmol CYP/mg MSP) *
(mg MSP/gram liver) = product formed/min/gram liver (1-2)
(product formed/min/mg MSP) * (mg MSP/gram liver) =
product formed/minute/gram liver (1-3)
Equation 1-2 would be best suited when in vitro metabolic rates are expressed per unit of
the CYP form demonstrated to be responsible, and the data are deemed sufficient to represent the
variance of the MSP content of human liver. Because of the demonstrated variance of CYP
forms within the isolated MSP and because of the content of endoplasmic reticulum (which
contains the microsomal enzymes) may vary among humans, Equation 1-2 seems to better
1-10
-------
capture the variance of the expression of CYP enzyme expression in the intact liver. However,
Equation 1-3 is perhaps most widely applicable at present, given that in vitro metabolic rates
(and rate constants) are most commonly expressed per unit subcellular protein. Both equations
demonstrate the need for data describing the content of microsomal protein in intact liver.
The mathematical combination of individual data sets describing 1) enzyme activity, 2)
enzyme content of MSP, and 3) MSP content of the intact liver must be undertaken through
statistically defensible procedures. When data sets are demonstrated (or assumed) to be
independent and the observations are lognormally distributed within the sets, the statistical
method of moments (addition of errors) can be combined to produce the overall distribution. In
this method, the critical information is the geometric mean and geometric standard deviation of
the distributions. Most publications describing the variance of enzyme activity and enzyme
expression are devoid of descriptions of distribution and the measurements necessary to facilitate
this recombination of data. The recombination of these data sets will produce a distribution,
whose parameters can be used to identify metabolic rates at specifically identified points in the
resultant distributions in the population (i.e., those representing the 5tn an(j 95th percentiles of
the distribution).
Most PBPK models are constructed assuming that all metabolism occurs in the liver, and
the metabolic rate constants (apparent Vmax) are expressed as mg substrate metabolized/hr/kg
body weight. Therefore the in vitro rate constants must be converted to those units. This is
accomplished by correcting for molecular weight, time, the fractional composition of the body
mass accounted for by liver (approximately 2.6% in humans), and body weight. The PBPK
models can be expanded to incorporate additional information on extrahepatic metabolism or
protein binding as this information becomes available.
1-11
-------
REQUIRED CONDITIONS
Several conditions must first be met for this strategy to be successful:
• Information must exist to identify the target organ (and should identify the target cells in
that organ), the mechanism of toxic action, and the metabolic species responsible for
toxicity (parent compound or metabolite). For example, available information may
identify the parent chemical and its concentration in the brain as responsible for the noted
nervous system toxicity.
• When a metabolite is responsible for toxicity, the identity of the metabolite must be
known and enzyme(s) responsible for its formation must be identified. For example, the
epoxide metabolite formed by CYP1 A2-catalyzed oxidation may be responsible for
toxicity.
• The kinetic mechanism of metabolism should be known. Most CYP enzymes follow
Michaelis-Menten saturation kinetics. Ideally, information on metabolic rate should be
expressed per unit enzyme. For example, the Vmax is most suitable when expressed as
product formed/time/pmol CYP2E1.
• Data quantifying the expression and variance of the enzyme in the intact liver tissue
should be available. Sometimes information exists quantifying metabolic rate per unit of
individual enzyme, and those data may be best employed when other information is
available describing their expression in MSP (Snawder and Lipscomb, 2000). Most
often, metabolic rates are expressed as product per unit subcellular fraction (i.e., product
formed per mg MSP), and so the distribution of the pertinent subcellular fraction (and not
the expression of the individual enzymes) to the liver is required. Equation 1-3
represents this condition. Regardless, not only should the variance be known, but also
the type of distribution. The relationships between the individual data sets must be
known to enable the most valid statistical recombination of the sets.
• An adequately characterized PBPK model must be available for adaptation.
APPLICABILITY
With the availability of large banks of well-characterized subcellular fractions (mainly
hepatic MSP) derived from the livers of human organ donors comes the opportunity to determine
several measures of human interindividual biochemical variance. Although several
investigations have failed to identify a consistent inverse relationship between post mortem cold-
clamp time and microsomal enzyme activity, the assumption that the activity of these enzymes in
vitro represents their activity in vivo must be applied. From these samples, we can measure
1-12
-------
interindividual differences in enzyme activity and differences in enzyme content in isolated
MSP. The in vitro metabolism of several CYP2E1 substrates, such as furan (Kedderis et al.,
1993; Kedderis and Held, 1996), perchloroethylene (Rietz et al., 1996), and trichloroethylene
(Lipscomb et al., 1998) have been successfully extrapolated to the in vivo setting through
application of adequately developed and validated PBPK models. The additional validation of
the extrapolation procedure for metabolic activity based on enzyme recovery data is important.
This demonstrates the applicability of the methodology to determine the interindividual variance
of risk-relevant PK outcomes (i.e., the amount of metabolite formed in the liver for a
bioactivated hepatotoxicant) for xenobiotics to which humans cannot be safely exposed for the
generation of experimental data. It is anticipated that toxicological data can be generated in test
species in vivo and in vitro to determine the metabolic species responsible for toxicity, the PK of
the xenobiotic and metabolite(s), and the identity of the enzyme responsible for metabolism.
With this information, an adequate test animal-based PBPK model can be extrapolated to
humans, using human tissue partition coefficients and the appropriate physiological parameters.
Data on human enzyme recovery could be used to develop appropriate bounds on the distribution
of metabolic activity for evaluation with the PBPK model.
CONCLUSIONS
1. Human interindividual PK variance is important both for chemicals with adequate human PK
data and for those chemicals to which humans cannot be experimentally exposed.
2. The described approach makes maximal use of data on the variance of enzyme content in the
intact liver and enzyme specific activities derived in vitro with samples prepared from human
organ donors.
3. The statistical combination of multiple data sets can be performed using widely accepted
statistical methods.
4. This methodology can be employed to quantify the human interindividual variance of risk-
relevant PK outcomes, when sufficient data are available to identify underlying fundamental
1-13
-------
conditions such as the toxicologically-active metabolic species, the target organ for toxicity,
the enzyme responsible for its formation/degradation, and the distribution of that enzyme to
the intact liver.
5. Additional studies to characterize the distribution of MSP in intact liver in humans, in test
animal species, and across developmental stages will benefit future efforts.
REFERENCES
Iyer, K.R. and M.W. Sinz. 1999. Characterization of phase I and phase II hepatic drug
metabolism activities in a panel of human liver preparations. Chem. Biol. Interact. 118:151-
169.
Kedderis, G.L. 1997. Extrapolation of in vitro enzyme induction data to humans in vivo. Chem.
Biol. Interact. 107:109-121.
Kedderis, G.L. and S.D. Held. 1996. Prediction of furan pharmacokinetics from hepatocyte
studies: Comparison of bioactivation and hepatic dosimetry in rats, mice, and humans. Toxicol.
Appl. Pharmacol. 140:124-13 0.
Kedderis, G.L., M.A. Carfagna, S.D. Held, R. Batra, I.E. Murphy and M.L. Gargas. 1993.
Kinetic analysis of furan biotransformation by F-344 rats in vivo and in vitro. Toxicol. Appl.
Pharmacol. 123:274-282.
Lipscomb, J.C., J.W. Fisher, P.D. Confer and J.Z. Byczkowski. 1998. In vitro to in vivo
extrapolation for trichloroethylene metabolism in humans. Toxicol. Appl. Pharmacol.
152:376-387.
Reitz, R.H., M.L. Gargas, A.L. Mendrala and A.M. Schumann. 1996. In vivo and in vitro
studies of perchloroethylene metabolism for physiologically based pharmacokinetic modeling in
rats, mice and humans. Toxicol. Appl. Pharmacol. 136:289-306.
Shimada, T., H. Yamazaki, M. Mimura, Y. Inui and P.P. Guengerich. 1994. Interindividual
variations in human liver cytochrome P450 enzymes involved in the oxidation of drugs,
carcinogens and toxic chemicals: Studies with liver microsomes of 30 Japanese and 30
Caucasians. J. Pharmacol. Expt. Ther. 270:414-423.
Snawder, I.E. and J.C. Lipscomb. 2000. Interindividual variance of cytochrome P450 forms in
human hepatic microsomes: Correlation of individual forms with xenobiotic metabolism and
implications in risk assessment. Reg. Toxicol. Pharmacol. 32:200-209.
NOTICE: The authors appreciate the constructive comments received from Dr Harlal
Choudhury. The views expressed in this paper (NCEA-C-0963J) are those of the individual
authors and do not necessarily reflect the views and policies of the US Environmental Protection
1-14
-------
Agency (EPA). This paper has been reviewed in accordance with EPA's peer and administrative
review policies and approved for presentation and publication. This manuscript was published in
Science of the Total Environment as part of the proceedings of the Spring Conference on
Toxicology, held April 2001, in Fairborn, Ohio. Lipscomb, J.C. and Kedderis, G.L. 2002.
Incorporating human interindividual biotransformation variance in health risk assessment. Sci.
Total Environ. 288:13-21.
1-15
-------
LIVER
Homogenation
and
Centrifugation
\
Microsomal
Protein
Enzyme Activity
Metabolite per:
• mg Microsomal Protein,
• pmol CYP2E1
Nuclei, Debris,
Cytoplasm
Enzyme Content
pmolesCYP2El/mg
Microsomal Protein
FIGURE 1-1
Isolation of Microsomal Protein from the Intact Liver. In this procedure, intact tissue is
homogenized and subjected to an initial centrifugation which results in the sedimentation of
cellular debris, mitochondria and nuclei. A subsequent higher speed centrifugation results in
sedimentation of microsomal protein, the fraction enriched for the content of endoplasmic
reticulum and associated enzymes. The supernatant of this high-speed centrifugation contains
proteins distributed to the cytoplasm of the cell. A resuspension of the pelleted microsomal
protein is performed in limited volume to retain as much of the concentrating effect as possible.
1-16
-------
Liver
Isolated
Microsomal
Protein
Microsomal Protein
CYP3A
O CYP2E1
O CYP1A2
FIGURE 1-2
Relationship Between Intact Liver, Microsomal Protein and Some CYP Forms. The isolation of
microsomal protein from intact liver via homogenation of tissue and differential centrifugation
results in a 100,000 x g pellet which is enriched for endoplasmic reticulum content. The
enrichment results in an artificial increase in the concentration of biological components
associated with the endoplasmic reticulum. This isolation produces a fraction (microsomes;
MSP) which is subjected to in vitro investigations of metabolic activity and enzyme content.
However, a quantitative relationship to the intact liver is not possible without further information
on the distribution of microsomal protein to the intact liver.
1-17
-------
LoglO (mg MSP/gram Liver
LoglO (pmolCYP2EI/gram Intact Liver)
Specific Activity
vs. Chemical-^
D
LoglO (pmolCYP 2El/mg MSP)
PBPK Model
Relevant
Exposure Scenario
Magnitude of Variance in PK Output
FIGURE 1-3
Extrapolation and Incorporation of In Vitro Derived Metabolic Rates in PBPK Modeling. This
figure depicts the framework for deriving appropriate in vitro measures and their extrapolation
into a PBPK model. Basically, the metabolic rate should be expressed per unit of responsible
enzyme, and the distribution of that enzyme to the intact liver must be known. Because CYP2E1
and many other enzymes are expressed solely in the endoplasmic reticulum or MSP of the cell,
this figure depicts the distribution of MSP to the intact liver (panel A). Secondly, the
distribution of CYP2E1 within the MSP derived from liver tissue of human organ donors is
depicted (panel B). By statistically combining these two (independent) data sets (see Equation
(2)), an estimate of the CYP2E1 content of intact liver can be produced (panel C). Consistent
with estimating the variance surrounding 90% of the population, we chose to represent the 5th
and 95th percentiles of this distribution for evaluation. The enzyme activity representing the 5th
and 95th percentiles for enzyme content can be determined by multiplying the content (pmoles
CYP2El/gram liver) by the specific activity of CYP2E1 toward a selectively metabolized
substrate, and then correcting for molecular weight, time and fractional body weight attributed to
the liver (2.6% or 26 grams liver/kg body mass). Panel D demonstrates the specific activity
(turnover number) of the enzyme for the xenobiotic substrate as a point estimate, but a
distribution for that value might also be obtained and employed to capture additional
biochemical variance. These extrapolated upper and lower bound metabolic rates, when
corrected to those expressed as mg/hr/kg body weight, can then be incorporated in a human
PBPK model. The model should be exercised to simulate the exposure under relevant
conditions. Models may behave differently, and the impact of the variance in enzyme content
and activity may be different under different exposure conditions.
1-18
-------
2. HOW DIFFERENCES IN ENZYME EXPRESSION CAN TRANSLATE INTO
PHARMACOKINETIC VARIANCE AND SUSCEPTIBILITY TO TOXICITY
ABSTRACT
Advances in risk assessment methodologies invite the consideration of more
pharmacokinetic, pharmacodynamic and biochemical information than ever before. When
toxicity data are adequate, the identification of enzymes responsible for detoxication or
bioactivation invite their consideration in risk assessment. Consideration should be given not
only to the ontogeny of enzyme expression, but to other biochemical factors such as tissue lipid
content and changes in relative organ size and blood flow during development. Dogma that
enzyme alterations result in alterations of risk requires reexamination. Sometimes these do alter
risk, even though the enzymes form lexicologically active metabolites. Knowledge that a given
enzyme is responsible for metabolism of an environmental contaminant often stimulates the
search for and use of information on its variance in expression and/or activity. Separate data on
enzyme expression (content) and enzymatic activity can be useful in human health risk
assessment, but requires the application of physiologic constraints, which can vary with age.
Without these constraints, data from otherwise well-conducted studies have fueled incomplete
and misleading conclusions. This manuscript uses the generic example of variance in xenobiotic
metabolizing enzymes to make some points on the proper conduct and interpretation of in vitro
findings pertinent to risk and susceptibility, such as findings inferred from studies of genetic
polymorphisms. This information must be integrated in the context of the intact animal and/or
intact human. The technique of physiologically based pharmacokinetic modeling offers a
platform upon which to integrate findings from disparate areas of biochemical research, while
maintaining the constraints of the intact body.
2-1
-------
BACKGROUND
Biological variability is a key component of risk, serving as the basis for differences in
susceptibility between species, and among adult members of the human species and between
adults and children. Continued advances in risk assessment methodologies should capture the
refinements in our understanding of biological variability underlying differential susceptibility to
effects. Not all biological or biochemical differences modulate risk, and a critical task is to
discern which biochemical differences do and which do not contribute to susceptibility by
linking them with key causal events. Logically, the next step in the process is to evaluate those
which do alter risk and determine the basis for their contribution. Although biological variability
can relate to either pharmacokinetic (PK) or pharmacodynamic (PD) processes, the present
manuscript attempts to address variance in drug metabolizing enzymes, whether differences in
their expression relate to developmental, genetic or other factors, and the determination of their
contribution to risk through PK alterations. Of paramount importance is the availability of
xenobiotic substrate to the liver, and consideration of first-order metabolism. None of these
findings should be taken out of the context of the intact body:
Structure without function is a corpse, function without structure is a
ghost; in vitro findings unbounded by in vivo constraints are
conjecture.
TOXICITY AND RISK ASSESSMENT
The availability of advanced information on human genetics and proteomics, as well as
the increasing level of detail about toxic mechanisms, invites the inclusion of more technical
information in human health risk assessments. Risk assessments should be advanced to include
relevant data interpreted under appropriate constraints. In the absence of full considerations of
human anatomy, physiology and biochemistry, this can lead to some mischaracterizations. Risk
2-2
-------
assessors, and those developing advanced toxicokinetic and toxicodynamic information should
bear in mind that the ultimate location where genetic variance and human biochemical
individuality, regardless of developmental stage, impacts human health risk is the juxtaposition
of the toxic molecular species and the biochemical target. This interaction may be at a specific
site such as a receptor (Nesaretnam et al., 1996) or an individual protein (Soldin et al., 2001), or
it may result from the development of a non-specific condition, such as oxidative stress
(Umemura et al., 1998).
The derivation of human exposure limits often includes extrapolation of doses at
identified response levels in animal studies to corresponding doses in the general human
population and the further consideration of sensitive humans, requiring a second extrapolation of
the dose in the general human population to that in sensitive individuals. The uncertainty factors
governing these two extrapolations (UFA and UFn, respectively) have been further divided into
pharmacokinetic (PK) and pharmacodynamic (PD) components (IPCS, 2001; US EPA, 1994).
These components have different importance in the risk assessment paradigm (Figure 2-1).
While both are involved in the dose-response evaluation phase, only PK is involved in exposure
assessment, which is best developed to describe the internal (target tissue) dose rather than the
concentration of agent in an environmental contact medium.
METABOLISM AND PHARMACOKINETICS IN RISK ASSESSMENT
Recently, the default uncertainty values have begun to be replaced with chemical-specific
uncertainty factors in the derivation of reference doses (RfD) (Smallwood et al., 2001; Murray
and Andersen, 2001). This requires toxicity information sufficient to identify the chemical
species (e.g., parent or metabolite, which metabolite) responsible for toxicity and the target organ
or tissue.
2-3
-------
Genetic polymorphisms and the impact of enzyme variance between individuals and
among age groups should be evaluated in a chemical-specific manner. When differences in
enzyme activity are risk-relevant, some preliminary consideration should be given to the
potential impact of variance on the ultimate risk. A hypothetical case where a chemical is either
95% or 99% metabolized provides a useful example. When metabolism results in bioactivation,
the risk resides in the formation of the metabolite and the significance of that difference
(0.99/0.95 = 1.04-fold) in the lexicologically active species is rather small. However, when
metabolism represents a detoxication step, the risk is from the parent chemical, not the
metabolite, and the difference [(1 - 0.95) / (1-0.99) = 5-fold] becomes appreciable.
Consideration of enzyme kinetics, especially reactions governed by first-order kinetics,
becomes critical in considering whether differences in enzyme content and/or activity can be
important mediators of susceptibility. The parameters governing first-order reactions are Vmax
(the theoretical maximal initial rate of the reaction) and Km (the substrate concentration
necessary to drive the reaction at one-half maximal velocity). These constants are unique for
each reaction; even reactions for different substrates which are catalyzed by the same enzyme;
the constants are readily measurable for most substrates. Their characterization is complicated
by the choice of in vivo test system (selection of test species and dosing conditions) choice of in
vitro test system (subcellular fraction or isolated hepatocytes derived from animals or humans)
and method of detection employed to quantify metabolism (appearance of product or
disappearance of parent compound). Under the conditions used to characterize metabolism,
Vmax and Km are representative of "quantity" and "quality", respectively. Vmax is directly
dependent on the concentration of the enzyme in the test system. Km, on the other hand,
includes the affinity of the substrate for the enzyme (and vice-versa), and reflects the ability of
2-4
-------
the enzyme to catalyze the reaction at lower substrate concentrations. Thus, observed changes in
the Vmax for one reaction may be generalizable to the metabolism of other substrates for the
same enzyme, while the generalizability of changes in Km is complicated by chemical-specific
characteristics and, for the many lipophilic environmental contaminants, by age-dependent
differences in tissue lipid content. The relationship between substrate concentration (in vivo, and
for risk assessment purposes, this is best represented by the concentration in target tissues),
Vmax and Km is demonstrated in the Michaelis-Menten rate equation (Equation 2-1).
Rate = (Vmax * [s]) / (Km + [s]) (2-1)
Saturation of metabolism (when substrate is not limiting to the reaction rate) is not frequent in
environmentally-encountered toxicants, only seldom observed in occupational exposure, but is
frequently observed in experimental studies with research animals. When the concentration of
the enzyme in metabolically active tissues varies, the metabolism of the substrate is related to the
amount of enzyme present (note that physiological restrictions of the delivery of the substrate to
the enzyme have an important, sometimes limiting, effect on metabolism in vivo, this is
discussed later). Thus, under conditions where substrate delivery is not limiting, rates of
metabolism are directly proportionate to the amount of enzyme present (Vmax). However, the
rate of metabolism is not proportionately related to the Km value; results in Figure 2-2
demonstrate that when substrate is present at concentrations far above the Km value, that
metabolic rates are not particularly sensitive to changes in Km values. Thus, alterations of Km
and Vmax values have different effects on metabolic rates, dependent on the ratio of substrate
concentration to Km value. Figure 2-2 demonstrate the effect of 10-fold increases and 10-fold
decreases in Km and Vmax on metabolic rates when substrate is present below, at and above the
Km. Here, metabolic rates are proportionate to Vmax across the entire range of substrate
2-5
-------
concentrations. Changes like these are typical of enzyme induction - a term used to indicate an
in vivo up-regulation of enzyme content. In contrast, 10-fold changes in Km only produce
changes in metabolic rate roughly proportionate to the change in Km value at substrate
concentrations 100-fold lower than the Km value. A change in Km value results in only
moderate changes in metabolic rate when substrate is present at concentrations much higher than
the Km value. These results plainly demonstrate 1) the need to determine whether alterations in
metabolic rate are due to changes in the quantity of protein (enzyme) present (directly related to
Vmax), or due to changes in the qualitative nature of the enzyme (directly related to Km), and 2)
the need to ascertain the substrate concentration relative to the Km value. The predictive value
of in vivo findings relating polymorphisms (qualitative or quantitative changes) and quantitative
age-dependent changes in enzyme content of drug metabolizing enzymes to metabolic rates can
only be fully appreciated when the metabolic rates are related to 1) the concentration of substrate
attained in vivo, and 2) the Km for the reaction. An additional note of caution should be applied
to address the developmental pattern of drug metabolizing enzymes, especially so when multiple
enzymes man contribute to the metabolism of a given xenobiotic. Several data sets have
demonstrated the contribution of several P450 forms to the oxidation of trichloroethylene; these
forms have appreciably different affinities (Km values) and different developmental patterns of
expression.
USING ENZYME KINETIC DATA TO ESTIMATE PHARMACOKINETIC VARIANCE
The basis of differences in tissue concentrations of toxicants typically brings to mind
differences in the partitioning of the chemical from blood to tissue or differences in metabolism.
Removal of the xenobiotic from circulating blood reduces tissue concentrations, and the effect of
the xenobiotic, whether beneficial or toxic, on the biological system. For therapeutics, the
2-6
-------
concentration in blood (or plasma) is typically the target of predictions, as this fluid is available
for analysis during clinical and preclinical trials. The effect of interest is related to blood
(plasma) concentrations. These conditions differ from those in chemical risk assessment, in
which concentrations of toxic agent in, most often, non-blood target tissue are the object of
investigation. Under these two disparate conditions, different measures of clearance are
justified. To estimate the impact of variance in enzyme content and activity, the best approach is
to employ the aid of a physiologically based pharmacokinetic (PBPK) model (Lipscomb and
Kedderis, 2002). For extrapolation and inclusion in the model, the variance in enzyme content
should be statistically quantified (see Snawder and Lipscomb, 2000) and information on the
variance in catalytic activity (Michaelis-Menten kinetic parameters) should be available (see
Lipscomb et al., 1997).
Intrinsic clearance (CL;nt) of substrates (Equation 2-2) describes the removal of
substances from the circulation and is expressed in units of milliliters cleared per hour. This
measure has value in the pharmacokinetic assessment of therapeutics (Houston, 1994) where
target tissue concentration is not considered critical, and when the doses are somewhat
standardized. However, Clint does not account for the concentration of the substrate in the
metabolically active tissue (Table 2-1).
CLint = Vmax/Km (2-2)
Hepatic clearance (CLhep; Equation No. 2-3) of substrates is used to describe the
metabolism of substrates in physiologically based pharmacokinetic (PBPK) modeling, and
includes a term used to quantify the rate of delivery of the substrate to the liver via hepatic blood
flow. Delivery to liver is a function of the concentration of the chemical in the medium which
contacts blood, the partitioning of the chemical from the medium into blood, and the partitioning
2-7
-------
of the chemical from the blood to the liver (Figure 2-3). These three physiologic components are
not considered in the prediction of intrinsic clearance.
CLhep = (Vmax/Km) * QLiver / (Vmax/Km) + QLiver (2-3)
Table 2-1 demonstrates the impact of variances in Vmax and Km in the presence of
variances in the Quver term in a hypothetical but plausible example. The results from this
example clearly demonstrate the differential applicability of assumptions about the impact of
qualitative (Km) and quantitative (Vmax) differences in xenobiotic metabolizing enzymes
relative to exposure conditions. Under the conditions of a lower exposure (Quver= 0.005), a
10-fold increase in Vmax (enzyme content) results in virtually no change in CLhep. In
comparison, this same change in Vmax under the conditions of higher exposure (Quver= 50)
results in a roughly proportionate difference in CLhep. Because of the V/K term in the equation
for Hepatic Clearance, this same effect would be observed for a 10-fold decrease in Km, because
V/K would remain the same as that value for a 10-fold increase in Vmax. Note that for this
exercise, difference in the V/K term representing Intrinsic Clearance remains fixed at 10-fold,
regardless of substrate delivery. Figure 2-3 demonstrates the importance of substrate delivery
via hepatic circulation. Kedderis (1997) treated this concept in more detail, when he
demonstrated that the influence of enzyme induction on furan metabolism was dampened
significantly by limitations of metabolism imposed by delivery of substrate via blood flow.
The delivery of a substrate to the liver can be rate-limiting in metabolism and is
determined by factors including 1) contact of blood with the substrate, 2) the concentration of the
substrate in the medium contacting blood, 3) solubility of the substrate in blood, 4) the
distribution (partitioning) of the substrate from blood into extrahepatic tissues, 5) the rate of
blood flow to liver, 6) the duration that blood remains in contact with the liver, and 7) the
2-8
-------
partitioning of the substrate from the blood into liver tissue. The solubility of xenobiotics in
blood and solid tissues is a critical factor in pharmacokinetic analyses, and is one for which few
children-specific data sets exist. Metabolic capacity of the liver is characterized by the enzyme
content of the liver and by the activity of the enzyme present in the liver. The activity of an
enzyme functioning under first order conditions is best described by the Michaelis-Menten Rate
equations, integrating substrate concentration, Vmax and Km values. Figure 2-4 demonstrates
three general cases of chemical metabolism. Case A represents a flow-limited scenario, in which
the metabolic capacity of the liver is such that it quite efficiently metabolizes the chemical
delivered to the liver. Here, increases in the metabolic capacity will have no further effect on the
amount of chemical metabolized; increases in the amount of chemical metabolized will only
occur under conditions which increase the delivery of the chemical to the liver. Decreases in
metabolic capacity may decrease the amount of chemical metabolized, but such a decrease must
be rather marked. Case B represents a chemical which is well-metabolized: the amount of
chemical delivered to the liver is such that the metabolic capacity of the liver (a function of both
enzyme content and activity) is capable of metabolizing the amount of chemical delivered.
Increases in metabolic capacity will not increase the amount of chemical metabolized, decreases
in metabolic capacity are more likely to result in decreases in the amount of chemical
metabolized than in the flow-limited case. Case C represents a poorly-metabolized chemical, in
which delivery of the chemical to the liver exceeds the metabolic capacity of the liver; the
amount of metabolite formed is dependent on the metabolic capacity of the liver. Increases in
metabolic capacity will result in more metabolite formed; decreases in metabolic capacity will
reduce the amount of metabolite formed. This figure demonstrates the need to apply physiologic,
2-9
-------
anatomic and biochemical constraints as well as relevant exposure conditions to in vitro findings
of variance in xenobiotic metabolism parameters when determining their relevance to chemical
risk.
IMPACT OF STUDY DESIGN
Studies of the impact of differences in enzyme content or activity, including
polymorphisms, on risk can be carried out at several levels and, with respect to polymorphisms
in xenobiotic metabolizing enzymes, carry the most weight in this order:
1. Characterization of metabolic activity toward the substrate of interest in vivo
2. Characterization of metabolic activity toward the substrate of interest in vitro
3. Characterization of metabolic activity against a marker substrate in vivo
4. Characterization of variance in enzyme expression (qualitative or quantitative)
5. Characterization of metabolic activity against a marker substrate in vitro
6. Characterization of changes in mRNA content
7. Characterization of changes in DNA
While changes in DNA are critical underlying events, these changes may occur in
different positions, and with different effects. For instance, genetic alterations occurring in non-
transcribed portions of the genome (exons) are highly unlikely to have functional significance,
while changes in introns are more likely to have a functional effect. Likewise, changes in base
sequence may or may not change amino acid sequence. For these reasons, even when changes
are observed, but not characterized, in genes coding for xenobiotic-metabolizing enzymes, it is
critical to demonstrate their impact on metabolism. Just as xenobiotics can be either detoxicated
or bioactivated through metabolism, changes in the genes for these enzymes can increase or
decrease their activity. These changes are best evaluated with respect to their impact on
enzymatic activity, and that impact should be characterized fully, and with a keen eye to likely
exposure conditions for the substrate of interest.
2-10
-------
SUMMARY AND CONCLUSIONS
It is important to remember that, just as metabolism can alter PK, PK can alter
metabolism. Factors which increase or decrease the activity of metabolic enzymes can, but do
not always, alter tissue dosimetry, which should be the fundamental expression of dose for the
dose-response assessment phase of risk assessment. Extrahepatic factors, including age-
dependent differences in organ blood flows and tissue lipid content, which alter the delivery of
the substrate to the liver can, but do not always, change rates of bioactivation or detoxication. In
the consideration of age-related PK differences measured in vivo, one must give careful
consideration to substrate concentrations attained in the liver, extrahepatic factors and the age-
dependent expression of drug metabolizing enzymes. The platform offered by a physiologically
based pharmacokinetic model may offer the best mechanism through which to extrapolate in
vitro findings such as qualitative and quantitative changes in a drug metabolizing enzyme or
enzymatic activity to the intact human. These models can be constructed so that other age-
dependent extrahepatic factors can be modified to reflect requirements, however there are few
data sets available which address tissue partition coefficients in children. When considering
interindividual variance in enzyme content and activity as modulators of risk, that consideration
should be chemical-specific: metabolic variance in one direction can increase the susceptibility
to one chemical while decreasing the susceptibility to another.
REFERENCES
Houston, J.B. 1994. Utility of in vitro drug metabolism data in predicting in vivo metabolic
clearance. Biochem. Pharmacol. 47:1469-1479.
IPCS (International Programme on Chemical Safety of the World Health Organization). 2001.
Guidance Document for the Use of Data in Development of Chemical-Specific Adjustment
Factors (CSAFs) for Interspecies Differences and Human Variability in Dose/Concentration-
Response Assessment. Available at http://www.ipcsharmonize.org/csaf-intro.html
2-11
-------
Kedderis, G.L. 1997. Extrapolation of in vitro enzyme induction data to humans in vivo. Chem.
Biol. Interact. 107:109-121.
Lipscomb, J.C. and G.L. Kedderis. 2002. Incorporating human interindividual
biotransformation variance in health risk assessment. Sci. Total Environ. 288:13-21.
Lipscomb, J.C., C.M. Garrett and I.E. Snawder. 1997. Cytochrome P450-dependent metabolism
of trichloroethylene: Interindividual differences in humans. Toxicol. Appl. Pharmacol.
142:311-318.
Lipscomb, J.C., J.W. Fisher, P.D. Confer and J.Z. Byczkowski. 1998. In vitro to in vivo
extrapolation for trichloroethylene metabolism in humans. Toxicol. Appl. Pharmacol.
152:376-387.
Murray, FJ. and M.E. Andersen. 2001. Data-derived uncertainty factors: Boric acid (BA) as an
example. Hum. Ecol. Risk Assess. 7:125-138.
Nesaretnam K., D. Corcoran, R.R. Dils and P. Darbre. 1996. 3,4,3',4'-Tetrachlorobiphenyl acts
as an estrogen in vitro and in vivo. Mol. Endocrinol. 10:923-936.
Smallwood, C.L., J. Swartout and J.C. Lipscomb. 2001. Using data to replace default
uncertainty factors for boron reference dose. Presented at the annual meeting of the Society for
Risk Analysis, Seattle, WA, December.
Snawder, I.E. and J.C. Lipscomb. 2000. Interindividual variance of cytochrome P450 forms in
human hepatic hicrosomes: Correlation of individual forms with xenobiotic metabolism and
implications in risk assessment. Reg. Toxicol. Pharmacol. 32:200-209.
Soldin O.P., L.E. Braverman and S.H. Lamm. 2001. Perchlorate clinical pharmacology and
human health: A review. Ther. Drug. Monit. 23:316-331.
Umemura T., A. Takagi, K. Sai, R. Hasegawa and Y. Kurokawa. 1998. Oxidative DNA damage
and cell proliferation in kidneys of male and female rats during 13-weeks exposure to potassium
bromate (KBrO3). Arch. Toxicol. 72:264-269.
US EPA (U.S. Environmental Protection Agency). 1994. Methods for Derivation of Inhalation
Reference Concentrations and Application of Inhalation Dosimetry. Office of Research and
Development, Washington, DC. EPA/600/8-90/066F.
2-12
-------
NOTICE: This manuscript (NCEA-C-1348) has been reviewed and cleared for publication in
accordance with EPA/ORD policy; the author is grateful to Harlal Choudhury, Jeff Gearhart and
Gregory Kedderis for critical comments during manuscript development. The views contained
herein are those of the author and not necessarily those of the Agency. This paper contains
several key points made during a presentation to the workshop on Biological Variability in
Children and Implications for Environmental Risk Assessment: New Perspectives on the Roles
of Ethnicity, Race and Gender, held at the University of Maryland, March 2002. It was
subsequently published in the Journal of Children's Health: Lipscomb, J.C. 2003. How
differences in enzyme expression can translate into pharmacokinetic variance and susceptibility
to risk. J. Child. Health 1:189-202.
2-13
-------
TABLE 2-1
Effect of a 10-Fold Induction of Vmax on Hepatic Clearance over a Four-Log
Increase in Substrate Delivery
Vmaxa
5
50
5
50
5
50
5
50
5
50
n b
^Liver
0.005
0.005
0.05
0.05
0.5
0.5
5
5
50
50
Kmc
10
10
10
10
10
10
10
10
10
10
CLint
5
0.5
5
0.5
5
0.5
5
0.5
5
0.5
CLnep
0.00495
0.004995
0.045455
0.049505
0.25
0.454545
0.454545
2.5
0.49505
4.545455
Magnitude of
Change in CLHep
1.01
1.09
1.82
5.50
9.18
a Vmax in units of mg/hour/gram liver
b Quver in units of mg/hr
c Km in units of mg substrate/liter blood
d Intrinsic Clearance in units of milliliters/hr
e Hepatic Clearance in units of milliliters/hr
NOTE: Changes in enzymatic activity, resulting from changes in enzyme content (pmoles of
enzyme/gram tissue) and/or the specific activity of the enzyme (pmoles substrate metabolized /
time / pmol of enzyme) can alter chemical metabolism, regardless of substrate concentration.
However, the effect of changes in Km is dependent on substrate concentration, as demonstrated
in Figure 2-2. Similarly, the delivery of substrate to the liver is critical in vivo, and represents a
somewhat complicated and frequently under appreciated limitation in the extrapolation of in
vitro findings. The results in Table 2-1 demonstrate the impact of considering substrate delivery
in vivo (Qijver) and the choice of clearance model (Intrinsic versus Hepatic Clearance) when
extrapolating in vitro data on chemical metabolism. Intrinsic Clearance (Equation 2-2) does not
take into account substrate delivery and predicts a consistent 10-fold change in clearance
regardless of substrate delivery, while Hepatic Clearance (Equation 2-3) includes rate of
substrate delivery and demonstrates that clearance can be substantially influenced by substrate
delivery (see also Figures 2-3 and 2-4).
2-14
-------
Exposure Assessment
f v
Hazard pK pr> V-^ Estimation
Identification i \ f of Risk
Dose / Response
Relationship
PK*PD
PK*PD
UFA
Dose at Given
Response Level ""
in Animals
.V
Corresponding
Dose in the Gen'l
Human Population
Corresponding
Dose in the Sensitive
Human Population
FIGURE 2-1
Pharmacokinetics and the Risk Assessment Paradigm. Both pharmacokinetic (PK) and
pharmacodynamic (PD) data are useful in establishing the value for the two (UFA - animal-to-
human, and UFH - human interindividual) uncertainty factors used to extrapolate the dose-
response relationship from animals to sensitive humans. PK data are also useful in refining
human and animal exposures to reflect tissue dosimetry of the chemical species responsible for
the effect (PD).
2-15
-------
o. -
Increased Vmax
Decreased Km
Increased
Decreased Vmax
0
10
20
30
40
50
[S]
FIGURE 2-2
The Relationship Between Km Value and Substrate Concentration. This figure represents the
magnitude of change in the rate of the metabolic reaction following 10-fold increases and
decreases in Vmax and Km values, with substrate concentrations from 0.45 to 450 \iM.
Metabolic rates were predicted by the Michaelis-Menten reaction equation. Basal conditions
were: Km = 45 jiM; [s] = 45 jiM, Vmax = 1200 nmol es/min/mg protein. Vmax changes result in
proportionate changes in metabolic rate across substrate concentrations, while changes in Km do
not result in proportionate changes in metabolic rate, especially under saturating substrate
concentrations.
2-16
-------
Chemical
Availability
Metabolic Amount
Capacity Metabolized
Delivery^ [ ] J
Metabolite
[s]
\ \
Absorption into blood
Solubility in blood
Extrahepatic
distribution
Blood flow to liver
Enzyme content
Enzyme activity
Metabolic Rate =
Vmax * [s]
Km + [s]
FIGURE 2-3
The Relationship Between Substrate Delivery, Metabolic Capacity and the Amount of
Metabolite Formed. The "delivery" term is comprised of the amount of chemical present in the
blood, the rate of blood flow to the liver, and the equilibrium attained between blood and liver
tissue concentration of the substrate, which can be a function of transit time through the liver;
this may be affected by extrahepatic factors such as the fat compartment serving as a "sink" for
highly lipophilic xenobiotics. Metabolic Capacity is comprised of the enzyme content of the
liver and the activity of the enzyme, dictated by the Km and Vmax values of the enzyme for the
chemical substrate under investigation.
2-17
-------
Chemical Metabolic Amount
Availability Capacity Metabolized
(A)
Delivery
Metabolite
(B)
Delivery
> o
Metabolite
(C)
Delivery
Metabolite
FIGURE 2-4
The Impact of Substrate Delivery and Metabolic Capacity on the Formation of Metabolites in
Liver. This figure shows three of the possible cases of this relationship. Panel A demonstrates
the metabolism of a flow-limited chemical. Here, the metabolic capacity of the liver is very
large relative to the delivery of the substrate. In this condition, increases in enzyme activity (i.e.,
through induction) will not increase the amount of metabolite formed. Panel B demonstrates the
metabolism of a well-metabolized chemical. Here, the metabolic capacity of the liver is
sufficient to handle the amount of substrate delivered. Panel C demonstrates the metabolism of a
poorly metabolized chemical. Here, the metabolic capacity of the liver is such that it is not
sufficient to handle the amount of chemical available. Under this condition, increases in
metabolic capacity, due to an increase in Vmax and/or a decrease in Km, will result in higher
amounts of the chemical metabolized. Likewise, a decrease in Vmax or an increase in Km will
result in further reductions of metabolism.
2-18
-------
3. APPLICATION OF IN VITRO BIOTRANSFORMATION DATA AND
PHARMACOKINETIC MODELING TO RISK ASSESSMENT
ABSTRACT
The adverse biological effects of toxic substances are dependent upon the exposure
concentration and the duration of exposure. Pharmacokinetic models can quantitatively relate
the external concentration of a toxicant in the environment to the internal dose of the toxicant in
the target tissues of an exposed organism. The exposure concentration of a toxic substance is
usually not the same as the concentration of the active form of the toxicant that reaches the target
tissues following absorption, distribution, and biotransformation of the parent toxicant.
Biotransformation modulates the biological activity of chemicals through bioactivation and
detoxication pathways. Many toxicants require biotransformation to exert their adverse
biological effects. Considerable species differences in biotransformation and other
pharmacokinetic processes can make extrapolation of toxicity data from laboratory animals to
humans problematic. Additionally, interindividual differences in biotransformation among
human populations with diverse genetics and lifestyles can lead to considerable variability in the
bioactivation of toxic chemicals. Compartmental pharmacokinetic models of animals and
humans are needed to understand the quantitative relationships between chemical exposure and
target tissue dose as well as animal to human differences and interindividual differences in
human populations. The data-based compartmental pharmacokinetic models widely used in
clinical pharmacology have little utility for human health risk assessment because they cannot
extrapolate across dose route or species. Physiologically based pharmacokinetic (PBPK) models
allow such extrapolations because they are based on anatomy, physiology, and biochemistry. In
PBPK models, the compartments represent organs or groups of organs and the flows between
3-1
-------
compartments are actual blood flows. The concentration of a toxicant in a target tissue is a
function of the solubility of the toxicant in blood and tissues (partition coefficients), blood flow
into the tissue, metabolism of the toxicant in the tissue, and blood flow out of the tissue. The
appropriate degree of biochemical detail can be added to the PBPK models as needed.
Comparison of model simulations with experimental data provides a means of hypothesis testing
and model refinement. In vitro biotransformation data from studies with isolated liver cells or
subcellular fractions from animals or humans can be extrapolated to the intact organism based
upon protein content or cell number. In vitro biotransformation studies with human liver
preparations can provide quantitative data on human interindividual differences in chemical
bioactivation. These in vitro data must be integrated into physiological models to understand the
true impact of interindividual differences in chemical biotransformation on the target organ
bioactivation of chemical contaminants in air and drinking water.
TOXICOLOGY AND HUMAN HEALTH RISK ASSESSMENT
One of the most difficult problems in toxicology and human health risk assessment is the
use of appropriate animal models to predict the potential adverse effects of xenobiotic exposures
to human beings. Laboratory animals can differ from humans in both the pharmacokinetics and
pharmacodynamics of toxicant action. Differences in biochemistry and physiology can lead to
differences in the target tissue dose of a toxicant following exposure of animals and humans to
the same toxicant concentration. Significant species differences have been observed in the
metabolic activation and detoxication of chemical toxicants. Species differences have also been
observed in the pharmacodynamics of toxicant action at the cellular and molecular level. It is
very difficult to know in advance if the toxic response to a chemical observed in experimental
animals is predictive of the response in humans, or if humans would exhibit a toxicity not
3-2
-------
expressed in experimental animals. Thus, differences between species can be both qualitative
and quantitative in nature. With information on the mode or mechanism of toxicity (a descriptor
of the response), appropriate pharmacokinetic information can be collected and used to refine the
dose-response assessment so that concentrations of the toxic moiety in the target organ are
considered.
In the development of therapeutic compounds and the reassessment of approved drugs
and some occupationally important chemicals it is possible to derive pharmacokinetic data in
human subjects. These human pharmacokinetic data can readily be compared with those derived
from research animals. However, in the case of lexicologically uncharacterized compounds, and
compounds already thought to exert toxicity in humans, ethical considerations preclude the
exposure of humans. One approach to overcome these difficulties is the use of in vitro systems.
A general experimental approach is to develop predictive in vitro systems using animal tissues
and compare the results with those from in vivo studies using experimental animals. If the in
vitro system provides a reasonable prediction of the in vivo result (e.g., the kinetics of toxicant
bioactivation), then analogous in vitro systems can be developed using human tissues. The in
vitro human data can be integrated into pharmacokinetic models to predict the human target
tissue concentrations of toxicants and their metabolites following chemical exposure, along the
lines of the classic in vitro/in vivo parallelogram approach (Figure 3-1). In this article, we will
demonstrate the importance of pharmacokinetics in risk assessment and discuss important
experimental considerations for producing useful in vitro data, the extrapolation of that data to
humans through the use of pharmacokinetic models, and the incorporation of pharmacokinetic
information into human health risk assessments.
3-3
-------
PHYSIOLOGICALLY BASED PHARMACOKINETIC MODELS
Pharmacokinetic models have utility in toxicology because the adverse biological effects
of toxic substances are dependent upon the exposure concentration and the duration of exposure.
Pharmacokinetic models can quantitatively relate the external concentration of a toxicant in the
environment to the internal dose of the toxicant in the target tissues of an exposed organism. The
exposure concentration of a toxic substance is usually not the same as the concentration of the
active form of the toxicant that reaches the target tissues following absorption, distribution, and
biotransformation of the parent toxicant.
Biotransformation modulates the biological activity of chemicals through bioactivation
and detoxication pathways. Many toxicants require biotransformation to exert their adverse
biological effects. Considerable species differences in biotransformation and other
pharmacokinetic processes can make extrapolation of toxicity data from laboratory animals to
humans problematic. Species differences in biotransformation pathways can be qualitative or
quantitative. If a species exhibits a metabolic pathway that is not present in another species, that
difference is qualitative. If both species possess a given metabolic pathway but differ in the rate
or flux through that pathway, the difference is quantitative. Interindividual differences in
biotransformation exist within genetically diverse human populations and among segments of the
general population that are exposed to enzyme inducing agents through occupations or lifestyles
that can alter the expression and/or function of xenobiotic-metabolizing enzymes. Differences in
enzyme content and activity can lead to considerable variability in the pharmacokinetics and
bioactivation of toxic chemicals. Some enzymes exhibit genetic polymorphisms, leading to the
expression of aberrant enzyme forms with reduced activity. Many xenobiotic-metabolizing
enzymes are induced by exposure to chemicals in food, air, water, and the environment.
3-4
-------
Individual differences in genetics and exposure to inducing agents lead to considerable
variability in xenobiotic metabolizing enzymes in human populations.
Compartmental pharmacokinetic models of animals and humans are needed to understand
the quantitative relationships between chemical exposure and target tissue dose as well as animal
to human differences and interindividual differences in human populations. The data-based
compartmental pharmacokinetic models widely used in clinical pharmacology have little utility
for human health risk assessment because they cannot extrapolate across dose route or species.
Physiologically based pharmacokinetic (PBPK) models allow such extrapolations because they
are based on anatomy, physiology, and biochemistry (Clewell and Andersen, 1994). In PBPK
models, the compartments represent organs or groups of organs and the flows between
compartments are actual blood flows. The concentration of a toxicant in a target tissue is a
function of the solubility of the toxicant in blood and tissues (partition coefficients), blood flow
into the tissue, metabolism of the toxicant in the tissue, and blood flow out of the tissue. The
appropriate degree of biochemical detail can be added to the PBPK models as needed. The goal
of PBPK modeling is to define one set of parameters that describe the behavior of a chemical in
an animal. Comparison of model simulations with experimental data provides a means of
hypothesis testing and model refinement. The predictive power of PBPK models makes them
ideally suited for use in human health risk assessments.
EXTRAPOLATION OF IN VITRO DATA TO HUMANS
In vitro biotransformation data can be extrapolated to intact animals and humans because
the overall rate of an enzyme-catalyzed reaction is directly proportional to the total amount of
enzyme present in the system. Therefore data generated with subcellular fractions such as
microsomes or cytosols can be extrapolated to in vivo based on protein content. Studies with
3-5
-------
rodent and human liver microsomes have established the quantitative expression of microsomal
protein in liver tissue (Houston, 1994; Lipscomb et al., 2002c) and cytochrome P450 enzymes in
microsomal protein (Shimada et al., 1994; Snawder and Lipscomb, 2000). Data from intact
cellular systems such as hepatocytes can be extrapolated to in vivo based on cell number. There
are approximately 130xl06 hepatocytes per gram of mammalian liver (Seglen, 1976; Arias et al.,
1982). The liver is approximately 2.6% of human body weight (Snyder et al., 1992).
The isolation of intact and metabolically competent hepatocytes provides an excellent
model to study biotransformation in an in vitro system that most closely resembles the in vivo
setting. While subcellular fractions such as microsomes enrich the content of associated
enzymes, enzymes in isolated hepatocytes retain their physical relationship to the endogenous
lipid environment, the natural orientation toward soluble cofactors, and the unperturbed spatial
relationship of multienzyme systems such as cytochromes P450 and NADPH-cytochrome P450
reductase. Isolated hepatocytes provide the best intact cellular system to predict chemical
pharmacokinetics when used in a nutritive medium such as Williams' Medium E to maintain
biochemical homeostasis. The cells synthesize the necessary cofactors and the enzymes are
arrayed in membrane and cytoplasmic structures in the same manner as in the intact liver.
Although there is some limited mechanical damage to the hepatocytes upon isolation, the cells
repair the damage quickly upon incubation. In contrast to freshly isolated hepatocytes, which
maintain enzyme functions close to those of the intact liver (Billings et al., 1977), hepatocytes in
monolayer culture rapidly decrease expression of many xenobiotic metabolizing enzymes and do
not reflect the kinetics of xenobiotic metabolism in vivo (Sirica and Pitot, 1980). Immortalized
liver cell lines such as Hep G2 cells also have low metabolic capacity and are not appropriate
models for predicting chemical pharmacokinetics.
3-6
-------
We have previously shown that extrapolation of the kinetic parameters for the
bioactivation of the rodent hepatocarcinogen furan from freshly isolated rat hepatocyte
suspensions in vitro to whole animals in vivo accurately predicted the pharmacokinetics
(Kedderis et al., 1993). Studies with human liver microsomal protein and isolated hepatocytes
have been used to predict the in vivo pharmacokinetics of the industrial chemical and common
groundwater contaminant trichloroethylene (Lipscomb et al., 1997, 1998). Studies by Houston
(1994) showed that rat hepatocytes were more consistent than rat liver microsomes in accurately
predicting the intrinsic clearance of 11 therapeutic agents metabolized by various isoforms of
cytochromes P450. Isolated rat hepatocytes and rat hepatic microsomes were consistently more
accurate in predicting intrinsic clearance than precision cut liver slices (Worboys et al., 1996).
Drug concentration gradients within the liver slice do not allow every cell in the slice to
participate in metabolism, indicating that this in vitro system is not appropriate for estimation of
in vivo metabolic parameters.
One important factor in extrapolating in vitro data to predict in vivo pharmacokinetics is
knowledge of the actual concentration of the xenobiotic in the cell suspension or incubation from
measurement of the partition coefficient. Most xenobiotics are not very water-soluble, and the
nominal concentration added is not always the concentration in the medium. Any other
kinetically significant component that removes substrate from the system, such as specific
protein binding or chemical reactivity, needs to be accounted for in the kinetic description of the
in vitro system. Care should be taken that volatile chemicals are incubated in sealed flasks and
that transfer rates between the vapor phase and aqueous phase are taken into account. This can
be accomplished using a simple two-compartment kinetic model of the incubation system
(Kedderis et al., 1993).
3-7
-------
In order to extrapolate in vitro kinetic data to whole animals, the kinetic mechanisms of
the enzymes catalyzing the metabolic reactions need to be known (or assumed). Most
cytochrome P450-catalyzed reactions are known to follow an ordered sequential mechanism
displaying saturable Michaelis-Menten kinetics (Hollenberg, 1992):
Rate = (Vmax * [S]) / (KM + [S]) (3-1)
where Vmax is the maximal rate of the reaction at infinite substrate concentration, [S] is the
substrate concentration, and KM is the Michaelis constant for the reaction. The Michaelis-
Menten Equation 3-1 indicates that the initial velocity of the reaction will increase hyperbolically
as a function of substrate concentration. The Vmax is a horizontal tangent to the top (saturated)
part of the curve, while the tangent to the initial linear portion of the hyperbolic curve is the
initial rate of the reaction, V/K. The V/K is the pseudo-first-order rate constant for the reaction
at low substrate concentrations. The point where these two tangents intersect corresponds to the
KM (Northrop, 1983). The KM is defined as the substrate concentration that gives one-half the
Vmax-
The in vitro kinetic data can be incorporated into PBPK models after rearrangement to
the appropriate units. The units of the Vmax value need to be expressed relative to body weight
rather than per mg protein or cell number. As described above, these relationships have been
defined. Values of KM, the substrate concentration that gives one-half Vmax, have units of
concentration and can be used directly if the solubilities of the chemicals in the in vitro system
were measured. The PBPK model can be used to simulate a variety of toxicant exposure
scenarios, including workplace environment, inhalation exposure, drinking water exposure,
exposure in food, or dermal exposure.
3-8
-------
INCORPORATION OF PHARMACOKINETICS INTO RISK ASSESSMENTS
The development of PBPK modeling techniques was undertaken to better describe dose
in terms of tissue concentrations for use in human health risk assessments for methylene chloride
(Andersen et al., 1987). This early work recognized and incorporated species differences in
metabolic capacity and a bifurcation in the metabolic pathway for methylene chloride, and
represented a significant advance in risk assessment theory and practice. Since that time, several
other works have addressed the employment of pharmacokinetic information in risk assessment,
including the USEPA's guidance on the derivation of Reference Concentrations (RfC) for
inhaled substances (US EPA, 1994). A recent publication by Clewell et al. (2002) breaks down
the application of pharmacokinetic investigations into risk assessment into six steps, summarized
as follows:
1. Evaluate toxicity information to understand the mode of action for the critical effect in
the critical species;
2. Identify or develop a PBPK model parameterized for the test species;
3. Employ the PBPK model to calculate the relevant dose metric for the dose at the required
dose level (i.e., BMD orNOAEL);
4. Define the uncertainty factors to be employed to determine the numerical value for the
corresponding dose metric in humans;
5. Develop or identify an appropriately parameterized human PBPK model, and employ that
model to determine the exposure conditions which result in the value for the dose metric
determined in step 4;
6. Consider the dose metrics evaluated, and give priority to the dose metric which provides
the most plausible basis for estimating the biologically effective dose in humans; when
unclear, select the dose metric which is the most health-protective (conservative).
The incorporation of pharmacokinetics into human health risk assessment is essential for
the understanding of species differences in toxicant bioactivation when human risk is
extrapolated from animal models and for understanding the impact of interindividual differences
3-9
-------
in toxicant bioactivation among humans. Measurement of differences in toxicant bioactivation
by human tissue preparations is an important beginning to understanding pharmacokinetic
factors that are related to risk, but the in vitro data must be integrated into PBPK models to
understand the consequences of kinetic variability in the context of the whole organism. Our
studies with furan and trichloroethylene illustrate these points.
The accurate prediction of furan pharmacokinetics from kinetic studies with isolated
F-344 rat hepatocytes (Kedderis et al., 1993) suggested that kinetic studies with hepatocytes
from other species could be used to develop species-specific pharmacokinetic models for furan.
Therefore we determined the kinetics of furan bioactivation by cytochrome P450 2E1 (CYP2E1)
in hepatocytes from B6C3F1 mice and humans (Kedderis and Held, 1996). The KM values for
furan oxidation were in the low micromolar range for hepatocytes from all three species.
Hepatocytes from male mice oxidized furan at a greater rate than rat or human hepatocytes.
Hepatocytes from three different human liver donors oxidized furan at rates equal to or greater
than those of rat hepatocytes. Note that these results are not predicted by allometric scaling of
the rat or mouse data to humans since allometry predicts slower rates in larger mammals
(Mordenti, 1986). This underscores the importance of using human tissue samples to predict the
human bioactivation of toxicants. A greater than 2-fold variation in Vmax was observed among
the three preparations of human hepatocytes (Kedderis and Held, 1996). The human liver
preparations exhibiting the highest bioactivation rates were from donors who died in automobile
accidents involving alcohol consumption, consistent with the interpretation that the higher Vmax
values were due to the induction of CYP2E1 by ethanol (Perrot et al., 1989).
The kinetic data from hepatocytes were used to develop PBPK models for furan in rats,
mice, and humans (Kedderis and Held, 1996). Simulation of inhalation exposure to 10 ppm
3-10
-------
furan for 4 hours indicated significant species differences in the amount of furan absorbed. The
absorbed dose of furan (mg/kg; inhaled minus exhaled divided by body weight) and the
integrated exposure of the liver to the toxic metabolite were approximately 3.5-fold and 10-fold
greater in rats and mice, respectively, than in humans following the same inhalation exposure.
The reason for this species difference is that humans are larger and physiologically slower than
mice or rats (Mordenti, 1986). The volatile toxicant furan is metered into the blood stream via
the breathing rate and distributed throughout the organism at rates that are a function of body
size. Thus the inhalation exposure concentration of a toxicant is clearly not an appropriate
measure of the dose to the organism. In the case of furan, comparing the absorbed dose (mg/kg)
or the target organ (liver) exposure to the toxic metabolite among species is more appropriate.
These concepts should be applied when assessing interspecies differences to inhaled toxicants,
particularly when animal data are used to estimate human health risks from chemical exposure.
Simulations with the dosimetry models showed that steady-state blood concentrations of
furan were achieved by approximately 1 hour after inhalation exposure to 10 ppm and were
similar among species (0.7-0.8 jiM). Comparison of the initial rates of furan oxidation with the
rate of hepatic blood flow for rats, mice, and humans indicated that the rate of furan oxidation
was approximately 13- to 37-fold higher than the rate of furan delivery to the liver via blood
flow (Kedderis and Held, 1996). These results indicate that the bioactivation of furan is limited
by hepatic blood flow. Simulations of the liver concentration of the toxic metabolite of furan as
a function of the furan exposure concentration indicated that furan bioactivation in humans was
limited by hepatic blood flow through at least 300 ppm. The hepatic blood flow limitation in
rodents and humans was also evident following oral bolus administration of furan through
approximately 2 mg/kg (Kedderis, 1997). One important consequence of the limitation of
3-11
-------
bioactivation by hepatic blood flow is that increases in Vmax due to CYP2E1 induction have
little or no effect on the amount of the toxic metabolite formed in the liver. Pharmacokinetic
analyses of the bioactivation of several other hazardous chemical air pollutants also indicated a
hepatic blood flow limitation of bioactivation (Kedderis, 1997).
The hepatotoxicity of trichloroethylene (TCE) is mediated by acid metabolites formed by
CYP2E1 oxidation (Bull, 2000), and differences in CYP2E1 expression have been hypothesized
to affect susceptibility to liver injury by TCE. Therefore the contribution of variance in CYP2E1
content and activity on the risk of hepatotoxic injury among adult humans was investigated
(Lipscomb et al., 2002a,b). New and existing data sets describing the microsomal protein
content of human liver (Lipscomb et al., 2002c), the CYP2E1 content of human liver microsomal
protein (Snawder and Lipscomb, 2000), and the in vitro Vmax for TCE oxidation by humans
(Lipscomb et al., 1997) from 60 human liver samples were combined and subjected to statistical
analysis. The data were log-normally distributed. Analysis by the method of moments indicated
that the 5th and 95th percentiles of the distribution in human liver (TCE oxidized per minute per
gram liver) differed by approximately 6-fold. These values were converted to mg TCE
oxidized/hr/kg body mass and incorporated in a human PBPK model for TCE (Allen and Fisher,
1993). Simulations of 8 hour inhalation exposure to 50 ppm (the TLV) and oral exposure to 5 jig
TCE/L in 2L drinking water (the MCL) showed that the amount of TCE oxidized in the liver
differed by 2% or less even though the distribution of metabolic capacity varied 6-fold. The
results are presented conceptually in Figure 3-2 and shown in Table 3-1. These results indicate
that differences in CYP2E1 enzyme expression among the central 90% of the adult human
population account for only approximately 2% of the variance in the risk-relevant
pharmacokinetic outcome for TCE-mediated liver injury (amount of TCE oxidized in the liver).
3-12
-------
These data indicate that physiological processes such as hepatic blood flow limit the full impact
of the differences in CYP2E1 activity toward TCE mediating the formation of toxic metabolites
(Lipscomb et al., 2002c). Integration of in vitro metabolism information into PBPK models may
reduce the uncertainties associated with risk contributions of variance in enzyme expression and
the uncertainty factors that represent pharmacokinetic variance. The in vitro data must be
evaluated in the context of the intact animal. A framework describing this process has been
previously published (Lipscomb and Kedderis, 2002).
REFERENCES
Allen, B.C. and J.W. Fisher. 1993. Pharmacokinetic modeling of trichloroethylene and
trichloroacetic acid in humans. Risk Anal. 13(l):71-86.
Andersen, M.E., HJ. Clewell III, M.L. Gargas, F.A. Smith and R.H. Reitz. 1987.
Physiologically based pharmacokinetics and the risk assessment process for methylene chloride.
Toxicol. Appl. Pharmacol. 87:185-205.
Arias, I.M., H. Popper, D. Schachter and D.A. Shafritz. 1982. The Liver: Biology and
Pathobiology. Raven Press, New York, NY.
Billings, R.E., R.E. McMahon, J. Ashmore and S.R. Wagle. 1977. The metabolism of drugs in
isolated rat hepatocytes: A comparison with in vivo drug metabolism and drug metabolism in
subcellular liver fractions. Drug Metab. Dispos. 5:518-526.
Bull, RJ. 2000. Mode of action of liver tumor induction by trichloroethylene and its
metabolites, trichloroacetate and dichloroacetate. Environ. Health Perspect. 108(Suppl 2):
241-259
Clewell, HJ. and M.E. Andersen. 1994. Physiologically-based pharmacokinetic modeling and
bioactivation of xenobiotics. Toxicol. Ind. Health. 10(1-2): 1-24.
Clewell, H.J., M.E. Anderson and H.A. Barton. 2002. A consistent approach for the application
of pharmacokinetic modeling in cancer and noncancer risk assessment. Environ. Health
Perspect. 110(l):85-93.
Hollenberg, P.P. 1992. Mechanisms of cytochrome P450 and peroxidase-catalyzed xenobiotic
metabolism. FASEB J. 6(2):686-694.
Houston, J.B. 1994. Utility of in vitro drug metabolism data in predicting in vivo metabolic
clearance. Biochem. Pharmacol. 47:1469-1479.
3-13
-------
Kedderis, G.L. 1997. Extrapolation of in vitro enzyme induction data to humans in vivo. Chem
Biol Interact. 107:109-121.
Kedderis, G.L. and S.D. Held. 1996. Prediction of furan pharmacokinetics from hepatocyte
studies: Comparison of bioactivation and hepatic dosimetry in rats, mice, and humans. Toxicol.
Appl. Pharmacol. 140:124-130.
Kedderis, G.L., M.A. Carfagna, S.D. Held, R. Batra, I.E. Murphy and M.L. Gargas. 1993.
Kinetic analysis of furan biotransformation by F-344 rats in vivo and in vitro. Toxicol. Appl.
Pharmacol. 123:274-282.
Lipscomb, J.C. and G.L. Kedderis. 2002. Incorporating human interindividual
biotransformation variance in health risk assessment. Sci. Total Environ. 288:13-21.
Lipscomb, J.C., C.M. Garrett and I.E. Snawder. 1997. Cytochrome P450-dependent metabolism
of trichloroethylene: Interindividual differences in humans. Toxicol. Appl. Pharmacol.
142:311-318.
Lipscomb, J.C., J.W. Fisher, P.D. Confer and J.Z. Byczkowski. 1998. In vitro to in vivo
extrapolation for trichloroethylene metabolism in humans. Toxicol. Appl. Pharmacol.
152:376-387.
Lipscomb, J.C., L.K. Teuschler, J. Swartout and G.L. Kedderis. 2002a. Incorporation of human
interindividual enzyme expression and biotransformation variance into human health risk
assessments. Toxicol. Sci. 66(Supp. 1): 154-155.
Lipscomb J.C., L.K. Teuschler, J.C. Swartout, D. Popken, T. Cox and G.L. Kedderis. 2002b.
The impact of cytochrome P450 2El-dependent metabolic variance on a risk relevant
pharmacokinetic outcome in humans. Risk Anal, (submitted for publication)
Lipscomb, J.C., L.K. Teuschler and J.C. Swartout. 2003. Variance of microsomal protein and
cytochrome P450 2E1 and 3 A forms in adult human liver. Tox. Mech. Meth. 13:45-51.
Mordenti, J. 1986. Man versus beast: Pharmacokinetic scaling in mammals. J. Pharmacol. Sci.
75:1028-1040.
Northrop, D.B. 1983. Fitting enzyme-kinetic data to V/K. Analyt. Biochem. 132:457-461.
Perrot, N., B. Nalpas, C.S. Yang and P.H. Beaune. 1989. Modulation of cytochrome P450
isozymes in human liver by ethanol and drug intake. Eur. J. Clin. Invest. 19(6):549-555.
Seglen, P.O. 1976. Preparation of isolated rat liver cells. Meth. Cell Biol. 13:29-83.
3-14
-------
Shimada, T., H. Yamazaki, M. Mimura, Y. Inui and P.P. Guengerich. 1994. Interindividual
variations in human liver cytochrome P450 enzymes involved in the oxidation of drugs,
carcinogens and toxic chemicals: Studies with liver microsomes of 30 Japanese and 30
Caucasians. J. Pharmacol. Expt. Ther. 270:414-423.
Sirica, A.E. and H.C. Pitot. 1979. Drug metabolism and effects of carcinogens in cultured
hepatic cells. Pharmacol. Rev. 31(3):205-228.
Snawder, I.E. and J.C. Lipscomb. 2000. Interindividual variance of cytochrome P450 forms in
human hepatic microsomes: Correlation of individual forms with xenobiotic metabolism and
implications in risk assessment. Reg. Toxicol. Pharmacol. 32:200-209.
Snyder, W.S., MJ. Cook, E.S. Nasset, L.R. Karhausen, G.P. Howells and I.H. Tipton. 1992.
Report of the Task Group on Reference Man. Pergamon Press, New York, NY.
US EPA. 1994. Methods for Derivation of Inhalation Reference Concentrations and
Application of Inhalation Dosimetry. Office of Research and Development, Washington, DC.
EPA/600/8-90/066F.
Worboys, P.D., B. Brennan, A. Bradbury and J.B. Houston. 1996. Metabolite kinetics of
ondansetron in rat. Comparison of hepatic microsomes, isolated hepatocytes and liver slices,
with in vivo disposition. Xenobiotica. 26(9):897-907.
NOTICE: The views expressed in this paper are those of the individual authors and do not
necessarily reflect the views and policies of the US Environmental Protection Agency (EPA).
This manuscript has been reviewed in accordance with EPA/ORD policy and approved for
public release. This research was supported in part by US EPA/NCEA cooperative agreement
CR828047-01-0. This paper was presented at the Spring Toxicology Conference, April 2002,
Cincinnati, Ohio, and published in Toxicology and Industrial Health: Kedderis, G.L. and
Lipscomb, J.C. 2003. Application of in vitro biotransformation data and pharmacokinetic
modeling to risk assessment. Toxicol. Ind. Health. 17:315-321.
3-15
-------
TABLE 3-1
Variance in Human Hepatic CYP2E1 -Dependent TCE Oxidative Capacity and the Amount of
TCE Oxidized"
Liver Metabolites (mg/L)
Oxidation Rate
5th Percentile
(approx. 7 mg/hr/kg)
95th Percentile
(approx. 40 mg/hr/kg)
Magnitude of Difference
Inhalation
258.3
264.9
2%
Oralc
5.4
5.5
2%
aData from Lipscomb et al. (2002a,b)
bSimulations of 8 hr exposure to TCE (50 ppm) by a 70 kg male human
Simulations of exposure to TCE in drinking water (5 |ig/L; 2 L) over 24 hours by a 70 kg male
human
3-16
-------
Findings in Humans
in vitro
Findings in
Humans
Findings in
Animals
Findings in
animals
FIGURE 3-1
The in vitro/in vivo Parallelogram Approach. In this scheme, data are derived in vitro in test
animal species and considerations including those imposed by anatomy and physiology are used
to extrapolate those findings to the in vivo setting in the animal. Likewise, information on the
underlying biochemistry for the event are collected from animals and humans in vitro, and serve
as the basis upon which to modify expectations from in vitro studies with animal tissue
preparations to predict the results obtained from studies of human tissues in vitro. Lessons
learned from the in vitro to in vivo extrapolation within the species and lessons learned from the
extrapolation of results from in vitro studies across species are combined to better enable the
prediction of the in vivo consequences in the human.
3-17
-------
o
C
_o
"^
o
5
_o
"^
-------
4. THE IMPACT OF CYTOCHROME P450 2E1-DEPENDENT METABOLIC
VARIANCE ON A RISK-RELEVANT PHARMACOKINETIC
OUTCOME IN HUMANS
ABSTRACT
Risk assessments include assumptions about sensitive subpopulations, such as the
fraction of the general population that is sensitive and the extent that biochemical or
physiological attributes influence sensitivity. Uncertainty factors (UF) account for both
pharmacokinetic (PK) and pharmacodynamic (PD) components, allowing the inclusion of risk-
relevant information to replace default assumptions about PK and PD variance (uncertainty).
Large numbers of human organ donor samples and recent advances in methods to extrapolate in
vitro enzyme expression and activity data to the intact human enable the investigation of the
impact of PK variability on human susceptibility. The hepatotoxicity of trichloroethylene (TCE)
is mediated by acid metabolites formed by cytochrome P450 2E1 (CYP2E1) oxidation, and
differences in the CYP2E1 expression are hypothesized to affect susceptibility to TCE's liver
injury. This study was designed specifically to examine the contribution of statistically
quantified variance in enzyme content and activity on the risk of hepatotoxic injury among adult
humans. We combined data sets describing 1) the microsomal protein content of human liver, 2)
the CYP2E1 content of human liver microsomal protein, and 3) the in vitro Vmax for TCE
oxidation by humans. The 5th and 95th percentiles of the resulting distribution (TCE oxidized per
minute per gram liver) differed by approximately 6-fold. These values were converted to mg
TCE oxidized/hr/kg body mass and incorporated in a human PBPK model. Simulations of 8 hr
inhalation exposure to 50 ppm and oral exposure to 5 jig TCE/L in 2L drinking water showed
that the amount of TCE oxidized in the liver differs by 2% or less under extreme values of
4-1
-------
CYP2E1 expression and activity (here, selected as the 5th and 95th percentiles of the resulting
distribution). This indicates that differences in enzyme expression and TCE oxidation among the
central 90% of the adult human population account for approximately 2% of the difference in
production of the risk-relevant PK outcome for TCE-mediated liver injury. Integration of in
vitro metabolism information into physiological models may reduce the uncertainties associated
with risk contributions of differences in enzyme expression and the UF that represent PK
variability.
INTRODUCTION
Traditional non-cancer risk assessments in the US EPA apply uncertainty factors to
extrapolate the measures of effects between animals and humans, and among humans. These two
factors (UFA and UFH, respectively), may be further subdivided into their respective
pharmacodynamic (PD) and pharmacokinetic (PK) components (Bogdanffy and Jarabek, 1995;
Jarabek, 1995; Bogdanffy et al., 1999). WHO (1998) and the International Programme on
Chemical Safety (IPCS, 1994, 1998) have provided guidance and application of the separate
consideration of PD and PK, and the US EPA has also separately quantified PD and PK
variability in the UFA applied to reference concentration (RfC) values and reference dose (RfD)
values, where each has been ascribed a value of one-half log (1005, or 3.16) (U.S. EPA, 1999,
2000, 2001). In addition, PD and PK components of UFH for RfC values have also been
separately considered for some substances such as methyl methacrylate (U.S. EPA, 1998).
Studies with humans can be conducted to assess the pharmacodynamics or pharmacokinetics of
environmental or occupational chemicals. While human clinical trials can assess the PK and PD
of potential therapeutic substances, human studies with potentially toxic environmental or
occupational chemicals are not usually conducted over concentration ranges known or predicted
4-2
-------
to result in adverse effects. The limited information available from human studies with
environmental chemicals provides critical (but often limited) information which can be extended
by in vitro studies using preparations from human tissues. Care must be taken so that the in vitro
investigations are focused on risk-relevant endpoints, and are conducted with the relevant tissues,
tissue preparations, and chemical concentrations. It is critical that the concentrations used for in
vitro studies are within the range of tissue concentrations observed or predicted in vivo in
humans following chemical exposure. Studies with human subjects or human tissue preparations
in vitro can identify variability in PK outcomes such as the blood concentrations of parent
chemical and metabolites or the rates of metabolite production or elimination. When these PK
outcomes overlap with the PK outcomes most linked to risk (verified by results from mode of
action and PK studies with research animals), then additional information on the variability of
these PK outcomes will advance our understanding of susceptibility, and provide information
with which to replace default values for uncertainty in extrapolations of risk. Although data
from multiple human subjects may seem preferable as the basis from which to determine human
PK variability, those data seldom exist, and when they do the data usually offer little information
on risk-relevant PK outcomes such as target tissue dosimetry. Physiologically based PK (PBPK)
models allow the application of physiologic and anatomic constraints to clarify the linkage
between external concentrations and target tissue concentrations (Figure 4-1) and offer a
mechanism through which information obtained ex vivo or in vitro may be evaluated in proper
context. Results from PBPK model simulations of relevant exposure scenarios provide a useful
approach for estimation of PK variability between research animals and humans and among
humans when other data are limiting. This technique offers the opportunity to extrapolate
concentrations of bioactive chemical moieties in target tissues across doses and routes of
4-3
-------
exposure. The inclusion of data derived in vitro through the exposure of human tissue
preparations offers an advance over exposing humans to noxious agents, and several studies have
demonstrated the applicability of in vitro findings in refining PBPK models.
While in vitro measurements of specific biochemical reactions from multiple human
samples can yield qualitatively valuable data on human variance, they must be tied to human
anatomy and physiology, and the impact of their variance evaluated under real exposure
scenarios to be of quantitative value. This study was constructed on the framework for
extrapolation of in vitro metabolic rate information and PBPK model incorporation previously
suggested (Lipscomb and Kedderis, 2002).
Enzymes are protein molecules that catalyze chemical reactions (Lehninger, 1975). Over
100 years ago, the study of enzymes and their properties demonstrated that the rates of enzyme-
catalyzed reactions are directly proportional to the total enzyme present in the system
(Lehninger, 1975; Segel, 1975). This property of enzymes provides the basis for extrapolation of
in vitro biotransformation data to whole animals and humans (Kedderis, 1997). Therefore data
generated with subcellular fractions such as microsomes or cytosols can be extrapolated to in
vivo based on protein content (Snawder and Lipscomb, 2000). Human liver is approximately
2.6% of body weight (Arias et al., 1982).
In addition to enzyme content or activity and organ weight, the kinetic mechanism of the
enzyme (the comings and goings of substrates and products) needs to be taken into account to
extrapolate in vitro data to whole animals or humans (Kedderis, 1997). The CYP2E1-catalyzed
oxidation of TCE follows Michaelis-Menten saturation kinetics (Lipscomb et al., 1997):
v = (Vmax*[S])/(KM + [S]) (4-1)
4-4
-------
where v is the initial velocity of the reaction, Vmax is the maximal rate of the reaction at infinite
substrate concentration, [S] is the substrate concentration, and KM is the Michaelis constant for
the reaction. The Michaelis-Menten Equation 4-1 indicates that the initial velocity of the
reaction will increase hyperbolically as a function of substrate concentration. The Vmax is a
horizontal tangent to the top (saturated) part of the curve, while the tangent to the initial linear
portion of the hyperbolic curve is the initial rate of the reaction, V/K. The V/K is the pseudo-
first-order rate constant for the reaction at low substrate concentrations. The point where these
two tangents intersect corresponds to the KM (Northrop, 1983). The KM is defined as the
substrate concentration that gives one-half the Vmax- The KM for each substrate is an inherent
property of the enzyme (Lehninger, 1975). A lower KM for one substrate compared to a second
substrate indicates that the first substrate has a more rapid initial rate (V/K) of metabolism. The
value of KM can vary with the structure of the enzyme; for example, in the polymorphism of the
CYP2D6-mediated oxidation of debrisoquine and related drugs (Gut et al., 1986).
Experimentally, KM can vary with pH, temperature and ionic strength in vitro. Therefore in vitro
kinetic measurements intended for extrapolation to intact animals and humans should be done
under experimental conditions mirroring the in vivo situation as closely as possible (Kedderis,
1997).
The in vitro kinetic data can be incorporated into PBPK models after rearrangement of
the Vmax to the appropriate units. Values of KM have units of concentration and can be used
directly if the solubility of the chemical in the in vitro system is known. Incorporation of the
extrapolated kinetic parameters into a PBPK model for humans allows prediction of target tissue
dosimetry following a variety of exposure scenarios.
4-5
-------
We have adapted an existing PBPK model to predict the difference among humans in the
risk-relevant PK outcome for the hepatotoxicity of trichloroethylene (TCE) under the conditions
of human variance in the rates of TCE oxidation. We focused on the hepatotoxicity of TCE
because: 1) the PK of TCE have been characterized and modeled in research animals and
humans (reviewed in Fisher, 2000); 2) more than 95% of an absorbed dose of TCE is oxidized in
research animals and humans (Bloemen et al., 2001); 3) CYP2E1 has been demonstrated to be
the enzyme responsible for the oxidation of TCE in research animals and humans and in vitro
preparations at low concentrations (Nakajima et al., 1988); 4) the hepatotoxicity of TCE has been
demonstrated to depend on acid metabolite(s) derived from oxidative metabolites of TCE
(Barton and Clewell, 2000; Bull, 2000); 5) the CYP2E1-mediated oxidation of TCE is rate
limiting in the further formation of acid metabolite(s) (Ikeda et al., 1980); 6) the expression of
CYP2E1 is modulated by genetic, environmental and lifestyle factors; and 7) large numbers of
human liver tissue and human liver tissue preparations are currently available in contrast to
preparations and tissues from other human organs. The results on the variance of the distribution
of CYP2E1 in adult human liver will be especially applicable to other environmental
contaminants which are also substrates for this enzyme.
The investigation was accomplished by first characterizing the variance about the
CYP2E1 mediated oxidation of TCE among human samples in vitro, second by quantifying the
variance of human hepatic CYP2E1 content, and third by extrapolating the bounds of variance of
TCE oxidation among the adult human population to a human PBPK model. Two separate
statistical analyses were conducted - one based on convenience and one based on technical
accuracy. The amount of TCE metabolized (oxidized) in the liver was simulated as a dose
surrogate for the hepatotoxicity of TCE. The goal of the present investigation was to quantify
4-6
-------
the variability in a risk-relevant PK outcome for the hepatotoxicity of TCE, and to demonstrate
the usefulness of advanced data on human biochemical individuality in quantifying the
variability of risk-relevant PK outcomes for inclusion in risk assessments. Data on the
distribution of CYP2E1 in the intact liver has not been used to estimate the degree of
susceptibility to risk for metabolized chemicals. We hypothesized that the degree of natural
variance in human hepatic levels of CYP2E1 would result in similar differences in the oxidation
of TCE in the intact human.
METHODS
Several sets of information describing or based on microsomal protein (MSP) were
collected for assimilation, extrapolation and incorporation into a PBPK model. The objective of
the extrapolation was to transition expression of apparent Vmax from units of "pmoles TCE
oxidized/min/mg MSP" to units of conventional PBPK modeling, mg/hr/kg body mass.
Necessary data were compiled from multiple sources, and used to describe the various
parameters, whose distributions were analyzed and combined. Table 4-1 demonstrates the
relationship between those data sets and parameters. TCE is oxidized by CYP2E1, and that
metabolic rate had been previously measured and presented in units of MSP (nmol/min/mg
MSP). Thus, the need to express apparent Vmax as pmol TCE oxidized/min/pmol CYP2E1.
CYP2E1 is isolated in MSP, thus the need to quantify CYP2E1 in MSP (pmoles CYP2El/mg
MSP). Because MSP is one constituent of liver, the amount of MSP per gram liver tissue (mg
MSP/gram liver) needed computing. These data facilitate the extrapolation of in vitro metabolic
capacity (comprised of enzyme activity and enzyme content) to the intact liver. Units of the
calculation cancel (pmol TCE oxidized/min/pmol CYP2E1) * (pmol CYP2El/mg MSP) * (mg
MSP/gram liver), leaving units of pmol TCE oxidized/min/gram liver. Correction for molecular
4-7
-------
weight of TCE, 60 min/hr and assumptions about the fractional composition of body mass
attributed to the liver compartment (liver = BW*0.026) results in units of mg TCE
oxidized/hr/kg.
Human Samples and Quantification of CYP Proteins. Both prepared MSP and intact liver
tissue were obtained for this investigation from various sources (Human Cell Culture Center,
Laurel, MD; International Institute for the Advancement of Medicine, Exton, PA; Vitron,
Tucson, AZ; Tissue Transformation Technologies, Inc., Edison, NJ). All tissues and
preparations were derived from adult human organ donors who were devoid of antibodies
directed against infectious diseases. The MSP content of CYP2E1 and other CYP forms was
previously investigated and reported for 40 donors (Snawder and Lipscomb, 2000). In the
present analysis, 20 samples of intact tissue were obtained, and MSP prepared via the method of
Guengerich (1989) (Figure 4-2). CYP2E1 content of aliquots of (post 100 x g) homogenate
protein and MSP were determined by enzyme-linked immunosorbent assay (ELISA) following
the method of Snawder and Lipscomb (2000).
Distribution of CYP2E1 to Human Hepatic MSP. Data on the CYP2E1 content from 40
samples of human hepatic MSP were available from Snawder and Lipscomb (2000), and data
derived from an additional 20 samples of human hepatic MSP were combined to yield a total
sample of 60 adult human organ donors for which data on the CYP2E1 content of MSP
(CYP2Msp, parameter B, data set 1) were available (Lipscomb et al., 2003) (Table 4-2, described
in the following section). Of this set of 60 samples of MSP (representing 60 organ donors), 15
were used to estimate the in vitro metabolic parameters for TCE and CYP2E1 content of MSP;
45 were subjected only to the determination of CYP2E1 in MSP (and 20 of that 45 were paired
with liver homogenate to determine the MSP content of liver.
4-8
-------
Estimation of Proteins in Intact Liver. In this analysis, twenty samples of intact liver tissue
were assayed (Table 4-2). The total amount of protein (CYP and non-CYP; microsomal and
cytosolic proteins) in intact liver (PROuv) was empirically determined based on the protein
content of the post 100 x g liver supernatant, after correcting for volume according to Equation
4-2. It was assumed that no protein was lost during the sedimentation of nuclei and debris at
100 xg.
(mg protein/ml homogenate) * (ml homogenate/gram tissue) (4-2)
= (mg homogenate protein/gram tissue)
The content of CYP2E1 in total hepatic protein (CYP2Pro) and in MSP (CYP2Msp;
parameter B) was determined so that a measure of the liver content of MSP (MSPuv; parameter
C) could be derived. The content of CYP2E1 in liver (pmoles CYP2El/gram liver; CYP2Liv;
parameter A) was derived empirically by combining two data sets (PROuv CYPpRo), described in
Equation 4-3, and was estimated via the statistical method of moments (Analysis via Method of
Moments section) and by computational statistics (Analysis via Computational Statistics
section), below. For the 20 individual organ donors, the separate amounts of CYP2E1 per gram
liver was empirically determined according to the following equation:
(pmol CYP2El/mg homogenate protein)*(mg homogenate protein/gram tissue) (4-3)
= (pmol CYP2El/gram tissue)
The amount of MSP per gram liver was estimated according to Equation 4-4. This is the data set
(MSPiJv; parameter C, data set 2) which will be combined with information on the distribution of
CYP2E1 to MSP (CYP2Msp; parameter B, data set 1) to determine the distribution of CYP2E1 to
the intact liver (CYPLiv; parameter A).
(pmol CYP2El/gram tissue)/(pmol CYP2El/mg MSP) (4-4)
= (mg MSP/gram tissue)
4-9
-------
CYP2El-Dependent Oxidation of TCE. The Michaelis-Menten kinetic constants were
available for 23 samples of MSP from Lipscomb et al. (1997). The metabolism of TCE to
chloral hydrate, representing oxidation by CYP2E1, was quantified by measuring the formation
of chloral hydrate. Apparent Vmax was expressed as pmol TCE oxidized/min/mg MSP. From
this set of 23 original samples, 15 remained, and CYP2E1 content of those microsomal protein
samples was quantified by ELISA (Snawder and Lipscomb, 2000). We sought to develop a
more technical description of Vmax (the theoretical maximal initial rate of the reaction in the
presence of unlimiting substrate concentration), and one which would be more readily
extrapolable to the in vivo setting through incorporation of the information on the hepatic content
of CYP2E1. To accomplish this, the Vmax values (pmoles/min/mg MSP) available from the
previously published study (Lipscomb et al., 1997) were divided by the CYP2E1 content of MSP
(pmoles CYP2El/mg MSP) to yield Vmax values expressed as pmoles TCE
oxidized/minute/pmol CYP2E1 (Table 4-3). This measure (parameter D) and its distribution are
referred to as data set 3 and are described as TCEcyp2.
Statistical Analysis. Tables 4-2 and 4-3 summarize the data employed in the statistical analyses.
Probability distributions were fitted by the SAS® 8.0 Analyst routine to data describing the
following variables (with mnemonic variable name): A = pmol CYP2El/gram liver (CYP2L;V); B
= pmol CYP2El/mg MSP (CYP2Msp); C = mg MSP/gram liver (MSPLiv) = A/B. StatFit
software was used to determine an optimal distribution fit to the 15 observations for parameter
D, pmoles TCE oxidized, min/pmol CYP2E1 (TCEcYP2). The LogNormal distribution was
selected with parameters (j, = 3.4812 and a = 0.4156 for the imbedded normal distribution,
implying a geometric mean and standard deviation of 32.5 and 1.515, and an arithmetic mean
4-10
-------
and standard deviation of 35.4 and 15.4. This distribution was accepted via Chi-Squared,
Kolmogorov-Smirnov, and Anderson-Darling statistical tests at the a = 0.05 (95%) confidence
level.
Three sets of data were available: a set of n = 60 samples, for which laboratory
measurements were available on B = (CYP2Msp), an n = 20 subset of the 60 samples, for which
several additional laboratory measurements were available (PROuv, CYP2Pro, CYP2Msp), and a
set of n=15 samples, for which one laboratory measurement was available (TCEcYP2). These
three sets of available data were first analyzed separately. The additional variables (CYP2Liv,
and C = MSPuv) were calculated from the measurement data.
For all variables, normal, lognormal, exponential and Weibull distributions were fit using
standard statistical tests of goodness-of-fit (Kolmogorov-Smirnoff, Cramer-von Mises &
Anderson-Darling) and a visual examination of quantile-quantile plots. The null hypothesis was
that the distribution fit the data well, with a rejection of the null at p < 0.10. All these analyses
were performed using SAS®. Each of the distributions was adequately approximated by a
lognormal distribution, the parameters of which are the mean (|j,) and standard deviation (s) of
the logarithms of the observations.
Analysis via Method of Moments - For convenience, ignoring the dependence between
data set 2 and the n=20 (matched subset) of data set 1, and because the consistency of goodness-
of-fit of the data to the lognormal distributions (excluding data set 3: TCEcYP2; pmol TCE
oxidized/min/pmol CYP2E1), we applied the statistical method of moments (addition of errors,
Equation 4-5 to combine data sets 1-3 to estimate parameter E, pmol TCE oxidized/minute/gram
liver. All goodness-of-fit p-values were greater than 0.15. As a convenience, the lognormal
4-11
-------
parameter will be represented by the geometric mean (GM = e^) and geometric standard
deviation (GSD = es), respectively, in this paper. Equation 4-5 demonstrates the method used to
estimate the distribution of Vmax values, where the distributions for parameters B (pmol
CYP2El/mg MSP), C (mg MSP/gram liver), and D (pmol TCE oxidized/min/pmol CYP2E1) are
combined mathematically. The values at the 5th (X0s) and 95th (X95) percentiles for the resulting
distribution (parameter E, pmoles TCE oxidized/min/gram liver) were calculated by Equations
4-6 and 4-7, respectively.
Lnorm[|j, = |j,i + |j,2 + (j,3, s = sq root(si2 + s22 + S32)] (4-5)
where:
Hi = mean of logs of observations
s; = standard deviation of logs of observations
1 = data set 1 - (CYP2Msp)
2 = data set 2 - (MSPLiv)
3 = data set 3 - (TCECYP2)
XW"-1645**] (4-6)
X95 = e[»+1645*Sl (4-7)
Analysis via Computational Statistics - We next sought to model the distribution of A
= pmol CYP2El/gram liver with greater precision by using all of the available data, including
the correlation (Figure 4-3) on variables B = pmol CYP2El/mg MSP and C = mg MSP/gram
liver. Since A = B*C, these three variables are not statistically independent. Moreover, it is
perhaps not obvious how or whether the 40 measurements of B that are not matched to
measurements on A and C (observations 21-60 in Table 4-2) can be used to improve estimation
of the distribution of A. However, we were able to synthesize and apply two techniques from
computational statistics - mixture distribution modeling (Titterington et al., 1985) and
4-12
-------
classification trees (Zhang and Singer, 1999; Breiman et al., 1984) - to use all of the B and C
data, including the 40 unmatched measurements on B, to model the distribution of A.
The methodology for estimating the frequency distribution of A using all available
measurements (i.e., using the joint distribution of A and B, as well as the derived variable C) was
as follows.
1. The frequency distribution for A can be expressed using marginal and conditional
probabilities as follows:
= a) = Z(b;C)Pr(A
= Z(b;C)Pr(A
B = b & C = c)Pr(B = b & C = c)
B = b & C = c)Pr(B = b)Pr(C = c | B = b)
where the sum (or integral) is taken over all (b, c) pairs of values. Thus, A is interpreted
as having a distribution that depends on the (perhaps unobserved) values of B and C.
2. The terms Pr(A | B = b & C = c), Pr(B = b), and Pr(C = c | B = b) are estimated
empirically from all of the available data using a classification tree estimator. Figure 4-4
shows the classification tree fit to the first 20 cases in Table 4-2, i.e., those with data on
both A and B (and hence C). This tree provides an estimate of the distribution of A
conditioned on the values of B and C. The fit was performed using the
KnowledgeSeeker™ package (http://www.angoss.com/ProdServ/indexH.html.)
Interpretively, the distribution of A is modeled as a finite mixture distribution
(Titterington et al., 1985) with a number of components to be estimated from the data. These
components correspond to leaves in the classification tree in Figure 4-4. The conditional
distribution of A depends on which component distribution a case belongs to. Figure 4-4
shows the sample means and standard deviations for four component distributions (leaves),
estimated from the first 20 cases in Table 4-2. Using only these data, the estimated
distribution of A would correspond to the following finite (4-component) mixture
distribution:
Cluster 1: weight = 9/20, sample mean = 2029.23, sample standard deviation = 614.41
Cluster 2: weight = 4/20, sample mean = 2416.88, sample standard deviation = 182.68
Cluster 3: weight = 5/20, sample mean = 3093.62, sample standard deviation = 496.97
4-13
-------
Cluster 4: weight = 2/20, sample mean = 4728.3, sample standard deviation = 389.9
Here, the four components are termed "clusters" since they correspond to sets of cases for which
the distribution of A values is approximately the same (i.e., the classification tree algorithm is
unable to find any statistically significant difference among them.)
3. This initial tree based on the first 20 (full-data) cases in Table 4-2 was refined by using
the remaining 40 observations of B values in Table 4-2 (i.e., cases 21-60) to better
estimate the fraction of all cases for which B < 53 (the defining characteristic of Cluster
1). The pooled estimate from all 60 cases is that 32/60 (= 0.53, 95% CI = 0.40 to 0.66) of
A values are drawn from Cluster 1. The revised cluster weights using all 60 observations
on B are: 0.53 for Cluster 1; 0.17 for Cluster 2; 0.21 for Cluster 3; and 0.09 for Cluster 4.
While the cluster-specific sample sizes are very small (n = 2 for Cluster 4), this
decomposition of the distribution of A into a weighted mixture of component
distributions actually decreases the variance in estimates of the true mean (and other
statistics) of A compared to using a single estimated distribution (Feller, 1968).
The methodology summarized in steps 1-3 can be further refined, e.g., by using
resampling to establish robust boundaries for the classification tree splits, or by using a Bayesian
posterior distribution for the fraction of cases belonging to different clusters. However, given the
small number of cases (n = 20) with full data, additional refinements of the tree estimator in
Figure 4-4 with the cluster weights obtained from all 60 measurements for B are not expected to
greatly improve the estimation of the distribution of A.
Combination of Data Sets. A program was developed in the MATLAB software (see
Appendix A) to produce A, D and A*D random variates in accordance with the distributions for
A and D derived above. The distribution of A was taken from that determined by computational
statistics. This program identified 100,000 random variates. Eight thousand of the generated
A*D values were selected at random and were subjected to a further analysis to find an optimal
distributional fit (StatFit has a limit of 8000 values).
4-14
-------
PBPK Model. Human metabolism of TCE was simulated using the PBPK model of Allen and
Fisher (1993) and SimuSolv software (Dow Chemical Co., Midland, MI). The model structure
consisted of four tissue compartments (liver, rapidly perfused tissues, slowly perfused tissues,
and fat) and a gas exchange compartment (lung) connected by blood flows. All TCE
biotransformation was assumed to take place in the liver and follow Michaelis-Menten kinetics.
The liver was described as a well-stirred homogeneous compartment. Previous studies have
demonstrated that more complex heterogeneous models of the liver, such as the parallel tube and
dispersion models, were not better than the simple well-stirred model at predicting the in vivo
clearance of 28 drugs from in vitro data (Houston and Carlile, 1997). The model was set to
simulate two extreme exposure scenarios - 1) simulating a higher, but permitted, occupational
exposure at the Threshold Limit Value (TLV) for TCE, which is 50 ppm in air for an 8-hour
working day (ACGIH, 2001), and 2) simulating a low-dose environmental exposure via drinking
water containing the maximally allowable concentration of TCE (5 |ig/L) (U.S. EPA, 1997).
With knowledge that the hepatic metabolism of TCE in vivo is limited by blood flow, the model
was set to simulate a "worst case" scenario of an oral bolus dose, because this (rather than a
slower oral ingestion rate) would be the more likely scenario to produce differences in the
amount of TCE metabolized (AML). AML is presented in units of mg TCE metabolized over
the course of the simulation per L of liver tissue. Simulations of AML (amount of TCE
metabolized in the liver compartment) were evaluated, because the CYP2E1-dependent
oxidation of TCE is a required step in the formation of the hepatotoxic metabolite, trichloroacetic
acid (TCA), in vivo. Chloral hydrate, the oxidative metabolite of TCE, does not have a
measurable half-life in vivo following its formation from TCE, but is immediately converted to
4-15
-------
TCA and trichloroethanol. For these simulations, the model incorporated both extremes of the
distribution of the Vmax for TCE oxidation (5th and 95th percentiles).
RESULTS
Distribution of CYP2E1 to Human Hepatic MSP. Analysis of 60 samples of MSP
derived from individual adult human organ donors for the content of CYP2E1 (pmoles
CYP2El/mg MSP, parameter B, data set 1) indicated that the Lognormal distribution adequately
represented the set of observations. The geometric mean and geometric standard deviations
required to reconstruct the overall distribution and simulate the value for a percentile of interest
are presented in Table 4-4. These values agree well with those reported by Shimada et al.
(1994). Variance between the values at the 5th and the 95th percentiles of the distribution
approximated 4-fold.
Distribution of CYP2E1 to Intact Human Liver. Three types of analytical procedures
were used to determine the distribution of CYP2E1 to intact liver tissue (pmoles CYP2El/gram
liver, parameter A) derived from adult human organ donors. First, the most direct measure, but
one for which only 20 observations are available, is depicted in equations 1 and 2 and involved
the application of the ELISA technique to liver homogenate (post 100 x g) protein. The
empirical distribution of the 20 observations indicates that the magnitude of variance between
the observations representing approximately the 5th and approximately the 95th percentiles of the
empirical distribution is approximately three-fold. These raw data indicate a mean value of 2643
and a standard deviation of 962 pmoles CYP2El/gram liver.
Second, the application of the statistically-limited method of moments required the
characterization of the two underlying distributions: 1) parameter B, pmoles CYP2El/mg MSP,
4-16
-------
data set 1, and 2) parameter C, mg MSP/gram liver, data set 2. Because the observations in
these two sets of data were adequately fit by a lognormal distribution, the values for the
geometric mean and geometric standard deviations for each data set (Table 4-4) were combined
(Equation 4-4). The lognormal distribution of the liver content of MSP (parameter C) was
characterized by a geometric mean of 52.9 mg MSP/gram liver and a geometric standard
deviation of 1.476 (arithmetic mean and standard deviation of 57 + 23 mg MSP/gram liver). The
MSP content of CYP2E1 (Parameter B) was determined in a sample set of 60, and demonstrated
a lognormal distribution, with a geometric mean of 48.9 pmoles CYP2El/mg MSP and a
geometric standard deviation of 1.6. The arithmetic mean + standard deviation was 54 + 23
pmoles CYP2El/mg MSP. When combined, the results indicated a geometric mean of 2587
pmoles CYP2El/gram intact adult human liver, with a geometric standard deviation of 1.48.
From analysis via Equations 4-5 and 4-6, these parameter values indicate values at the 5th and
95th percentile of the distribution to be 949 and 7053 pmoles CYP2El/gram. These values are
similar to those indicated by the direct measurement of CYP2E1 in homogenate protein, above.
These data indicate that the central 90% of the population represented by these adult organ
donors expresses a CYP2E1 content which varies 7.4-fold.
Finally, a specific probability distribution for the parameter A was developed, based upon
the clusters derived in the Analysis via Computational Statistics section above. Recall that
clusters of values for A were identified, into which values for parameter B (pmoles CYP2El/mg
MSP) were segregated. In this manner, the influence of parameter B, or its determinant qualities,
on parameter A were characterized. A continuous probability distribution was not fit to the
individual clusters due to the small number of observations within each cluster; instead, the
distribution of A was assumed to be discrete and consist only of the observed values (4th
4-17
-------
column, parameter A, of Table 4-2). Clusters were assumed to occur with proportional
frequencies equal to the weights (0.53, 0.17, 0.21, 0.09} and to have counts of (9, 4, 5, 2} as
described previously. Within a cluster, each value belonging to that cluster is assumed to occur
with equal frequency. This approach provides probabilities of 0.53/9 (=0.0589) for values in
cluster 1, 0.17/4 (=0.0425) for values in cluster 2, 0.21/5 (=0.0420) for values in cluster 3, and
0.09/2 (=0.0450) for values in cluster 4. Cluster statistics are presented in Figure 4-4. If we use
x; to represent the ith value of parameter A, and p; to represent the probability of the ith value,
then the mean of the resulting distribution is:
£/>,.*,.= 2561.77 (4-8)
z=l
while the variance of the distribution is
,•*,• I =865>563-40 (4-9)
2=1
providing a standard deviation of 930.36.
Note that the mean of the raw data for parameter A (Table 4-2) is 2642.8 while the
standard deviation (using Equation 4-9) is 937.21 (the sample standard deviation is 962). We see
that accounting for the influence of parameter B has shifted our estimates of parameter A slightly
downward, and has slightly decreased the standard deviation. This is a result of the individual
probabilities being shifted slightly upward or downward from 0.05 in accordance with the
distribution shown in empirical distribution. This distribution was used in the recombination of
data describing parameter A with data describing parameter D, as discussed below.
In vitro Metabolic Rate Constant (Vmax). The Vmax for the oxidation of TCE by CYP2E1
(pmoles TCE oxidized/min/pmol CYP2E1, parameter D, data set 3, Table 4-3) was evaluated in
4-18
-------
a data set of 15 samples. The apparent Vmax observed in vitro (pmoles TCE oxidized/min/mg
MSP) was converted to the more specific units of pmol TCE oxidized/min/pmol CYP2E1 by
dividing the observed Vmax value by the content of CYP2E1 in the MSP (pmoles CYP2El/mg
MSP). The resulting set of 15 observations (pmol TCE oxidized/min/pmol CYP2E1) were fit
optimally with the lognormal distribution; its parameters were (J, = 3.4812 and a = 0.4156 for the
imbedded normal distribution, implying a geometric mean and standard deviation of 32.5 and
1.515, and an arithmetic mean and standard deviation of 35.4 and 15.4. This distribution was
accepted via Chi-Squared, Kolmogorov-Smirnov, and Anderson-Darling statistical tests at the a
= 0.05 (95%) confidence level. The difference between values at the 5th and 95th percentiles of
the distribution approximated 4-fold.
Determining the Metabolic Capacity of Intact Tissue and Extrapolation of Units.
MATLAB results of the simulations of 100,000 random variates of A*D revealed a plot (not
shown) suggestive of a lognormal distribution, and a normal distribution (not shown) of the logs
of values of A*D. StatFit analyzed 8000 of these variates, and indicated that the most likely
distribution was the lognormal, with parameters (J, = 11.2748 and a = 0.5466. The lognormal
distribution was accepted via Chi-Squared, Kolmogorov-Smirnov and Anderson-Darling
statistical tests at the a = 0.05 (95%) confidence level. The parameters of the lognormal
distribution indicate a geometric mean of 78,810 pmoles TCE oxidized/minute/gram liver, and a
geometric standard deviation of 1.7274. Applying these values in Equations 4-6 and 4-7 results
in 5th and 95th percentile values of 32,069 and 193,679 pmoles TCE oxidized/minute/gram liver,
respectively. These values were corrected for molecular weight, time and fractional composition
4-19
-------
of the body represented by the liver (liver weight = 2.6% body mass) to yield values of 6.6 and
39.7 mg/hr/kg at the 5th and 95th percentiles of the distribution, respectively.
PBPK Model Predictions. The PBPK model for TCE was used to simulate exposure of a 70 kg
male human to 50 ppm TCE for 8 hours. This exposure scenario represents the maximum
recommended exposure of an individual to TCE in the workplace. The extremes of expression
of the CYP2E1-mediated oxidation of TCE in the liver used here (approximately 6-fold) resulted
in a difference in TCE hepatic metabolism of approximately 2% (Table 4-5). For this
simulation, the amount of TCE oxidized over the exposure period per volume of liver (|ig/L) was
used as the dose metric most linked with hepatotoxic injury/risk. Simulation of the oral ingestion
of TCE (5 |ig/L) in 2 L of drinking water using the 5th and 95th percentiles of the TCE oxidation
rate gave similar results (Table 4-5). These data indicate that physiological processes limit the
full impact of the differences in CYP2E1 activity toward TCE mediating the formation of toxic
metabolites. Previous PK analyses of the effect of enzyme induction on the bioactivation of TCE
and other volatile organic compounds indicated a hepatic blood flow limitation of the
bioactivation process (Kedderis, 1997). The rate of blood flow delivery of these substances to
the liver is much slower than the rate of bioactivation in the liver, limiting the impact of enzyme
induction or interindividual variability. This study focused on the issue of whether enzymic
variance alone could contribute substantially to susceptibility to hepatotoxic injury, and did not
attempt to capture or examine the effect of other factors (e.g., differences in hepatic blood flow)
on the PK of TCE.
DISCUSSION
Advanced PK studies and PBPK modeling allow the development of the linkage between
external dose and target tissue dosimetry. PBPK models can predict target tissue concentrations
4-20
-------
associated with specific levels of response in animals or humans (LOAEL, NOAEL or BMD-
derived level of response). Human PBPK models can examine the risk-relevant PK outcome of
chemical exposure (i.e., tissue levels of bioactive metabolite) and predict the external exposure
(i.e., mg/kg/day) required to produce this PK outcome at the same level observed in research
animals at the corresponding level of toxicity. When adequate information is available to
quantify the metabolism to or from the bioactive chemical form, the human PBPK model can be
further refined to include data on enzyme (metabolic) variance in human tissues. Then the PBPK
models can be exercised to examine not only animal to human differences in the risk-relevant PK
outcome, but also the human interindividual variance in the expression of that PK outcome.
Biotransformation is a critical determinant of both PK and risk since metabolism is
involved in the bioactivation and detoxication of xenobiotics. Genetic polymorphisms and
enzyme induction due to environmental and lifestyle factors can affect the level of expression of
xenobiotic metabolizing enzymes. Thus, genetic polymorphisms become critical to risk only
when they alter PK outcomes. The refinement of human health risk assessments for chemicals
metabolized by the liver to reflect data on human interindividual PK variability can be
accomplished through 1) the characterization of enzyme expression in large banks of human
liver samples, 2) the employment of appropriate techniques for the quantification and
extrapolation of metabolic rates derived in vitro, and 3) the judicious application of PBPK
modeling.
Numerous PK outcomes may be simulated by PBPK modeling; the identification of the
risk-relevant PK outcome(s) from toxicity studies allows the study of their variability through
adequately constructed PBPK models. When PK models are constructed to include metabolic
rates (and rate constants) derived in vitro, several extrapolations are necessary, not the least of
4-21
-------
which is the extrapolation of enzyme content (Figure 4-5). PBPK models include the apparent
Vmax expressed as mg/hr/kg body mass, while typical in vitro studies express Vmax in terms of
nmoles product formed/minute/mg microsomal protein. Accurate extrapolation requires initially
that enzyme content be expressed per unit intact liver (i.e., pmoles CYP2El/gram liver), and the
extrapolation has usually included a numerical estimation of the MSP content of liver (i.e, 50 mg
MSP/gram liver). The MSP content of intact liver has been measured and used to extrapolate in
v/Yro-derived metabolism kinetic constants for use in PBPK modeling efforts in humans (Reitz et
al., 1996; Lipscomb et al., 1998) and to infer measures of intrinsic clearance (Clint) in traditional
rat-based PK models (Carlile et al., 1997). In the previous PBPK based approach for TCE (Reitz
et al., 1996), samples expressing extreme values for kinetic constants (Km and Vmax) were
chosen for extrapolation to a PBPK model. Those extrapolations were based on the
hepatocellularity of intact liver tissue, and on microsomal protein content of liver, rather than the
content of CYP2E1 of liver. The Vmax value was not previously extrapolated on the basis of
CYP2E1 content as no data existed at the time through which to quantify the distribution of the
key metabolic enzyme within human liver. With respect to the distribution of the cytochromes
P450 in one preparation of human liver (microsomes), several investigations (Lipscomb et al.,
1997; Shimada et al., 1994; Iyer and Sinz, 1999) have revealed quantitative information about
the content of these multiple enzyme forms in this preparation, but reveal no direct information
on the type of distribution (i.e., lognormal) of the enzymes within MSP, their content or
distribution to the intact liver in situ. In the present study, we developed measures of the liver
content of microsomal protein, of which the CYP enzymes (and other important xenobiotic-
metabolizing enzymes, i.e., glucuronyl transferases) are a constituent. This key piece of
information is necessary to estimate the content of the enzyme(s) in the intact liver. By
4-22
-------
combining the two data sets on 1) the MSP content of CYP (pmoles CYP/mg MSP), and 2) the
liver content of MSP (mg MSP/gram liver), we derived the liver content of CYP (pmoles
CYP/gram liver), and developed measures of that variance, employing a total number of 60
samples derived from adult human organ donors. The approach also included the determination
and inclusion of the human interindividual variance in metabolic activity toward TCE derived in
an additional set of 15 samples. This analysis allowed the variance of that critical enzyme
kinetic parameter (Vmax) to be examined among humans, and expressed as pmoles TCE
oxidized per minute per pmol CYP2E1. This parameter (pmol TCE oxidized/min/pmol
CYP2E1) did vary among the human samples evaluated, not surprisingly. This may be
explained, in part, by potential underlying genetic differences impacting CYP2E1 activity,
differences in the presence of other CYP forms which also metabolize TCE at higher
concentrations and human to human interindividual differences in the lipid composition (both
qualitative and quantitative) of isolated MSP preparations. The activity of isolated enzymes
represents the functional status of their respective donors. The stability of these enzymes upon
isolation and storage seems not to be a major contributor to this variance. The level of detail in
this expression of Vmax allowed for a direct combination with information on the variance of
CYP2E1 in intact human liver, which enabled the resulting PBPK analysis of the impact of that
variance on the risk-relevant pharmacokinetic (PK) outcome, amount of TCE metabolized in the
liver, among adult humans. Together, these data sets separately describing enzyme activity and
enzyme content combine to describe the metabolic capacity of the liver. The resulting combined
distribution for the Vmax value demonstrated that this parameter (mg/hr/kg) differed more than
6-fold between the values at the 5th and 95th percentiles of the distribution. When these values
were separately integrated into the PBPK model, resulting estimates of the amount of TCE
4-23
-------
oxidized over the exposure period differed by only 2%. Thus, widely divergent values for
apparent Vmax, resulting from both variance in enzyme content and activity, had little effect on
the in vivo metabolism of TCE, and will have little effect on the hepatotoxic injury following
TCE exposure in humans.
The present work demonstrates a significant advantage over earlier studies in that
statistically valid and robust measures of enzyme content and enzyme activity have been
developed and incorporated into the PBPK-based approach. This advance allows the application
of the approach to estimate population distributions of risk, when chemical dose-response
parameters (e.g., slope factors) are available. With the availability of large banks of well-
characterized subcellular fractions (mainly hepatic MSP) derived from the livers of human organ
donors comes the opportunity to determine several measures of human biochemical
individuality, which will be applicable to many environmental, occupational and therapeutic
compounds. Although several investigations have failed to identify a consistent inverse
relationship between post mortem cold-clamp time (the time interval between the perfusion,
removal and refrigeration of liver tissue; and the freezing of the tissue or microsomal protein
isolation) and microsomal enzyme activity, the assumption that the activity of these enzymes in
vitro represents their activity in vivo must be recognized as such. From these samples, we can
measure interindividual differences in enzyme activity and differences in enzyme content in
isolated MSP. The in vitro metabolism of several CYP2E1 substrates, such as furan (Kedderis et
al., 1993; Kedderis and Held, 1996), perchloroethylene (Reitz et al., 1996), and trichloroethylene
(Lipscomb et al., 1998) have been successfully extrapolated to the in vivo setting through
application of adequately developed and validated PBPK models. The additional validation of
the extrapolation procedure for metabolic activity based on enzyme recovery data is important.
4-24
-------
This demonstrates the applicability of the methodology to determine the interindividual
differences of risk-relevant PK outcomes (i.e., the amount of metabolite formed in the liver for a
bioactivated hepatotoxicant) for xenobiotics to which humans cannot be safely exposed for the
generation of experimental data. It is anticipated that toxicological data can be generated in test
species in vivo and in vitro to determine the metabolic species responsible for toxicity, the PK of
the xenobiotic and metabolite(s), and the identity of the enzyme responsible for metabolism.
With this information, an adequate test animal-based PBPK model can be extrapolated to
humans, using human tissue partition coefficients and the appropriate physiological parameters.
Data on human enzyme recovery could be used to develop appropriate bounds on the distribution
of metabolic activity for evaluation with the PBPK model to represent predefined proportions of
the population.
The successful application of this approach requires the avoidance of several pitfalls. It
requires the following:
1. The metabolic process under investigation must be as directly linked to the risk-relevant
PK outcome as possible. The correct identification of the critical toxic effect, against
which protection is warranted, or toward which susceptibility requires evaluation. In the
absence of a defined link between this effect and it's most closely related and measurable
or predictable PK outcome (e.g., AML), then further effort will not advance the goals of
the approach.
2. The tissues/preparations included in the experiments must be viable. The reliance on
human tissues of research grade can be troublesome; the comparison of in v/Yro-derived
metabolic rates and rate constants, especially in humans, requires some justification that
these ex vivo or in vitro systems maintain the metabolic capacity they possess in vivo.
The isolated hepatocyte model is more closely related to the in vivo situation than the
isolated microsomal protein preparation, but metabolic rates from both systems require
extrapolation based on recovery information, to the in vivo situation. Reliance upon data
derived from compromised in vitro systems can lead to under predictions of in vivo
metabolism. The inclusion of data from compromised systems (i.e., lengthy 37°
incubations of microsomal protein, the application of immortalized cell lines, etc.) must
be avoided. The evaluation of metabolic activity toward recognized marker substrates
and assessment of cellular viability provide some evidence of in vitro system stability.
4-25
-------
3. There must be sufficient data to enable extrapolation based on protein recovery. Lack of
data or uncertainty in the available data quantifying the relationship between the in vitro
system and the in vivo situation greatly complicate the extrapolation procedure. Values
for hepatocellularity in rats and humans, and values for microsomal protein content of
rats and humans are available for use in extrapolation procedures.
4. The derivation of metabolic rate constants must be accomplished under valid
experimental conditions. Rate constants must be derived under conditions where rate is
proportionate to an increase in protein content, over time and with increasing substrate
concentrations (for first-order reactions). The value of such data is enhanced when rate
constants are tied specifically to the enzyme, rather than the subcellular fraction (e.g.
pmoles/min/pmol CYP2E1 vs pmoles/min/mg MSP).
5. Data should be used to identify the pertinent enzyme; the presence of more than one
enzyme complicates enzyme kinetic evaluations. Additional uncertainty is encountered
in the metabolic evaluation of substrates, toward which multiple enzymes are active.
Given the human interindividual variability on enzyme expression as a result of genetic,
dietary and lifestyle choices, different ratios of two potentially active enzymes may be
observed. In this instance, the approach to in vitro enzyme kinetic investigations must be
robust enough to separately identify the kinetic constants applicable to each of the
enzymes. Kinetic constants derived for the preparation, without regard to the pertinent
enzymes can falsely indicate that the apparent Vmax value is shifted upward due to the
contribution of a low affinity form, when in vivo substrate concentrations would not be
sufficient to drive an appreciable contribution of this enzyme to the reaction.
6. This approach relies on the availability of a "validated" PBPK model. While
generalization of model structure and physiological components across chemicals is often
the case, the models must include parameters demonstrated or judged to be chemical-
specific. In addition to metabolic rate constants, tissue partition coefficients (PC) are
highly chemical-specific, and differ for the same tissue type among species. The
application of PC values derived in other species or adapted from PC values of related
chemicals requires justification.
7. Finally, the approach is aimed specifically at quantifying human interindividual
differences in metabolic capacity. This approach is specifically not aimed at quantifying
human interindividual PK difference for TCE oxidation; it was developed to test the
hypothesis that variability in metabolic capacity alters susceptibility to hepatotoxic injury
from TCE exposure. The approach, here, demonstrates the applicability of the statistical
bounds established for the population under investigation. In that regard, the application
of a representative sample set is required. The data must support the identification of
distribution type and include enough observations so that confidence can be placed in the
identification of values at predetermined points of the distribution. While these are some
of the general pitfalls, investigators trained in the disciplines of the individual
investigatory steps will be quite familiar with many of the more technical pitfalls.
4-26
-------
To members of the risk assessment community who are advocating the development of
approaches which provide more information than just "safe exposure limits" (e.g., RfC and RfD
values), the present approach may be useful. The approach is centered on the identification of
the risk-relevant PK outcome through evaluation of toxicity and PK investigations, not
necessarily through PBPK investigations. Under optimal conditions, the linking of PBPK
modeling approaches with data describing human biochemical individuality (enzyme content and
enzyme activity) will allow the quantification of the PK component of UFH. The collection of
advanced measures of human biochemical individuality (e.g., differences in the liver's content of
critical xenobiotic metabolizing enzymes) will broaden the applicability of this approach to
other chemicals whose PK are modulated by the same enzyme. It is conceivable that when this
parameter (enzyme content and activity) modulates the production of the risk-relevant PK
outcome, information about the population distribution of the parameter (i.e., hepatic content of
CYP2E1) will lead to applications demonstrating the fraction of the population which will be
protected by regulations which specify a given level of chemical exposure. Similarly, with
carcinogenicity slope factors, risk-relevant PK outcomes can be converted directly to measures
of risk, indicating the level of risk corresponding to a given level of enzyme content and activity.
By converting exposure to tissue dose, and having information to link tissue dose to risk, the
PBPK modeling approach may be usefully employed to develop distributions of risk, rather than
simply assessing or demonstrating the health protective nature of a given exposure.
The purpose of the present study was to explore the potential advantage of including
additional, specific, information on human biochemical individuality as a process to refine the
human health risk assessment process. Because an ever-increasing amount of data is being
developed through the analysis of tissues derived not only from surgical resections, but also from
4-27
-------
human organ donors, these sources of tissues offer a unique potential to increase our knowledge
about human biochemical individuality.
SUMMARY AND CONCLUSIONS
Human interindividual PK variability is important both for chemicals with adequate
human PK data and for those chemicals to which humans cannot be experimentally exposed.
Because CYP2E1 activity limits (in vitro) the production of oxidative, hepatotoxic metabolites of
TCE, we evaluated the distribution of that enzyme in liver from up to 75 adult human organ
donors by applying published and accepted biochemical and statistical methods. The
extrapolation of in vitro data captured both the variance in enzyme content and enzyme activity
among adult humans.
CYP2E1 content and metabolic activity toward TCE are described by lognormal
distributions. The central 90% of the human population represented by these adult organ donors
differs by less than 4-fold in the hepatic content of this critical xenobiotic metabolizing enzyme;
that same fraction of the population differed by approximately 6-fold with respect to the
oxidation of TCE. The finding and additional information to be gained from the now-
characterized distribution of CYP2E1 to intact human liver will be useful not only to the
assessment of risk from TCE exposure, but also to the assessment of risks from other
environmental chemicals that are also metabolized by this enzyme including chloroform, carbon
tetrachloride, benzene, toluene, and styrene. Because the metabolism of TCE is limited by blood
flow to the liver, divergent values of Vmax do not result in appreciable differences in the risk-
relevant PK outcome, the amount of TCE metabolized in the liver. Therefore, factors which
increase the hepatic expression of CYP2E1 and/or its metabolic activity will not always result in
4-28
-------
proportionate changes in key PK outcomes. This is because of the relatively low solubility of
TCE in blood, and the relatively high capacity of the liver to metabolize TCE (due to a relatively
high level of expression of the enzyme and the relatively high metabolic activity of the enzyme
toward TCE), the limiting factor, in vivo, for TCE oxidation becomes the rate at which TCE is
delivered to liver tissue by hepatic blood flow. In this situation, increases in TCE metabolic
capacity, even from the 5th to the 95th percentiles of the distribution, result in only a 2% increase
in the amount of TCE metabolized. With respect to the hepatotoxicity of TCE resulting from
exposure scenarios similar to those employed in this analysis, these data indicate that the amount
of PK variability attributed to enzymic variance among humans is approximately 2%. The
approach described here is especially applicable to chemicals to which humans cannot be
experimentally exposed for ethical reasons. The application of actual, not hypothesized, bounds
of variance and the definition of the distribution of enzyme content and activity among humans
can allow the calculation of finite levels of risk (when dependent on the PK outcome) at different
chosen percentiles of the distribution of enzyme content and activity.
Several conditions must first be met for this strategy to be successful:
• The target organ, mode or mechanism of action, and metabolic species responsible for
toxicity must be known.
• The target tissue-toxic chemical species dose response relationship must be known.
• The biotransforming enzyme must be known and information on the variance and
type of its distribution among humans must be known.
• The kinetic mechanism of metabolism must be known and expressed per unit of
enzyme.
• An adequately characterized PBPK model must be available for adaptation.
4-29
-------
We have quantified the extent of variance in enzyme content of a critical xenobiotic
metabolizing enzyme, CYP2E1, and the variance in the hepatic biotransformation of a key
environmental contaminant, trichloroethylene. The parameters of the resulting lognormal
distributions can be used to identify the bounds of biochemical and pharmacokinetic variance
(e.g., 90% of the population), within which susceptibility can be determined and allows the
replacement of hypothesized magnitudes of difference with actual measurement of such when
determining the impact of enzyme variance on risk.
This manuscript identifies the conditions and types of data required, communicates and
applies a logical approach, and describes the limitations of the approach in estimating the human
interindividual variance of risk-relevant PK outcomes which may signify susceptibility to
chemical injury. While data set 3 is unique to TCE, data set 2 will be useful in estimating the
hepatic content of all enzymes contained in the microsomal fraction, when their distribution
characteristics are known, and the information derived from the combination of data sets 1 and 2
are directly applicable to other environmental contaminants that are also substrates for CYP2E1.
REFERENCES
ACGIH (American Conference of Governmental Industrial Hygienists). 2001. Threshold Limit
Values for Chemical Substances and Physical Agents and Biological Exposure Indices.
Cincinnati, OH.
Allen, B.C. and J.W. Fisher. 1993. Pharmacokinetic modeling of trichloroethylene and
trichloroacetic acid in humans. Risk Anal. 13:71-86.
Arias, I.M., H. Popper, D. Schachter and D.A. Shafritz. 1982. The Liver: Biology and
Pathobiology. Raven Press, New York, NY.
Barton, H.A. and HJ. Clewell, III. 2000. Evaluating the noncancer effects of trichloroethylene:
Dosimetry, mode of action, and risk assessment. Environ. Health Perspect. 108(Supp. 2):
323-334.
4-30
-------
Bloemen, L.J., A.C. Monster, S. Kezic et al. 2001. Study on the cytochrome P-450- and
glutathione-dependent biotransformation of trichloroethylene in humans. Int. Arch. Occup.
Environ. Health. 74:102-108.
Bogdanffy, M.S. and A.M Jarabek. 1995. Understanding mechanisms of inhaled toxicants:
Implications for replacing default factors with chemical-specific data. Toxicol. Lett.
82-83:919-932.
Bogdanffy, M.S., R. Sarangapani, D.R. Plowchalk, A.M. Jarabek and M.E. Andersen. 1999. A
biologically-based risk assessment for vinyl acetate-induced cancer and noncancer inhalation
toxicity. Toxicol. Sci. 51:19-35.
Breiman L., J. Friedman, R. Olshen and C. Stone. 1984. Classification and Regression Trees.
Wadsworth Publishing.
Bull, RJ. 2000. Mode of action of liver tumor induction by trichloroethylene and its
metabolites, trichloroacetate and dichloroacetate. Environ. Health Perspect. 108(Supp.
2):241-259.
Carlile, D.J., K. Zomorodi and J.B. Houston. 1997. Scaling factors to relate drug metabolic
clearance in hepatic microsomes, isolated hepatocytes, and the intact liver: Studies with induced
livers involving diazepam. Drug Metab. Dispos. 25:903-911.
Feller W. 1968. An Introduction to Probability Theory and Its Applications, Vol. 1, 3rd ed.
Wiley, New York, NY. p. 231.
Fisher, J.W. 2000. Physiologically based pharmacokinetic models for trichloroethylene and its
oxidative metabolites. Environ. Health Perspect. 108, Supp. 2:265-273.
Guengerich, F.P. 1989. Analysis and characterization of enzymes. In: Principles and Methods
of Toxicology, A.W. Hayes, Ed. Raven Press, New York, NY. p. 777-814.
Gut, J., T. Catin, P. Dayer, T. Kronbach, U. Zanger and U.A. Meyer. 1986.
Debrisoquine/sparteine-type polymorphism of drug oxidation: Purification and characterization
of two functionally different human liver cytochrome P-450 isozymes involved in impaired
hydroxylation of the prototype substrate bufuralol. J. Biol. Chem. 261:11734-11743.
Houston, J.B. and D.J. Carlile. 1997. Prediction of hepatic clearance from microsomes,
hepatocytes, and liver slices. Drug Metab. Rev. 29:891-922.
Ikeda, M., Y. Miyake, M. Ogata and S. Ohmori. 1980. Metabolism of trichloroethylene.
Biochem. Pharmacol. 29:2983-2992.
4-31
-------
IPCS (International Programme on Chemical Safety). 1994. Environmental Health Criteria 170.
Assessing human health risks of chemicals: Derivation of guidance values for health-based
exposure limits. World Health Organization, Geneva.
IPCS (International Programme on Chemical Safety). 1998. Environmental Health Criteria 204:
Boron. World Health Organization, International Programme on Chemical Safety, Geneva,
Switzerland. ISBN 92 4 157204 3.
Iyer, K.R. and M.W. Sinz. 1999. Characterization of phase I and phase II hepatic drug
metabolism activities in a panel of human liver preparations. Chem. Biol. Interact.
118:151-169.
Jarabek, A.M. 1995. Interspecies extrapolation based on mechanistic determinants of chemical
disposition. Human Ecol. Risk Assess. 1:641-662.
Kedderis, G.L. 1997. Extrapolation of in vitro enzyme induction data to humans in vivo. Chem.
Biol. Interact. 107:109-121.
Kedderis, G.L. and S.D. Held. 1996. Prediction of furan pharmacokinetics from hepatocyte
studies: Comparison of bioactivation and hepatic dosimetry in rats, mice, and humans. Toxicol.
Appl. Pharmacol. 140:124-130.
Kedderis, G.L., M.A. Carfagna, S.D. Held, R. Batra, I.E. Murphy and M.L. Gargas. 1993.
Kinetic analysis of furan biotransformation by F-344 rats in vivo and in vitro. Toxicol. Appl.
Pharmacol. 123:274-282.
Lehninger, A.L. 1975. Biochemistry, 2nd ed. Worth, New York, NY. p. 183-216.
Lipscomb, J.C. and G.L. Kedderis. 2002. Incorporating human interindividual
biotransformation variance in health risk assessment. Sci. Total Environ. 288:13-21.
Lipscomb, J.C., C.M. Garrett and I.E. Snawder. 1997. Cytochrome P450-dependent metabolism
of trichloroethylene: Interindividual differences in humans. Toxicol. Appl. Pharmacol.
142:311-318.
Lipscomb, J.C., J.W. Fisher, P.D. Confer and J.Z. Byczkowski. 1998. In vitro to in vivo
extrapolation for trichloroethylene metabolism in humans. Toxicol. Appl. Pharmacol.
152:376-387.
Lipscomb, J.C., L.K. Teuschler, J.C. Swartout, C.A.F. Striley and I.E. Snawder. 2003. Variance
of microsomal protein and cytochrome P450 2E1 and 3 A forms in adult human liver. Toxicol.
Mech. Meth. 13:45-51.
4-32
-------
Nakajima, T., T. Okino, S. Okuyama, T. Kaneko, I. Yonekura and A. Sato. 1988. Ethanol-
induced enhancement of trichloroethylene metabolism and hepatotoxicity: Difference from the
effect of phenobarbital. Toxicol. Appl. Pharmacol. 94:227-237.
Northrop, D.B. 1983. Fitting enzyme-kinetic data to V/K. Analyt. Biochem. 132:457-461.
Reitz, R.H., M.L. Gargas, A.L. Mendrala and A.M. Schumann. 1996. In vivo and in vitro
studies of perchloroethylene metabolism for physiologically based pharmacokinetic modeling in
rats, mice and humans. Toxicol. Appl. Pharmacol. 136:289-306.
Segel, I.H. 1975. Enzyme Kinetics. Wiley, New York, NY. p. 1-17.
Shimada, T., H. Yamazaki, M. Mimura, Y. Inui and P.P. Guengerich. 1994. Interindividual
variations in human liver cytochrome P450 enzymes involved in the oxidation of drugs,
carcinogens and toxic chemicals: Studies with liver microsomes of 30 Japanese and 30
Caucasians. J. Pharmacol. Expt. Ther. 270:414-423.
Snawder, I.E. and J.C. Lipscomb. 2000. Interindividual variance of cytochrome P450 forms in
human hepatic microsomes: Correlation of individual forms with xenobiotic metabolism and
implications in risk assessment. Regul. Toxicol. Pharmacol. 32:200-209.
TitteringtonD.M., A.F.M. Smith and U.E. Makov. 1985. Statistical Analysis of Finite Mixture
Distributions. Wiley, New York, NY.
U.S. EPA. 1987. National Primary Drinking Water Regulations; Synthetic Organic Chemicals;
Monitoring for Unregulated Contaminants. Federal Register. 52(130):25690.
U.S. EPA. 1998. Toxicological Review of Methyl Methacrylate (80-62-6) in Support of
Summary Information on the Integrated Risk Information System (IRIS). National Center for
Environmental Assessment, Washington, DC. Available online at http://www.epa.gov/iris.
U.S. EPA. 1999. Toxicological Review of Ethylene Glycol Monobutyl Ether (111-76-2) in
Support of Summary Information on the Integrated Risk Information System (IRIS). National
Center for Environmental Assessment, Washington, DC. Available online at
http://www.epa.gov/iris.
U.S. EPA. 2000. Toxicological Review of 1,3-Dichloropropene (542-75-6) in Support of
Summary Information on the Integrated Risk Information System (IRIS). National Center for
Environmental Assessment, Washington, DC. Available online at http://www.epa.gov/iris.
U.S. EPA. 2001. Toxicological Review of Hexachlorocyclopentadiene (77-47-4) in Support of
Summary Information on the Integrated Risk Information System (IRIS). National Center for
Environmental Assessment, Washington, DC. Available online at http://www.epa.gov/iris.
4-33
-------
WHO (World Health Organization). 1998. Guidelines for Drinking Water Quality, Geneva,
Switzerland, 2nd ed., Addendum to Volume 2, Health Criteria and Other Supporting
Information. Geneva, Switzerland, p. 15-30. ISBN 92 4 1545143.
Zhang H. and B. Singer. 1999. Recursive Partitioning in the Health Sciences. Springer, New
York, NY.
NOTICE: The views expressed in this paper (NCEA-C-0956) are those of the individual
authors and do not necessarily reflect the views and policies of the US Environmental Protection
Agency (EPA). This research was supported by an interagency agreement between US
EPA/NCEA and CDC/NIOSH (No. DW75851501; John C. Lipscomb, Project Officer) and
through a cooperative agreement between US EPA/NCEA and Dr Gregory Kedderis (No.
CR828047-01-0, John C. Lipscomb, Project Officer). The authors wish to express their
gratitude to Glenn Suter and Bob Bruce (EPA, NCEA, Cincinnati) and Ulrike Bernauer (BgVV,
Berlin, Germany) for insightful comments during preparation of the manuscript. Preliminary
results from this investigation were presented at the annual meeting of the Society for Risk
Analysis, December 2000, Arlington, VA and December 2001, Seattle, WA; at the annual
meeting of the Society of Toxicology, March 2001, San Francisco, CA and March 2002,
Nashville, TN; and at the Spring Toxicology Conference April 2001, Wright-Patterson Air Force
Base, OH. We are sincerely grateful to organ donors and their families; these and other studies
would not be possible without their generous contributions. This paper has been published in
Risk Analysis:
Lipscomb, J.C., L.K. Teuschler, J. Swartout, D. Popken, T. Cox, and G.L.
Kedderis. 2003. The Impact of Cytochrome P450 2El-dependent Metabolic
Variance on a Risk Relevant Pharmacokinetic Outcome in Humans. Risk Anal.
23:1221-1238.
4-34
-------
TABLE 4-1
Identification of Data Sets and Parameters for Statistical Evaluation
Information
CYP2E1
content of intact
liver
CYP2E1
content of MSP
MSP content of
intact liver
TCE
metabolized per
unit CYP2E1
TCE
metabolized per
unit of intact
liver
Data set
Data Set 1
(n=60)
Data Set 2
(n=20)
Data Set 3
(n=15)
Units
pmol CYP2El/gram
liver
pmol CYP2El/mg
MSP
mg MSP/gram liver
pmol TCE/min/
pmol CYP2E1
pmol TCE
oxidized/minute/gram
liver
Parameter
A
B
C
D
E
Notes
Directly measured via
ELISA (n=20); and
separately predicted
statistically
Directly measured via
ELISA
Derived: C = A/B
Original measurements
from (11) corrected by
CYP2E1 content from
(18)
Statistically estimated:
E = B*C*D;
extrapolated to Vmaxc
4-35
-------
TABLE 4-2
Liver Enzyme Data*
SAMPLE
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
mg homog.
protein per gram
liver
134
101
137
100
113
151
154
148
115
137
181
180
126
124
122
137
152
130
69
126
CYP2E1 permg
Homog. Protein
16.1
25.6
17.9
15.2
10.9
12.2
25.0
12.4
21.5
20.9
24.6
27.8
21.4
21.9
22.3
24.4
16.3
24.1
25.2
14.0
Parameter A = pmol
CYP2E1 per gram
liver
2157.4
2585.6
2452.3
1520.0
1231.7
1842.2
3850.0
1835.2
2472.5
2863.3
4452.6
5004.0
2696.4
2715.6
2720.6
3342.8
2477.6
3133.0
1738.8
1764.0
Parameter B = pmol
CYP2E1 permg
MSP
85.5
99.8
83.4
23.0
34.0
36.3
76.8
46.0
69.0
58.5
54.0
64.0
68.0
53.0
46.0
66.0
41.0
24.0
42.0
41.0
52.5
94.0
46.5
90.0
11.0
64.0
41.0
64.0
30.0
57.5
53.5
55.0
52.0
29.0
39.0
39.0
73.0
70.0
130.0
34.0
31.0
48.0
29.5
19.0
77.0
91.0
37.0
74.0
Parameter C=
mg MSP per gram
liver = A/B
25.2
25.9
29.4
66.1
36.2
50.7
50.1
39.9
35.8
48.9
82.5
78.2
39.7
51.2
59.1
50.6
60.4
130.5
41.4
43.0
4-36
-------
TABLE 4-2 cont.
SAMPLE
49
50
51
52
53
54
55
56
57
58
59
60
mg homog.
protein per gram
liver
CYP2Elpermg
Homog. Protein
Parameter A = pmol
CYP2E1 per gram
liver
Parameter B = pmol
CYP2E1 permg
MSP
44.0
50.0
75.0
29.5
69.0
91.0
48.0
36.0
41.0
26.0
26.0
53.0
Parameter C=
mg MSP per gram
liver = A/B
*Source: Lipscomb et al., 2003
4-37
-------
TABLE 4-3
CYP2E1 Content and TCE Metabolic Activity Used to
Produce Data Set 3, Describing Parameter D
SAMPLE
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
Sample
number
(from 17)
HHM67
HHM84
HHM86
HHM88
HHM55
HHM60
HHM77
HHM78
HHM81
HHM82
HHM89
HHM144
HHM58
HHM61
HHM79
pmol TCE oxidized/
min/mg MSP
(from 17)
1113
1724
1039
1432
1422
1746
943
1627
1416
2353
890
1584
2078
2623
3455
pmol CYP2E1/
mgMSP
(from 15)*
11
64
44
50
52.5
91
19
77
37
74
30
30
94
90
91
Geometric Mean
Geometric Standard Deviation
pmol TCE oxidized/
min/pmol CYP2E1
(Parameter D)
101.2
26.9
23.6
28.6
27.1
19.2
49.6
21.1
38.3
31.8
29.7
53.7
22.1
29.1
38.0
32.5
1.538
*Data not used in estimation of parameter B.
4-38
-------
TABLE 4-4
Distributions of TCE Metabolism Rate Constant, Microsomal Protein
and CYP2E1 Content of Adult Human Liver
Parameter
Description
Distribution
GM
GSD
Range
5th Percentile
95th Percentile
A
CYP2Llva
(derived)
Discrete
2562
930
1232 - 5004
1232
4453
B
CYP2MSpb
(data set 1)
Log Normal
48.9
1.6
11-130
22.5
106
C
MSPLlvc
(data set 2)
Log Normal
52.9
1.476
27-108
27.9
100
D
TCEcYP2
(data set 3)
Log Normal
32.5
1.515
19.2-101.2
16.4
64.4
E
TCELlve
(derived)
Log Normal
78,810
1.7274
—
32,069
193,679
apmoles CYP2El/gram liver; data are presented as the arithmetic mean, arithmetic standard
deviation and values at the 5.89th and 95.5th percentiles, n=20.
bpmoles CYP2El/mg MSP, n = 60.
cmg MSP/gram liver, n = 20.
dVmax of CYP2E1 in human liver MSP toward TCE, pmoles/min/pmol CYP2E1, n = 15
eVmax as pmoles TCE oxidized/min/gram liver, derived via computational statistics
TABLE 4-5
Effect of Human Hepatic CYP2E1 Activity Distribution on the Bioactivation of TCE
Following an Inhalation and an Oral Exposure
Oxidation Rate
5th Percentile (6.6 mg/hr/kg)
95th Percentile (39.7 mg/hr/kg)
Difference
Liver Metabolites (ug/L)
Inhalationa
258.3
264.9
2%
Oralb
5.4
5.5
2%
Simulations of 8 hr exposure to TCE (50 ppm) by a 70 kg male human as described under
Methods.
bSimulations of exposure to TCE in drinking water (5 |ig/L; 2 L) over 24 hr by a 70 kg male
human as described under Methods.
4-39
-------
Toxicity and TK Testing
Defined
Mode of
Action
Exposure
Internal Dose
T
Distribution
Metabolism
T
Concentration in
Target Tissue
T
Response
(LOAEL, NOAEL,
BMD)
Risk Assessment
Exposure
Internal Dose
Distribution
A
Metabolism
A
Concentration in
Target Tissue
A
Response
FIGURE 4-1
Application of PBPK Modeling to Link External Dose with Concentration of Toxicant in Target
Organs. The approach builds upon information on mode of action, which demonstrates the
relationship between tissue response, a PD phenomenon, and the PK outcome directly related to
that response. PBPK models are then developed and employed to define the relationship
between the external exposure and the target organ toxicant concentration, usually performed in
test animal species. Once completed, and based on assumptions about the similarity in the
qualitative and quantitative nature of the PD effect (mode of action) between the test species and
humans, parallel PBPK models are developed for the human, and are exercised to "back-
track'the toxicant from a predetermined concentration in the target tissue to the corresponding
exposure concentration (external dose).
4-40
-------
Isolation of Microsomal Protein
LIVER
Homogenation
and
Centrifugation
Nuclei, Debris,
Cytoplasm
V
MICROSOMAL
PROTEIN
Enzyme Activity
Metabolite per:
• mg Microsomal Protein,
• pmol CYP2E1
Enzyme Content
pmolesCYP2El/mg
Microsomal Protein
FIGURE 4-2
Relationship Between Intact Liver, Microsomal Protein and Some CYP Forms. The isolation of
microsomal protein from intact liver via homogenation of tissue and differential centrifugation
results in a 100,000 x g pellet which is enriched for endoplasmic reticulum content. The
enrichment results in an artificial increase in the concentration of biological components
associated with the endoplasmic reticulum. This isolation produces a fraction (microsomes;
MSP) which is subjected to in vitro investigations of metabolic activity and enzyme content.
However, a quantitative relationship to the intact liver is not possible without further information
on the distribution of microsomal protein to the intact liver.
4-41
-------
0)
>
140
120 -
100 -
E 80 -
2
o
51 60 -
E 40 H
20 -
r2 = -0.293
20 40 60 80 100
pmoles CYP2E1/mg MSP
120
FIGURE 4-3
Correlation Between mg MSP/gram and pmol CYP2El/mg MSP. The slight, but statistically
significant, correlation between the two parameters dictated the choice of statistical methods.
Data from Lipscomb et al. (2003).
4-42
-------
Average A
# of records
2642.77
20
B
[23, 53)
Avg=2029.23
std=614.41
n=9
Cluster 1
[25.233,36.226)
avg=2416.88
std=182.68
n=4
Cluster 2
[53,99.8]
Avg =3144.75
n=ll
[36.226,59.143)
avg=3093.62
std=496.97
n=5
Cluster 3
[59.143, 130.542]
avg=4728.3
std=389.9
n=2
Cluster 4
FIGURE 4-4
Classification Tree Model for the Distribution of A = pmol CYP2El/gram Liver. The
distribution of A is modeled as a finite mixture distribution with components corresponding to
the leaves in the depicted classification tree. The conditional distribution of A depends on which
component distribution a case belongs to. The components are bounded by breakpoints in the
observed values for B and C. The sample means and sample standard deviations for the four
component distributions are estimated from the first 20 cases in Table 4-2.
4-43
-------
LoglO (pmol CYP2El/m
MSP), n = 60
Convert Units
Relevant
xoosure Scenario
LoglO (mg MSP/gram
Liver), n = 20
PBPK Model!
Logio (pmol TCE/Min/ | Magnitude of Variance in PK Output
pmol CYP2E1), n = 15 ^^^^^^^^^^~^^^^^^^^^^~
FIGURE 4-5
Extrapolation and Incorporation of in vitro Derived Metabolic Rates in PBPK Modeling. This
figure depicts the framework for deriving appropriate in vitro measures and their extrapolation
into a PBPK model. The model was exercised to simulate environmentally and occupationally
relevant exposures.
4-44
-------
APPENDIX
MATLAB Code to Generate AxD
numReps = 250000;
cluster{l}= [1520
cluster{2}=
cluster!3}=
cluster!4}=
clusterCDF=
1231.7 1842.2
4 2585.6 2452,
2863.3 2696.4
6 5004] ;
70 .91 1.0];
1835.2 2720.
3 2472.2] ;
2715.6 3342.
[2157
[3850
[4452
[.53
A = zeros(1,numReps);
result = zeros(1,numReps);
for i = 1:numReps
D = exp(3.4812 + .4156*randn);
clusterNum = min(find(rand < clusterCDF))
clusterSize = length(cluster{clusterNum})
clusterlndex = ceil(rand*clusterSize);
A(i) = cluster{clusterNum}(clusterlndex);
result(i) = A(i)*D;
end
writeArray = result';
save Dvalues.txt writeArray -ASCII
disp('A parameters')
disp([mean(A) std(A)])
disp('AxD parameters')
disp([mean(result) std(result)])
hist(result,100)
title('Empirical Distribution of A x D')
figure(2)
hist (log(result),100)
title('Empirical Distribution of In (A x D) ' )
6 2477.6 3133 1738.
8]
1764];
4-45
-------
5. PHARMACOKINETIC ANALYSES TO SUPPORT AN INHALATION RfC FOR
CHLOROFORM
BACKGROUND
Chloroform (CHCb) is a problematic drinking water contaminant and byproduct of the
disinfection (DBF) of drinking water with chlorine (Richardson, 1998; Rook, 1974). Chloroform
is carcinogenic to rats and mice; and the assessment of its risk involves nonlinear concentration
extrapolation because of the involvement of cytotoxicity. Cytotoxicity appears to be dependent
on the formation of a reactive metabolite, phosgene, from chloroform and its interaction with
cellular components. When the damage due to the formation of reactive phosgene exceeds the
normal repair capacity, cytotoxicity occurs, predisposing the tissue to a carcinogenic response.
Chloroform metabolism has been assessed through various experimental designs, in
several mammalian species, and with differing degrees of specificity. Studies with rodent liver
preparations have indicated that cytochrome P450 2E1 (CYP2E1) is the major enzyme in
metabolizing low concentrations of chloroform. Other enzymes can contribute to chloroform
metabolism at higher chloroform concentrations (Testai et al., 1996); however, these higher
concentrations are unlikely to be found in the tissues of humans exposed to chloroform through
drinking water or by inhalation, even at occupational exposure limits (threshold limit value and
time-weighted average value of 10 ppm; ACGIH, 2003).
Physiologically based pharmacokinetic (PBPK) modeling has been usefully applied to
extrapolate tissue concentrations of toxicants across dose, route, and species for health risk
assessment. It also allows for the interjection of realistic measures of human interindividual
biochemical and physiologic variability into the determination of tissue doses of toxicants
(Lipscomb and Kedderis, 2002). This report employs PBPK modeling to combine measures of
5-1
-------
biochemical, physiologic, and anatomic variability between test animal species and humans, as
well as among human life stages, to quantify the risk-relevant pharmacokinetic outcome, i.e.,
liver metabolism of chloroform, under inhalation exposure scenarios. The results are considered
in the context of evaluating the role of species and human interindividual toxicokinetic (TK)
variability in determining uncertainty factors for risk assessment (Figure 5-1).
Purpose. At present, the risk assessment for chloroform does not include a Reference
Concentration (RfC) for inhaled chloroform. The purpose of this analysis is to provide
information on tissue dosimetry (concentration of the metabolite in liver), including differences
between rats and humans as well as variability among humans, to support the derivation of the
uncertainty factors governing inter- and intraspecies TK for subsequent derivation of the RfC.
Objective. The objective of this analysis is to identify several anatomic, physiologic, and
biochemical parameters associated with the distribution and metabolism of chloroform and to
apply PBPK modeling in the rat and human to quantify species and human-interindividual
differences in the amount of chloroform metabolized at steady state during a continuous
exposure.
RISK ASSESSMENT APPLICATION
This approach was developed to reduce uncertainty in the derivation of the RfC for
chloroform by applying PBPK modeling to quantify species extrapolation and to provide a
framework through which the impact of quantified human interindividual variability in important
model parameters could be evaluated with respect to the risk-relevant internal dose metric.
Animal bioassay data were evaluated via benchmark dose modeling. After identifying a no-
effect intermittent exposure in mice, PBPK modeling was applied to transform the external
concentration to an internal dose metric, the value of which was used as a point of departure for
5-2
-------
Toxicity and TK Testing
Exposure
i
Internal Dose
Defined
Mode of
Action
Distribution
*
Metabolism
Concentration in
Target Tissue
Response ^—^—
(LOAEL, NOAEL, BMD)
Risk Assessment
Exposure
Internal Dose
Distribution
t
Metabolism
Concentration in
Target Tissue
.Response
FIGURE 5-1
A Conceptual Presentation of the Application of TK Information and the TK Approach to
Human Health Risk Assessment. Data from animal studies can identify the target organ or
tissue, the lexicologically relevant form of the chemical (parent or metabolite), and can describe
the link between metabolism and response. With this information, and information sufficient to
develop TK models for the test species and the human, some linkage can be made based on
knowledge or assumptions about species sensitivity to the lexicologically active form of the
chemical. With this information, and these models in hand, the animal model can be exercised to
simulate the appropriate dose metric under NOAEL exposure conditions, and the human model
can be "run backwards" to correlate the level of the appropriate dose metric with an exposure
concentration.
5O
-------
further modeling in humans, with specific attention devoted to assessing age-dependent and adult
interindividual variability of several key physiologic and biochemical parameters. These
parameters include blood:air partitioning of chloroform, hepatic blood flow (HBF), and
CYP2E1-dependent bioactivation of chloroform. The results are expressed relative to external
concentrations, consistent with EPA policy regarding the derivation of the Human Equivalent
Concentration for inhaled substances.
SCOPE AND LIMITATIONS
This analysis was constrained for several reasons. The mode of action of chloroform
involves metabolism to phosgene, and chloroform metabolism (disappearance of parent
chemical) was measured using a closed and recirculating system gas uptake with rat and human
microsomal proteins (separately). As such the assumption is that the formation of reactive
metabolites from chloroform is accurately indicated by the disappearance of parent chemical.
Second, the liver was chosen as the site of toxicity by considering the prevalence of liver injury
noted in animal studies and humans, and the lack of epidemiologic information indicating
respiratory (nasal) toxicity in chloroform-exposed humans (results not here presented). Third,
the analysis was undertaken to evaluate only two of the five non-cancer uncertainty factors, those
relating to inter- and intraspecies uncertainty/variability. Further, the analysis was conducted to
evaluate only the toxicokinetic component, and did not include a technical consideration of inter-
and intraspecies variability in response (toxicodynamics). Any presentation of reference
concentrations derived from suggested uncertainty factor values should be considered as
illustrative examples. In conclusion, the present work was aimed at developing a human
equivalent concentration for the inhalation exposure to chloroform and developing and
5-4
-------
presenting results that will be useful in establishing a value for the toxicokinetic component of
intraspecies uncertainty factor.
APPROACH
The approach is based on toxicological information identifying the metabolism of
chloroform as a causative factor of damage leading to cytotoxicity in the liver. Recently, an
expert panel was convened by the International Life Sciences Institute / Health and
Environmental Sciences Institute to evaluate the mode of action of chloroform (Andersen et al.,
2000; ILSI, 1997). This panel concluded that the body of available evidence was sufficient to
warrant confidence in oxidatively-derived chloroform metabolites as causative in tissue
insult/injury, and that carcinogenicity was secondary to chloroform-induced cytotoxicity. The
panel concluded that the "likely mode of action was the accumulation of reactive, oxidized
metabolites, leading to cell toxicity" but was unable to distinguish between the amount of
metabolite formed per unit tissue or the rate of metabolism (Andersen et al., 2000; ILSI, 1997).
They recommended that critical studies of chloroform cytotoxicity should be conducted on the
basis of rates or amounts of metabolite produced per volume of tissue, and that a PBPK
modeling-based approach including "specific parameter (value)s related to metabolic rates,
enzyme affinities and enzyme tissue distribution" be undertaken. The panel indicated that
"measures of effective dose related to the rate of metabolism, or rate of formation of phosgene,
would be expected a priori to be meaningful", and that the mean daily rate of chloroform
metabolism in tissue would be an appropriate dose measure. The approach taken in this
document is consistent with the recommendations of the ILSI expert panel.
While some investigators have begun to address "rate" of chloroform oxidation, this
investigation has relied on the accumulation of metabolite, specifically the cumulation of
5-5
-------
metabolism over a 24-hour period, as the dose metric for evaluation. This approach is absolutely
consistent with the recommendations developed by the ILSI-convened panel of experts. Here,
chloroform metabolism is assessed as CM24, the integrated amount of chloroform metabolized
over a 24-h period, expressed in milligrams per liter (mg/L) of liver tissue. The 24-h time frame
is based on two considerations: (1) chloroform is cleared from mouse liver before the end of a
24-h period beginning at the initiation of a 6-h exposure period (5 ppm), i.e., each 24-h period is
independent and (2) the 24-h period is applied in humans, due to the relationship between the
intent of RfC values to reflect risk/safety under continuous-exposure conditions. Determining
CM24 under steady state conditions avoids underestimating metabolism, which may occur when
extrapolating the results of an intermittent exposure to a chronic-exposure situation. For
extrapolating animal effects, the CM24 resulting from "no-effect" exposure scenarios identified
from inhalation bioassays will be used as the point of departure.
Following a practice for which precedent exists in IRIS files for other chemicals, the
mode of action for chloroform has been determined and related to the formation of
lexicologically active metabolites via oxidation by cytochrome P450. While other metabolites
are formed, and are also reactive in nature, those metabolites formed by and following the
oxidation of chloroform make up more than 90% of metabolized chloroform. Hence, this
analysis will focus on the metabolic pathway catalyzed by cytochrome P450 2E1 (CYP2E1).
The internal dose metric, CM24, will be the basis of comparison of exposures between and
among species in describing the human equivalent concentration (HEC) and selection of values
for the TK component of the uncertainty factor addressing human interindividual
variability/uncertainty (UFn) (Figure 5-2).
5-6
-------
This analysis includes specific information on several factors that vary within the human
population in order to address the human interindividual differences in chloroform TK. These
factors are evaluated by quantifying their effect on the external concentrations required to derive
the CM24 values of interest. Because CYP2E1 catalyzes the oxidation of many low molecular
weight solvents and environmental contaminants, and because its expression among humans
varies appreciably, this analysis includes quantitative information on CYP2E1 variance among
humans. Furthermore, because delivery of chloroform to the liver tissue through HBF can be a
limiting factor in its metabolism at low blood concentrations, the analysis also includes
quantitative information on the variability of HBF among humans. Specific human models have
been constructed to simulate the adult male, the obese adult male, the adult female, the 1-year-
old and the 9-year-old child. These characteristics (quantified variability of chemical-specific
metabolic capacity, and quantified variance of FffiF) set the present analysis apart from its
predecessors. This is the first time, to our knowledge, that these data have been combined in the
analysis of tissue dosimetry in humans for conducting a human health risk assessment.
Briefly, the analysis will follow several steps. Steps 1 to 4 are depicted in Figure 5-3.
1. A PBPK model will be developed using the structure for styrene (Ramsey and Andersen,
1984; Figure 5-4) and parameter values for chloroform, including partition coefficients
and metabolic parameters deemed appropriate, or developed, for rats, mice, and humans.
2. The PBPK model for mice and rats will be used to simulate CM24 based on the exposure
conditions identified from inhalation bioassays. The lowest CM24 value will be used for
extrapolation to humans.
3. The PBPK model for the adult male human will be parameterized to include human-
specific parameters including body weight, organ volumes and flows, and metabolic
parameters.
5-7
-------
UF
A-TK
UF
A-TD
UF
H-TK
UF
H-TD
Animal
Exposure
_^ General.
Human
Human
Equivalent
Concentration
Interspecies
Extrapolation
Complete
_». Sensitive
Human
FIGURE 5-2
Subdivision of Uncertainty Factors into Toxicokinetic and Toxicodynamic Components. The
International Programme on Chemical Safety (IPCS, 2001) has released draft guidance on the
development of Chemical-Specific Adjustment Factors (CSAF), which demonstrates the
separation of UFA and UFn into respective TK and TD components, allowing replacement of
default values with data-derived values. With respect to animal-to-human extrapolation, this is
consistent with U.S. EPA policy governing the development of HECs in the inhalation RfC
methodology, which addresses species differences in toxicokinetics.
-------
Use of PBPK Models to Demonstrate
Exposure Conditions in Mice and
Humans Anticipated to Result
in the Same CM24 Value
Mouse
Human
PBPK
1
Internal Dose
I
Distribution
I
Metabolism
I
Exposure
Response
Models
Internal Dose
t
Distribution
t
Metabolism
t
_^. Continuous Exposure
At Steady State,
Human Equivalent
Concentration
FIGURE 5-3
Application of PBPK Modeling to Extrapolate Internal Dosimetry Between Species. The
development of the Human Equivalent Concentration from mouse no-effect levels was
accomplished based on CM24, the integrated amount of chloroform metabolized over a 24-h
period, expressed per L of liver tissue. Studies with mice indicated a NOAEL exposure of 5
ppm, which was not duration-adjusted. The mouse PBPK model was exercised to reveal CM24
in the mouse, and the human model was exercised over a range of continuous exposure
concentrations to reveal a linear relationship between CM24 and exposure concentration. This
relationship was used to determine the human exposure condition necessary to develop the CM24
value obtained from simulations of the mouse exposure conditions (see Figure 5-6).
5-9
-------
Q
alv
C.
inh
Q
ven
'vf
vs
vr
'vl
Q
ALVEOLAR SPACE
LUNG BLOOD
alv
'alv
cv=cf/Pf
FAT
Vf
cw = c / P
v s s
POORLY PERFUSED
TISSUES Vs
c = c / P
v r r
RICHLY PERFUSED
TISSUES Vr
Cv=C,/P,
LIVER
V,
V
max
Q
'art
Q.
'art
Q.
'art
Q.
'art
Q,
'art
METABOLISM
FIGURE 5-4
Five-Compartment PBPK Model. A five-compartment PBPK model was developed to address
chloroform dosimetry and metabolism in rats, mice, and humans, based on structure developed
and originally published for styrene (Ramsey and Andersen, 1984). This structure has been
successfully applied for a range of volatile organic compounds.
5-10
-------
4. Using values for these parameters that are representative of the median values for adult
male humans, the model will be used to simulate an inhalation exposure to chloroform in
inspired air. Through an iterative process, the model will determine the duration of
exposure and the concentration of chloroform in inspired air required to reach the steady
state that produces a value for CM24 equivalent to that observed in the identified animal
exposure scenario. This will represent the HEC.
5. The adult male human PBPK model will be re-parameterized to reflect human
interindividual differences in the Vmax for CYP2E1-dependent chloroform metabolism,
HBF, and blood:air partitioning of chloroform.
6. The adult male human PBPK model will again be used to demonstrate the effects of
parameter variability on CM24. Model simulations will demonstrate differences in
exposure concentrations necessary to obtain a CM24 value equivalent to that
demonstrated at exposure to (HEC/3) ppm. The adult female model will be exercised to
simulate this exposure to compare CM24 values.
7. Finally, the human PBPK model will be re-parameterized to include the values for
parameters known to differ between adults and children. Those include fractional
contribution of the fat and liver compartments to body mass, metabolic parameters for
chloroform, and values for tissue partition coefficients. Because discussions with the
Office of Water have identified the reliance on 10 kg as a representative body mass for
some risk analyses with chemicals, the model will specifically include parameters on
body composition derived from reference values for that body mass. Additional analyses
will address the 30 kg child. These body masses are representative of children aged 1
and 9 years, respectively.
8. The child PBPK models will be used to compare exposure concentrations necessary to
reproduce the value for CM24 observed at the HEC in the adult human model containing
median values for parameters whose values are varied.
APPLICATION OF PHYSIOLOGICALLY BASED PHARMACOKINETIC MODELING
PBPK modeling will be applied to transform external concentrations to measures of
internal doses for species extrapolation in derivation of the inhalation RfC. The model structure
(Figure 5-4) is that of Ramsey and Andersen (1984) developed for styrene and successfully
applied to numerous substances since its publication two decades ago; source code is presented
as Appendix A. The derivation of the FIEC follows EPA methodology on RfCs (U.S. EPA,
1994) related as differences in external exposure under the conditions of maintaining the internal
dose metric between species. Here, the value of CM24 obtained from model predictions based on
5-11
-------
bioassay exposure conditions is standardized between species. The human PBPK model predicts
the continuous exposure, under steady state conditions, that would be required to produce that
value for CM24. Likewise, variability within humans is evaluated as the difference in the
concentration of chloroform in inspired air required to produce the same CM24 value. These
differences can be considered during the process of estimating the value for the TK component
ofUFH.
This approach differs in several areas from the traditional use of default uncertainty
factors to extrapolate animal findings to humans. Under the traditional NOAEL approach, a no-
or lowest-observed-adverse-effect level is determined in animals and adjusted for duration
(subchronic to chronic) to account for animal-to-human differences as well as human
interindividual differences. Each of these three factors has a default value of 10. In the present
approach, division of the animal-to-human uncertainty factor (UFA) and human interindividual
uncertainty factor (UFH) into their respective TK and toxicodynamic (TD) components provides
the structure for the inclusion of TK information on chloroform metabolism between species
(rats and humans) and among humans into the model. Sufficient information describes a mode
of action and its causative chemical moiety (i.e., the oxidative metabolite) toward which PBPK
modeling may be aimed. Pharmacokinetic and species-specific information are sufficient to
describe the development of species-specific PBPK models to allow this more comprehensive
approach to be used in developing values for uncertainty factors.
The approach undertaken develops an HEC based on pharmacokinetic information
pertinent to liver toxicity as the critical effect. The derivation of the HEC obviates the TK
component of the animal-to-human uncertainty factor (UFA-TK). The remaining TD portion of
the uncertainty factor (UFA-TD) remains, and will be addressed using a default value of 3 (half an
5-12
-------
order of magnitude). Additional data demonstrating human interindividual differences in HBF,
which is a major limiting factor in chloroform metabolism, will be incorporated in the adult
PBPK model. Additional data describing metabolic variability available from measured levels of
CYP2E1 in adult and child liver tissues quantitative measures of CYP2E1 activity toward
chloroform and human variability in blood:air partitioning of chloroform will be used to
construct models useful in addressing human interindividual variability. This representation of
intraspecies differences, as well as sex- and age-dependent variability in blood flows, ventilation,
body weight and compartment sizes, will be incorporated in the model to ascertain their
influence on chloroform metabolism in vivo. This process will be used to support a value to
replace the default uncertainty factor addressing the TK component of the human interindividual
uncertainty factor (UFH-TK). Figure 5-1 illustrates the application of TK information and
concepts in human health risk assessment. Figure 5-3 demonstrates the concept of using such
information to develop the HEC for chloroform. We will develop this FIEC based on CM24 of
the most sensitive species (i.e., the mouse) as the point of departure.
PBPK modeling for the human will be conducted at steady state, based on the concept
that the RfC is for a chronic, lifetime-continuous exposure. Because data demonstrate species-
dependent differences in biochemical parameters (e.g., tissue partitioning and metabolism)
exposure influences the time it takes to reach steady state. Model simulations of species
differences in intermittent exposures will be complicated by species-dependent differences in
biochemical parameters. Thus, species differences in the amount of chemical metabolized (i.e.,
the risk-relevant TK outcome) may be influenced by exposure duration. However, because
chloroform is cleared from the mouse liver within 18 hours after the end of a 6-h, 5 ppm
exposure, CM24 will be estimated in mice based on a single cycle of a 6-h-on/18-h-off exposure.
5-13
-------
This model is constructed to demonstrate the CM24, the 24-h integrated amount of chloroform
metabolized, normalized per liter of liver tissue. It is expressed in units of concentration, mg
chloroform metabolized per liter liver (mg/L). In that respect, it differs in output metric and
application from models developed and employed to simulate intermittent exposures (i.e., Tan et
al., 2003).
METHODS: PBPK MODEL STRUCTURE AND PARAMETERS
Model Structure. The five-compartment structure developed by Ramsey and Andersen (1984)
for styrene was employed as the template. This model structure (Figure 5-4) assumes that the
concentration of chemical in blood coursing through tissues of the body comes to equilibrium
with concentrations of the chemical in tissues and air during the duration of perfusion.
Partition Coefficient Derivation. Blood:air PC (B:A PC) and tissue:air (T:A PC) values were
developed using the vial equilibration method (Sato and Nakajima, 1979), combined with gas
chromatographic evaluation of chemical concentration, quantified by an external standard curve
of known concentrations. This method was employed to determine T:A PC values for
chloroform in blood of rats and humans, and in solid tissues from rats for this study. Because
initial intentions were to rely on the rat as the most sensitive species, because of the larger tissue
masses available from rats and because of the marked similarity across species in the solid-
tissue:air partition coefficients (Thomas, 1975), tissue:air partition coefficients were derived
using liver, kidney, muscle and adipose tissues from rats. Complete results, and a comparison of
observed values to values predicted using structure-activity relationships, are contained in
Appendix B. Samples of blood from rats, mice, and humans were exposed to a mixture of
chloroform and five other volatile organic compounds; the concentrations of the individual
organic compounds in the headspace were determined at equilibrium. This approach has been
5-14
-------
determined to produce PC values equivalent to those obtained with single chemicals in the vial
equilibration method (Beliveau et al., 2001). All studies were conducted through institutionally
reviewed protocols. Human blood samples were remnant samples (unperturbed blood remaining
after analytical procedures had been performed) obtained from the Division of Clinical
Laboratory Services, Wright-Patterson Regional Medical Center, OH and coded to ensure
anonymity of the subjects. The study protocol was reviewed and approved by the Wright-
Patterson Institutional Review Board and the EPA Human Studies Coordinator. Samples were
provided from 11 adult males, aged 36 to 80 years, 10 adult females aged 22 to 87 years, and 11
children aged 3 to 7 years.
PBPK models incorporate partitioning information as tissue:blood partition coefficient
(T:B PC) values, which represent the ratio of chemical concentrations in tissue:blood at
equilibrium. Like most PBPK models, the model structure used in this examination was based
on the assumption that the duration of perfusion of tissues (organs) is sufficient for tissue and
blood concentrations of parent chemical to come to equilibrium, at concentrations dictated by the
T:B PC value for that tissue. Tissue:Blood PC values are determined by Equation 5-1:
T:APC/B:APC = T:BPC (5-1)
Mice. Partition coefficient values used to populate the mouse PBPK model were taken from Tan
et al. (2003), which relied, in part, on PC values originally reported by Corley et al. (1990).
These values can be found later in Table 5-7. Because they were not experimentally derived for
the present investigation, they are not here demonstrated.
Adult Rats. The T:A PC values derived from tissues obtained from 10 adult male rats are
presented in Table 5-1, and the resulting T:B PC values are presented in Table 5-2. The B:A and
T:B PC values were used to parameterize the rat PBPK model for chloroform.
5-15
-------
TABLE 5-1
Adult Rat Blood: Air and Tissue: Air Partition Coefficient Values*
Mean
S.D.
Blood:Air
17.7
2.5
LiverAir
17.6
3.2
Kidney:Air
14.8
2.7
Muscle:Air
16.9
10.1
Fat:Air
351
48
*Ten adult male Fischer 344 rats were employed in this investigation. B:A PC and T:A PC
values were derived using tissues from individual rats.
TABLE 5-2
Adult Rat Tissue:Blood Partition Coefficient Values Derived from Paired Tissues*
Mean
S.D.
LiverBlood
1.0
0.2
Kidney :Blood
0.8
0.1
Muscle:Blood
1.0
0.8
FatBlood
19.9
2.0
*B:A PC and T:A PC values were derived using tissues from individual rats, and rat-specific T:B
PC values developed using animal-matched samples were used in the rat PBPK model.
5-16
-------
Adult Humans. Blood: Air PC values were derived for chloroform in 21 remnant blood samples
from adult humans (Table 5-3). The mean value of 11.34 was combined with rat T: A PC values
to compute T:B PC values for inclusion in the adult human PBPK models for chloroform.
Thomas (1975) previously demonstrated that, for methylene chloride, T:A PC values across
species (rat to human) are more consistent (rathuman ratios of 1.0 to 1.2) for liver, kidney,
muscle, and fat, than for blood (rathuman ratio of 1.9). Therefore, the human B: A PC value was
combined with adult rat T:A PC values (presented in Table 5-1) to compute T:B PC values
representative of the adult human (Table 5-4). These values were used to parameterize the adult
human PBPK model for chloroform.
TABLE 5-3
Adult Human Blood:Air Partition Coefficient Values
Males, n = 1 1
Females, n = 10
Combined, n = 21
B: A PC value
11.9 + 0.9a
10.7 + 2.1
11.34+ 1.65b
Range of Observations
9.7- 13
6.9- 13.3
aData presented as mean + S.D. Mean values for men and women were not significantly
different as determined by one-tail t-tesi assuming unequal variance (p = 0.063); and by two-
tailed Mest assuming unequal variance (p = 0.126).
b Because there was no statistical difference between males and females, we chose to combine
the data.
TABLE 5-4
Tissue:Blood Partition Coefficient Values for Adult Humans*
Mean
S.D.
LiverBlood
1.6
0.3
Kidney :Blood
1.3
0.2
Muscle:Blood
1.5
0.9
FatBlood
31.0
4.2
*Blood from 21 adult humans and solid tissues from 10 adult male Fischer 344 rats were used in
this investigation. A B: A PC value of 11.34 (the human mean) was combined with individual
T:A PC values to develop 10 estimates of individual T:B PC values (via equation 5-1). Data are
presented as the mean and S.D. of those 10 values.
5-17
-------
Children. Remnant blood samples obtained from 7 male and 4 female pediatric patients (age 3
to 7 years) were used to determine B:A PC value for children. The B:A PC values demonstrated
a mean and standard deviation of 12.41 + 1.17. Samples of liver, kidney, muscle, and fat were
taken from 8 male and 8 female rat pups on postnatal day (PND) 10 and used to determine T:A
PC values (Table 5-5). This report relies on B:A PC values derived from blood from children
aged 3-7 years to inform T:B PC values used in PBPK models constructed for the 1-year-old and
the 9-year-old child (Table 5-6). No data describing the B:A PC values in these specific ages
were available; no age-specific information on blood composition suitable to inform structure-
activity-based predictions (i.e., water content, total lipid content, neutral lipid content) were
available in Reference Man or its update (ICRP, 1975, 2002). The validity of this generalization
is based on several pieces of information and some assumptions. Malviya and Lerman (1990)
determined the blood:air PC values for halothane, sevofluorane and isofluorane (two-carbon
halogen substituted anesthetics) in blood from preterm neonates, full-term neonates and adults.
Their results indicated that B: A PC values increase from the neonatal period to adulthood and
that the values increase less than 14% between the neonatal period (average age 39 weeks) and
adulthood. Lerman et al. (1984) demonstrated B:A PC values for halothane, and isofluorane in
infants at delivery, children aged 3-7 years and adults. Those results indicated B:A PC values in
children aged 3-7 that were intermediate between values observed in infants and in adults. Given
the magnitude of increase between children (aged 3-7 years) and adults, it seems that the B:A PC
values characterized for this group are sufficient for children aged 9 years.
Parameter Values: For this evaluation, the model structure has been reparameterized for rats,
mice, and humans and to include age-specific parameter values for humans (Table 5-7) with
newly available chloroform-specific partitioning data and chloroform-specific metabolism
5-18
-------
TABLE 5-5
Tissue: Air PC Values in PND 10 Rat Pups
Mean
SD
Liver: Air
17.4
1.3
Kidney:Air
12.6
0.9
Muscle:Air
34.3
24.7
Fat:Air
243
48
TABLE 5-6
Tissue :Blood PC Values Used for Children PBPK Model*
Mean
SD
LiverBlood
1.4
0.1
Kidney :Blood
1.0
0.1
Muscle:Blood
2.8
2.0
FatBlood
19.6
3.9
*Results were obtained by combining the mean B:A PC value (12.41) determined from 11
pediatric patients with T:A PC values for solid tissues presented in Table 5-5 via equation 5-1.
5-19
-------
TABLE 5-7
Age and Species Dependent Model Parameter Values
BW
KM
vmaxc
— 5th Percentile
vmaxc
— Geo. Mean
vmaxc
— 95th
percentile
QPC
QCC
VFC
VLC
QLC
QFC
PB
PL
PR
PS
PF
Adult Male
70
0.012
4.306
8.956
15.56
15
15
0.19
0.026
0.25
0.09
11.34
1.6
1.6
1.5
31
Adult
Female
60
0.012
4.306
8.956
15.56
15
15
0.21
0.026
0.25
0.09
11.34
1.6
1.6
1.5
31
Child, 9 yr
30
0.012
3.595
7.866
17.212
15
15
0.158
0.0268
0.25
0.09
12.41
1.4
1.4
2.8
19.6
Child, 1 yr
10
0.012
3.304
7.23
15.82
15
15
0.1245
0.03425
0.25
0.09
12.41
1.4
1.4
2.8
19.6
Rat
0.25
0.012
ND
5.218
ND
14
14
0.07
0.04
0.25
0.09
17.7
1.0
1.0
1.0
19.9
Mouse
0.025
0.352a
ND
22.8a
ND
23
23
0.10
0.055
0.25
0.09
24. lb
0.7b
0.7b
0.54a
10.04a
aFrom Corley et al. (1990).
bFrom Tan et al. (2003), PL = PR.
ND = not determined.
PB = Blood: Air PC Value; PL = LiverBlood PC Value; PR = PC Value for rapidly perfused
tissues, derived using liver tissue; PS=PC Value for slowly perfused tissues, derived using
muscle tissue; PF = FatBlood PC value. Values for the body composition of fat (VFC) and liver
(VLC) of children were obtained from ICRP (1975).
Variability in hepatic blood flow (QLC) is presented in Table 5-8.
5-20
-------
TABLE 5-8
Parameters and Selected Percentile Values for the Fraction of Cardiac Output
as Hepatic Blood Flow
Distribution
Normal
Lognormal*
Beta
Parameter and Value
Mean = 0.273,
Standard Deviation =
0.087
Geometric Mean = 0.258,
Geometric Standard
Deviation = 1.411
a = 6.865,
P = 18.269
Percentile
5th
0.130
0.147
0.141
50th
0.273
0.259
0.267
95th
0.417
0.456
0.427
*Values from the lognormal distribution were included in the pharmacokinetic analysis.
5-21
-------
information. As demonstrated in the table, the adult female differed from the adult male with
respect to body weight and fraction of body weight represented by the fat compartment. In
addition, a model was constructed for the obese adult male human by modifications made
relating to the fat compartment.
One measure of obesity in adults is the body mass index (BMI) (CDC, 2005). The BMI
is defined as BW in kg divided by the square of height in meters. The normal range of BMI in
adults is 18.5 to 24.9. The CDC (2005) defines an obesity as a BMI greater than 30. This
information was used to develop physiological parameters for an obese male human.
A 70 kg male with a BMI of 21.7 is defined as a "normal" individual. From the
definition of BMI, the "normal" individual has a height of 1.796 meters (70/(1.796)2 = 21.7).
Assuming that the obese individual has the same height as the "normal" individual but has a BMI
of 30, the obese individual weighs 97 kg (97/(1.796)2 = 30). Assuming that the increase in BW
in the obese individual is solely due to fat accumulation, the liver in the obese individual would
weigh the same as the liver of the "normal" individual, 1.82 kg (VLC = 0.026*BW, Table 1). In
the "normal" individual, the fat compartment (VFC = 0.19*BW, Table 1) would weigh 13.3 kg.
If the additional weight (27 kg) in the 97 kg obese individual was entirely due to an increased fat
compartment, then the fat compartment in the obese individual would be 13.3 + 27 = 40.3 kg.
These calculations indicate that an obese individual would have a VLC of 0.019 * BW and a
VFC of 0.415 *BW.
The model has been further modified to allow for the naturally occurring variance in the
HBF in humans (Table 5-8). The model contains parameter values and details for the mouse that
differ from some other reports. Many of the values for physiological parameters used in Corley
et al. (1990), such as organ volumes and flow rates, are nonstandard. That model also included a
5-22
-------
pathway for CYP destruction in male mice that Amman et al. (1998) showed does not occur, and
which was not included in another recent chloroform PBPK model (Tan et al., 2003). Therefore,
we used recommended standard mouse physiological values from Arms and Travis (1988), as
had been used by Kedderis and Held (1996) for their study of furan pharmacokinetics in rats,
mice, and humans. The B:A and liverair (L:A) partition coefficients from Tan et al. (2003) were
used, along with the values for fat and muscle partition coefficients from Corley et al. (1990), to
calculate T:B PC values. Partition coefficients used in the models for rats and humans are based
on B:A and T:A PC values determined using samples from young and mature animals/subjects
for this investigation. For the mouse model, PC values for B:A, liverair and kidney:air were
taken from Tan et al. (2003) and values for fatair and muscle:air were taken from Corley et al.
(1990). Note that these values from Corley are also used in the Tan et al. work. Both Tan and
Corley relied on values for VmaxC and Km that were optimized by fitting to separately-
developed gas uptake data. Because of the number of successful models based on or adapted
from the Corley et al. (1990) model and/or its parameter values (i.e., Constan et al., 2002;
Corley et al., 2000; Delic et al., 2000; Evans et al., 2002; Gearhart et al., 1993; Reitz et al.,
1990), we chose to incorporate the VmaxC and Km values from the Corley model, rather than
from the Tan model. More recently derived PC values for blood, liver and kidney were available
from the work of Tan, and those values were employed in the present work.
Paired measurements (obtained simultaneously from the same person) of cardiac output
(CO) and HBF were available for 35 human subjects and databased predictions of CO were
developed for an additional 234 subjects for which hepatic blood flow measurements were
available (see Appendix C). Only measures of FffiF determined by indocyanin green (ICG)
clearance were employed. A total of 35 sets of paired (individual) observations were available in
5-23
-------
the open literature (Caesar et al., 1961; Wiegand et al., 1960; Feruglio et al., 1964; Reemtsma et
al., 1960), and an additional 24 paired data sets were graciously provided by the authors of
publications in which mean values were available (lijima et al., 2001; Sakka et al., 2001).
Commercially available software (Statistical Analysis System, SAS) was applied to the above 59
individual values for HBF/CO to determine the parameters for the normal, lognormal, and beta
distributions according to the method of moments. Results are presented in Table 5-8. The
derivation of these values is described in detail in Appendix C. Among the 35 sets for which CO
was measured, HBF and CO appeared uncorrelated.
METABOLIC VARIABILITY
This section provides a brief description of the methodology employed; for a detailed
description, see Appendix D. Metabolic capacity is a function of the separate contributions of
enzyme expression (content) and enzyme activity. Historically, differences in enzyme content of
microsomal protein (MSP) isolated from liver have been used to infer differences among humans
in the liver enzyme content. However, those investigations ignored the distribution of MSP
content of intact liver as an additional contributor to human interindividual variability. This
investigation used ELISA-based measures of CYP2E1 in liver tissues from 60 adult human
organ donors and 10 child organ donors to determine the range of variability of CYP2E1
expression (CYP2E1 content) among adults and children. An additional in vitro metabolism
study was also performed to determine the activity of CYP2E1 toward chloroform. Valid
statistical procedures were applied to determine the metabolic capacity of the human liver,
incorporating measures of variability of enzyme content as well as in v/Yro-derived metabolic
rate constants. Finally, metabolic capacity was extrapolated to the intact liver, and scaled by
5-24
-------
body weight for inclusion in PBPK modeling efforts designed to assess the contribution of the
variance in metabolic capacity to chloroform metabolism.
Statistical analyses determined that the distribution of CYP2E1 in adult human liver was
lognormal. The values at the 5th percentile, the geometric mean, and 95th percentile of the
distribution are presented in Table 5-9.
To measure of the activity of human CYP2E1 toward chloroform, 3 samples of adult
human MSP were evaluated in an in vitro gas uptake system developed and employed by
EPA/ORD/NHEERL. Although genetic polymorphisms of CYP2E1 are known, no evidence
exists to support their involvement in altering the catalytic activity of the enzyme. Should such
evidence be demonstrated in the future, the applicability of genetic polymorphisms to in vivo, as
opposed to in vitro, chloroform metabolism must be demonstrated to substantiate their inclusion
as risk-modifying factors.
With knowledge of the CYP2E1 content of the 3 MSP samples used to study chloroform
metabolism, in v/Yro-derived Vmax values could be converted from units of mass/time/mg MSP to
units of mass/time/pmol CYP2E1. The results demonstrated a specific activity of human
CYP2E1 of 5.24 pmoles chloroform metabolized/minute/pmol CYP2E1. The value in rats was
5.29 pmoles chloroform metabolized/minute/pmol CYP2E1. The similarity in values among
these species is not surprising, given the conservation of CYP2E1 across mammalian species.
This more specific measure of enzyme activity was then combined with measures of enzyme
content to produce measures of metabolic capacity (Vmax), expressed per gram liver tissue, and
converted to units of body weight by accounting for the fractional composition (for adults, 0.026)
of body weight accounted for by liver mass. Results of the extrapolation to intact tissue are
presented in Table 5-9.
5-25
-------
TABLE 5-9
Distribution of Chloroform Metabolic Rate Constants in Adults and Children
Parameter
Adult
Childa
Enzyme Content — pmoles CYP2El/gram Liver
5th percentile
Geometric mean
95th percentile
Specific Activity — pmoles
CHC13/min/pmol CYP2E1
1232b
2562b
4453b
5.24
1288
2818
6167
5.24C
Metabolic Capacity (Vmax) — ug CHC13/h/gram Liver
5th percentile
Geometric mean
95th percentile
46.3
96.3
167.3
48.35
105.8
231.5
Metabolic Capacity (Vmax) — mg CHC13/hd
5th percentile
Geometric mean
95th
Adult
84.27
175.27
304.49
9-year-old Child 1 -year-old Child
38.87 16.56
85.06 36.24
186.13 79.29
a Child organ donors ranges 10 months to 17 yrs of age, refer to Appendix D, Table D-2.
bFrom Lipscomb et al. (2003).
c Measured in adult samples (Lipscomb et al., 2004).
dmg CHC13/h = ug CHC13/h/gram liver x (1000 ug/mg) x BW x VLC x (1000 grams/kg);
difference between values derived for the 9- and 1-year-old child are based on differential
contributions of liver mass to body mass (VLC = body composition of liver).
5-26
-------
For incorporation into PBPK modeling, the above measures of Vmax (mg/h) were scaled
by BW °'7, and the parameter was redescribed as VmaxC by Equation 5-2. The results are
demonstrated in Table 5-10.
VmaxC = Vmax/BW °'7 (5-2)
Simulated Exposure Conditions. Although the mouse is considered to be the most
sensitive species and CM24 is to be determined using a 24-h period with intermittent exposure, a
comparison of chloroform metabolism among rats, mice, and adult male humans can provide
valuable insight. Three sets of simulations were performed, with the goals of (1) evaluating the
temporal responses of chloroform metabolism in all species, (2) characterizing species
differences in chloroform metabolism to derive the HEC, and (3) characterizing human
interindividual differences in chloroform metabolism that may be used to estimate the value for
the TK component of the UFn. The first two sets of simulations used the rat, the mouse, and the
"general" adult human models (that developed for the adult male human and including median
values for parameters whose values varied), while the third set used the adult male human PBPK
model that had been parameterized to reflect age-dependent and human interindividual
differences in enzyme activity, HBF, and blood:air partitioning. The first set of simulations
examined the amount of time required to attain steady state, defined chloroform:liver
concentrations, and defined species differences in chloroform metabolism at the same
concentrations under steady state conditions. Figure 5-5 illustrates the chloroform concentration-
time profile for the first 24 hours of exposure to 5 ppm chloroform in inspired air in rats, mice,
and humans. Steady state in humans is not reached until approximately 7 ds of exposure (not
shown). To provide an additional comparison, the rat, mouse, and general adult human models
were programmed to describe CM24 under steady state at exposure concentrations of 0.1, 0.5,
5-27
-------
TABLE 5-10
Derivation of VmaxC Values for Inclusion in PBPK Modeling for Adults, Children, and Rats
Mouse
Rat
Adult Human, 5th percentile
Adult Human, Geometric Mean
Adult Human, 95th percentile
9-year-old Child, 5th percentile
9-year-old Child, Geometric Mean
9-year-old Child, 95th percentile
1 -year-old Child, 5th percentile
1 -year-old Child, Geometric Mean
1 -year-old Child, 95th percentile
'max
(mg/h)
1.73
1.98
84.27
175.27
304.49
38.87
85.07
187.13
16.56
36.24
79.29
BW
(kg)
0.025
0.25
70
30
10
BW07
0.076
0.38
19.57
10.81
5.01
' max'-'
(mg/h/kg)
22.8a
5.218
4.36
8.956
15.56
3.595
7.866
17.212
3.304
7.230
15.82
NOTE: For comparison, Corely et al. (1990) employed a VmaxC value of 15.7 mg/h/kg for
humans, and 6.8 mg/h/kg for rats to model the disposition of chloroform.
a From Tan et al. (2003).
5-28
-------
0)
0)
c
E
o
M—
§
.2
O
c
O
"+J
2
*j
c
0)
o
c
o
o
Rat
Mouse
20 -
10 -
Human
10 15
Time (h)
20
25
FIGURE 5-5
Achieving Steady State in Liver Tissue. Each of the models was used to simulate a continuous
exposure to 5 ppm chloroform in inspired air. Results demonstrate that the concentration of
chloroform in liver reaches steady state in mice more rapidly than humans or rats.
5-29
-------
1.0, 1.5, 2, 5, 10, and 20 ppm. Once steady state for chloroform concentration in liver was
attained, each of the models was used to simulate chloroform metabolism over the ensuing 24
hours to provide general information on species differences in metabolism (Figure 5-6). This
simulation differs from that in which 24-h chloroform metabolism was predicted using rats and
mice exposed for 6 hours as the basis for quantitative dose adjustment for risk assessment (refer
to Figure 5-3). The present simulation was conducted to examine the exposure-metabolism
relationship.
The second, and least extensive, set of simulations involved only the mouse model in
estimating CM24 under an intermittent exposure. To demonstrate that chloroform concentrations
do not accumulate with repeated daily 6-h exposures and that a single exposure cycle with an
intermittent exposure was sufficient to estimate CM24 for species extrapolation, the model
simulated a 6-h, 5 ppm exposure repeated daily for 7 ds. Results (not shown) demonstrated that
hepatic chloroform concentrations returned to baseline within 24 hs of exposure (6 hs on
exposure, 18 hs off). Therefore, CM24 was determined using a 6-h, 5 ppm chloroform exposure.
The resulting CM24 value was used as a point of departure for species comparison. The
concentration-metabolism relationship developed for the general adult human model (Figure 5-7)
was used to identify the continuously encountered concentration resulting in the CM24 value
observed in the 6-h, 5-ppm exposed mouse.
The third set of simulations examined changes in chloroform metabolism related to
human interindividual differences in metabolic capacity, tissue:air partitioning, and FffiF, as well
as differences in sex- and age-related parameter values. To demonstrate age-dependent
differences, three continuous exposure concentrations (0.9, 2.5, and 5 ppm) were used, and the
5-30
-------
10000
Mouse
Rat
Human
0.1
1 10
Concentration in Air, ppm
100
FIGURE 5-6
Relationship Between Exposure Concentration and CM24 in Rats, Mice, and Humans at Steady
State. These results demonstrate a marked linearity in chloroform metabolism (CM24) over the
exposure range 0.1 to 20 ppm in both rats and humans under steady state, ind infer species
differences useful in estimating the HEC.
5-31
-------
10000
1000
100
o
0.1
1 10
Concentration in Air, ppm
100
FIGURE 5-7
Derivation of the Human Equivalent Concentration (HEC). CM24 resulting from the no-effects
exposure (5 ppm x 6 hours) in mice was determined. Results from the human PBPK model,
under a continuous exposure at steady state indicate that an equivalent exposure would be 8.1
ppm.
5-32
-------
human model was adapted to include "general" values for each parameter, including those for
children.
Demonstration of the HEC accounts for species-dependent TK differences. To account
for species differences in sensitivity (the TD component of the animal-to-human uncertainty
factor; UFA-TD) and because of the linear relationship between exposure and metabolism order of
magnitude, e.g., 3.16. This half-order of magnitude value was applied because TK accounts for
only part of the interspecies differences, and the EPA has recently retained a partial default (one-
half order of magnitude, 10E°5) for residual TD differences. Because the default value for UFA
is 10, a value of 3.16 (one-half order of magnitude) was assigned to represent the TD
"uncertainty" for UFA- The CM24 in the general adult human model resulting from an exposure
to HEC/3.16 was determined and used as a benchmark against which to evaluate human
interindividual differences. The adult male human model was reparameterized to include values
for VmaxC, HBF, and blood:air partitioning (with resulting changes in blood:tissue partitioning)
that were representative of observed variability. The model was adapted for children of 30 kg
and 10 kg body mass (aged nine and one years, respectively) by including (external
concentration and CM24, the HEC (rather than the CM24 value) was divided by one-half
weight/age-related child organ volumes and enzyme content, B:A PC values (derived using
remnant blood samples from children), and T:A PC values (derived from PND 10 rats). The
CM24 resulting from exposure of the general adult to HEC/3.16 was determined at steady state.
Within each model, one parameter was varied at a time to demonstrate the effect of its variance
on CM24, and the magnitude of impact of varying each parameter was characterized by the
change in external concentration resulting in the same CM24 value. To determine the impact of
sex-dependent changes in body composition, the model for the adult human female was
5-33
-------
exercised to simulate exposure to (HEC/3.16) ppm until steady state was reached, and CM24 was
compared to that attained in the adult human male model, exercised to simulate the same
exposure conditions. To examine the impact of obesity, the "general" adult human male model
and the obese adult male human model were exercised to simulate steady state exposures to 0.75,
2.56 and 10 ppm chloroform.
RESULTS
Derivation of the Rodent NOAEL Values. Four important studies (Larson et al., 1996;
Templin et al., 1996, 1998; Constan et al., 2002) provide data useful in estimating a NOAEL for
liver and/or kidney effects in mice. This section reviews the exposure concentrations and
estimates a duration-adjusted exposure concentration for each exposure concentration to compare
results among slightly differing exposure protocols and to demonstrate the use of conventional
risk assessment methods in estimating an RfC value with default uncertainty factor values.
However, no duration adjustment will be made in the exposure extrapolation between animals
and humans, because the extrapolation will be based on an internal dose metric describing
chloroform metabolism in the liver.
Constan et al. (2002) exposed female B6C3F1 mice to chloroform for 7 ds at
concentrations of 0, 10, 30, and 90 ppm for different daily durations (Table 5-11). These authors
concluded that duration, as well as concentration, played a critical role in toxicity. They
concluded that exposures at 10 ppm for durations of approximately 6 hours or longer could lead
to an increased LI, which is evidence of tissue regeneration, in the liver of B6C3F1 mice. They
also applied a PBPK model for mice and humans into which they incorporated their derived
values for mouse chloroform partition coefficients and the values for human chloroform tissue
partition coefficients from Corley et al. (1990). The B:A PC values used in Constan et al. (2002)
5-34
-------
TABLE 5-11
Exposure Conditions, Relative Liver Weight, and Hepatocyte Labeling Index
from Constan et al. (2002)
Exposure
Cone., Duration
0 ppm, 6 hours
0 ppm, 1 8 hours
10 ppm, 6 hours
10 ppm. 18 hours
30 ppm, 2 hours
30 ppm, 6 hours
30 ppm, 12 hours
90 ppm, 2 hours
90 ppm, 6 hours
C x T
0
0
60 ppm x h
180 ppm x h
60 ppm x h
180 ppm x h
360 ppm x h
180 ppm x h
540 ppm x h
Relative liver
weighta
4.72 + 0.32
4.45 + 0.21
5.54 + 0.46d
5.55 + 0.31d
4.93 + 0.14
5.07 + 0.25
5.97 + 0.37d
5.16 + 0.17d
6.27 + 0.42d
Labeling Index
0.55 + 0.27 b
0.70 + 0.22C
0.93 + 0.32
1.83 + 1.05
1.50+1.23
2.94+ 1.95d
10.04 + 3. 96d
2.82+ 1.10d
19.92 + 4.00d
Values are presented as mean + S.D., n = 5 mice.
a Percentage weight gain at necropsy from preexposure body weight. Average initial body
weights were 12.1 + 1.2 grams.
b Control for the 2- and 6-h exposure groups.
c Control for the 12- and 18-h exposure groups.
d Significantly greater than respective control,/? < 0.05.
5-35
-------
were 24.3 for mice and 7.3 for humans. Their model results demonstrated that the human would
have to be exposed to 110 ppm to drive chloroform metabolism at the rate per gram liver
observed in the 10-ppm exposed mouse.
Templin et al. (1998) exposed BDF1 mice to chloroform 6 hs/d, 5 ds/wk via inhalation
for up to 13 wks (Table 5-12). Targeted exposure concentrations were 0, 1, 5, 10, 30, and 90
ppm. Daily checks verified that chamber concentrations were within 0.5% of targeted
concentrations for the 5-, 30-, and 90-ppm concentrations, and within 8% of the 1-ppm
concentration. Male mice were exposed for 3, 7, or 13 wks. Female mice were exposed for 3 or
13 wks. Male mice exposed to 30 and 90 ppm were "stepped up" to the final concentration.
Male mice in the 30-ppm group were exposed to 5 ppm for 2 wks, to 10 ppm for another 2 wks,
then to 30 ppm beginning on wk 5. Male mice in the 90-ppm group were exposed to 5 ppm for 2
wks, to 10 ppm for the next 2 wks, to 30 ppm for wks 5 and 6, then to 90 ppm beginning on wk7.
Female mice were exposed for 3 and 13 wks to 5, 30, or 90 ppm without the step-up procedure.
Templin et al. (1998) found that concentrations above 5 ppm produced histological
changes in liver and kidney of male and female mice, but histological changes were not evident
in liver or kidney of either males or females exposed to 5 ppm chloroform. This data set
establishes 5 ppm as a NOAEL for these effects in BDF1 mice (Table 5-13).
Larson et al. (1996) examined the response of B6C3F1 mice exposed to chloroform at 0,
0.3, 2, 10, 30, and 90 ppm by inhalation. Measured chloroform average concentrations were 0.3,
1.99, 10.0, 29.6, and 88 ppm. Exposures were conducted for 6 hs/d, daily for 4 ds for 3, 6, or 13
wks; or for 6 hs/d, 5 ds/wk for 13 wks (Table 5-12). Histopathology and LI of the nasal
passages, kidney, and liver were examined. In female mice, that nasal turbinate LI was
significantly elevated after 4-d exposure at 10 ppm and above, after 3-wk exposure at 2 ppm and
5-36
-------
TABLE 5-12
Exposure Conditions and Endpoints Examined in Rodent Inhalation Bioassays
Study
Constan et al.
(2002)
Templin et al.
(1998)
Larson et al.
(1996)
Templin et al.
(1996)
Species
B6C3F1
mice
BDF1
mice
B6C3F1
mice
F344
rats
Concentration
0, 10, 30, 90
ppm
0, 1, 5, 30, 90
ppm
0, 0.3, 2, 10,
30, 90 ppm
0, 10, 90 ppm
0, 2, 10, 30,
90, 300 ppm
0, 30, 90, 300
ppm
Duration
2,6, 12, 18hs/dx 7 ds
6hs/d, 5ds/wk, 3, 7, or 13
wks
6 hs/d, 7 ds/wk, 4 ds, 3, 6
or 13 wks;
6 hs/d, 5 ds/wk, 13 wks
13 wks, 6 hs/d,
5 ds/wk
6 hs/d, 7 ds/wk, 4 ds, 3, 6
or 13 wks
6 hs/d, 5 ds/wk, 13 wks
Endpoint
LW:BW,
hepatocyte LI
Liver and kidney
histopathology
Nasal passages,
liver, and kidney
histopathology
Liver and kidney LI
Nasal passages,
liver, and kidney
histopathology and
LI; LW:BW;
KW:BW
5-37
-------
TABLE 5-13
Endpoints and Response Levels Identified in Rodent Inhalation Bioassays
Study
Constan et al.
(2002)
Templin et al.
(1998)
Larson et al.
(1996)
Templin et al.
(1996)
Duration
7ds
13 wks, 5 ds/wk
13 wks, 7 ds/wk
13 wks, 5 ds/wk
13 wks, 7 ds/wk
13 wks, 5 ds/wk
13 wks, 7 ds/wk
13 wks, 7 ds/wk
13 wks, 5 ds/wk
13 wks, 7 ds/wk
13 wks, 5 ds/wk
13 wks, 7 ds/wk
13 wks, 7 ds/wk
13 wks, 5 ds/wk
NOAEL
30 ppm, 2 hs/d;
lOppm, 18 hs/d
90 ppm male
30 ppm female
5 ppm male
90 ppm female
5 ppm male
30 ppm female
30 ppm male
10 ppm female
10 ppm male
10 ppm female
10 ppm male
ND male
90 ppm male
90 ppm female
90 ppm male
90 ppm female
90 ppm male
90 ppm female
10 ppm male
10 female
30 ppm male
30 ppm female
30 ppmb
2 ppm male0
10 ppm male0
LOAEL
30 ppm,
6 hs/d
ND male
90 ppm female
30 ppm male
ND female
30 ppm male
90 ppm female
90 ppm male
30 ppm female
90 ppm female
90 ppm female
30 ppm male
10 ppm male
ND male
ND female
300 ppm male
300 ppm female
300 ppm male
300 ppm female
300 ppm male
30 ppm female
90 ppm male
90 ppm female
90 ppmb
10 ppm male0
30 ppm male0
Endpoint
Hepatocyte LI
Liver LI
Kidney LIa
LW:BW
Liver LI
Liver LI
Kidney LIa
Nasal turbinate
LI
Liver LI
Kidney LI
LW:BW and
KW:BW
Nasal ULLI
a Chloroform exposure did not induce histologic changes in kidneys of female mice.
b Reported in general without regard to sex.
0 Data for females stated to be similar, but not shown.
ND = effect level not demonstrated.
5-38
-------
above; and after 6-wk exposure only at 90 ppm; but it was not significantly elevated in any group
examined after 13 wks of exposure. Nasal lesions occurred in mice exposed to 10 ppm and
above; they were transient and only observable after 4 ds of exposure. These authors concluded
that after 6 wks of exposure, only exposure to 30 ppm and higher produced biologically
significant liver alterations in the female mice. After 7-ds/wk exposure for 13 wks, "very mild
degenerative changes" were noted in the livers of mice exposed to 10 ppm. In contrast, no
significant liver changes were found in the mice exposed for 13 wks to 10 ppm chloroform but
only for 5 ds/wk. The authors concluded that 10 ppm was a NOAEL value for liver effects.
Renal effects were not observed in female mice exposed under any of these experimental
exposure regimens (Table 5-13).
Larson et al. (1996) indicated that nasal lesions and regenerative cell proliferation in
males were similar to those in female mice and observed no other respiratory tract effects. The
incidence of "mild" liver lesions in male mice was elevated to 30 ppm and higher over controls
after 6 wks of exposure. The hepatocyte LI was significantly elevated only in male mice that
were exposed to 90 ppm, at both 6 and 13 wks of exposure. Among mice exposed to 90 ppm
chloroform for 13 wks, those exposed for 5 ds/wk demonstrated a significantly elevated
hepatocyte LI versus controls; however, the LI in those exposed mice was significantly lower
than that observed in male mice exposed to 90 ppm for 7 ds/wk. The effect of exposure duration
(continuous versus 2 ds off) was also evident in the liver changes observed in female mice
exposed to 30 ppm chloroform, indicating the importance of duration of exposure as a
determinant of effect. Renal lesions (nephropathy) in male mice were confined to the proximal
convoluted tubules (PCT) epithelial cells. After 3 wks of exposure, concentration-dependent
increases in nephropathy were evident in the 30- and 90-ppm groups. Enlarged PCT epithelial
5-39
-------
cell nuclei were observed in male mice exposed for 13 wks to chloroform at 10, 30, and 90 ppm;
enlarged nuclei and scattered areas of regenerating foci were found in male mice exposed for 13
wks, 5 ds/wk. For renal effects, the 5-ds/wk, 13-wk concentration of 10 ppm represents a
LOAEL; and the 2-ppm exposure represents the NOAEL (Table 5-13). The authors concluded
that, for kidney effects, B6C3F1 mice were more sensitive to a 5-ds/wk regimen than a
continuous regimen, indicating that a "days-off' exposure might better allow for cellular
regeneration. Furthermore, Larson et al. (1996) found that the nasal lesions demonstrated in
these mice occurred only at 10 ppm and higher concentrations and were transient. They
concluded that the nasal response may be only a minor concern for environmentally exposed
people, given the lower concentrations found in the environment.
Templin et al. (1996) examined the formation and distribution of histologic changes and
LI in liver, kidney, and nasal tissues of F-344 rats exposed to chloroform vapor for up to 13 wks.
Male and female rats were exposed for 6 hs/d, 7 ds/wk to chloroform at concentrations of 0, 2,
10, 30, 90, or 300 ppm. Renal effects such as scattered vacuolization, individual tubular cellular
necrosis, and enlarged nuclei were observed in male rats exposed to 90 and 300 ppm for 7 ds/wk,
but not in those exposed to 30 or 90 ppm for 5 ds/wk. A significantly increased LI in renal
cortex was observed in male and female rats exposed to 30 ppm 7 ds/wk, but not in those
exposed 5 ds/wk, for 13 wks. Livers of male rats were affected by 13 wks of chloroform
exposure at the 90 and 300 ppm exposure level, but not at the lower concentrations. The LI was
only elevated at the 300 ppm level. Rats exposed for 13 wks at 90 ppm demonstrated scattered
vacuolated hepatocytes (6 of 15 rats) and single-to-multiple hepatocytes necrosis (9 of 15 rats).
Mitotic figures and diffuse vacuolization of liver were evident in male rats exposed to 300 ppm
5-40
-------
for 13 wks. Thus, as determined for male F-344 rats, 30 ppm is the NOAEL for male rat liver
effects, and 10 ppm is the NOAEL for renal effects (Table 5-13).
In female F-344 rats exposed 13 wks to 30 ppm and above, Templin et al. (1996)
observed histologic changes consisting of vacuolization, whereas comparable male rats
demonstrated vacuolization and necrosis. As in the male rats, the LI in renal cortex was
significantly elevated in 30-ppm exposed female rats. In the liver, areas of mild vacuolization
and, infrequently, hepatocyte degeneration were observed in female rats exposed for 13 wks to
90 ppm chloroform. Centrilobular-to-midzonal degeneration was evident in 300-ppm exposed
female rats. Similar to previous observations in male rats, female rats exposed to 90 ppm for 5
instead of 7 ds/wk demonstrated liver profiles that were not significantly different from control
rats. Among the 13-wk exposed female rats, the hepatocyte LI was increased only in the 300-
ppm exposure group. This concentration, but no other, produced increases in the LI after 3 wks
of exposure. Thus, female F-344 rats appeared to demonstrate fewer of the severe consequences
of chloroform exposure, even though they demonstrated a NOAEL for renal effects at 10 ppm.
The NOAEL for female rat liver effects was 90 ppm (Table 5-13).
In summary, liver effects can serve as the critical effect, and NOAEL values can be
developed for both the mouse and rat from studies by Templin et al. (1998) and Templin et al.
(1996), respectively. Data from these studies demonstrate a 6-hs/d exposure to 5 ppm in mice as
a NOAEL.
Model Response. The model was successfully parameterized with mouse, rat, and adult male
human values for partitioning, body composition, blood flows, and metabolic rates extrapolated
from in vitro studies with hepatic MSP (rat and human) or extrapolated from in vivo gas uptake
studies with mice. To examine the likelihood of the rapid attainment of steady state, as measured
5-41
-------
by the concentration of chloroform in liver tissue, each model was used to simulate a continuous
exposure to 5 ppm chloroform in inspired air. Results over the initial 24 hours (see Figure 5-3)
show that several hours are necessary to reach steady state, and that steady state is not achieved
in a similar time frame between species.
An example of model function at a fixed, continuously encountered concentration was
provided by simulating exposure to 5 ppm chloroform, which represents a concentration one-half
that established as the TLV for the workplace. Results indicated that 5-ppm exposure resulted in
roughly 540 mg/L (in liver) of metabolites formed over a 24-h period in the mouse and that 234
mg/L (in liver) of metabolites formed over a 24-h period in the rat. Exposure of the general adult
(male) human to 5 ppm under steady-state conditions resulted in a CM24 of roughly 75 mg/L (in
liver), about one-third that of the similarly exposed rat and nearly 14% that of the similarly
exposed mouse (see Table 5-14, Figure 5-6). In comparison, simulated exposure to 16 ppm in
the general adult (male) human was required to bring the CM24 value (238.8 mg/L) near the
CM24 value attained in the 5-ppm exposed rat (234.39 mg/L). Exposure of the human to 36 ppm
was required to attain the CM24 value (540 mg/L) demonstrated in the 5-ppm exposed mouse.
These results demonstrate that, compared to rodents, the human converts much less of a dose of
chloroform to toxic metabolites. The results presented in Table 5-14 indicate that the formation
of metabolites was proportionately related to exposure concentration over the range surrounding
this concentration, in adults and in children. Likewise, differences among age groups were
consistent over this range.
Chloroform Model Verification. This report presents the application of a single model
structure to characterize metabolism of chloroform in mice, rats and humans. Confidence in the
model's predictive ability is increased when predictions of pharmacokinetic outcomes are similar
5-42
-------
TABLE 5-14
Chloroform Metabolized in Liver in Mice, Rats, and General Adult (Male) Humans
Species
CM24 (mg/L)
Percent of General Adult Human
5 ppm at Steady State
Mouse
Rat
General Adult Human
General 9-year-old Child
General 1 -year-old Child
540.4
234.39
74.64
93.97
100.56
724
314
100
126
135
2.5 ppm at Steady State
Mouse
Rat
General Adult Human
General 9-year-old Child
General 1 -year-old Child
270.18
117.15
37.31
46.99
50.21
724
314
100
126
135
0.9 ppm at Steady State
Mouse
Rat
General Adult human
General 9-year-old Child
General 1 -year-old Child
97.29
42.22
13.43
16.91
17.97
724
314
100
126
134
5-43
-------
to observed values for pharmacokinetic measurements made in vivo in the species studied. There
are no established guidelines for acceptability, though increased emphasis should be placed on
regions of the exposure-response continuum relevant to the intended use. For example,
confidence in model predictions verified only by fit to observations made at exposure
concentrations that are orders of magnitude above the range of exposures where the model will
be used in species extrapolation will do little to instill confidence in model predictions at low
concentrations. Increased confidence in model predictions can be obtained when comparisons
between observed values and predicted values can be made using data more closely related to the
dose metric of interest, rather than further from it. In the present application, the identified dose
metric was the integrated exposure of the liver to chloroform metabolites over 24 hours,
basically, an AUC measurement. Several pharmacokinetic outcomes could have been selected
for efforts to verify the model. These could have included disappearance of chloroform from
closed-chamber studies, exhalation of absorbed chloroform, concentration-time profiles for
chloroform in circulating blood or plasma, etc. However, this effort identified and employed
observations of chloroform metabolism because they are more closely related to the risk-relevant
dose metric for the toxicity studies. Here, comparisons between observations of chloroform
metabolism in mice (Corley et al., 1990) and adult male and female humans (Fry et al., 1972)
and model predictions are made for the purpose of model verification.
Initially, it was thought that the rat would serve as the sensitive species, instead of the
mouse, and thus represent the species from which human predictions would be extrapolated. For
that reason, partition coefficients were determined using blood, liver, kidney, muscle, and fat
from the rat as well as blood from humans; and metabolic rate constants were derived in vitro
using liver preparations from rats and humans. Inasmuch as the approach was ultimately revised
5-44
-------
to employ the mouse as the sensitive species, and because the human model contains
tissue:blood partition coefficient values derived from rat tissue:air and human blood:air partition
coefficient values, it seems only reasonable that efforts be undertaken to compare model
predictions to observations of chloroform metabolism in rat, mouse and humans, where possible.
Additionally, we will examine implications of the appreciable difference between the values for
Km which we derived in vitro and applied in the present modeling approach versus values for
Km optimized by modeling software used in previous PBPK approaches to chloroform
dosimetry, including that of Corley et al. (1990). This will serve four purposes: 1) model
structure will be justified, 2) the application of values for partition coefficients developed for rat
tissues will be justified, 3) greater confidence can be placed in model predictions from the human
model, which contain tissue partition coefficients based in part on observations made with rat
tissues, and 4) sufficient confidence can be placed in the approach of extrapolating and
incorporating in v/Yro-derived metabolic rate constants as accomplished in the present work.
Comparisons to Mouse and Rat Pharmacokinetic Data. A data set described by
Corley et al. (1990) contains information obtained from rats and mice exposed to chloroform in
an open-chamber study design. In that modeling approach, whole-body chloroform metabolism
was simulated by developing metabolic rate constants using data describing whole-body
metabolism. Metabolic rate constants (Km and VmaxC) were optimized by allowing the model
software to vary them until adequate fit was observed between experimentally-collected data and
model predictions. In vivo data on chloroform metabolism were collected from mice and rats by
conducting open-chamber exposures were for six hours. Mice were exposed to 10, 89 and 399
ppm; rats were exposed to 93, 356 and 1041 ppm. These exposure concentrations are higher
5-45
-------
than those used in the simulations here conducted for species extrapolation of dose, especially so
for the rat.
Differences between the modeling approaches used (Corley et al. versus the present),
especially those relating to developing model predictions of total (whole-body) chloroform in the
former and application only of liver-specific metabolic capacity to predict chloroform
metabolism in the latter, require specific considerations of fit between observed and predicted
values for metabolism. Additionally, the source for metabolic rate constants for the rat in the
two modeling approaches (Corley used model-optimized values; the present approach
extrapolated in v/Yro-derived metabolic rate constants) and the tissue-specificity of the rate
constants must also be considered. Chloroform metabolism is saturable, and Michaelis-Menten
rate descriptions of metabolic rate are appropriate. In vivo, tissue concentrations of chloroform
serve as substrate concentrations in the Michaelis-Menten rate equation.
The model structure and exposure scenarios simulated in our investigation are suitable to
simulate open-chamber results simulated by Corley et al. (1990). Corley et al. (1990) used radio
labeled chloroform to facilitate metabolite quantification; loss rates from atmosphere are not
collected in the open-chamber design. The evaluation of these data offer advantage over closed-
chamber data describing loss of chloroform from the finite amount available in the atmosphere,
because Corley measured chloroform metabolism in rats and mice, rather than disappearance
from headspace/atmosphere. This pharmacokinetic outcome (chloroform metabolized) is more
closely related to the dose metric of interest (chloroform metabolite per unit liver mass; CM24)
than is loss of chloroform from atmosphere. Inasmuch as Corley observed saturation of whole-
body (total) chloroform metabolism between 93 and 356 ppm concentrations in the rat, and
because predictions of metabolic rates under saturating exposure conditions are sensitive to
5-46
-------
values for VmaxC, but insensitive to values for Km, and because the rat is not the species used
for extrapolation, fit to the rat data is not so critical as fit to the mouse data.
Corley's predicted values were 0.88 to 2.08 fold different from observed values, and
these differences were more pronounced in rats and mice, each, at the lower range of
concentrations evaluated. Table 5-15 demonstrates the observed amounts of chloroform
metabolized from rats and mice (Corley et al., 1990), amounts of chloroform metabolized as
predicted by the model developed by Corley et al. (1990) and the amounts of chloroform
metabolized as predicted by the present model. As presented, the fit between Corley's
predictions and the observations is not appreciably different for the mouse, but differ to some
extent for the rat. However, both are adequate.
Although the species extrapolation here employed involves mice and humans, emphasis
on rat model comparison is made because 1) human tissue partition coefficients are based in part
on rat tissue:air partition coefficients, which are used in the rat model, and 2) both the rat and
human models rely on the extrapolation of in v/Yro-derived metabolic rate constants for
chloroform derived with liver preparations, and are expressed solely in terms of hepatic
metabolism. Confidence developed for the adequacy of the rat model and its basis should carry
over directly over directly to the human model, and should be tempered with information
describing the comparison of human predictions to observations of chloroform metabolism in the
human.
With respect to mice, the present model offers a slightly improved fit to the available
observations for chloroform metabolism seen at the lowest concentrations. This lends
confidence to its application to simulate chloroform metabolism in mice, used as the point of
5-47
-------
TABLE 5- 15
Results from Open Chamber Metabolism Studies in Rats and Mice a
Species
Mouse
Rat
ppm
10
89
366
93
356
1041
Total Metabolized
Observed
0.22 mg
2.14
6.76
12.10
19.40
35.03
Corley -Predicted
0.46 mg (2.08)b
3.99 (1.86)
6.18 (0.91)
13.41 (1.11)
22.74 (1.17)
30.79 (0.88)
Present - Predicted
0.37 mg (1.68)b
3.25 (1.51)
9.95 (1.47)
9.49 (0.78)
11.75 (0.61)
11.83 (0.34)
From Corley et al. (1990).
1 (Predicted: Ob served)
5-48
-------
extrapolation and the basis for quantification for species difference in chloroform metabolism
that will be used to develop the HEC.
The values for model parameters employed in the Corley et al. (1990) model fit the data
better, but Km values in that study were determined by model optimization, rather than by a
more direct in vitro evaluation, as done in the present study. Thus, it is not surprising that the
model parameters employed in the Corley model result in a better "fit" to the observations than
those in the present study. However, for the reasons discussed in following paragraphs, this
apparent difference in model fit (for rat data, especially) should not diminish confidence in the
predictive capabilities of the present model. Smith and Evans (1995) have also addressed
confidence placed in model-optimized values for metabolic rate constants.
The observed difference between observations and predictions of Corley and of the
present work of chloroform metabolism in the rat should not diminish confidence placed in the
model, inasmuch as exposures of interest are much below 25 ppm, and because the rat does not
serve as the basis for species extrapolation and HEC determination. Reasons for differences may
relate to Km values employed and the contributions of extrahepatic tissues to chloroform
metabolism. While argument may be made that the observation of metabolic saturation in the
present model may be a result of a lower Km value being employed in the present model, versus
the application of a model-optimized Km value, as done in the previous (Corley) model, this
seems to be poorly founded, for reasons discussed in following paragraphs.
Model Sensitivity. The sensitivity of the CF PBPK model to changes in selected
parameter values was determined from simulations of inhalation exposure of an adult male to
2.56 ppm CF by varying the values of PB (blood:air partition coefficient for CF; unitless), QLC
(hepatic blood flow; L/hr/kg), Vmaxc (maximal rate of CF metabolism; mg/hr/kg) and KM
5-49
-------
(Michealis constant for CF metabolism; mg/L) individually and examining the effect on CM24,
the integrated exposure of the liver to CF metabolites over 24 hours.
Figure 5-8 shows the effect of varying PB, the blood:air partition coefficient for CF, on
CM24. The PB reflects the solubility of CF in the blood that carries CF to the internal organs. As
expected, the CM24 was low when the partition coefficient (solubility in blood) was low. CM24
increased nonlinearly when the value of the partition coefficient increased, reaching a plateau at
a value of approximately 50 (Figure 5-8). Figure 5-9 shows the effect of changing the liver
blood flow rate (QLC) on CM24 for CF. CM24 increased nonlinearly with increasing QLC. The
dramatic decreases in CM24 were observed at QLC values below approximately 20% of cardiac
output.
Figure 5-10a shows the relationship between Vmaxc, the maximal rate of CF metabolism,
and CM24. CM24 was strongly dependent upon the rate of metabolism and rapidly increased up
to a Vmaxc value of approximately 0.5 mg/hr/kg. Figure 5-10b shows an expanded scale of the
relationship between CM24 and Vmaxc up to values of 2 mg/hr/kg, clearly illustrating the plateau
in CM24 above a Vmaxc of 0.5 mg/kg/hr. The CM24 values above this Vmaxc were essentially
the same, consistent with the interpretation that the rate of CF metabolism in the liver is limited
by the rate of hepatic blood flow. Figure 5-10 clearly demonstrates that Vmaxc values greater
than 0.5 mg/hr/kg do not significantly increase CM24. Thus, interindividual variability in Vmaxc
(a reflection of variability in enzyme content) at values greater than approximately 0.5 mg/hr/kg
does not translate to interindividual variability in CF metabolism or CM24 (a risk-related
endpoint), as has been demonstrated in the simulations of the CF PBPK model (U.S. EPA, 2004).
5-50
-------
CM
O
60
50
40
30
20
10
20
40 60 80
B:APC
100
120
FIGURE 5-8
Relationship Between the Blood:air Partition Coefficient (B:A PC) for CF and CM24. The arrow
indicates the B:A PC value used in the PBPK model.
5-51
-------
CM
O
60
50
40
30
20
10
0.1 0.2 0.3 0.4 0.5 0.6 0.7
QLC
FIGURE 5-9
Relationship Between Hepatic Blood Flow (QLC) and CM24 for CF. The arrow indicates the
QLC value used in the PBPK model.
5-52
-------
50
40
r
30
CM
o i
20
10
20
40 60 80
Vmaxc
100
120
FIGURE 5-10a
Relationship Between the Maximal Rate of CF Metabolism (Vmaxc) and CM24. The arrow
indicates the Vmaxc value used in the PBPK model.
5-53
-------
CM
S
o
20
10
0.5
1
Vmaxc
1.5
FIGURE 5-1 Ob
Expanded Scale Showing the Relationship Between the Maximal Rate of
CF Metabolism (Vmaxc) and CM24
5-54
-------
Figures 5-1 la and 5-1 Ib show the effect of varying KM, the Michaelis constant for CF
metabolism, on CM24. The KM is defined as the CF concentration that gives one-half the
maximal rate (Vmaxc) of CF metabolism. The KM is a complex kinetic constant that contains
terms for both the binding and catalytic oxidation of CF by CYP2E1. Changes in KM will affect
the initial rate (V/K) of CF metabolism, with smaller KM values yielding higher initial rates and
hence larger CM24 values (Figure 5-11). As KM increases, the initial rate of CF metabolism
decreases, resulting in lower values of CM24 (Figure 5-11).
Comparisons to Human Pharmacokinetic Data. Corley et al. (1990) also reported
findings from Fry et al. (1972), in which the cumulative amount of chloroform metabolized 8
hours following oral dosing was evaluated in human subjects. Fry et al. (1972) dosed human
volunteers with 500 mg chloroform in olive oil encased in a gelatin capsule and measured
pulmonary exhalation of 13C-labeled chloroform and carbon dioxide, and unchanged chloroform
in urine. Subjects 1-5 were males with an average body weight of 70 kg (range 61.8-80 kg), and
data from these subjects were pooled for a single representative example. For this group of five
individuals, 1) average dose (mg/kg) was calculated, 2) average fraction of dose accounted for by
pulmonary excretion of unmetabolized chloroform from average dose was determined, 3)
average fraction of dose metabolized was determined by subtracting the fraction excreted
unchanged in expired air from 1.0 to yield the average fraction of dose metabolized, and 4) the
average fraction of dose metabolized was multiplied by the average dose to yield the average
amount of chloroform metabolized, expressed as mg/kg. Subjects 6-8 were females with body
weights 62.7, 59 and 58 kg, respectively. The PBPK model for the adult female was
reparameterized to reflect these body weights and the amount of chloroform metabolized was
calculated as above. Original data for body weight, total dose, and fraction of dose exhaled
5-55
-------
10
20
30
KM
40
50
60
FIGURE 5-1 la
Relationship Between the Michaelis Constant (KM) for CF Metabolism and CM24
5-56
-------
45
40
35 H
30
•* 25 H
CM
O 20
15
10
5 H
0
0.1
0.2
0.3
0.4
0.5
KM
0.6
0.7
0.8
0.9
FIGURE 5-lib
Expanded Scale of the Relationship Between the Michaelis Constant (KM) for CF Metabolism
and CM24. The arrow indicates the KM value used in the PBPK model.
5-57
-------
unchanged are reported by Fry et al. (1972). Corley et al. (1990) reported the comparison of
their prediction of chloroform metabolized to observations of Fry et al. (1972) for Subject 9, a
65-kg male receiving a dose of 1000 mg (15.4 mg/kg); for mean values from the group of five
males; and for mean values from the group of three 500 mg-dosed females.
These comparisons were used to develop confidence in the original publication by Corley
et al., and represent the best available data through which to develop comparisons of model
predictions of chloroform metabolism in humans. Results presented in Tables 5-16 and 5-17
demonstrate a reasonable fit between observed amounts of chloroform metabolized and
predictions, whether from the model developed previously by Corley and coworkers or the
presently developed model, though predictions for the single 65 kg male are not as good as for
the group. Based on the magnitude of the difference between observations and predictions, the
comparison of predictors developed by the model here presented and the observations made by
Fry et al. (1972) instills confidence in the predictive ability of the human models developed.
These results, in the face of 5- to 15-fold different values for Km and results presented in Figures
5-1 la and 5-1 Ib, underscore the relative insensitivity of this model to values for Km and should
belay concerns about the impact of Km values as a sensitive model parameter when predicting
cumulated amounts of chloroform metabolized (e.g., CM24).
Choice of Values for Km. The modeling of chemicals whose metabolism is limited by
hepatic blood flow is somewhat complicated. In this endeavor, cumulated amounts of metabolite
formed may not track closely with rates of metabolism. The present modeling endeavor has been
scrutinized concerning the appreciable difference between the Km values employed here and Km
values employed in previously published PBPK models for chloroform. Present values for Km
(0.012 mg/L) were extrapolated from in vitro investigations of subcellular preparations from
5-58
-------
TABLE 5-16
Results from Studies with Male Humans
Human
Male
70kg BW (avg, n=5)
65 kg BW (one
individual)
Oral dose
7.14mg/kg
15.4mg/kg
Total Metabolized
Observed
268.8 mg
317.2mg
Corley - Predicted
305.9(1.14)
501.8(1.58)
Present - Predicted
328.4(1.22)
495.6(1.56)
a Metabolite formation calculated from observations reported to 8 hours post-exposure (Fry et al.,
1972).
b (Predicted: Ob served)
TABLE 5-17
Results from Studies with Female Humansa
Subject
Subject 6
Subject 7
Subjects
Corley et al (1990) Average15
Fry et al (1972) Average13
BW(kg)
62.7
59.0
58.0
60.0
60.0
Oral Dose (mg/kg)
7.975
8.475
8.621
8.357
8.357
CF Metabolized (mg)
317.8
312.6
311.1
295.8
324.0C
a Experiment described by Fry et al. (1972).
b Reported in Table 10 of Corley et al. (1990).
c Actual measurement of excreted CO2 by Fry et al. (1972).
5-59
-------
human liver (see Appendix D), whereas the publication of Corley et al. (1990) developed and
employed a model-optimized Km values ranging from 0.3 to 0.5 mg/L.
In a criticism of developing point estimates for values used for metabolic rate constants in
PBPK modeling efforts, Smith and Evans (1995) undertook a systematic evaluation of
chloroform PBPK modeling approaches to simulating open chamber, closed chamber and iv
chloroform data sets. These authors employed the model structure of Corley et al. (1990), but
like our present work, they opted not to include a term describing metabolic inhibition in mice.
This is based on findings of Gearhart et al. (1993) in which liver samples obtained from male
mice exposed to chloroform were subsequently subjected to in vitro analysis and demonstrated
no loss of cytochrome P450-dependent activity.
Using maximum likelihood techniques, Smith and Evans (1995) demonstrated that a
range of Km values, extending downward from values of approximately 2 mg/L even to negative
values, were capable of simulating chloroform metabolism equally well. These authors indicate
that this is possible because "when metabolism is blood-flow limited, any model of metabolism
with rate constants capable of metabolizing 100% or more of delivered substrate will describe
the data equally well." Using several data sets, including that from Corely et al. (1990), Smith
and Evans (1995) demonstrated that Km values approximating 0.05 mg/L fell well within the
95th percent confidence limits for Km developed from simulations of gas uptake data originally
reported by Gargas et al. (1986). For gas uptake data derived in mice, there was no appreciable
difference in fit when Km values were varied from approximately 0.05 to 0.9 mg/L. Thus the
values for Km employed in the present investigation fall well within the range of values
determined plausible and demonstrated to effectively simulate chloroform metabolism. While
optimized point estimates used in other modeling efforts may be higher than those presently
5-60
-------
employed, there is no reason to suspect that the approximately forty-fold lower value for Km
employed here for rats and humans should serve as a basis upon which to lower confidence in
model predictions. We have here demonstrated a high degree of similarity in our model
predictions to observed data, particularly for mice and humans.
Human Equivalent Concentration. To adjust the animal no-response exposure metrics to an
anticipated human NOAEL, this analysis was focused on developing a CM24 value in the general
human that correlated with CM24 in the animal models for each test species. This was done by
developing and exercising PBPK models for each species. Figure 5-3 illustrates the process and
Figure 5-7 shows the results. The exposure scenario chosen was the 6-h, 5-ppm exposure in the
mouse. It was not duration-adjusted, because exposure adjustment was based on an internal dose
metric, CM24. The relationship between CM24 and exposure concentration (Figure 5-6)
demonstrated near perfect linearity for a continuous exposure to concentrations between 0.1 and
20 ppm in mice, rats, and humans. Some degree of similarity in the internal dose metric in rats
and mice occurs when the internal dose metric is developed from the respective duration adjusted
NOAEL values (see Table 5-18). Of note is that a difference of 6-fold in external NOAEL
values corresponds with a difference of 2.6-fold when evaluated for the internal dose metric.
TABLE 5- 18
Derivation of the Human Equivalent Concentration from Liver Effects Observed
in Mice and Rats
Species
Mousea
Ratb
NOAEL
5 ppm
30 ppm
CM24 (mg/L liver
over 24 hours)
136
355
Human Equivalent Concentration
8.1 ppm
~ 25 ppm
lTemplinetal. (1998).
'Templinetal. (1996).
5-61
-------
This difference suggests that mice are "more sensitive" than rats for both TK and TD
reasons. The HEC developed from studies with mice is 8.1 ppm, and the HEC developed from
studies with rats is approximately 25 ppm.
Human Variability. Because the mode of action for chloroform involves hepatic metabolism
and because the metabolism of chloroform is known to be limited by its delivery to the liver via
HBF, studies were designed to simulate the effect of biologically plausible variability in
metabolism and blood flow. In the adult, 9-year-old child, and 1-year-old child simulations,
metabolic capacity (VmaxC) and blood flow to the liver (QLC, fraction of CO) were varied
according to the values determined for the 5th and 95th percentiles of each distribution.
Deriving CM24 at Which to Determine Human Variability. The following calculation was
performed to reduce the mouse internal dose metric to account for potentially greater sensitivity
(a pharmacodynamics consideration) in humans. Standard EPA practice has been to allow a one-
half order of magnitude (10E°5 = 3.16) uncertainty factor to account for species TK or TD
differences. With the aid of pharmacokinetic information, or only B:A PC values, it is possible
to effectively reduce the mouse TK component to 1 and adjust the animal NOAEL by 3.16 to
account for PD differences. Thus, the HEC of 8.1 ppm was divided by 3.16 to yield a
concentration of 2.56 ppm. From the relationship described by the general human model for
inspired concentration and CM24, 2.54 ppm corresponds to a CM24 value of 42.75 mg/L. This
value was established as a benchmark, and the various human models were used to determine the
concentration of chloroform in inspired air that was necessary to produce this CM24 value.
Human Variability: Metabolic Capacity and Age. The results of the simulation experiments
using actual values for the percentiles of the distribution for chloroform metabolism in human
adults and children indicated that variability in this parameter had little effect on overall
5-62
-------
chloroform metabolism within a given age group. The results presented in Table 5-19
demonstrate the exposure concentrations, under the influence of varying metabolic capacities,
required to produce a CM24 value of 42.75 mg/L of liver, the value predicted in the general
human model under an exposure to 2.56 ppm. In the general adult model, changes in VmaxC,
corresponding to values at the 5th and 95th percentile of the distribution did not require that
exposure concentrations be adjusted more than 0.5% to bring CM24 values back to the value
observed when VmaxC represented the median value. This lack of impact of VmaxC variability on
CM24 values was also observed in the 9-year-old and 1-year-old child.
Among age groups, an equivalent CM24 value (42.75 mg/L) was reached at 2.56 ppm in
the general adult, at 2.07 ppm in the general 9-year-old child, and at 1.99 ppm in the general
1-year-old child. With concentrations in inspired air equivalent between adults and children,
CM24 values in children were nearly 35% higher than CM24 values in the adult (Table 5-14).
Human Variability: Hepatic Blood Flow. Data from multiple human studies were used in
evaluating variability in the fraction of CO directed to the liver, HBF/CO. The PBPK model for
the human was varied to include different rates of hepatic perfusion. Because the liver represents
a richly perfused tissue, blood flow to the liver (QLC) was balanced against flow to other richly
perfused tissues (QRC). From the available 59 measures of HBF (as a fraction of CO), statistical
analysis was unable to differentiate a goodness of fit between the lognormal distribution type and
the beta distribution type. Additionally, the base model was parameterized to include HBF as
0.25 of CO; the geometric mean for HBF was 0.259 from the lognormal distribution and 0.267
from the beta distribution. Therefore, the examination of the impact of blood flow will not
produce 42.75 mg/L as the central point for CM24. Instead, that point will be slightly different
because the model was reparameterized to include distribution-specific measures of central
5-63
-------
TABLE 5-19
The Impact of CYP2E1 -Dependent Metabolic Parameters on Chloroform Metabolism Among
Selected Segments of the Human Population
Subject Group
Adult — 5th percentile
General adult
Adult — 95th percentile
9-year-old — 5th percentile
General 9-year-old
9-year-old — 95th
percentile
1 -year-old — 5th percentile
General 1 -year-old
1 -year-old — 95th
percentile
CM24
42.75 mg/L
ppm
2.56
2.56
2.56
2.07
2.07
2.07
1.99
1.99
1.99
Magnitude of difference
(ppm)*
0.002
1.0
-0.001
0.003
1.0
-0.001
0.003
1.0
-0.001
*The model output was used to predict the exposure concentration required to produce a CM24
value of 42.75 mg/L. Difference in this exposure concentration within an age group is presented.
A higher metabolic capacity (95th percentile) requires lowering the exposure concentration
slightly (0.1%) to produce an equivalent CM24 value. Individuals with lower metabolic capacity
(VmaxC value) can tolerate slightly higher (0.3% higher) concentrations while maintaining the
same CM24 value.
5-64
-------
tendency. Table 5-20 demonstrates the impact of variance in HBF on chloroform metabolism
among adult humans. For this simulation, metabolic capacity was held constant at the geometric
mean value of 8.956 mg/h-kg for VmaxC. Chloroform was present in inhaled air at 2.56 ppm.
These results indicate that the difference in CM24 values between the mean of the general
distribution for FffiF and the CM24 in individuals at the 95% percentile for the distribution is less
than 17%. Thus, while FffiF plays a deterministic role in regulating chloroform metabolism, the
bounds of the rather wide, but naturally occurring, variance in HBS (a physiologic limitation)
restrict its impact (quantified as the difference in external exposure concentrations (2.56 versus
2.19 ppm) producing the same CM24 value) to less than 17% among adults. Differences in Table
5-20 that were in the negative range were not considered, because a negative difference in CM24
would confer a reduction in risk.
Human Variability: BloodrAir PC Value. In experiments to support this study, the range of
B:A PC values observed in actual samples of adult blood was 6.6 to 13.3. The model was
parameterized to simulate the conditions of VmaxC and FffiF representing the values at the mean
for their respective distributions. Observed maximal and minimal values for blood:air
partitioning were used to recompute the liverblood PC (L:B PC) value. It was observed that
incorporating the lowest observed B:A PC value (6.6) resulted in a CM24 value of 37.62, a
decrease of 12.7%. On the other hand, increasing the B:A PC value to the maximally observed
value (13.3) resulted in a slight increase in CM24 value (to 44.46, or 104% of the previous value).
Combined Variability in the Adult. To estimate the combined impact of increased blood flow
in conjunction with increased metabolic capacity and high blood:air partitioning, the model was
populated with values corresponding to those at the 95th percentiles of the distribution for HBF
from the lognormal (0.456), with the VmaxC value representing that at the 95th percentile of the
5-65
-------
TABLE 5-20
Effect of Variance in Hepatic Blood Flow on Chloroform
Metabolism Among Adult Humans
HBF(%ofCO)
Chloroform (ppm)
CM24
Magnitude of Difference
in CM24
Lognormal distribution
0.147
0.259
0.456
2.56
35.79
42.75
49.95
-0.163
1.0
1.12*
*With respect to external concentrations producing an equivalent CM24 value in humans,
maximal difference was noted in the lognormal distribution values, between the HBF values at
the geometric mean and the 95th percentile of the distribution. Model results describing the
relationship between CM24 and external chloroform concentrations indicate that an exposure to
2.19 ppm in this segment of the population (HBF = 0.456 CO) would be required to produce a
CM24 value of 42.75 mg/L (see Figure 5-12).
5-66
-------
Mouse,
6-hr @ 5 ppm
Adult Female,
2.46 ppm
Sensitive Adult Male,
2.23 ppm •«-
Obese Adult Male, ^_
2.05 ppm
9-yr old, 2.07 ppm •«-
1-yrold, 1.99 ppm •*•
O)
E
CD
CO
o
O)
10
CN
II
21
O
Adult Male Human, 8.1 ppm continuous
Exposure = HEC
Divide by 3.2 to account
forTD
(HEC/3.2) = 2.56 ppm
FIGURE 5-12
Application of Toxicokinetic Analysis to Address Animal-to-Human and Human Interindividual
Differences for Chloroform Metabolism and Risk Assessment. CM24 was used as the basis to
adjust concentration between and among species. With CM24 set to 136 mg/L from mouse
studies, the HEC is 8.1 ppm, and dividing that concentration by 3.16 to account for TD
differences between test animals and humans results in a concentration of 2.56 ppm, for a
continuous exposure at steady state (in male humans) CM24 value of 42.75 mg/L. The CM24
value in adult females exposed to 2.56 ppm was 44.50.
The lower portion of the figure depicts the process used to ascertain the impact that
measured differences in blood flow to the liver and metabolic activity of CYP2E1 toward
chloroform, as well as age-dependent parameter values, have on the development of GVk/t.
These differences, as best expressed for risk assessment purposes, are expressed in units of
exposure concentration (ppm). Those concentration differences may be considered in
developing a quantitative value for the TK portion of UFH. This analysis demonstrated that
differences in age (adults versus children), differences in CYP2E1 content/activity, and
differences in HBF (as the difference between the geometric mean value and the value at the 95th
percentile of the distribution) result in differences of up to 1.3-fold (2.56 ppm/1.99 ppm) in this
measure. No data are available to describe the magnitude of TD variability among humans.
5-67
-------
distribution (15.56 mg/h-kg), and with a B:A PC value of 13.3 (and resulting L:B PC value of
1.3) and used to simulate an exposure to 2.56 ppm. The results indicated that CM24 increased by
approximately 15% (to 49.19 mg/L) above the CM24 value observed in the general adult human
population. Under these conditions (high HBF, high metabolic capacity, high B:A PC value),
exposure to 2.23 ppm would result in a CM24 value of 42.75 mg/L.
Obesity in Adult Males. The adult male PBPK model was reparameterized to reflect a state of
obesity as defined by CDC. The model was exercised to determine CM24 at steady state under
exposures to 0.75, 2.56 and 10 ppm. At those concentrations, CM24 values of 15.662, 53.458
and 208.8 mg/L were observed, respectively. Because of the near linearity of the relationship
between external concentrations and CM24, linear regression was employed to determine what
external concentration equated to a CM24 value of 42.75 mg/L. Results indicated that the obese
adult male exposed to 2.05 ppm at steady state would be expected to attain the targeted value for
CM24.
Summary of Results. These results demonstrate the successful application of PBPK modeling
of chloroform metabolism in multiple species. The mouse appears to be more sensitive than the
rat, and the model employed CM24 to equalize concentrations of chloroform in inspired air
between mice and humans and between rats and humans. The HEC, developed from results with
mice, is 8.1 ppm. A downward adjustment by a factor of 3.16 was performed to account for TD
sensitivity of humans compared with rats. The resulting concentration of chloroform was 2.56
ppm. The CM24 value (42.75 mg/L of liver) corresponding to this concentration in the adult
model was used to simulate human interindividual differences in CM24 under the conditions of
varying age, HBF, and chloroform metabolic capacity. Results indicated that children were
somewhat more sensitive to CM24 production at concentrations equivalent to those in adults.
5-68
-------
CM24 values equivalent to those in the 2.56-ppm exposed adult were reached at 2.07 ppm in the
9-year-old child and at 1.99 ppm in the 1-year-old child — a differential sensitivity of 1.28-fold
(i.e., 2.56 ppm/1.99 ppm). The impact of variation in HBF among adults, assessed by comparing
the concentrations of chloroform required to produce equivalent CM24 values was 1.17-fold
whether HBF was set at the geometric mean or at the 95th percentile of the distribution (2.56 ppm
and 2.19 ppm, respectively). While the impact of blood flow on chloroform metabolism is not
surprising, it does represent the initial quantitative demonstration of biologically defensible
values for variability in liver blood flow, and its impact on bioactivation, as assessed via PBPK
modeling. The sensitivity of CM24 to changes in HBF is roughly 0.75, meaning that quite less
than a 1:1 relationship exists between blood flow and chloroform metabolism, implicating the
additional impact of other factors on chloroform metabolism. Changes in liver blood flow had a
more appreciable impact on chloroform metabolism than changes in metabolic capacity (VmaxC)
when assessed within an age group. Within an age group, changes in the VmaxC parameter had
less than 1% influence on CM24, when the difference was quantified between the CM24 values at
the geometric mean and at the 95th percentile for metabolic capacity. The impact of combining
high HBF with high metabolic capacity and high blood:air partitioning in the adult resulted in a
1.15-fold difference (2.23 versus 2.56 ppm) in the external concentrations required to produce a
CM24 value of 42.75 mg/L of liver.
DISCUSSION
These results indicate that chloroform is absorbed by rats, mice, and humans and that
substantial hepatic metabolism occurs. Because the metabolites of chloroform are responsible
for its mode of action, this analysis focused on the metabolism of chloroform in the liver. In
spite of nasal epithelium being a rather sensitive tissue in rodents exposed to chloroform by
5-69
-------
inhalation, a 13-wk inhalation study identified the liver as the critical target of toxicity and a 6-h,
5-ppm exposure was identified as the mouse no-effect level. Similarly, the rat no-effect level
was identified as a 6-h, 30-ppm exposure in a 13-wk inhalation study. The apparent lack of
sensitivity of human nasal tissues to inhaled chloroform was also taken into consideration. Thus,
the PBPK models employed were structured to quantify liver metabolism of chloroform in order
to enable comparisons of sensitivity across species and within segments of the human
population. Units of comparison were those of the integrated amount of chloroform metabolized
over 24 hours, identified as CM24. The use of CM24 allows a direct comparison of liver
concentrations across species and within the human species, as opposed to simply comparing the
amounts metabolized.
Because this analysis was constructed to support the derivation of an RfC value based on
liver effects, indicative of risk from lifetime exposure, CM24 was determined using intermittent
exposures in the mouse (replicating the response-related exposure), and the human model was
then run until steady state was reached. Thus, human CM24 values were demonstrated under
steady state during continuous exposure. Differences between rats, mice, and humans exposed to
the same conditions demonstrate that rats and mice absorb and metabolize chloroform more
efficiently than equally exposed humans. Furthermore, rats appear to be less sensitive to
chloroform, based on no-effect levels expressed as external concentrations. However, this
differential sensitivity is not based solely on differences in chloroform absorption and
metabolism. The no-effect level in rats (30 ppm) is 6-fold higher than that observed in mice (5
ppm), and the CM24 values at the no-effect level are roughly 3-fold higher in the rat versus the
mouse, as well — indicating the existence of species differences in response to metabolite
production. Thus, issues of sensitivity among species (rats, mice, and humans) should include
5-70
-------
specific consideration of internal dose metrics best linked with the mode of action, rather than
consideration solely of external exposure measurements. This is why dosimetric analyses of
CM24 were performed in mice, rats, and humans via PBPK modeling. The CM24 value of 136
mg/L (derived in mice) was chosen as the value for the dose metric upon which to base species
extrapolation, and derive the HEC (Figure 5-7). Through an iterative analysis, the human PBPK
model indicated that under internal steady-state the continuously-encountered external exposure
concentration resulting in a CM24 value of 136 mg/L in the human was 8.1 ppm. Consistent with
U.S. EPA guidance, this HEC was divided by a factor of 3 to account for remaining uncertainty
(i.e., toxicodynamics, etc) in the animal to human extrapolation step. Because of the near-
linearity observed in the relationship between external concentrations and internal doses (CM^)
in this range of concentrations, the same species-adjusted concentration (e.g., 2.56 ppm, Figures
5-7 and 5-12) would have resulted had the remaining factor of 3 been used to downwardly adjust
the internal dose metric (a process not covered by U.S. EPA guidance).
Delic et al. (2000) also performed a PBPK model-based analysis of chloroform
metabolism in mice and humans. However, the focus of that investigation was for intermittent,
occupational exposures, rather than continuous exposures representative of environmental
exposures. Another difference is that that investigation compared of rates of metabolism among
species (Delic et al., 2000) versus integrated amount of metabolites formed under the conditions
of continuous exposure and at steady state in the human. Furthermore, the results of Delic et al.
(2000) are based on a human model that includes values for organ volumes, blood flows,
partition coefficients, and metabolic rate constants that differ from those employed in our
investigation. Interestingly, the KM value used in the human model by Corley et al. (1990) —
which served as the basis for Delic's human model — was derived as the mean value of
5-71
-------
optimized KM values from rats and mice, while the KM value employed in the present work was
extrapolated from in vitro experiments conducted specifically for this purpose using samples of
human hepatic microsomal protein (Lipscomb et al., 2004). Delic et al. (2000) reported that a
peak metabolic rate of 391 nmol/h/gram liver was observed (at 3.7 hs into the 6-h exposure) in
mice exposed to 10 ppm, but that this metabolic rate was not observed in humans until the
exposure concentration reached 130 ppm — 13-fold higher. While credible, using such a
comparison of species differences in dose metrics instantaneously determined under conditions
that were not representative of environmental conditions is ill-advised.
In contrast to the interpretations of Delic et al. (2000), our results based on the integrated
amount of metabolite formed between species indicate a difference at this (mouse 6-h) exposure
concentration of approximately 3-fold between mice and humans who are subjected to a
continuous exposure. Our results may differ from those of Delic et al. (2000) for several
reasons, including differences in modeling approaches and in values for model parameters, such
as for tissue partitioning and metabolism. In contrast to results from rodents, adequate human
data on chloroform disposition in humans that could be used to verify the present human PBPK
model and its predictions of chloroform metabolism do not exist. An attempt was made to find
information against which to compare the model predictions of human interindividual TK
variability — specifically data that relate directly to hepatic chloroform metabolism following
inhalation exposure. However, even in studies which sound relevant, complications and
limitations of study design, evaluation, and the level of detail in reporting limit their value in
verifying our results.
The PBPK modeling approach to chloroform dosimetry has been undertaken by several
groups (Corley et al., 1990, 2000; Xu and Weisel, 2004). In developing and evaluating these
5-72
-------
models, the respective authors cite no data sets involving humans except that of Fry et al. (1972)
that would be useful as comparators for estimations of human interindividual differences in
chloroform metabolism. Corley et al. (1990) developed PBPK models for chloroform
disposition in rats, mice, and humans and used the human data available (Fry et al., 1972) against
which model predictions were compared. A comparison of the present model predictions to
those data are in Tables 5-15 and 5-16. Corley et al. (1990) also examined human interindividual
differences in hepatic microsomal metabolism of chloroform, but reported metabolism at only
one substrate concentration. They also reported the value as velocity over substrate
concentration (V/S), a measure that is not informative of human differences in metabolism
kinetic constants that are required as parameter values in models designed to evaluate the impact
of metabolic variance on tissue dosimetry. Furthermore, Corley et al. (1990) made no mention
of human chloroform data obtained from an inhalation exposure. More recently, Corley et al.
(2000) developed a PBPK model to assess the dermal absorption of chloroform in human.
Multiple subjects were exposed via the dermal route, but the only measures of exposure were
obtained in exhaled breath. The sampling strategy, combined with model structure and
parameterization, hinder the application of these data to assess human interindividual variability
in chloroform metabolism. Of the multiple studies cited by Corley et al. (2000), none reported
data from humans in which study design and individual human characteristics were sufficient to
aid in the understanding of chloroform metabolism.
On a number of occasions, Weisel and colleagues have undertaken and reported results of
modeling investigations designed to determine chloroform uptake by humans. In a recent
investigation (Xu and Weisel, 2004), human subjects were exposed to chloroform vapor via the
dermal and inhalation routes, and samples of expired breath were analyzed to ascertain route-
5-73
-------
dependent fractional uptake of chloroform. However, these results (concentrations of
chloroform in expired air) are not useful benchmarks against which to compare the metabolism
of chloroform, as we report. In conclusion, no data were located describing chloroform
disposition in humans against which the predictions of the model developed in the present study
could be verified.
Our present work quantifies species differences in chloroform metabolism and has
applied it to rodent no-effect exposure values in order to demonstrate values for the HEC for
application in the derivation of chloroform's RfC. We have used data from mice to demonstrate
an HEC of 8.1 ppm. Confidence in this species extrapolation is increased, given the ability of
the model to adequately simulate chloroform metabolism in both test species and humans.
An additional analysis of the impact that human interindividual variance exerts on the
risk-relevant TK outcome, CM24, was also performed. This was accomplished first by
quantifying the extent to which adult humans vary in the fraction of CO delivered to the liver.
To avoid the introduction of operator-based confounding factors that may occur in ultrasound
studies as well as the natural variability in bifurcation of the hepatic vasculature, we employed
only data derived from studies that developed ICG-based measures of HBF in the supine position
among fasted adults. Because ICG is a dye effectively cleared by the liver, measures of ICG can
be used to determine total HBF from the hepatic artery, splanchnic vein, and hepatic portal vein.
Data on a total of 59 subjects were available, either from values published in the open literature
or as individual data graciously provided by investigators who had published mean values from
that sample of subjects.
A second measure of human variance was quantified as the extent to which cytochrome
P450 2E1 (CYP2E1) content and activity toward chloroform (two separate measures) varied
5-74
-------
among adults. The variance of CYP2E1 content among samples of liver from child organ donors
was also determined. Because the metabolic activity of CYP2E1 is not a function of age (as is
content), the activity of adult CYP2E1 was also employed in analysis of children via PBPK
modeling.
The influence of age was estimated by incorporating values for body composition (i.e.,
VLC, volume of the liver compartment) developed for adults and children of different ages,
incorporating values for tissue partitions developed for adult humans and for children, using
adult and child blood as well as solid tissues from adult and young rats, and by incorporating
age-dependent variations in the hepatic content and activity of CYP2E1. The impact of body
weight and fat compartment size was investigated by developing models for adult females, adult
males and for obese adult males. Finally, for adults, variability in HBF was incorporated into the
model developed for the 70 kg adult male. These modifications represent a natural, but not
heretofore accomplished, extension of PBPK modeling to address biochemical, as well as
physiologic, variation among humans.
Varying chloroform metabolic capacity between the geometric mean and 95th percentile
for the adult population resulted in less than 1% change in chloroform metabolism. Biologically
plausible changes in CYP2E1 content and/or activity toward chloroform did not change CM24.
On the other hand, increasing FffiF from the geometric mean (approximately 26% of CO) to the
value at the 95th percentile of the distribution in adults (approximately 45% of CO) resulted in a
17% increase in CM24. No data on variability of FffiF were available for children. Chloroform
metabolism (CM24) in the 9- and 1-year-old child at the geometric mean for VmaxC was 126 and
135% of the value at the geometric mean for VmaxC in the adult population. Increasing VmaxC to
values at the 95th percentile of the distributions for the 9- and 1-year-old child resulted in less
5-75
-------
than a 1% increase in CM24. These results within age groups indicate that differences in
CYP2E1 content and activity toward chloroform, which may reach approximately 2-fold (VmaxC
when extrapolated from in vitro data), have little or no effect on the risk-relevant measure, CM24,
in intact humans. This is consistent with findings of the impact of CYP2E1 variability and its
influence on the metabolism of trichloroethylene (Lipscomb et al., 2003) and is a function of
blood flow delivery to the liver as the rate-limiting step in metabolism. On the other hand,
consistent with blood flow as the rate-limiting step in metabolism, differences in HBF result in
alterations of CM24, but the sensitivity of CM24 to HBF is only about 75%.
In a previous investigation of PK differences among adults and between adults and
children, Pelekis et al. (2001) used predicted values for chloroform partition coefficients in adult
and child tissues, based on age-dependent differences in lipid and water content of tissues. In
that analysis, the concentration of parent chloroform, but not its metabolism or metabolites, was
addressed following a simulated 30-d exposure. Those adult-child differences in the
concentration of parent compound in tissues (i.e., liver, muscle, fat, blood) were less than 10, but
were consistently higher in the child. Although our present analysis was not designed to capture
tissue concentrations of parent compound, this difference is consistent with our demonstration
that the child converts a higher dose of chloroform to toxic metabolites than the adult.
To investigate the impact of sex- and obesity-related changes in body composition,
related to the fat compartment, models were additionally constructed for the adult female and the
obese adult male. Results indicated an approximate 4% difference between CM24 values attained
in adult males and adult females; and an approximate 25% difference in CM24 values attained in
the obese adult male compared to the 70 kg adult male.
5-76
-------
Recently, Tan et al. (2003) published the results of modeling efforts designed to link
chloroform metabolism with hepatocyte repair and regeneration. Data on hepatocyte LI
originally published by Constan et al. (2002) were used as the basis for their model development
and its evaluation. Their pharmacodynamic model employed a PBPK model to transition an
external (i.e., inhaled) concentration to a measure of chloroform metabolism (rate per cubic
centimeter of liver) during an intermittent exposure regimen. The Tan et al. (2003) model was
based on the female B6C3F1 mouse and employed the base model of Corley et al. (1990), with
the exception that the parameters governing the inhibition of CYP2E1-dependent metabolism by
chloroform metabolites were not included, as Tan et al. indicated that they were not necessary.
Corley's model was based on the structure developed by Ramsey and Andersen (1984) for
styrene. Tan et al. (2003) developed B:A, L:A, and kidney:air (K:A) PC values for this sex and
strain of mouse using Sato and Nakajima's vial equilibration technique (Sato and Nakajima,
1979). B:A, L:A, and K:APC values were 24.1, 16.9, and 12.2, respectively, and resulted in L:B
and K:B PC values of 0.7 and 0.5, respectively. The PBPK model was used, after fitting to in
vivo chloroform gas uptake data for female B6C3F1 mice, to optimize VmaxC, which the model
predicted to be 10.06 mg/h-kg. The PD model was constructed to define "virtual damage" by
linking chloroform exposure to cytolethality. Virtual damage was not directly correlated with
any measured endpoint, and the rate of damage formation was proportional to the rate of
chloroform metabolism per volume of tissue. The parameter, Rmet, was defined as the rate of
chloroform metabolism per volume tissue, and was expressed in units of mg/h-cm3. The model
was constructed around the premise that a liver-derived factor circulating in blood (a signal
kinetic factor) serves as a stimulus to existing hepatocytes to divide and replace dead cells and
that a saturable rate exists for repair of cytotoxic damage from oxidative metabolites of
5-77
-------
chloroform. The authors indicated that, "The rate of damage repair, the form of the linkage
between damage and cellular death rate, and the signal kinetics that drive regenerative
proliferation constitute a structural alternative to an empirical correlation between a target tissue
dose surrogate and the toxicological endpoint." Results demonstrated a good fit between model
predictions of hepatocyte LI (from Constan et al., 2002) and model predictions for 2-, 6-, 12-,
and 18-h exposures to 10, 30, and 90 ppm. In their publication (Tan et al., 2003), Figure 4
demonstrates that a 12 h/d exposure to 10 ppm for 7 consecutive days produced a hepatocyte LI
roughly equivalent to the control (nonexposed) hepatocyte LI of 0.7 + 0.2 %.
Uncertainty factors can be revised from default values or obviated by TK evaluations.
Three levels of specificity can be used to develop RfC values. The first is the application of
default methodologies to external concentrations. The second is the application of TK
evaluations to develop the HEC, and the application of the default value for UFn. The third
method is to employ available information (here TK information) to both develop the HEC and
describe the value of the TK component of UFn. Each of these approaches would result in
appreciably different values for the RfC for chloroform.
In the first case, the application of default methodology to external concentrations
employs duration adjustment. The no-effect exposure of 5 ppm x 6 hours, 5 ds/wk is adjusted to
yield a duration-adjusted concentration of 0.9 ppm, to which default values of UFA (10) and
(10) are applied, resulting in a concentration of 0.009 ppm adjusted for inter and intra species
uncertainty (i.e., 0.9 ppm/100).
The application of TK information, whether only for species extrapolation or for the
additional extrapolation within humans, results in appreciably different values. The second
approach would quantify TK variability between species to develop the HEC and apply the
5-78
-------
default value for UFH to develop the RfC. This approach develops and applies species
differences in the translation of external concentrations to the internal dose metric in order to
extrapolate a no-effect mouse exposure to an HEC (inhalation) of 8.1 ppm. Applying default
values for UFA-ID (3) and for UFn (10), would result in a concentration of 0.27 ppm adjusted for
inter and intra species uncertainty (i.e., 8.1 ppm/[3 x 10]).
Finally, TK information may be used to develop the HEC and to quantify human
interindividual variability, yielding a "data-derived" value for UFH-TK. In this approach, the HEC
is adjusted downward by one-half order of magnitude to adjust for the increased TD sensitivity
of the human compared to animals, resulting in a concentration of 2.56 ppm. Consistent with the
application of TK evaluation to adjust for external exposures, the internal dosimeter, CM24, was
again employed to address human interindividual variability. Analogous to addressing species
differences, the TK evaluation precedes the application of TD uncertainty factors within the
human species. When evaluations were conducted at steady state and PBPK models for adults
and children were varied to include values for FffiF, blood:air partitioning of chloroform,
metabolic rate constants and body composition, the largest difference in external concentrations
was observed between the 70 kg adult male model and the model for the 1-year-old child. When
standardized on a CM24 value of 42.75 mg/L, the magnitude of that difference (expressed in
terms of the external concentration) was 1.3 (i.e., 2.56 ppm/1.99 ppm). This difference exceeded
the difference between the 70 kg adult male human and the adult female (1.04); the adult human
concomitantly "sensitized" for B:A PC value, FffiF, and metabolic capacity (i.e., 2.56 ppm/2.23
ppm = 1.15-fold); and the obese adult male (1.25). The difference between these external
exposure conditions (maximally observed difference of 2.56 ppm/1.99 ppm = 1.3) should be
included in discussions of the value to be ascribed to UFH-TK. In the absence of data directing the
5-79
-------
selection of the value for UFH-TD, the UFH-TD should remain at the default value of one-half order
of magnitude (3.16). With enough confidence in the present evaluation of human interindividual
TK variability (child:adult = 1.3-fold), quantitative reliance in these results may support
replacement of the default uncertainty factor value. Following the development of the HEC
(inhalation) of 8.1 ppm, a default value of 3 for UFA-TD, the data-derived value of 1.3 for UFH-TK
and a default value of 3 for UFn-TD could be applied in the development of an RfC value.
CONCLUSIONS
PBPK modeling was successfully applied to convert external exposure concentrations to
internal measures of chloroform metabolites in the mouse, rat, and human. These models were
used to transition no-effect values from toxicity studies with rats and mice into HECs, based on
an internal dose metric quantifying chloroform metabolism in each species/model. Humans
convert less of a dose of chloroform to metabolites than do either rats or mice.
Separate human models were constructed to represent the adult, the 9-year-old and the
1-year-old child. Models were populated with age-specific organ volumes and partition
coefficients. CYP2E1 content of liver was determined for adults and children. Chloroform
metabolic rate constants were derived in vitro, extrapolated to the intact organ, scaled by BW°7,
and incorporated into the PBPK models. Laboratory studies demonstrated variability of the
blood:air partitioning of chloroform into human blood. Variability of HBF among adults was
determined from previously published results. Under no conditions was the difference between
chloroform metabolism under mean conditions in the adult human and under conditions
representing extreme values (or values at the 95th percentile for a distribution) elevated more
than 25%. Among the factors assessed within the adult human the impact on variability in
5-80
-------
chloroform metabolism was greatest: HBF > B:A PC value > Vmax. Children converted as much
as 30% more of a dose of chloroform to metabolites than adults.
This report demonstrates the feasibility of using PBPK modeling as a platform to
integrate anatomic variance (i.e., relative organ volumes), and physiologic variance (i.e., HBF)
with biochemical variance (e.g., in CYP2E1 content, enzyme kinetic parameters, and blood:air
partitioning) to assess their separate impact on the formation of a risk-relevant pharmacokinetic
outcome. These results indicate the potential application of this approach to quantitatively assess
human interindividual variability at the level of the risk-relevant pharmacokinetics outcome,
which for chloroform is the amount of chloroform oxidized in the liver. This is significant
because other conditions (e.g., diseases) can alter basic physiology and biochemistry in humans,
and these results demonstrate the applicability of PBPK modeling to translate those changes into
harbingers of susceptibility, when they, themselves, can be quantified.
The present report follows Agency RfC guidance in the sequence of extrapolating animal
internal dosimetry to a human equivalent concentration (i.e., the external concentration resulting
in the same internal dose), then continuing on with application of the toxicodynamic uncertainty
factor. A more technically correct procedure may have been to have downwardly adjusted the
animal internal dose (CM24) by a default factor of 3 for animal to human differences in response
(toxicodynamics), then employed the PBPK model to predict an external concentration resulting
in a value for the internal dose that could be used as the starting point for application of the
human interindividual variability uncertainty factor (UFH). This is an important point. When
the relationship between external concentration and internal dosimetry is non-linear in the range
of the animal point of departure, this issue becomes critical. However, because the system here
5-81
-------
studied demonstrates marked linearity in this range, the results will be the same, whether the
downward adjustment for TD is applied to the external concentration or the internal dose.
REFERENCES
ACGIH (American Conference of Governmental Industrial Hygienists). 2003. Threshold Limit
Values for Chemical Substances and Physical Agents and Biological Exposure Indices.
American Conference of Governmental Industrial Hygienists, Cincinnati, OH.
Amman, P., C.L. Laethem and G.L. Kedderis. 1998. Chloroform-induced cytolethality in
freshly isolated male B6C3F1 mouse and F-344 rat hepatocytes. Toxicol. Appl. Pharmacol.
149:217-25.
Andersen, M.E., M.E. Meek, G.A. Boorman et al. 2000. Lessons learned in Applying the U.S.
EPA proposed cancer guidelines to specific compounds. Toxicol. Sci. 53:159-172.
Arms, A.D. and C.C. Travis. 1988. Reference physiological parameters in pharmacokinetic
modeling. U.S. Environmental Protection Agency, Washington, D.C. EPA/600/6-88/004.
Beliveau, M., G. Charest-Tardif and K. Krishnan. 2001. Blood:air partition coefficients of
individual and mixtures of trihalomethanes. Chemosphere. 44:377-381.
Caesar, J., S. Shaldon, L. Chiandussi, L. Guevara and S. Sherlock. 1961. The use of indocyanin
green in the measurement of hepatic blood flow and as a test of hepatic function. Clin. Sci.
21:43-57.
CDC (Centers for Disease Control and Prevention). 2005. BMI - Body Mass Index: BMI for
Adults. Centers for Disease Control and Prevention. Available at
http ://www. cdc.gov/nccdphp/dnpa/bmi/bmi-adult.htm.
Constan, A.A., B.A. Wong, J.I. Everitt and B.E. Butterworth. 2002. Chloroform inhalation
exposure conditions necessary to initiate liver toxicity in female B6C3F1 mice. Toxicol. Sci.
66:201-208.
Corley, R.A., A.L. Mendrala, F. A. Smith et al. 1990. Development of a physiologically based
pharmacokinetic model for chloroform. Toxicol. Appl. Pharmaocl. 103:512-527.
Corley, R.A., S.M. Gordon and L.A. Wallace. 2000. Physiologically based pharmacokinetic
modeling of the temperature-dependent dermal absorption of chloroform by humans following
bath water exposure. Toxicol. Sci. 53:13-23.
Delic, J.I., P.O. Lilly, AJ. MacDonald and G.D. Loizou. 2000. The utility of PBPK in the
safety assessment of chloroform and carbon tetrachloride. Reg. Toxicol. Pharmacol.
32:144-155.
5-82
-------
Evans, M.V., W.K. Boyes, I.E. Simmons, O.K. Litton and M.R. Easterling. 2002. A
comparison of Haber's rule at different ages using a physiologically based pharmacokinetic
(PBPK) model for chloroform in rats. Toxicology. 176:11-23.
Feruglio, F.S., F. Greco, L. Cesano, D. Indovina, G. Sardi and L. Chiandussi. 1964. Effect of
drug infusion on the systemic and Splanchnic circulation: I. Bradykinin infusion in normal
subjects. Clin. Sci. 26:487-491.
Fry, B.J., R. Taylor and D.E. Hathaway. 1972. Pulmonary elimination of chloroform and its
metabolite in man. Arch. Int. Pharmacodyn. 196:98-111.
Gargas, M.L., M.E. Andersen and HJ. Clewell III. 1986. A physiologically based simulation
approach for determining metabolic constants from gas uptake data. Toxicol. Appl. Pharmacol.
86:341-352.
Gearhart, J.M., C. Sekel and A. Vinegar. 1993. In vivo metabolism of chloroform in B6C3F1
mice determined by the method of gas uptake: The effects of body temperature on tissue
partition coefficients and metabolism. Toxicol. Appl. Pharmacol. 119:258-266.
ICRP (International Commission on Radiological Protection). 1975. Task Force on Reference
Man, Vol 25.
ICRP (International Commission on Radiological Protection). 2002. Basic Anatomical and
Physiological Data for Use in Radiological Protection: Reference Values. ICRP Publication 89.
J. Valentin, Ed. Pergamon Press, New York, NY. ISBN 0080442668.
lijima, T., F. Ohishi, T. Tatara and Y. Iwao. 2001. Effect of continuous infusion of
prostaglandin El on hepatic blood flow. J. Clin. Anesth. 13:250-254.
ILSI (International Life Sciences Institute). 1997. An Evaluation of EPA's Proposed Guidelines
for Carcinogen Risk Assessment Using Chloroform and Dichloroacetate as Case Studies. Report
of ILSI HESI Expert Panel. International Life Sciences Institute, Washington, DC.
IPCS (International Programme on Chemical Safety). 2001. Guidance Document for the Use of
Data in Development of Chemical-Specific Adjustment Factors (CSAFs) for Interspecies
Differences and Human Variability in Dose/Concentration-Response Assessment. World Health
Organization, Geneva. Available at http://www.who.int/pcs/harmon site/harmonize/index.htm.
Kedderis, G.L. and S.D. Held. 1996. Prediction of furan pharmacokinetics from hepatocyte
studies: Comparison of bioactivation and hepatic dosimetry in rats, mice, and humans. Toxicol.
Appl. Pharmacol. 140:124-30.
Larson, J.L., M.V. Templin, D.C. Wolf et al. 1996. A 90-day chloroform inhalation study in
female and male B6C3F1 mice: implications for cancer risk assessment. Fundam. Appl.
Toxicol. 30:118-37.
Lerman, J., G.A. Gregory, M.M. Willis and E.I. Eger Jr. 1984. Age and solubility of volatile
anesthetics in blood. Anesthesiology. 61(2):139-143.
5-83
-------
Lipscomb, J.C. and G.L. Kedderis. 2002. Incorporating human interindividual
biotransformation variance in health risk assessment. Sci. Total Environ. 288:13-21.
Lipscomb, J.C., L.K. Teuschler, J. Swartout, D. Popken, T. Cox and G.L. Kedderis. 2003. The
impact of cytochrome p450 2el-dependent metabolic variance on a risk relevant
pharmacokinetic outcome in humans. Risk Anal. 23:1221-1238.
Lipscomb, J.C., H. Barton, R. Tornero-Velez et al. 2004. The metabolic rate constants and
specific activity of human and rat hepatic cytochrome P450 2E1 toward chloroform. J. Toxicol.
Environ. Health. 67:537-553.
Malviya, S. and J. Lerman. 1990. The blood/gas solubilities of sevoflurane, isoflurane,
halothane, and serum constituent concentrations in neonates and adults. Anesthesiology.
72(5):793-796.
Pelekis, M., L.A. Gephart and S.E. Lerman. 2001. Physiological-model-based derivation of the
adult and child pharmacokinetic intraspecies uncertainty factors for volatile organic compounds.
Reg. Toxicol. Pharmacol. 33:12-20.
Ramsey, J.C. and M.E. Andersen. 1984. A physiologically based description of the inhalation
pharmacokinetics of styrene in rats and humans. Toxicol. Appl. Pharmacol. 73:159-175.
Reemtsma, K., G.C. Hottinger, A. DeGraff, Jr. and O. Creech. 1960. The estimation of hepatic
blood flow using indocyanin green. Surg. Gynecol. Obstet. 110:353-356.
Reitz, R.H., A.L. Mendrala, R.A. Corley et al. 1990. Estimating the risk of liver cancer
associated with human exposure to chloroform using physiologically based pharmacokinetic
modeling. Toxicol. Appl. Pharmacol. 105:443-459.
Richardson, S.D. 1998. Identification of drinking water disinfection by-products. In: The
Encyclopedia of Environmental Analysis & Remediation, Vol. 3, R.A. Meyers, Ed. John Wiley,
New York, NY. p. 1398-1421.
Rook, J.J. 1974. Formation of haloforms during chlorination of natural waters. Water Treat.
Exam. 23:234-243.
Sakka, S.G., K. Reinhart, K. Wegscheider and A. Meier-Hellmann. 2001. Variability of
splanchnic blood flow in patients with sepsis. Intensive Care Med. 27:1281-1287.
Sato, A. and T. Nakajima. 1979. A vial-equilibration method to evaluate the drug-metabolizing
enzyme activity for volatile hydrocarbons. Toxicol. Appl. Pharmacol. 47:41-46.
Smith, A.G. and J.S. Evans. 1995. Uncertainty in fitted estimates of apparent in vivo metabolic
constants for chloroform. Fundam. Appl. Toxicol. 25:29-44.
Tan, Y.-M., B.E. Butterworth, M.L. Gargas, and R.B. Conolly. 2003. Biologically motivated
computational modeling of chloroform cytolethality and regenerative cellular proliferation.
Toxicol. Sci. 75:192-200.
5-84
-------
Templin, M.V., J.L. Larson, B.E. Butterworth et al. 1996. A 90-day chloroform inhalation study
in F-344 rats: Profile of toxicity and relevance to cancer studies. Fund. Appl. Toxicol.
32:109-125.
Templin, M.V., A.A. Constan, D.C. Wolf, B.A. Wong, and B.E. Butterworth. 1998. Patterns of
chloroform-induced regenerative cell proliferation in BDF1 mice correlate with organ specificity
and dose-response of tumor formation. Carcinogenesis. 19:187-193.
Testai, E., V. de Curtis, S. Gemma, L. Fabrizi, P. Gervasi, and L. Vittozzi. 1996. The role of
different cytochrome P-450 isoforms in in vitro chloroform metabolism. J. Biochem. Toxicol.
11:305-312.
Thomas, V. (V. Fiserova-Bergerova). 1975. Biological-Mathematical Modeling of Chronic
Toxicity. U.S. Air Force, Aerospace Medical Research Laboratory, Aerospace Medical
Division, Wright-Patterson AFB, Ohio, AMRL-TR-75-5.
U.S. EPA. 1994. Methods for Derivation of Inhalation Reference Concentrations and
Application of Inhalation Dosimetry. Office of Research and Development. October.
EPA/600/8-90/066F.
Weigand, B.D., S.G. Ketterer, and E. Rapaport. 1960. The use of indocyanin green for the
evaluation of hepatic function and blood flow in man. Am. J. Dig. Dis. 5:427-436.
Xu X and C.P. Weisel. 2004. Human respiratory uptake of chloroform and haloketones during
showering. J Expo Anal Environ Epidemiol. 2004:1-11. Available at:
http://www.nature.com/cgi-
taf/DvnaPage.taf?file=/iea/iournal/vaop/ncurrent/full/7500374a.html&filetype=pdf
NOTICE: This chapter has been subjected to internal and external peer review in accord with
EPA/ORD policy. The authors wish to thank Hugh Barton, Justin Teeguarden and Jeff Gearhart
for review comments. The work has been further improved through insightful comments from
Harvey Clewell, Michael Gargas, Janusz Byczkowski, Dick Bull, Moiz Mumtaz and Gary
Ginsberg.
5-85
-------
APPENDIX A
COMPUTER CODE FOR THE CHLOROFORM PBPK MODEL
The PBPK model for chloroform was coded in acslXtreme software (Aegis Technologies,
Huntsville, AL), versions 1.3.2 and 1.3.19. The parameter values used for chloroform are shown
in Table Y. The code for the chloroform PBPK model follows below. The ! separates comments
from computer code.
PROGRAM volatiles
! PBPK model for volatile organic chemicals
! 6/24/03 GLK based on furan 5/11/94
INITIAL
CONSTANT QPC = 14. lalveolar ventilation rate (L/hr/kg)
CONSTANT QCC = 14. ! CARDIAC OUTPUT (1/HR/KG)
CONSTANT QLC = 0.25 IFRACTIONAL BLOOD FLOW TO LIVER
CONSTANT QFC = 0.09 IFRACTIONAL BLOOD FLOW TO FAT
CONSTANT BW = 0.25 IBODY WEIGHT (KG)
CONSTANT VLC = 0.04 IFRACTION LIVER TISSUE
CONSTANT VFC = 0.07 IFRACTION FAT TISSUE
CONSTANT BVC = 0.06 IFRACTION BLOOD VOL
CONSTANT VABC = 0.35 IFRACTION ARTERIAL BLOOD VOL
CONSTANT VVBC = 0.65 IFRACTION VENOUS BLOOD VOL
CONSTANT PL = 0.90 ILIVER/BLOOD PARTITION COEFF
CONSTANT PF = 9.72 IFAT/BLOOD PARTITION COEFF
CONSTANT PS = 0.64 I SLOWLY PERFUSED/BLOOD PART COEFF
CONSTANT PR = 0.90 IRICHLY PERFUSED/BLOOD PART COEFF
CONSTANT PB = 6.59 IBLOOD/AIR PARTITION COEFF
CONSTANT MW = 68.07 IFURAN MOLECULAR WEIGHT (G/MOL)
CONSTANT VMAXC = 4.86 IMAXIMAL VELOCITY (MG/HR/KG)
CONSTANT KM = 0.136 IMICHAELIS-MENTEN CONSTANT (MG/L)
CONSTANT ODOSE = 0. ! ORAL DOSE (MG/KG)
CONSTANT KA = 2.0 I ORAL UPTAKE RATE (/HR)
CONSTANT IVDOSE = 0. IIV DOSE (MG/KG)
CONSTANT CONC = 0. I INHALED CONC (PPM)
I TIMING COMMANDS
CONSTANT TSTOP = 6. ILENGTH OF EXPT (HR)
CONSTANT TCHNG = 4. ILENGTH OF EXPOSURE (HR)
CONSTANT TINF = 0.002 ILENGTH OF IV INFUSION (HR)
CONSTANT POINTS = 500 INUMBER OF POINTS IN PLOT
CINT = TSTOP/POINTS I COMMUNICATION INTERVAL
I SCALED PARAMETERS
QC = QCC*BW**0.74
QP = QPC*BW**0.74
QL = QLC*QC
QF = QFC*QC
5-86
-------
QS = 0.24*QC - QF
QR = 0.76*QC-QL
VL = VLC*BW
VF = VFC*BW
VS = 0.82*BW-VF
VR = 0.09*BW - VL
BV = BVC*BW
VAB = BV*VABC
VVB = BV*VVBC
VMAX = VMAXC*BW**0.7
DOSE = ODOSE*BW
IVR = IVDOSE*BW/TINF
END ! OF INITIAL
DYNAMIC
IALG = 2 IGEAR METHOD FOR STIFF SYSTEMS
DERIVATIVE
!CI = CONC IN INHALED AIR (MG/L)
CIZONE = RSW(T.GT.TCHNG,0.,1 )
CI = CIZONE*CONC*MW/24450
! AI = AMOUNT INHALED (MG)
RAI = QP*CI
AI = INTEG(RAI,0.)
!MR = AMOUNT REMAINING IN STOMACH (MG)
RMR = -KA*MR
MR = DOSE*EXP(-KA*T)
!CA = CONC IN SYSTEMIC ARTERIAL BLOOD (MG/L)
CA = (QC*CV + QP*CI)/(QC + (QP/PB))
AUCB = INTEG(CA,0.)
! AX = AMOUNT EXHALED (MG)
CX = CA/PB
CXPPM = (0.7*CX + 0.3*CI)*24450./MW
RAX = QP*CX
AX = INTEG(RAX,0.)
! AS = AMOUNT IN SLOWLY PERFUSED TISSUES (MG)
RAS = QS*(CA - CVS)
AS = INTEG(RAS,0.)
CVS = AS/(VS*PS)
CS = AS/VS
! AMOUNT IN RAPIDLY PERFUSED TISSUES (MG)
RAR = QR*(CA - CVR)
AR = INTEG(RAR,0.)
CVR = AR/(VR*PR)
CR = AR/VR
! AF = AMOUNT IN FAT (MG)
RAF = QF*(CA-CVF)
AF = INTEG(RAF,0.)
5-87
-------
CVF = AF/(VF*PF)
CF = AF/VF
! AL = AMOUNT IN LIVER (MG)
RAL = QL*(CA - CVL) - RAM1 + RAO
AL = INTEG(RAL,0.)
CVL = AL/(VL*PL)
CL = AL/VL
AUCL = INTEG(CL,0.)
! AMI = AMOUNT METABOLIZED, P450 SATURABLE PATHWAY (MG)
RAM1 = (VMAX*CVL)/(KM + CVL)
RAM1M = RAM1*1000./MW
AMI = INTEG(RAM1,0.)
CAM1 = AM1/VL
DM = CAM1*1000./MW
! AO = TOTAL MASS INPUT FROM STOMACH (MG)
RAO = KA*MR
AO = DOSE-MR
!IV = IV INFUSION RATE (MG/HR)
IV = IVR*(1. - STEP(TINF))
!CV = MIXED VENOUS BLOOD CONC (MG/L)
CV = (QF*CVF + QL*CVL + QS*CVS + QR*CVR + IV)/QC
ITMASS = MASS BALANCE (MG)
TMASS = AF + AL + AS + AR + AMI + AX + MR
TERMT(T.GE.TSTOP)
END !OF DERIVATIVE
END ! OF DYNAMIC
END ! OF PROGRAM
5-88
-------
APPENDIX B
DERIVATION OF CHLOROFORM PARTITION COEFFICIENTS
TISSUE PARTITION COEFFICIENTS
Partition coefficients (PC) is a unitless number which describes the partitioning of a
substance between two dissimilar media. The solubility of chemical substances in tissues and
the body is a primary driver of tissue concentrations of the chemical upon exposure from an
environmental medium. In PBPK modeling, the T:B PC and B:A PC values are included in the
file of anatomic, physiologic, and biochemical parameters used to develop the model. The
derivation first requires the characterization of B:A PC and T:A PC values; T:B PC values are
then computed by Equation B-l.
T:A PC value / B:A PC value = T:B PC value (B-l)
Some more recent PBPK models employ PC values derived using only human tissues.
However, the availability of solid tissues from humans, and their application in determining
these PC values is the exception, rather than the rule. Many models employ PC values
determined using tissues from rats only, given the similarity of tissue types and composition
among mammalian species. However, it is rather well accepted that differences, especially in the
B:A PC, exist between rats and humans. For chloroform's B:A PC value, previous
investigations demonstrated this species-dependent variability. Gargas et al. (1989) presented
values of 20.8 and 6.85 for adult male F-344 rats and humans, respectively; Beliveau et al.
(2001) reported a B:A PC value of 21.3 for adult male F-344 rats; Batterman et al. (2002)
reported a B:A PC value of 10.7 for human blood. In 1990, Corley et al. published a PBPK
model for chloroform in humans, which was based on PC values derived for rats and humans.
Rat PC values were determined specifically for the study, and PC values for human tissues were
5-89
-------
obtained from Steward et al. (1973). In that publication, previously available published findings
were condensed and reported for several solvents and anesthetic agents. Steward's report
indicates a B:A PC value of 7.43 in humans. This paper also reported PC values for chloroform
in human blood, liver, kidney, muscle, and brain tissue based on original observations and in fat
based on predictions based on partitioning into olive oil. For comparison, the PC values
employed by Corley et al. (1990) are repeated in Table B-l. Because of the availability of these
data, chloroform was also included in a suite of chemicals, for which PC values in rat blood and
tissues and human blood has also recently been re-investigated (reported in the following
sections).
TABLE B-l
Rat and Human Blood: Air and Tissue:Blood Partition Coefficient Values as reported in
Corley etal. (1990)
Blood:Air
Rata
Humanb
20.8
7.43
Blood:Air
Rat
Human
20.8
7.43
LiverAir
21.1
17.0
LiverBlood
1.01
2.28
Kidney:Air
11.0
11.0
Kidney :Blood
0.53
1.48
Muscle:Air
13.9
12.0
Muscle:Blood
0.67
1.61
FatAir
203
280
FatBlood
9.75
37.68
a Generated by and reported in Corley et al. (1990).
bFrom values collected in Steward et al. (1973), reported in Corley et al. (1990).
The availability of laboratory resources or tissues and the need for more automation has
also driven the development of quantitative structure-activity methods. Several approaches have
been developed, some of which are tailored to fit the characteristics of specific classes of
chemicals. These methods have been employed to predict T: A PC values, based on the
5-90
-------
physicochemical properties of the chemical (i.e., octanol-water partition coefficient and water
solubility) and tissue-specific characteristics (i.e., lipid and water content).
DETERMINATION OF VALUES
Blood: Air and T: A PC values were developed using the vial equilibration method (Sato
and Nakajima, 1979), combined with gas chromatographic evaluation of chemical concentration,
quantified by an external standard curve of known concentrations. Chloroform and five other
volatile organic compounds were exposed in a mixture, and their concentrations in headspace
were determined at equilibrium. This approach has been determined to produce PC values that
are the same as those obtained using single chemical exposures. All studies were conducted
through institutionally reviewed protocols. Human samples were remnant samples obtained
from the Division of Clinical Laboratory Services, Wright-Patterson Regional Medical Center,
OH and coded to ensure anonymity of the subjects. Study protocol was reviewed and approved
by the Wright-Patterson Institutional Review Board and the EPA Human Studies Coordinator.
Samples were provided from 11 adult males, aged 36 to 80 years, and 10 adult females aged 22
to 87 years.
PREDICTION OF PC VALUES
Blood:air and tissue:air PC values were predicted according to Equation A-2.
PB:A or PT:A = [PO:w * PW:A(VnlB)] + [PW:A * VwB] (B-2)
where, PB:A is the B:A PC value, PT:A is the T:A PC value, PO:W is the octanol:water partition
coefficient, PW:A is the waterair partition coefficient, VnlB is the volume fraction of neutral
lipids in blood, and VwB is the volume fraction of water in blood. PQ:W and PW:A were predicted
by the freeware, KOWWIN and HenryWIN, respectively (U.S. EPA, 2003). The fractional
5-91
-------
content of water in blood was available (ICRP, 1975) and the fat content of blood as reported
(ICRP, 1975) was employed for the value for VnlB (Table B-2).
TABLE B-2
Fractional Fat and Water Content of Adult Human Tissues (ICRP, 1975)
Blood
Water
Fat
0.806
0.0065
Liver
0.71
0.069
Kidney
0.76
0.05
Muscle
0.79
0.022 (male)
0.029 (female)
Fat
0.15
0.80
Several data sets of PC values for chloroform were available for this investigation. We
have combined those data and predictions from QSAR approaches to develop and compare T:B
PC values for chloroform. They are based on (1) B:A and T:A PC values derived using rat
tissues, (2) B:A PC values derived using human blood and T:A PC values using solid tissues
from rats, (3) QSAR predictions of B:A and T:A PC values using information characterizing the
fat and water content of human tissues, and (4) a hybrid approach which employs QSAR
predictions of B:A and T:A PC values for human tissues, but scaled to account for the
differences between the observed human B:A PC value and the predicted human B:A PC value.
The results of these are presented in Tables B-3 through B-7.
The final approach (Table B-8) to estimating T:B PC values combined observed human
B:A values with predictions of B:A and T:A values to refine estimates of T:A PC values is based
on the ratio of observed:predicted values for chloroform's B:A PC value in human blood. The
predicted B:A PC value was 3.94, and the observed value was 11.34 — 2.88-fold higher than
predicted. Thus, the predicted T:A PC values for humans were multiplied by 2.88 and combined
with the human mean B:A PC value of 11.34 to estimate human T:B PC values. Table B-9
5-92
-------
TABLE B-3
Blood: Air and Tissue:Air Partition Coefficient Values Derived from Studies with
Rat Tissues
Rat
1
2
3
4
5
6
7
8
9
10
Mean
S.D.
Blood:Air
17
17.5
15.2
20.3
12.3
18.5
17.4
18.4
20.7
20
17.7
2.5
LiverAir
17.2
17.4
17.8
18.3
13.6
25.7
16.3
15.6
16.2
18.2
17.6
3.2
Kidney:Air
16.8
17.6
21.1
16.9
8.7
14.8
14
16.3
15
16
14.8
2.7
Muscle:Air
10.9
10.7
10.1
32.5
38.1
9.5
13.8
17.7
12
13.5
16.9
10.1
Fat:Air
365
352
346
405
229
327
365
378
350
379
351
48
Ten adult male Fischer 344 rats were used in this investigation. B:A PC and T:A PC values were
derived using tissues from individual rats (USAF, 2005).
NOTE: Tissues from rat 4 demonstrates a generally higher B:A PC and T: A PC values than
other rats, this is also true to some extent for tissues from rat 5. Chloroform PC values were
determined in a cocktail approach, and this same pattern is observed for other chemicals in this
exposure suite, including benzene, methyl ethyl ketone, trichloroethylene, and perchloroethylene
(not shown). Because this pattern is demonstrated for multiple tissues (brain not shown) from
the same rats and for multiple chemicals, this indicates that the data are not spurious.
5-93
-------
TABLE B-4
Tissue:Blood Partition Coefficient Values Derived from Studies with Rat Tissues
Rat
1
2
3
4
5
6
7
8
9
10
Mean
S.D.
LiverBlood
1.01
0.99
1.17
0.90
1.11
1.39
0.94
0.85
0.78
0.91
1.0
0.2
Kidney :Blood
0.99
1.01
0.80
0.83
0.71
0.80
0.80
0.89
0.72
0.8
0.8
0.1
Muscle:Blood
0.64
0.61
0.66
1.60
3.10
0.51
0.79
0.96
0.58
0.68
1.0
0.8
FatBlood
21.5
20.1
23.9
19.9
18.6
17.7
20.9
20.5
16.9
19.0
19.9
2.0
Ten adult male Fischer 344 rats were employed in this investigation. B:A PC and T:A PC values
were derived using tissues from individual rats, and rat-specific T:B PC values were developed
(USAF, 2005).
TABLE B-5
Blood: Air Partition Coefficient Values Derived from Studies with Adult Human Blood
B: A PC value
Males, n = 1 1
Females, n= 10
Combined, n = 21
11.9 + 0.9
10.7 + 2.1
11.34+1.65
Range of Observations
9.7-13
6.9-13.3
6.9-13.3
Data presented as mean + S.D. Mean values for men and women were not significantly different
as determined by one-tail t-test assuming unequal variance (p = 0.063) (USAF, 2005).
5-94
-------
TABLE B-6
Tissue:Blood Partition Coefficient Values Derived by Combining Mean Human B: A PC
Values With Individual Rat T: A PC Values
Rat
1
2
3
4
5
6
7
8
9
10
Mean
S.D.
LiverBlood
1.52
1.53
1.57
1.61
1.20
2.27
1.44
1.37
1.43
1.60
1.6
0.3
Kidney :Blood
1.48
1.55
1.07
1.49
0.77
1.31
1.23
1.44
1.32
1.41
1.3
0.2
Muscle:Blood
0.96
0.94
0.89
2.86
3.36
0.84
1.22
1.56
1.06
1.19
1.5
0.9
FatBlood
32.2
31.0
32.1
35.7
20.2
28.8
32.1
33.3
30.8
33.4
31.0
4.2
Blood from 21 adult humans and solid tissues from 10 adult male Fischer 344 rats were used in
this investigation. A B:A PC value of 11.34 (human mean) was combined with individual T:A
PC values to develop estimates of individual T:B PC values (USAF, 2005).
TABLE B-7
Predictions of Human Blood: Air, Tissue: Air, and Tissue:Blood Partition Coefficient Values
Blood:Air
3.94
LiverAir
12.38
LiverBlood
3.14
Kidney: Air
9.98
Kidney :Blood
2.53
Muscle: Air male)*
6.27
Muscle:Blood
1.59
Fat:Air
110.15
FatBlood
27.95
*The muscle:air PC and Muscle:blood PC values for females were 7.23 and 1.83, respectively.
5-95
-------
TABLE B-8
Human Tissue:Air and Tissue:Blood Partition Coefficient Values Based on Adjusted
Predictions of Tissue: Air PC Values
Blood:Air*
11.34
LiverAir
35.63
Liver:Blood
3.14
Kidney:Air
28.72
Kidney :Blood
2.53
Muscle: Air (male)
18.05
Muscle:Blood
1.59
FatAir
317
Fat:Blood
27.95
*Mean value for human B:A PC, observed in humans.
TABLE B-9
Comparison of B:A and T:B PC Values for Mature Individuals
Method
1. Rat Only (this study)
Rat (Corley et al.,
1990)
2. Human B : A, Rat T: A
(this study)
3. Predicted
4. Adjusted
Human only (Stewart
etal., 1973)
Blood:
Air
17.7
20.8
11.34
3.94
(to) 11.34
7.43
Liver:
Blood
1.0
1.01
1.6
3.14
3.14
2.28
Kidney:
Blood
0.8
0.53
1.4
2.53
2.53
1.48
Muscle:
Blood
1.0
0.67
1.2
1.59
1.59
1.61
Fat:
Blood
19.9
9.75
33.4
27.95
27.95
37.68
Methods 3 and 4 yield the same values for T:B PC value because Method 4 adjusts each solid
tissue T: A PC value to account for differences between the observed and predicted value for the
B:A PC. The difference between values derived in these methods is the B:A PC value. In
addition, Gargas et al. (1989) demonstrated B:A PC values of 6.85 and 20.8 in blood from
humans and rats, respectively, for chloroform.
5-96
-------
presents a comparison of available PC values for chloroform, including (1) empirically derived
T:A PC values in rat tissues, (2) T:A PC values from rat tissues combined with B:A values from
humans, (3) values determined by estimating partitioning into lipid and water in combination
with information on the water and lipid composition of tissues, and (4) a hybrid approach using
observed values for B:A PC values to adjust those predicted from oil and water partitioning.
ESTIMATION OF PC VALUES IN CHILDREN
For this investigation, B:A PC values for chloroform were collected in remnant samples
of children's blood. Because age-specific values for tissue lipid and water content were
available (ICRP, 1975), predictions of T:A PC values for solid tissues were developed using
adult- and child-relevant lipid and water contents (Table B-10). When data described a range of
fat or water content, the median of the range was employed. ICRP (1975) presented no data for
the water content of child kidney, so the water content of adult kidney was included. The ratio of
these values was used to adjust the T: A PC values for adult human tissues reported by Steward et
al. (1973) to values reflective of child T:A PC values (Table B-l 1). These T:A PC values were
combined with child-specific B:A PC values to develop T:B PC values for inclusion in the model
(Tables B-12 and B-13).
Pelekis et al. (2001) also used the tissue contents of water, neutral lipids, and
phospholipids to estimate T:A PC values and to derive T:B PC values for the hypothetical 10 kg
child, the median body weight between 1 and 2 years of age. Water and lipid content was
obtained from literature summarized in Pelekis et al. (1995).
DERIVATION OF PC VALUES IN PND RAT PUPS
Individual blood and tissue samples were obtained from 10 male and 10 female PND 10
rat pups. B:A and T:A PC values were derived for each tissue sample individually, and T:A and
5-97
-------
TABLE B- 10
Fractional Fat and Water Content of Children's Tissues (ICRP, 1975)
Liver
Water
Fat
0.799
0.036
Kidney
0.76a
0.0273
Muscle
0.785b
0.02
Fat
0.475
0.555
a Value from adult kidney.
b Value at 4 to 7 months of age.
TABLES- 11
Predicted T: A PC Values in Humans
Liver: Air
Adult
Child
AdultChild
Ratio
Observed
Adult value
Adjusted
Child Value
12.38
8.23
0.66
17.0
11.30
Kidney:Air
9.98
6.87
0.66
11.0
7.31
Muscle:Air
6.27
5.98
0.66
12.0
7.98
Fat:Air
110.15
77.95
0.66
280
186
Child-Specific Blood: Air and Tissue:Blood PC Values
Blood:Air
12.41
LiverBlood
0.91
Kidney :Blood
0.59
Muscle:Blood
0.64
FatBlood
15.0
TABLE B-12
Child-Specific Blood: Air and Tissue:Blood PC Values
Blood:Air*
LiverBlood
Muscle:Blood
FatBlood
14.9
5.79
4.79
86.4
*B:A PC value was predicted by the PBPK model employed.
5-98
-------
TABLE B-13
Adult-Specific Blood:Air and Tissue:Blood PC Valuesa
Blood:Airb
7.43
Liver:Blood
3.0(H)
2.8 (L)
Muscle:Blood
3.69(H)
2.56 (L)
Fat:Blood
77.0 (H)
23. 6 (L)
aT:B PC Values presented as high (H) and low (L) values predicted from extreme values for
water and lipid content.
bB:A PC value was predicted by the PBPK model employed.
5-99
-------
B:A values from each pup were paired to generate individual-specific values for T:B PCs. Table
B-14 presents mean and ranges for chloroform B:A and T:B PC values for PND 10 rats.
TABLE B-14
Blood: Air and Tissue:Blood PC Values in PND 10 Rat Pups
Blood:Air
Mean
SD
Min
Max
13.3
0.84
12.2
15.0
LiverBlood
1.31
0.13
1.03
1.53
Kidney :Blood
0.95
0.09
0.81
1.05
Muscle:Blood
2.54
1.61
1.24
7.95
FatBlood
18.27
3.44
10.96
23.7
(USAF, 2005)
DERIVATION OF BLOODrAIR PC VALUES IN CHILDREN
Remnant blood samples obtained from pediatric patients were used to determine B:A PC
value for children. The B:A PC values derived using samples from 7 males and 4 females (aged
3 to 7 years) demonstrated a mean and standard deviation of 12.41 + 1.17.
REFERENCES
Batterman, S., L. Zhang, S. Wang and A. Franzblau. 2002. Partition coefficients for the
trihalomethanes among blood, urine, water, milk and air. Sci. Total Environ. 284:237247.
Beliveau M., G. Charest-Tardif and K. Krishnan. 2001. Blood:air partition coefficients of
individual and mixtures of trihalomethanes. Chemosphere. 44:377-381.
Corley, R.A., A.L. Mendrala, F. A. Smith et al. 1990. Development of a physiologically based
pharmacokinetic model for chloroform. Toxicol. Appl. Pharmacol. 103:512-527.
Gargas, M.L., RJ. Burgess, D.E. Voisard, G.H. Cason and M.E. Andersen. 1989. Partition
coefficients of low-molecular weight volatile chemicals in various liquids and tissues. Toxicol.
Appl. Pharmacol. 98:87-99.
ICRP (International Commission on Radiological Protection). 1975. Task Force on Reference
Man, Vol. 25.
Pelekis, M., P. Poulin and K. Krishnan. 1995. An approach for incorporating tissue composition
data into physiologically based pharmacokinetic models. Toxicol. Ind. Health. 11:511-522.
5-100
-------
Pelekis M., L.A. Gephart and S.E. Lerman. 2001. Physiological-model-based derivation of the
adult and child pharmacokinetic intraspecies uncertainty factors for volatile organic compounds.
Reg. Toxicol. Pharmacol. 33:12-20.
Sato A. and T. Nakajima. 1979. A vial-equilibration method to evaluate the drug-metabolizing
enzyme activity for volatile hydrocarbons. Toxicol. Appl. Pharmacol. 47:41-46.
Steward, A., P.R. Allot, A.L. Cowles and W.W. Mapleson. 1973. Solubility coefficients for
inhaled anesthetics for water, oil, and biological media. Brit. J. Anaesth. 45:282-293.
USAF (United States Air Force). 2005. Determination of Partition Coefficients for a Mixture of
Volatile Organic Compounds in Rats and Humans at Different Life Stages. Air Force Research
Laboratory Human Effectiveness Directorate, Applied Biotechnology Branch, Wright-Patterson
AFB, OH. AFRL-HE-WP-TR-2005-0012.
U.S. EPA. 2003. Estimation Program Interface (EPI) Suite. Available at:
http://www.epa.gov/oppt/exposure/docs/episuite.htm.
NOTE: Data in Tables B-3, B-4, B-5, B-6 and B-14 were developed through IAG No.
DW-57-93960601-0 with the U.S. Air Force, Air Force Research Laboratory/HEST, Wright-
Patterson AFB Ohio (USAF, 2005). Results in this report may vary from those in the Air Force
Technical report in that the latter were derived from a more full sample set.
5-101
-------
APPENDIX C
VARIABILITY OF HEPATIC BLOOD FLOW IN ADULT HUMANS
BACKGROUND
Many xenobiotics have characteristics that limit their in vivo rates of metabolism by the
rate at which they are delivered to the liver. Pharmacokinetic modelers have investigated this
phenomenon, termed it "flow-limited metabolism," and have demonstrated that several
xenobiotics important both industrially and as ubiquitous environmental contaminants fall into
this category. Flow-limited metabolism can be easily conceptualized as the condition in which
the metabolic capacity of the liver exceeds the capacity for delivery to the liver via hepatic blood
flow (HBF). This may be based on factors including (1) low concentration of the xenobiotic in
the environmental medium, (2) poor solubility of the xenobiotic in blood, (3) poor ability of the
xenobiotic to leave the blood and enter the liver, or (4) a very high capacity of the liver to
metabolize the xenobiotic. Regardless, for xenobiotics whose metabolism is flow-limited, blood
flow to the liver is rate-limiting, and warrants further consideration, especially when examining
PK differences among humans.
As a biologically active organ responsible for the bulk of xenobiotic metabolism, the liver
requires adequate delivery of oxygen and nutrients. The liver is unique among organs, in that it
receives both arterial and venous blood in-flow. The hepatic artery (HA) branches and perfuses
the liver with oxygen-rich blood. The hepatic portal vein (PV) courses along the intestine and
stomach, thence supplies the liver with blood rich in absorbed nutrients. Recognized variance in
the branching of the hepatic artery occurs among humans, and abnormalities of this vasculature
complicate, and may preclude, organ transplantation. The liver is drained by out-flow through
the hepatic vein (HV). The conservation of mass indicates that measures of HBF may either
5-102
-------
measure the rate of delivery through the hepatic artery and portal vein or the rate of return to the
systemic circulation (hepatic vein).
The goal of this investigation was to quantify human interindividual differences in HBF,
quantified as fraction of cardiac output (CO). The available scientific literature was searched
and queries to the authors of several key original investigations revealed a marked interest in
cooperation and the availability of several important data sets containing individual-specific
measures of CO and HBF. Because of the dependence of HBF on CO, PBPK models have been
structured to incorporate HBF as a fraction of CO. To reduce uncertainties in this measurement,
only studies reporting both CO and HBF in a pair-wise fashion were considered for inclusion.
Multiple techniques have been applied to measure HBF, but few published studies report
both HBF and CO. HBF has been measured by xenon computer-aided tomography, by echo and
Doppler ultrasound, by pulsed-dye elimination, and by indocyanin green (ICG) clearance.
Noninvasive (e.g., ultrasound) measurements involve significant input by the operator in
carefully aiming the device, including assumptions about HA bifurcations. Specific
investigations have identified inter-observed differences in U.S. investigations as a significant
source of variability in measurements of HBF. The same may apply to CAT scans. Therefore,
we decided to focus only on measurements taken from dyes, including ICG. An additional
benefit with this decision is that dye elimination measures both HA and PV deliveries.
Routinely, dye content is simultaneously measured from a peripheral site and from a catheter
placed into the HV. Difference in dye concentration is adjusted by the hepatic extraction
efficiency to yield a measure of HBF in units L/min. Because of the uncertainties in measuring
and reporting of HBF via Doppler measurements, these were not included. This investigation
relied only on HBF measured through the clearance of the dye, ICG. The application of HBF
5-103
-------
measurements determined via ICG clearance allowed the measurement of total HBF, both
arterial and venous. The underlying data from six reported investigations were available.
DATA AND EVALUATION
The oft-cited publication by Brown et al. (1997) indicates a reliance on the review article
by Williams and Leggett (1989) as the basis for derivation of a representative value for the
fraction of CO directed to the liver. In order to be as technically accurate as possible, the eight
sources of original data in which ICG was used as a measure of HBF, cited in Williams and
Leggett (1989) were evaluated. Because they did not present individual data and there is a low
probability of gaining such information (based on publication date), four of these studies were
not used in the present estimation of variability of HBF. In addition, two more recent
publications were located, and data from those studies were provided by the study authors.
Whereas the older data sets often did not contain measures of CO or CI, Williams and Leggett
apparently adjusted CI by fixed values or proportions of fixed values scaled to BSA, rather than
by age-adjusted measures of CI. It should be noted that age plays a critical role in determining
cardiac index. Because several of these data sets did not contain measures of CO or CI, CI was
estimated with information on subject age, by interpolating the age-adjusted CI (in units of
L/min/m2 body surface) from the graph presented in Guyton (1986, Figure 23-2), and
multiplying the interpolated CI by the body surface area presented in the original studies.
Guyton's figure demonstrates a greater than proportionate decline in CI from a value of
approximately 3.6 L/min/m2 at age of 35 to a value approximating 2.3 L/min/m2 at an age of 90.
From this method, it was demonstrated that the mean value for the fraction of CO as HBF
from nine samples from Caesar et al. (1961) (Table C-l, "Controls" [n = 4] and "Miscellaneous"
5-104
-------
TABLE C-l
Hepatic Blood Flow (HBF) and Fraction of Cardiac Output as HBF (CO as HBF),
Estimated from Data of Caesar et al. (1961)
Patient
Control 3
Control 4
Control 6
Control 8
Misc. 33
Misc. 34
Misc. 35
Misc. 36
Misc. 37
Sex
M
F
M
M
M
F
M
M
F
Age
(years)
50
45
52
56
60
48
55
51
31
BSA (m2)
1.63
1.56
1.80
2.00
1.68
1.50
1.76
1.72
1.28
EHBF
(mL/min)
1210
1470
1970
1190
1500
1390
1420
2000
950
CO*
(mL/min)
2863
2991
2789
2716
2679
2899
2753
2881
3487
Mean
S.D.
C.V.
CO as HBF
0.259
0.315
0.392
0.219
0.333
0.320
0.293
0.404
0.213
0.305
0.068
0.223
*Estimate of age-specific CI read from Guyton (1986, Figure 23-2), multiplied by BSA. Body
weight was not reported.
5-105
-------
cases [n = 5]) were not significantly different, and so they were pooled to give a sample of n = 9,
with a value of 0.305 + 0.067 (mean + S.D.).
Individual values for the fraction of CO directed to the liver (HBF/CO) were determined
from 35 paired of observations of CO and HBF based on measurements of ICG clearance. A
total of 35 sets of paired (individual) observations were available in the open literature (Caesar et
al., 1961; Wiegand et al., 1960; Feruglio et al., 1964; Reemtsma et al., 1960) (Tables C-2
through C-4) and an additional 24 paired data sets were graciously provided by the authors of
more recent publications in which mean values were available (lijima et al., 2001; Sakka et al.,
2001) (Tables C-5 and C-6).
TABLE C-2
Hepatic Blood Flow (HBF) and Fraction of Cardiac Output as HBF (CO as HBF), Estimated
From Data of Wiegand et al. (1960)
Patient
15
16
17
18
19
20
21
22
23
Sex
M
M
F
M
M
F
F
M
F
Age (years)
41
17
16
17
39
36
29
14
57
BSA
(m2)
1.84
1.96
1.67
1.65
2.10
1.56
1.92
1.74
1.54
EHBFa
(ml/min)
881
1741
1338
1508
1655
827
778
1732
921
COb (ml/min)
5639
7733
6510
6628
7247
6742
5067
6694
5143
Mean
S.D.
C.V.
CO as HBF
0.156
0.225
0.206
0.227
0.228
0.123
0.154
0.259
0.179
0.195
0.045
0.231
a Estimated hepatic blood flow, converted from reported units of mL/min/m2 to mL/min by
multiplying by the stated BSA.
b Estimate of age-specific CI interpolated from Guyton (1986, Figure 23-2), multiplied by BSA
to obtain CO in units of mL/min. Body weight was not reported.
5-106
-------
TABLE C-3
Hepatic Blood Flow (HBF) and Fraction of Cardiac Output as FffiF (CO as HBF), as Reported
by Feruglio et al. (1964), "Before" Condition
Case
1
2
3
4
5
6
7
8
9
10
11
Sex
F
M
M
F
M
M
M
M
F
M
M
Age (years)
42
48
41
32
18
39
29
36
38
43
60
HBF (mL/min)
855
1260
1390
1480
910
1005
1180
1265
1200
1460
1520
CO (mL/min)
4380
6970
4020
4200
6260
3160
6400
5800
5100
4600
6040
Mean
S.D.
C.V.
CO as HBF
0.181
0.346
0.145
0.318
0.184
0.218
0.317
0.251
0.195
0.352
0.235
0.249
0.073
0.293
NOTE: BSA not reported.
5-107
-------
TABLE C-4
Hepatic Blood Flow (HBF) and Fraction of Cardiac Output as HBF (CO as HBF),
Derived from HBF and Patient Characteristics Presented by Reemtsma et al.
(1960, Table III, constant infusion data)
Control
Subject
js
wl
er
dh
jd
tk
Sex
F
M
M
M
M
M
Age
(years)
14
26
36
42
50
56
Weight
(kg)
40
72
89
61.3
70
65.9
BSAa
(m2)
1.28
1.82
2.03
1.67
1.79
1.74
Mean
S.D.
C.V.
HBF
(mL/min)
1100
2400
2200
1800
1700
1000
1700
566
0.332
CO
(mL/min)
5410
6299
6786
5233
5261
4895
5647
730
0.129
CO as
HBFb
0.230
0.381
0.324
0.344
0.323
0.204
0.297
0.074
0.252
aBody Surface area not reported; determined by the formula to convert weight (in kg) to BSA in
m2 (ICRP, 1975, p. 17). BSA = [((4*W) +7)/(W + 90)]. Costeff (1966 — in ICRP, p. 243)
evaluated the accuracy of this formula over the weight range 1.5-100 kg and demonstrated that
the root mean square departure from the formulas is 12% or less.
b CO not reported, estimated by estimating age-specific CI (from Guyton, 1986, Figure 23-2) and
multiplying by BSA. Note for later: %CO as HBF was lower, but NSD in cirrhotics, at 0.237 +
0.101.
5-108
-------
TABLE C-5
Hepatic Blood Flow (HBF)and Cardiac Output (CO) Data Which Served as the Basis for
Results Presented in lijima et al. (2001)
Case
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Sex
M
M
M
M
F
M
F
F
M
M
M
M
F
F
M
F
Age (years)
53
58
50
36
47
64
74
48
53
58
49
35
48
47
64
74
Weight (kg)
67.5
44
62
59
47
51
55
49
67.5
44
62
59
49
47
54
55
Mean
S.D.
C.V.
HBF (mL/min)
1148
1122
2773
1616
1601
1271
1170
868
2030
1547
952
1214
871
2736
1652
914
1468
601
0.409
CO (mL/min)
3070
3290
6700
4290
5020
5690
4510
5160
6010
6450
3740
3980
4350
6100
5510
3720
4849
1146
0.236
CO as HBF
0.374
0.341
0.414
0.377
0.319
0.223
0.259
0.168
0.338
0.240
0.255
0.305
0.200
0.449
0.300
0.246
0.301
0.079
0.263
NOTE: BSA not reported.
Individual data describing patient characteristics, CO and HBF were graciously provided by Dr.
Takehiko lijima.
5-109
-------
TABLE C-6
Hepatic Blood Flow (HBF) and Cardiac Output (CO) Data Which Served as the Basis for
Results Presented in Sakka et al. (2001)
Case
1
2
3
4
5
6
7
8
Sex
F
F
M
F
M
F
F
M
Age
(years)
70
55
46
39
51
74
74
77
Weight
(kg)
110
85
60
40
95
85
70
60
BSA
(m2)
2.17
1.96
1.69
1.36
2.15
1.92
1.80
1.58
Mean
S.D.
C.V.
HBF
(mL/min)
89
176
331
376
222
169
128
315
226
104
0.460
CO
(mL/min)
10770
8460
8150
9020
9890
7090
6420
6550
8294
1570
0.189
CO as
HBF
0.082
0.208
0.405
0.417
0.224
0.239
0.199
0.480
0.282
0.136
0.483
NOTE: Individual data describing patient characteristics, CO and HBF were graciously
provided by Dr. Samir Sakka. HBF measurements were based on subject-specific hepatic
extraction ratios.
5-110
-------
lijima et al. (2001) graciously provided the original subject-specific data from their study,
and a statistical analysis (one-tailed ^-test, Excel) demonstrating that the mean values for CO,
HBF, and %CO as HBF were not different between males and females in this study. Therefore
all samples were pooled for further analysis (Table C-5). Consistent with other data sets, these
data also demonstrated that values for CO, HBF, and %CO as HBF were slightly lower in
females than in males.
SUMMARY
All of the above studies were conducted on overnight-fasted subjects, who were
evaluated in the morning in the supine position. This similarity among studies reduced the
impact which feeding state, posture, and exercise status may have on the direction of blood flow
to the liver. Commercially available software (Statistical Analysis System, SAS) was applied to
the above 59 individual values for HBF/CO to determine the parameters for the normal,
lognormal, and beta distributions according to the method of moments. Results are presented in
Table C-7.
TABLE C-7
Parameters and Selected Percentile Values for the Fraction of Cardiac Output (CO) as
Hepatic Blood Flow
Distribution
Normal
Log
Normal
Beta
Parameter and Value
Mean = 0.273,
Standard Deviation =
0.087
Geometric Mean = 0.258,
Geometric Standard
Deviation = 1.411
a = 6.865, p= 18.269
Percentile
5th
0.130
0.147
0.141
50th
0.273
0.259
0.267
95th
0.417
0.456
0.427
5-111
-------
REFERENCES
Brown, R.P., M.D. Delp, S.L. Lindstedt, L.R. Rhomberg and R.P. Bellies. 1997. Physiological
parameter values for physiologically based pharmacokinetic models. Toxicol. Ind. Health.
13:407-484.
Caesar, J., S. Shaldon, L. Chiandussi, L. Guevara and S. Sherlock. 1961. The use of indocyanin
green in the measurement of hepatic blood flow and as a test of hepatic function. Clin. Sci.
21:43-57.
Costeff, H. 1966. A simple empirical formula for calculating approximate surface area in
children. Arch. Dis. Child. 41(220):681-683.
Feruglio, F.S., F. Greco, L. Cesano, D. Indovina, G. Sardi and L. Chiandussi. 1964. Effect of
drug infusion on the systemic and Splanchnic circulation: I. Bradykinin infusion in normal
subjects. Clin. Sci. 26:487-491.
Guyton, A.C. 1986. Textbook of Medical Physiology. W.B. Saunders, Philadelphia, PA.
lijima T., F. Ohishi, T. Tatara and Y. Iwao. 2001. Effect of continuous infusion of
prostaglandin El on hepatic blood flow. J. Clin. Anesth. 13:250-254.
Reemtsma, K., G.C. Hottinger, A. DeGraff, Jr. and O. Creech. 1960. The estimation of hepatic
blood flow using indicyanin green. Surg. Gynecol. Obstet. 110:353-356.
Sakka, S.G., K. Reinhart, K. Wegscheider and A. Meier-Hellmann. 2001. Variability of
splanchnic blood flow in patients with sepsis. Intensive Care Med. 27:1281-1287.
Weigand, B.D., S.G. Ketterer and E. Rapaport. 1960. The use of indocyanin green for the
evaluation of hepatic function and blood flow in man. Am. J. Dig. Dis. 5:427-436.
Williams, L.R. and R.W. Leggett. 1989. Reference values for resting blood flow to organs of
man. Clin. Phys. Physiol. Meas. 10:187-217.
5-112
-------
APPENDIX D
EXTRAPOLATION OF IN VITRO DERIVED CHLOROFORM METABOLIC RATE
CONSTANTS AND VARIABILITY FOR PBPK MODELING
This section demonstrates the extrapolation of the rate constant, Vmax, from its units of in
vitro measurement (i.e., pmol product/minute/pmol CYP2E1) to units (e.g., mg/h/liver) which
can be scaled according to body weight and incorporated into PBPK model code, using adult-
specific information and based on a fractional liver mass of 2.6% of body weight and a 70 kg
adult (with a resulting liver mass of 1820 grams). This approach requires several conditions,
including that the responsible enzyme has been identified, its content in the metabolic
preparation (here, microsomal protein) is known, and that the concentration of the enzyme in
intact liver is known (Lipscomb and Kedderis, 2002). The last point may also be addressed by
measuring the content of the enzyme in liver homogenate and correcting for volume (not shown).
The first condition necessary for extrapolation is that the content of the responsible
enzyme be known in the preparation used in metabolism studies and in the intact liver. This was
accomplished via the Enzyme-Linked ImmunoSorbent Assay (ELISA; Snawder and Lipscomb,
2000) using antibody against CYP2E1, the enzyme responsible for chloroform metabolism.
Samples of human hepatic MSP were screened for CYP2E1 content and three samples
expressing roughly 80 to 100 pmoles CYP2El/mg MSP were selected for use in chloroform
metabolism studies (Lipscomb et al., 2004). From that study, the specific activity of human
CYP2E1 was determined to be 5.24 pmoles chloroform oxidized/minute/pmole CYP2E1. The
specific activity of CYP2E1 expressed in samples of Fischer 344 rat liver microsomal protein
was demonstrated to be quite similar, at 5.29 pmoles chloroform oxidized/minute/pmole
CYP2E1. Because there are no data demonstrating an influence of age on metabolic activity,
5-113
-------
when expressed as product per unit enzyme, the specific activity determined in samples of adult
MSP was also employed for the child.
EXTRAPOLATION OF VMAX IN ADULTS
The concentration of CYP2E1 was determined in homogenate prepared from samples of
liver from 20 different adult organ donors (Lipscomb et al., 2003b) and was volumetrically
extrapolated to the intact liver (pmoles CYP2El/gram liver). CYP2E1 content was also
determined in MSP (pmoles CYP2El/mg MSP) derived from those homogenates. A
mathematical combination of these data sets was accomplished to determine the concentration of
MSP in intact liver for these 20 adult organ donors (Equation D-l).
(pmol CYP2El/gram tissue)/(pmol CYP2El/mg MSP) = (mg MSP/gram tissue) (D-l)
Because an additional 40 samples of adult human MSP had also been analyzed for
CYP2E1 content (Snawder and Lipscomb, 2000), their values were combined with the values
from the 20 samples to yield a samples set of n = 60. Statistical analysis indicated a slight
negative correlation between the CYP2E1 content of MSP and the MSP content of intact liver, in
the set of 20 paired analyses (Lipscomb et al., 2003b), and so a more rigorous statistical analysis
was performed to estimate the distribution type, parameters, and values at given percentiles of
the distribution for the concentration of CYP2E1 in adult human liver (Lipscomb et al., 2003a).
To extrapolate the in vitro derived value for the apparent Vmax for the reaction, the
specific activity (in units of pmoles chloroform oxidized/minute/pmole CYP2E1) was
incorporated with the content of CYP2E1 in intact liver (in units of pmoles CYP2El/gram liver
tissue) via Equation D-2. The resulting extrapolated rates were then corrected for units of time
and molecular weight to yield rate values in units of mg/h/liver. Because each body contains
5-114
-------
only one liver, the resulting extrapolated maximal metabolic rate (Vmax, in units of mg/h) was
scaled by body weight (in units of kg) to the 0.7 power.
(product formed/min/pmol CYP) x (pmol CYP/gram liver) x (grams/liver)
= product formed/min/liver (D-2)
To determine these values at multiple points in the distribution of CYP2E1 content, the
amount of CYP2E1 in intact liver was determined at the geometric mean, and at the 5th and 95th
percentiles of the (lognormal) distribution, by multiplying the concentration (pmoles
CYP2El/gram liver) by the mass of the liver (1820 grams). The geometric mean value for
CYP2E1 content is 2562 pmoles CYP2El/gram liver, or 2562 nmoles CYP2El/kg liver.
If liver = 2.6% BW, if BW = 70 kg, then liver weight (mass) = 1.82 kg.
1.82 Kg x 2562 nmoles CYP2El/kg liver = 4662.84 nmoles CYP2El/liver.
These values were then combined with the specific activity toward chloroform, as follows.
The specific activity (apparent Vmax) for CYP2E1 toward chloroform is 5.24 pmoles
CHC13/min/pmol CYP2E1, and information in Table D-l is used to extrapolate this measure to
units representative of the intact liver.
5.24 nmoles CHC13 x 4662.84 nmoles CYP2El/liver = 24433 nmoles CHC13/min/liver.
24433 nmoles CHC13/min/liver x 60 mins/h = 1465997 nmoles CHC13/h/liver.
1465997 nmoles x 119.4 ng/nmole CHC13 = 175,040,029 ng CHC13/h/liver
The extrapolated Vmax value is 175 mg/h/liver.
TABLE D-l
Information Used to Extrapolate In Vitro Derived Vmax Value to the Intact Liver
Specific Activity
CYP2E1 Concentration
Body Mass
Liver Mass
CYP2E1 Amount
Units of Time
Formula Weight
5.24 pmoles CHCl3/min/pmole CYP2E1
2562 pmoles CYP2El/gram liver
70kg
0.026 x 70 kg BW = 1820 grams liver
4662840 pmoles CYP2El/liver
60 minutes/h
119.4gCHCl3/mole
5-115
-------
Equation 5-2 demonstrates the Vmax scaling procedure, necessary to incorporate
metabolic rate constants into PBPK models. In this procedure, 175 mg/h is divided by BW °7
(or, 70 °'7 = 19.57 kg) to yield a VmaxC value of 8.9 mg/h/kg. In comparison, other PBPK models
developed for chloroform have used 7 (Delic et al., 2000) or 15.7 mg/h/kg (Corley et al.,
1990), and Allen et al. (1996) indicated that a reasonable range for the mean VmaxC was 11 to 35
mg/h, scaled to a 1-kg animal.
EXTRAPOLATION OF VMAX FOR CHILDREN
The extrapolation of metabolic rate in children followed the procedure as for the adult.
Data in Tables D-2 and D-3 describe the sample of available children's tissues and the enzyme
content of those tissues. Table D-4 presents the results of a statistical analysis to determine the
parameters of the distribution of CYP2E1 to intact liver tissue from these donors.
TABLE D-2
Donor Demographic Information for 10 Child Organ Donors
Donor
1
2
3
4
5
6
7
8
9
10
Age (yrs)
7
13
<10
6
2
12
9
1 1 months
8
17
Sex
M
F
Not stated
F
F
F
M
M
M
M
Race
Not stated
Caucasian
Not stated
Caucasian
Caucasian
Black
Caucasian
Hispanic
Black
Not Stated
Source
HCCC
Vitron
Vitron
Vitron
Vitron
Vitron
Vitron
Vitron
Vitron
HCCC
Case No.
430
646
559
549
550
565
538
635
555
356
HCCC = Human Cell Culture Center Laurel, MD.
Vitron = Vitron, Inc., Tucson, AZ.
5-116
-------
TABLE D-3
Determination of Cytochrome P450 Enzyme in Liver Tissue of 10 Child Organ Donors
Donor
1
2
3
4
5
6
7
8
9
10
Mean
SD
Parameter
Mg Homog.
Protein/ Gram
Liver
122
234
289
190
302
249
172
193
205
408
236
81
Pmoles CYP2E1/
mg Homog.
Protein
22
16
32
13
23
23
22
19
19
16
Pmoles
CYP2E1/ gram
liver
2691
3745
9250
2431
6936
5747
3787
3663
3891
6529
4470 (GM)
1.5393(GSD)
Pmoles
CYP2El/mg
MSP
58
33
53
35
31
49
40
62
36
31
42.8
11.7
Mg
MSP/gram
Liver
46.4
113.5
174.5
69.5
223.7
117.3
94.7
59.1
108.1
210.6
55 ± 15% of total homogenate protein is MSP in children.
GM = Geometric Mean, GSD = Geometric Standard Deviation.
5-117
-------
TABLE D-4
Distribution of CYP2E1 to Adults and Children
Parameter
Adult
Childa
Enzyme Content — pmoles CYP2El/gram Liver
5th percentile
Geometric mean
95th percentile
Specific Activity — pmoles
CHC13/min/pmol CYP2E1
1232b
2562b
4453b
5.24
1288
2818
6167
5.24C
Metabolic Capacity — jig CHC13/h/gram Liver
5th percentile
Geometric mean
95th percentile
46.3
96.3
167.3
48.35
105.8
231.5
a Values apply to the 9-year-old and the 1-year-old child, but are further adjusted by a liver mass
of 804 grams and 342.5 grams in the 9- and 1-year-old child, respectively.
bFrom Lipscomb et al. (2003a).
c Measured in adult samples (Lipscomb et al., 2004).
5-118
-------
Metabolic capacity was converted from units of |ig/h/gram liver, to mg/h, based on a
liver mass of 342.5 grams in the 1-year-old child (10 kg body mass x 0.03425 body mass as
liver) and a liver mass of 804 grams in the 9-year-old child (30 kg body mass x 0.0268 body
mass as liver). Resulting Vmax values at the geometric mean were 36.24 mg/h and 85.06 mg/h
for the 1-year-old and 9-year-old child, respectively. When scaled by BW°'7, the resulting VmaxC
values were 7.23 and 7.87 mg/h/kg for the 1-year-old and 9-year-old child, respectively.
EXTRAPOLATION OF KM VALUES FOR ADULT HUMANS
The PBPK model developed for chloroform employs a Km value of 0.012 mg/L. In the
PBPK model, Km is expressed in venous blood at equilibrium with liver tissue (CVL), and the
derivation and extrapolation of Km value is consistent with the concept that Km is best described
in the aqueous compartment of the liver. The Km value for chloroform oxidation in humans was
derived from in vitro measures as demonstrated below.
Lipscomb et al (2004) reported values for three separate human microsome samples in
which Km values were reported as concentrations in the aqueous suspension, rather than
concentration in headspace. This was accomplished in the original study by determining the Km
value in headspace and correcting by the gas:misrosomal suspension partition coefficient
measured for that study. The Km values were 0.256, 0.113 and 0.090 jiM, and averaged 0.153
|iM in suspension. That value was corrected for the formula weight of chloroform (119.4
|ig/|imol), yielding an average concentration of 18.27 |ig/L. That value represents the
concentration of chloroform in solution driving reaction rates at half of their theoretical maximal
initial rates. To extrapolate this value into units useful in PBPK modeling, it was necessary to
express concentration in venous blood at equilibrium with liver, rather than in liver itself. The
concentration in the in vitro suspension was assumed to adequately represent the concentration in
5-119
-------
liver resulting in half-maximal rates of metabolism, and was adjusted to reflect the concentration
of chloroform in blood at equilibrium with liver. This was accomplished by dividing 18.27 |ig/L
by the liverblood partition coefficient value used in the model (PL=1.6). This resulted in a value
of 11.4 |ig/L, which was incorporated in the model as 0.012 mg/L.
REFERENCES
Allen, B.C., T.R. Covington and HJ. Clewell. 1996. Investigation of the impact of
pharmacokinetic variability and uncertainty on risks predicted with a pharmacokinetic model for
chloroform. Toxicology. 111:289-303.
Corley, R.A., A.L. Mendrala, F. A. Smith et al. 1990. Development of a physiologically based
pharmacokinetic model for chloroform. Toxicol. Appl. Pharmacol. 103:512-527.
Delic, J.I., P.O. Lilly, AJ. MacDonald and G.D. Loizou. 2000. The utility of PBPK in the
safety assessment of chloroform and carbon tetrachloride. Reg. Toxicol. Pharmacol.
32:144-155.
Lipscomb, J.C. and G.L. Kedderis. 2002. Incorporating human interindividual
biotransformation variance in health risk assessment. Sci. Total. Environ. 288:13-21.
Lipscomb, J.C., L.K. Teuschler, J. Swartout, D. Popken, T. Cox and G.L. Kedderis. 2003a. The
impact of cytochrome P450 2El-dependent metabolic variance on a risk relevant
pharmacokinetic outcome in humans. Risk Anal. 23:1221-1238.
Lipscomb, J.C., L.K. Teuschler, J.C. Swartout, C.A.F. Striley and I.E. Snawder. 2003b.
Variance of microsomal protein and cytochrome P450 2E1 and 3 A forms in adult human liver.
Toxicol. Mech. Methods. 13:45-51.
Lipscomb, J.C., H. Barton, R. Tornero-Velez et al. 2004. The metabolic rate constants and
specific activity of human and rat hepatic cytochrome P450 2E1 toward chloroform. J. Toxicol.
Environ. Health. 67:537-553.
Snawder, I.E. and J.C. Lipscomb. 2000. Interindividual variance of cytochrome P450 forms in
human hepatic microsomes: Correlation of individual forms with xenobiotic metabolism and
implications for risk assessment. Reg. Toxicol. Pharmacol. 32:200-209.
5-120
------- |