UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON D.C. 20460
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
May 30, 2006
EPA-SAB-06-006
The Honorable Stephen L. Johnson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, N.W.
Washington, D.C. 20460
Subject: SAB Review of EPA's Draft Risk Assessment of Potential Human
Health Effects Associated with PFOA and Its Salts
Dear Administrator Johnson:
In response to a request from EPA's Office of Pollution Prevention and Toxics (OPPT),
the Science Advisory Board (SAB) convened an expert panel to conduct a peer review of EPA's
Draft Risk Assessment of Potential Human Health Effects Associated with Perfluorooctonoic
Acid (PFOA) and Its Salts (dated January 4, 2005). PFOA is a synthetic (man-made) chemical
used in the manufacture of several commercially important products. PFOA has been detected in
the blood of the general U.S. population although it is not fully understood how individuals are
exposed to the chemical. To determine whether environmental exposure to PFOA might pose a
risk to human health, EPA's draft assessment provided an evaluation of available information on
the health effects and human exposure to PFOA. The draft assessment also compared measured
human blood levels with the estimated PFOA blood levels that are not anticipated to produce (or
can produce minimal) toxicities based on data in tested laboratory animals.
The SAB was asked to comment on: (a) EPA's analysis of how PFOA causes tumors in
rats and its relevance for human health and the weight-of- evidence conclusion about the
potential for PFOA to cause cancer in humans; (b) the selection of health effects endpoints for
risk assessment; (c) the adequacy of available data to provide information on exposure of the
general population to PFOA; and (d) EPA's risk assessment approach including the use of
kinetic models to estimate PFOA blood levels in available laboratory animals studies.
In general, the SAB Panel endorsed EPA's risk assessment approach, particularly, the
inclusion of multiple non-cancer health endpoints for risk assessment, and the use of PFOA
blood levels as a measure of estimated dose in place of the administered dose in toxicologic
studies. The Panel recommended the inclusion of additional non-cancer health endpoints for risk
assessment, and the use of the Benchmark Dose method to better estimate the lowest observed
-------
effect levels and no observed effect levels for risk assessment. Three-quarters of the Panel
judged that the weight-of-evidence conclusion for the potential of PFOA to cause cancer in
humans was more aligned and consistent with the hazard descriptor of "likely to be
carcinogenic" as described in the Agency's cancer guidelines (i.e., 2003 EPA Guidelines for
Carcinogen Risk Assessment). They also recommended that a risk assessment be conducted for
carcinogenic effects. About one-quarter of the Panel agreed with EPA's conclusion regarding the
potential cancer hazard of PFOA to humans and the designation of the cancer descriptor of
"suggestive evidence of carcinogenicity, but not sufficient to assess human carcinogenic
potential". Three quarters of the Panel considered the available human biomonitoring studies
adequate to characterize environmental risk to PFOA for the general population. However,
about one-quarter of the Panel believed that the available studies are inadequate for risk
assessment of subpopulations possibly more highly exposed to PFOA. The scientific rationales
for these viewpoints along with specific recommendations on these issues are detailed in the
Panel's report.
The SAB strongly urges the Agency to strengthen its risk assessment by considering
verified and peer reviewed new information found to be relevant and critical to the assessment.
We look forward to receiving your response to this review and appreciate the opportunity to
provide EPA with advice on this important subject. We stand ready to assist the Agency in any
future efforts in updating the draft risk assessment.
Sincerely,
/signed/ /signed/
Dr. M. Granger Morgan Dr. Deborah Cory-Slechta
Chair Chair
EPA Science Advisory Board PFOA Risk Assessment Review Panel
EPA Science Advisory Board
-------
NOTICE
This report has been written as part of the activities of the EPA Science Advisory Board
(SAB), a public advisory group providing extramural scientific information and advice to the
Administrator and other officials of the Environmental Protection Agency. The SAB is
structured to provide balanced, expert assessment of scientific matters related to problems facing
the Agency. This report has not been reviewed for approval by the Agency and, hence, the
contents of this report do not necessarily represent the views and policies of the Environmental
Protection Agency, nor of other agencies in the Executive Branch of the Federal government, nor
does mention of trade names of commercial products constitute a recommendation for use.
Reports of the SAB are posted on the EPA website at http;//www\epa,,gQWgab,
-------
U.S. Environmental Protection Agency
Science Advisory Board
Perfluorooctanoic Acid Review Panel
CHAIR
Dr. Deborah Cory-Slechta, Director, Environmental and Occupational Health Sciences
Institute, Robert Wood Johnson Medical School, University of Medicine and Dentistry of New
Jersey and Rutgers State University, Piscataway, NJ
MEMBERS
Dr. Ernest Abel, Professor of Obstetrics and Gynecology; Professor of Psychology; Director,
Reproductive Toxicology, Obstetrics and Gynecology Department, School of Medicine / C.S.
Mott Center for Human Growth and Development, Wayne State University, Detroit, MI
Dr. Melvin Andersen, Director, Department of Biomathematics and Physical Science, Centers
for Health Research, CUT, Research Triangle Park, NC
Dr. George Corcoran, Chairman and Professor, Department of Pharmaceutical Sciences,
School of Pharmacy & Health Sciences, Eugene Applebaum College, Wayne State University,
Detroit, MI
Dr. Norman Drinkwater, Professor and Chair of Oncology, McArdle Laboratory for Cancer
Research, University of Wisconsin Medical School, Madison, WI
Dr. William L. Hayton, Professor and Associate Dean, Division of Environmental Health
Science, College of Pharmacy, School of Medicine and Public Health, Ohio State University,
Columbus, OH
Dr. Michael A. Kamrin, Professor Emeritus, Institute of Environmental Toxicology, Michigan
State University, Haslett, MI
Dr. James Kehrer, Head, Division of Pharmacology and Toxicology, College of Pharmacy,
University of Texas at Austin, Austin, TX
Dr. James E. Klaunig, Professor and Director, Department of Pharmacology and Toxicology,
School of Medicine, Indiana University, Indianapolis, IN
Dr. Matthew P. Longnecker, MD ScD, Division of Intramural Research National Institute of
Environmental Health Sciences, Epidemiology Branch, National Institute of Environmental
Health Sciences, Research Triangle Park, NC
Dr. Ronald Melnick, Director of Special Programs, Environmental Toxicology Program,
National Institute of Environmental Health Sciences, Research Triangle Park, NC
11
-------
Dr. Franklin L. Mink, President, MAI, Lake Orion, MI
Dr. David M. Ozonoff, Professor, Department of Environmental Health, School of Public
Health, Boston University, Boston, MA
Dr. Stephen Roberts, Professor, Center for Environmental and Human Toxicology, University
of Florida, Gainesville, FL
Dr. Anne Sweeney, Associate Professor, Department of Epidemiology and Biostatistics, School
of Rural Public Health, Texas A&M University, Bryan, TX
Dr. Thomas T. Zoeller, Professor, Biology Dept, Morrill Science Center, University of
Massachusetts at Amherst, Amherst, MA
SCIENCE ADVISORY BOARD STAFF
Dr. Sue Shallal, Designated Federal Official, Science Advisory Board Staff Office, Washington,
DC
in
-------
TABLE OF CONTENTS
EXECUTIVE SUMMARY 1
ISSUE 1: RODENT PPAR-ALPHA MODE OF ACTION FOR HEPATOCARCINOGENESIS 1
ISSUE 2: DESCRIPTOR FOR CARCINOGENIC POTENTIAL 3
ISSUES: SELECTION OF ENDPOINTS 4
ISSUE 4: RISK ASSESSMENT APPROACH 5
ISSUE 4A: PHARMACOKINETIC MODELING AND USE OF AUC AS A MEASURE OF INTERNAL DOSE 5
Issue 4b: Cross Species Extrapolation 6
Issue 4c: Human Biomonitoring Data 7
INTRODUCTION 8
BACKGROUND 8
CHARGE QUESTIONS 8
RESPONSES TO THE CHARGE QUESTIONS 13
ISSUE 1: RODENT PPAR-ALPHA MODE OF ACTION FOR HEP ATOCARCINOGENESIS AND LIVER TOXICITY 13
Question 1. Please comment on the weight of evidence and adequacy of the data available to identify the key
events for the PPAR alpha agonist induced rodent liver toxicity and hepatocarcinogenesis for PFOA. Discuss
whether the uncertainties and limitations of these data have been adequately characterized. 13
ISSUE 2: DESCRIPTOR FOR CARCINOGENIC POTENTIAL 15
Question 2. Please comment on the proposed descriptor for the carcinogenic potential of PFOA 15
ISSUES: SELECTION OF ENDPOINTS 20
Question 3. Please comment on the selection of these toxicity endpointsfor the risk assessment. 20
Question 4. Given the available data to date, please comment on the most appropriate
lifestage/gender/species for assessing human risk. 22
Question 5. Please comment on the appropriateness of the available animal models. Please comment on
whether additional animal models should be investigated, and if so, what information would better enable us
to ascertain potential human risks 23
ISSUE 4A: PHARMACOKINETIC MODELING AND USE OF AUC AS A MEASURE OF INTERNAL DOSE 24
Question 6. Please comment on use of the one-compartment pharmacokinetic model 24
Question 7. Please comment on the use of the A UC as a measure of internal dose for rats and humans for
calculation oftheMOE. 25
ISSUE 4B: CROSS SPECIES EXTRAPOLATION 27
Question 8. Please comment on the need to use or modify the default value of 10 for cross species
extrapolation given the pharmacokinetic analysis 27
ISSUE 4c: HUMAN BIOMONITORING DATA 28
Question 9. Please comment on the adequacy of the human exposure data for use in calculating a MOE. ....28
REFERENCES 31
IV
-------
EXECUTIVE SUMMARY
EPA's Office of Pollution Prevention and Toxics (OPPT) requested that the Science
Advisory Board review the "Draft Risk Assessment of the Potential Human Health Effects
Associated with Exposure to Perfluorooctanoic Acid (PFOA) and its Salts" (hereafter referred to
as the "draft PFOA risk assessment document") which was made available publicly in January
2006. The PFOA Review Panel of the EPA Science Advisory Board met in February 2005 at
which time nine charge questions raised by OPPT were deliberated. These questions focused on
four issues including, a peroxisome proliferator-activated receptor • »(PPAR-alpha) mode of
action (MOA) for rodent liver tumors, carcinogenicity descriptors, useful models for evaluation
of health effects, toxicokinetic considerations and reliance on currently available human
biomonitoring exposure data for calculation of margins of exposure (MOEs). Further
discussions of the entire Panel were held during a conference call in July 2005.
This Executive Summary highlights the outcome of the Panel's deliberations. It includes
the context for the charge questions and issues raised for consideration by EPA, and the
conclusions reached by the SAB Review Panel. The Panel reviewed and discussed the draft risk
assessment and the data referenced therein with the understanding that further risk assessment
will proceed as more data on PFOA health effects become available. In instances where the
views of Panel members diverged, the summary and charge question responses focus to a greater
extent on the view expressed by about three quarters of the Panel members since the view of
about one-quarter of Panel members coincided with that already expressed in EPA's Draft Risk
Assessment. During the review period, new information1 was presented to the Panel for their
consideration. The Panel encourages EPA to consider new information that has been verified and
peer-reviewed prior to use in their revision of the Draft Risk Assessment.
Issue 1: Rodent PPAR-alpha Mode of Action for Hepatocarcinogenesis
In rats, PFOA induces liver adenomas, Leydig cell tumors (LCT) and pancreatic acinar
cell tumors (PACT). The draft document concludes that these tumors constitute a triad and are
the result of a PPAR-alpha agonism MOA. In this MO A, activation of PPAR-alpha leads to cell
proliferation and decreased apoptosis, clonal expansion of preneoplastic foci and subsequent
tumors. The draft document premises its conclusions about this MOA on studies showing that
PFOA is a potent peroxisome proliferator in liver of rats and mice and, like other peroxisome
proliferators, induces hepatomegaly in rats. In addition, requisite dose-response and temporal
associations for some key events for this MOA have been reported.
Comment on the Weight of Evidence and Adequacy of the Data Available to Identify the
Key Events for the PPAR-alpha agonist-induced Rodent Liver Toxicity and
Hepatocarcinogenesis for PFOA.
1 This information included, for example: A) a report entitled, "Pathology Peer Review and Pathology Working
Group Review of Mammary Glands from a Chronic Feeding Study in Rats with PFOA Report" conducted by
Experimental Pathology Laboratories, Inc. and submitted to the SAB by Dr. Larry Zobel of 3M Medical Department
and B) data and documents submitted to the SAB by Mr. Robert Bilott of Taft, Stettinius & Hollister LLP.
-------
The Panel's charge was to determine whether it agreed with the weight of evidence
supporting a PPAR-alpha MOA for rodent liver toxicity and hepatocarcinogenesis. Panel
members agreed that, considered collectively, evidence to date was consistent with an
interpretation that liver tumor induction likely results from a PPAR-alpha MOA. This is based
on the observations that PFOA activates the receptor, results in peroxisome proliferation,
increases beta-oxidation and produces hepatomegaly, with dose and temporal responses
consistent with the PPAR-alpha MOA. These events, moreover, depend upon a functional
PPAR-alpha receptor, and no other known MOA, e.g., DNA reactivity or mutagenicity, has been
identified.
However, with respect to uncertainties and limitations related to concluding that PPAR-
alpha is the sole MOA for rodent liver tumor induction and toxicity, Panel views diverged.
About three quarters of the Panel members believed that at the current time, sufficient
uncertainties and limitations of the data still exist with respect to reaching such a conclusion,
given that: 1) In contrast to what would be predicted, administration of PFOA, but not the
prototype PPAR-alpha agonist WY-14,643, increased liver weights in PPAR-alpha receptor
knockout mice, i.e., in mice where PPAR-alpha activation was precluded, raising the possibility
that PFOA-induced liver tumors could occur by PPAR-alpha independent effects. The
significance of this finding currently remains uncertain in the absence of a corresponding
assessment of histopathology or replication by another laboratory. 2) There is as yet no
published evidence that the induction of PPAR-alpha by PFOA results in clonal expansion of
pre-neoplastic foci which is considered a critical step in the proposed MOA. 3) There are no data
demonstrating increased cell proliferation and/or decreased apoptosis in the liver of PFOA-
treated rats, key causative events in the proposed MOA.
These Panel members also viewed two additional issues as requiring further
consideration. One is the relevance of the PPAR-alpha MOA to humans. Given that human
exposures to PFOA and related chemicals appear ubiquitous, uncertainties and limitations of the
data for children have not been adequately characterized to be able to conclude that the PPAR-
alpha MOA is not operative in this young age group. A secondary issue thought to require
additional characterization in the PFOA response was the potential role of Kupffer cells, resident
macrophages in the liver that do not express PPAR-alpha, but are activated by peroxisome
proliferators.
A different view expressed by the remaining one quarter of the Panel members was that
the observation in PPAR-alpha knockout mice of increased liver weights in response to PFOA,
but not to the prototype PPAR-alpha agonist WY-14,643, was not sufficiently significant to
undermine the view that PPAR-alpha agonism is the sole MOA for PFOA-induced rodent liver
tumors.
In summary, Panel members agreed that collectively the weight of evidence supports the
hypothesis that liver tumor induction in rodents by PFOA is mediated by a PPAR-alpha agonism
MOA. Most Panel members, however, also felt, based on current evidence, that it is possible that
PPAR-alpha agonism may not be the sole MOA for PFOA, that not all steps in the pathway of
PPAR-alpha activation- induced liver tumors have been demonstrated, that other
-------
hepatoproliferative lesions require clarification, and that extrapolation of this MOA across the
age range in humans is not supported. A few panel members did not share these reservations
about a PPAR-alpha agonism MOA for PFOA-induced rodent liver tumors.
Issue 2: Descriptor for Carcinogenic Potential
The draft document reaches the conclusion of 'suggestive' evidence of carcinogenicity
but not sufficient to assess human carcinogenic potential of PFOA. This conclusion was based
upon: 1) a PPAR-alpha MOA for liver tumors in rodents that was considered not relevant to
humans because of their decreased sensitivity to PPAR-alpha agonism when compared to
rodents, 2) the absence of hepatic cell proliferation in a 6 month study of PFOA administration in
cynomologous monkeys, the species considered closest in physiology to humans; 3) the absence
of a strong association between PFOA exposure and tumors in human studies as interpreted in
the draft document; 4) the belief that the LCT and PACT tumors produced by PFOA in rats were
probably not relevant to humans based on the lower levels of expression of the mediators
leutinizing hormone (LCT) and choleocystokinin growth factor receptors (PACT) in humans, as
well as differences in quantitative toxicodynamics between rats and humans; and 5) the view that
mammary fibroadenomas reported in female rats are equivocal based on their comparable rates
of occurrence relative to a historical control group.
Comment on the Proposed Descriptor for the Carcinogenic Potential of PFOA
About three quarters of the Panel members concluded that the experimental weight of
evidence with respect to the carcinogenicity of PFOA was stronger than proposed in the draft
document, and suggested that PFOA cancer data are consistent with the EPA guidelines
descriptor 'likely to be carcinogenic to humans'. According to EPA's Guidelines for Carcinogen
Risk Assessment2 (also known as EPA's Cancer Guidelines), this descriptor is typically applied
to agents that have tested positive in more than one species, sex, strain, site or exposure route,
with or without evidence of carcinogenicity in humans. Conclusions of these Panel members
were based on the following:
• While human data are ambiguous, two separate feeding studies in rats demonstrate that
PFOA is a multi-site carcinogen.
• Uncertainties still exist (see Issue 1 comments) as to whether PPAR-alpha agonism
constitutes the sole MOA for PFOA effects on liver. This was based on the fact that
PFOA, but not the prototypical PPAR-alpha agonist, WY-14,643, increases liver
weights in PPAR-alpha knockout mice, a finding of uncertain significance in the
absence of liver histopathology and replication of this finding. Further, mitochondrial
proliferation was suggested in the document as a basis of liver toxicity in monkeys
exposed to PFOA.
• The exclusion of mammary tumors in the draft document based on comparisons to
historical control levels from other laboratories was deemed inappropriate, since the
most appropriate control group is a concurrent control group. Using that comparison,
In March 2005, EPA published final Cancer Guidelines and Supplemental Guidance which can be found at the
following URL: http://cfpub.epa.gov/ncea/raf/recordisplay.cfm?deid=l 16283.
-------
increases in both fibroadenomas (22%, 42% and 48% for rats treated with 0, 30 and
300 ppm APFO (ammonium perfluorooctanoate or C8, the ammonium salt of PFOA),
respectively) and adenocarcinomas (15, 31% and 11%, respectively) were seen in the
Sibinski et al. (1987) 2 yr PFOA feeding study.
• Insufficient data are currently available to determine the MOA for the observed Leydig
cell tumors, pancreatic acinar cell tumors and mammary gland tumors. In the absence
of a defined MOA for these tumor types, they must be presumed to be relevant to
humans, as suggested by EPA's Cancer Guidelines.
Given the current data base, these Panel members were not willing to ascribe an
associated probability value to the potential for PFOA-induced carcinogencity. Nevertheless,
based on available evidence to date, most Panel members believed that risk assessments for each
of the PFOA-induced tumors are appropriate at the current time.
A different view expressed by the remaining one-quarter of the Panel members was that
currently available evidence does not exceed the descriptor "suggestive" of carcinogenicity,
based on the belief that PPAR-alpha agonism does serve as the sole MOA for PFOA-induced
rodent liver tumors (Issue 1) and that mammary tumors were not demonstrated in animals when
compared to historical controls. Thus, these members did not believe the evidence exceeded the
draft document descriptor of "suggestive".
Issue 3: Selection of Endpoints
The draft document proposes the use of multiple endpoints from several life stages,
species and gender for risk assessment. No specific recommendations on the most appropriate
parameters are stipulated at the current time.
Comment on the:
Selection ofToxicity Endpoints for the Risk Assessment
The Most Appropriate Lifestage/Gender/Species for Assessing Human Risk
The Appropriateness of the Available Animal Models
The Panel agreed with the current approach of inclusivity, given the current uncertainties
noted above with respect to carcinogenicity, as well as the paucity of information on potential
PFOA effects on non-cancer endpoints. Similarly, no exclusion of species should be considered
at present, and differences between genders as demonstrated in rat studies again suggest multiple
MO As for PFOA. The use of multiple animal models is appropriate particularly in light of the
reported differences in toxicokinetics in rats, non-human primates and humans. Resolution of
most appropriate parameters must await additional research, but the process will be facilitated by
the ability to measure internal dose.
Panel members did not reach full agreement as to endpoints that should be included for
risk assessment and the significance of occupational biomonitoring data. About three quarters of
the Panel members supported the inclusion of multiple cancer endpoints and liver histopathology
as well as consideration of the data from occupational and epidemiological studies. While the
draft document notes that the occupational studies suffer from the fact that they involve
-------
multiplicity of exposures, other studies have shown a high correlation among fluorinated
compounds in biological samples from the general population and occupational cohorts.
Therefore, these human studies could be advantageous for assessing potential interactions among
these compounds that may be associated with adverse human health effects. These Panel
members also believed that epidemiologic and occupational studies could not be disqualified
without disqualifying virtually all such studies in the risk assessment process. Moreover, it is
clear that occupationally-exposed populations have experienced the highest levels of exposure
and therefore reported health effects in these studies merit consideration.
A contrasting view expressed by the remaining one-quarter of the Panel members was
that the outcomes from studies of human health effects of PFOA were equivocal, and thus these
endpoints should not be incorporated into the risk assessment process.
Panel members agreed on the need for additional research, including PPAR-alpha
mediated and independent effects of PFOA. Non-carcinogenicity endpoints merit additional
attention for several reasons. It is not yet known whether carcinogenicity will represent the most
sensitive endpoint for PFOAs. Immunotoxicity has been reported, and derivations of MOEs for
such effects were encouraged by many Panel members. Given the prevalence in brain of PPAR
receptors, including PPAR-alpha, effects on nervous system structure and function warrant
attention. Moreover, no information currently exists with respect to critical periods; therefore, it
is important to evaluate effects across age groups. The observations of hormonal alterations in
treated animals also deserve further study to assess their importance.
Issue 4: Risk Assessment Approach
Issue 4a: Pharmacokinetic Modeling and Use of AUC as a Measure of Internal Dose
The draft document compares internal dose metrics from animal toxicology studies and
human biomonitoring studies for purposes of ultimately generating margin of exposure (MOE)
information. Area under the concentration curve (AUC) was calculated from PFOA serum levels
in human biomonitoring studies assuming a steady state. In some of the rat studies, serum PFOA
concentrations were available, or it was considered that sufficient pharmacokinetic information
was available to estimate serum levels. For this purpose, AUC was estimated from a
pharmacokinetic model. Specifically, compartmental modeling of serum concentrations using
single dose rat oral exposure studies were used to estimate internal dosimetry for the longer term
dosing studies based upon the premise that pharmacokinetic information for rats and humans is
sufficient for this purpose and that this approach does not exceed the limits of the available data.
Comment on the Use of the One Compartment Pharmacokinetic Model
The Panel concluded that the empirical model used in the draft document was adequate
for predicting blood levels resulting from repeated dosing, but that this fitting procedure is
specific to this limited data set and this particular application. Concern was expressed, therefore,
that use of the descriptor "one compartment" to describe PFOA pharmacokinetics in the draft
document is misleading, given the actual complexities in many of the available datasets, and the
term should be removed or replaced unless carefully qualified.
-------
Comment on the use ofthe AUC as a Measure of Internal Dose for Rats and Humans for
Calculation of the MOE
The Panel observed that while calculating blood AUC may be an appropriate method to
estimate internal dose, it is important to note that at the current time information on PFOA health
effects is limited. As additional data become available, other measures may also be appropriate,
such as the Cmax, the integrated dose above a minimum concentration, etc. Regardless of the
choice for the measure of internal dose, a clearer rationale needs to be presented for the approach
taken, and, importantly, for any choice adopted, the impact of the internal dose measure on the
magnitude of the MOE should be described. The Panel also believes that caution should be
exercised in assuming that the form of PFOA in blood, i.e., free compound or PFOA bound to
various proteins or lipids is constant in serum across the period of observation, given the current
information on metabolism.
Issue 4b: Cross Species Extrapolation
In extrapolating data from animal experiments to humans, a default value of 10 is
typically applied, with a factor of 3 for differences in toxicodynamics and a value of 3 for
toxicokinetic differences. In the PFOA draft risk assessment document, internal doses from
animal toxicology studies and human biomonitoring studies were compared. Derivation of data
from animal toxicology studies included both measured PFOA serum levels from non-human
primates and derived values from pharmacokinetic modeling from rat studies. The reliance on
internal dose metrics was considered by OPPT to be sufficient to reduce uncertainties and
therefore raised the question of the ability to either eliminate or reduce the default values for
cross species extrapolation.
Comment on the Need to Use or Modify the Default Value of 10 for Cross Species
Extrapolation Given the Pharmacokinetic Analysis
The use of internal dose metrics in this analysis was considered by the Panel to be a
significant step toward reducing uncertainty related to the toxicokinetic uncertainty associated
with interspecies extrapolation. Nevertheless, it did not believe that the direct use of blood
concentration in the assessment sufficiently reduced the overall uncertainly to eliminate or
modify the current default value. Significant uncertainties still remain, including the measured
internal dose that best predicts adverse effects in human and other species, the bias inherent in
measurement/modeling errors, the lack of information about non-cancer endpoints,
developmental vulnerability and the impact of gender, and the multiple PFOA environmental
exposures that occur in humans vs. animals, among others. The assumption that PFOA serum
levels are at steady state in children 2-12 years of age has not been tested and may not be valid.
The Panel likewise stressed that bench mark dose methodologies would be preferable to the
reliance in the draft document on LOAEL-driven MOE calculations.
-------
Issue 4c: Human Biomonitoring Data
Currently available data on PFOA levels in humans includes occupational biomonitoring
studies as well as three population studies within the U.S. The measurements from the population
studies come from: 1) samples from 6 American Red Cross blood banks; 2) a study of
Streptococcal A infection in children; and 3) elderly volunteers in a cognitive study in Seattle.
The draft EPA document only utilizes data derived from 1 and 2 above in its calculation of the
MOE. Occupational biomonitoring data were excluded from the calculation because it was stated
that sample sizes were small, data on gender were not available, and that blood monitoring data
obtained from 2000 would overestimate current serum levels, since PFOA exposure of this group
ceased in 2002. Measured levels from the elderly population were not utilized because values
were considerably lower, for unknown reasons, than those reported in the other population
studies for adults and children. From the other two population studies utilized in the draft
document, geometric means and 90th percentiles were calculated across genders for calculation of
MOEs.
Comment on the Adequacy of the Human Exposure Data for Use in Calculating a MOE
Panel members were not in full agreement as to the adequacy of the human exposure data for
inclusion in the MOE calculation. Many Panel members shared concerns about the approach adopted in
the draft document. One concern related to the generality of the populations currently included in the
MOE calculation. It was noted, for example, that use of the blood donor and pediatric biomonitoring
data may be acceptable if the purpose is to assess whether there is a potential health effect to the
"general" population, although there is some question as to the size of other non-occupational
populations that might be more highly exposed and the assumption that PFOA serum levels are at steady
state may not be valid for children or fetuses. About three quarters of Panel members agreed that
existing subpopulations of the general public are likely to be more highly exposed than those previously
reported and results from occupational studies should be included in the MOE calculation. A differing
view expressed by the remaining one quarter of the Panel members was that the human biomonitoring
data are equivocal and thus not useful to MOE calculation.
Three different summary statistics are presented in the draft document and used in the
calculation of the MOE. Of these, the Panel questioned the use of mean values, particularly
geometric means in the calculations. Additionally, no rationale was provided for the choice of
the 90th percentile as a summary statistic, rather than the use of a higher value. Whatever the
approach adopted, justification must be provided for the chosen summary measure and an
explicit objective for the MOE analysis described.
-------
INTRODUCTION
Background
This report was prepared by the Science Advisory Board (SAB) PFOA Risk Assessment
Review Panel (the "Panel") in response to a request by EPA's Office of Pollution Prevention and
Toxics (OPPT) to review their Drgj|J??^
Associated With Exposure to Perfliiorooctanoic Acid (PFOA) Its Salts. According to the
document, OPPT has been investigating PFOA and its salts to try to understand the health and
environmental issues presented by fluorochemicals, in the wake of unexpected toxicological and
bioaccumulation discoveries with respect to perfluorooctane sulfonates (PFOS). PFOA and its
salts are fully fluorinated organic compounds that can be produced synthetically or through the
degradation or metabolism of other fluorochemical products. PFOA is primarily used as a
reactive intermediate, while its salts are used as processing aids in the production of
fluoropolymers and fluoroelastomers and in other surfactant uses. PFOA and its salts are
persistent in the environment.
OPPT identified 4 issues where they were seeking the SAB's advice and
recommendations. These included the proposed mode of action, carcinogenicity descriptors,
toxicological endpoints selected and the pharmacokinetic modeling methods used in the risk
assessment. OPPT's assessment focused on the potential human health effects associated with
exposure to PFOA and its salts. Several toxicological endpoints and hypothesized modes of
action were considered. Internal dose metrics were estimated for animal toxicology studies with
pharmacokinetic modeling, and were obtained from human biomonitoring studies, assuming
steady state. Margin of Exposure (MOE) values were calculated from the internal dose metrics.
The SAB PFOA Review Panel was asked to comment on the scientific soundness of this risk
assessment.
The Panel deliberated on the charge questions during their February 22-23, 2005 face-to-
face meeting and during a conference call on July 6, 2005. The responses that follow represent
the views of the Panel. In all cases, there was agreement by a majority of the panel members as
to a particular recommendation. In some cases, there were one or more panel members that had
a differing point of view; these instances have been noted throughout the report. The specific
charge questions to the Panel are as follows:
Charge Questions
Issue 1: Rodent PPAR-alpha Mode of Action for Hepatocarcinogenesis
The postulated mode of action (MOA) of PPAR* 'agonist induced liver toxicity and liver
tumors in rodents involves four causal key events. The first key event is activation of PPAR* •
(which regulates the transcription of genes involved in peroxisome proliferation, cell cycle
control, apoptosis, and lipid metabolism). Activation of PPAR* 'leads to an increase in cell
proliferation and a decrease in apoptosis, which in turn leads to preneoplastic cells and further
-------
clonal expansion and formation of liver tumors. Of these key events, only PPAR* "activation is
highly specific for this MOA while cell proliferation/apoptosis and clonal expansion are
common to other modes of action. There are also several "associative" events that are markers
of PPAR* •agonism but are not directly involved in the etiology of liver tumors. These include
peroxisome proliferation (a highly specific indicator that this MOA is operative) and
peroxisomal gene expression.
Information that provides evidence that any specific chemical is inducing liver toxicity
and tumors via a PPAR* 'agonist MOA includes in vitro evidence of PPAR* *agonism (i.e.,
evidence from an in vitro receptor assay), in vivo evidence of an increase in number and size of
peroxisomes, increases in the activity of acyl CoA oxidase, and hepatic cell proliferation. The in
vivo evidence should demonstrate dose-response and temporal concordance between precursor
events and liver tumor formation. Other information that is desirable and may strengthen the
weight of evidence for demonstrating that a PPAR* "agonist MOA is operative includes data on
hepatic CYP4A1 induction, palmitoyl CoA activity, hepatocyte hypertrophy, increase in liver
weights, decrease in the incidence of apoptosis, increase in microsomal fatty acid oxidation, and
enhanced formation of hydrogen peroxide.
OPPT has proposed that there is sufficient weight of evidence to establish that the mode
of action for the liver tumors (and precursor effects) observed in rats following exposure to
PFOA is PPAR* •agonism.
Question 1 - Please comment on the weight of evidence and adequacy of the data available to
identify the key events for the PPAR* agonist-induced rodent liver toxicity and
hepatocarcinogenesis for PFOA. Discuss whether the uncertainties and limitations of these data
have been adequately characterized.
Issue 2: Descriptor for Carcinogenic Potential
Carcinogenicity studies in Sprague-Dawley rats show that PFOA induces a "tumor triad"
similar to a number of other PPARa agonists. This "tumor triad" includes liver tumors, Leydig
cell tumors (LCT), and pancreatic acinar cell tumors (PACT). OPPT has proposed that there is
sufficient evidence to conclude that the liver tumors are due to PPAR* «agonist MOA, and that
this MOA is unlikely to occur in humans based on quantitative differences between rats and
humans. In addition, the LCT and PACT induced in the rat by PFOA probably do not represent
a significant cancer hazard for humans because of quantitative toxicodynamic differences
between the rat and the human. Overall, based on no adequate human studies and uncertain
human relevance of the tumor triad (liver, Leydig cell and pancreatic acinar cell tumors) from
the rat studies, OPPT has proposed that the PFOA cancer data may be best described as
providing "suggestive evidence ofcarcinogenicty, but not sufficient to assess human
carcinogenic potential under the interim 1999 EPA Guidelines for Carcinogen Risk
Assessment, as well as the 2003 draft EPA Guidelines for Carcinogen Risk Assessment.
Question 2 - Please comment on the proposed descriptor for the carcinogenic potential of
PFOA.
-------
Issue 3: Selection of Endpoints
OPPT has proposed the use of several endpoints from several life stages, species and
gender for the risk assessment. For this draft assessment, OPPT has not made specific
recommendations on the most appropriate endpoint/lifestage/species/gender. Rather, all have
been presented to provide transparency.
For adults, endpoints were selected from the non-human primate and rat studies; the
endpoints included liver toxicity and possibly mortality for the non-human primates and
decreased body weight for rats.
For developmental endpoints, OPPT relied upon the definition of developmental toxicity
outlined in the Agency's Developmental Toxicity Risk Assessment Guidelines. These
guidelines state that the period of exposure for developmental toxicity is prior to conception to
either parent, through prenatal development and continuing until sexual maturation. (In contrast,
the period during which a developmental effect may be manifested includes the entire lifespan of
the organism). Based on this definition of developmental exposure, OPPT considered
developmental effects in the rat two-generation reproductive toxicity study to include reductions
in Fl mean pup body weight (sexes combined) on lactation days 1, 5 and 8, an increase in
mortality during the first few days after weaning (both sexes), a delay in the timing of sexual
maturation (both sexes), and a reduction in mean body weight postweaning (Fl males only).
Question 3 - Please comment on the selection of these toxicity endpoints for the risk assessment.
Question 4 - Given the available data to date, please comment on the most appropriate
lifestage/gender/species for assessing human risk.
Question 5 - Please comment on the appropriateness of the available animal models. Please
comment on whether additional animal models should be investigated, and if so, what
information would better enable us to ascertain potential human risks.
Issue 4: Risk Assessment Approach
A margin of exposure (MOE) approach can be used to describe the potential for human
health effects associated with exposure to a chemical. The MOE is calculated as the ratio of the
NOAEL or LOAEL for a specific endpoint to the estimated human exposure level. The MOE
does not provide an estimate of population risk, but simply describes the relative "distance"
between the exposure level and the NOAEL or LOAEL. In this risk assessment there is no
information on the sources or pathways of human exposure. However, serum levels of PFOA,
which are indicative of cumulative exposure, were available from human biomonitoring studies.
In addition, serum levels of PFOA were available for many of the animal toxicology studies or
there was sufficient pharmacokinetic information to estimate serum levels. Thus, in this
10
-------
assessment internal doses from animal and human studies were compared; this is analogous to a
MOE approach which uses external exposure estimates.
Issue 4a: Pharmacokinetic Modeling and Use of AUC as a Measure of Internal Dose
As noted above, internal dose metrics from animal toxicology studies and human
biomonitoring studies were compared in this draft assessment. For humans, the area under the
concentration curve (AUC) was calculated from measured PFOA serum levels in human
biomonitoring studies, assuming steady state. For the rat toxicology studies, the area under the
concentration curve (AUC) and C were estimated from a pharmacokinetic model. The
max
pharmacokinetic analysis could be done using a number of approaches including non-parametric
analysis, physiologically based pharmacokinetic (PBPK) modeling, and classical compartmental
modeling. Each has strengths and limitations given the available data. Non-parametric analyses
provide a description of the data that have been collected, but have fairly limited ability to make
predictions across species or to account for variations in exposures. PBPK modeling is perhaps
the ideal approach for addressing PFOA for purposes of cross-species extrapolation. Extensive
pharmacokinetic studies have been undertaken in rats demonstrating complex phenomena
including high tissue concentrations in liver, kidney and serum and enterohepatic recirculation of
the parent compound. These could be addressed using PBPK modeling for the rats, but the more
limited information in monkeys and humans would either require substantial assumptions or
preclude use of this approach. Classical compartmental modeling can be used to analyze the
existing data on blood concentrations in rats, monkeys, and humans. Currently, the available
pharmacokinetic information for rats and humans is sufficient to support compartmental
modeling. Comparisons of serum protein binding across species indicated a high degree of
binding in all species but also interspecies differences in the percentage of unbound PFOA in
plasma. In light of the documented differences in clearance of PFOA across sexes in rats and
across species, compartmental modeling of serum concentrations provides a sound approach for
estimating internal dosimetry without exceeding the limits of the available data, so this approach
was selected for this risk assessment.
Question 6 - Please comment on the use of the one compartment pharmacokinetic model.
Question 7 - Please comment on the use of the AUC as a measure of internal dose for rats and
humans for calculation of the MOE.
Issue 4b: Cross Species Extrapolation
Judgments about the "adequacy" of a MOE are based on many considerations including
uncertainty associated with cross species extrapolation. Typically, a value of 10 is considered
which consists of a value of 3 for toxicodynamics and a value of 3 for toxicokinetics. Each of
these can be decreased or increased if there are data to warrant it. In this draft assessment,
internal doses from animal toxicology studies and human biomonitoring studies were compared.
For humans, the internal doses were based on measured PFOA serum levels in human
biomonitoring studies. For the non-human primate toxicology studies, internal doses associated
with the NOAEL and/or LOAEL were based on measured PFOA serum levels. For the rat
11
-------
toxicology studies, pharmacokinetic modeling was used to estimate an internal dose metric
associated with a NOAEL or LOAEL.
Question 8 - Please comment on the need to use or modify the default value of 10 for cross
species extrapolation given the pharmacokinetic analysis.
Issue 4c: Human Biomonitoring Data
For this draft assessment, human biomonitoring data of PFOA serum levels were
available for adults and children. Similar analytical methods were used to measure the PFOA
levels in both sets of blood samples. The adult data included 645 U.S. adult blood donors (332
males, 313 females) from 2000-2001, ages 20-69, obtained from six American Red Cross blood
banks located in: Los Angeles, CA; Minneapolis/St. Paul, MN; Charlotte, NC; Boston, MA;
Portland, OR, and Hagerstown, MD. Each blood bank provided approximately 10 samples per
10-year age interval (20-29, 30-39, etc.) for each sex.
The children's data included a sample of 598 children, ages 2-12 years old, who had
participated in a study of group A streptococcal infections. The samples collected in 1994-1995
from children residing in 23 states and the District of Columbia were analyzed for PFOA in
2002.
Question 9 - Please comment on the adequacy of the human exposure data for use in calculating
aMOE.
12
-------
RESPONSES TO THE CHARGE QUESTIONS
Issue 1: Rodent PPAR-alpha Mode of Action for Hepatocarcinogenesis and Liver Toxicity
Question 1. Please comment on the weight of evidence and adequacy of the data available to
identify the key events for the PPAR alpha agonist induced rodent liver toxicity and
hepatocarcinogenesis for PFOA. Discuss whether the uncertainties and limitations of these data
have been adequately characterized.
As discussed in the EPA Draft Risk Assessment of the Potential Human Health Effects
Associated with Exposure to Perfluorooctanoic Acid and its Salts, a sequence of four key events
define the mode of action (MOA) by which PPAR-alpha agonists induce rodent liver tumors.
According to this MOA, the initial causal event is (1) activation of PPAR-alpha, which regulates
the expression of genes involved in peroxisome proliferation, cell cycle control, apoptosis, and
lipid metabolism. These transcriptional events lead to (2) increased cell proliferation and/or
decreased cell death. The chronic increase in cell growth occurs primarily in the preneoplastic
focal lesions in the liver resulting (3) in the clonal expansion of the preneoplastic lesions, which
ultimately results (4) in the development of hepatocellular neoplasms. In addition, there are
"associative" events that may or may not be causally linked to the PPAR-alpha MOA for
hepatocarcinogenesis which include blockage of cell to cell communication, an increase in
peroxisomes, an increase in peroxisomal enzymes, and liver and hepatocyte hypertrophy.
The Panel agreed that, considered collectively, the weight of evidence to date is
consistent with the assertion that PFOA is a PPAR-alpha agonist and can induce liver changes in
adult rats that have been associated with PPAR-alpha activation. As discussed in the draft PFOA
risk assessment, some of the key elements to establish this MOA have been demonstrated by
appropriate experiments. In vitro studies demonstrate that PFOA is a PPAR-alpha agonist, and
treatment of rats and/or mice results in peroxisome proliferation, increased beta-oxidation, and
hepatomegaly, with dose and temporal responses consistent with this MOA for liver tumor
induction. Studies comparing PPAR-alpha null and wild-type mice showed that PFOA-induced
peroxisome proliferation, beta-oxidation, and immunotoxicity depend on the presence of a
functional receptor. Further, no other established modes of action of liver cancer-induction have
been reported for PFOA. PFOA is neither DNA reactive nor mutagenic, and thus not involved in
a genotoxic mode of action; nor is the liver neoplastic effect due to the induction of repeated
hepatocyte death and compensatory regeneration (a cytotoxic mode of action). No PPAR-alpha
independent MOA for the rat liver tumor induction has been proposed.
With respect to the weight of evidence and the adequacy of consideration of uncertainties
and limitations, however, the Panel did not reach full agreement. About three quarters of the
Panel members believed that data gaps still exist and not all of the causal events in the PPAR-
alpha MOA have been demonstrated for PFOA. These include the induction of cell proliferation
in the liver at early times following PFOA treatment and/or modulation of apoptosis in
hepatocytes. They also shared the belief that while the PFOA Draft Risk Assessment in general
appropriately discusses the uncertainties and limitations of the data that support a PPAR-alpha
MOA for PFOA-induced liver tumors in adult rats, it fails to consider three issues contrary to
this MOA in sufficient detail.
13
-------
First, in a study by Yang et al. (2002) cited in the report in the context of the receptor
dependence of PFOA immunotoxicity, PPAR-alpha null mice exhibited >2-fold increases in
liver weight but no peroxisomal acyl CoA oxidase induction in response to PFOA. No increase
in liver weight was observed in PPAR-alpha null mice treated with the well-characterized
prototype PPAR-alpha agonist, WY-14,643. While this finding is of uncertain significance, due
to the lack of histopathology and the absence of a second study showing such an effect, it
nevertheless raises the possibility that PFOA may induce some of its effects in mouse liver by a
PPAR-alpha-independent pathway. This observation and the associated uncertainty were not
mentioned in the context of liver tumor induction in the draft PFOA risk assessment. Secondly,
uncertainties exist with respect to the relevance to exposed fetuses, infants and children of the
PPAR-alpha agonist MOA for induction of liver tumors in adults. Humans are refractory to
some, but not all, PPAR-alpha activation effects. Data from studies using PPAR-alpha receptor
knockout mice have shown that these receptors are essential for the rapid induction of liver
neoplasms after exposure to WY-14,643. However, humans do have functional PPAR-alpha
receptors, leaving unanswered the question as to why they respond so differently from rats and
mice to PPAR-alpha agonists. Available data suggest that the difference between humans and
rats or mice may be a consequence of a lower number of PPAR-alpha receptors such that the
PPAR-alpha MOA is not considered likely to yield a similar hepatic cancer response in adult
humans. However, exposures of fetuses, neonates and children to PFOA remain a potential
concern. Rat studies suggest similar PPAR-alpha receptor levels in neonates and adults, but
because adult humans have so few receptors, and information in fetuses, neonates and children is
minimal, this same extrapolation cannot be made in humans. Given that human exposures to
PFOA and related chemicals appear ubiquitous, uncertainties and limitations of the data for
children have not been adequately characterized to be able to conclude that the PPAR-alpha
MOA is not operative in this young age group.
Second, the current draft PFOA risk assessment states (page 76 lines 15-16) that the
"[a]ctivation of PPAR-alpha leads to an increase in cell proliferation and a decrease in apoptosis,
which in turn leads to preneoplastic cells ..." Questions were raised as to whether there is
available experimental evidence that the induction of PPAR-alpha results in an increase in the
number of preneoplastic foci. The effect of the PPAR-alpha activation appears to be at the level
of focal lesion clonal expansion (Klaunig et al., 2003), however clonal expansion of focal
lesions, which is not unique to a PPAR-alpha MOA, has not been shown in rats treated with
PFOA.
Thirdly, some Panel members felt that the role of Kupffer cells (shown in Figure 1, page
78 of the draft document) should be discussed in the text of the draft PFOA risk assessment.
There is an extensive literature on the essential role of Kupffer cells in signaling peroxisome
proliferator-induced hepatocyte proliferation. Studies have shown that hepatocyte proliferation
and peroxisome proliferation occur by different mechanisms. Parzefall et al. (2001) and Hasmall
et al. (2001) demonstrated that peroxisome proliferators had no effect on DNA synthesis but still
induced peroxisomal acyl CoA oxidase activity in cultured rat and mouse hepatocytes that had
been purified to remove contaminating Kupffer cells. Kupffer cells, which are resident
macrophages in the liver, are a major source of growth factors (tumor necrosis factor alpha,
interleukins) that induce DNA synthesis or suppress apoptosis in purified hepatocytes. A key
14
-------
finding relevant to the proposed MOA is that Kupffer cells do not express PPAR-alpha (Peters et
a/., 2000), but are activated by peroxisome proliferators. Prevention of Kupffer cell activation by
glycine inhibited, although not completely, the development of liver tumors by the potent
peroxisome proliferator, WY-14,643 (Rose etal., 1999). There are no data available on the
effects of peroxisome proliferators on human Kupffer cells. Recognizing the role of Kupffer cell
activation in the induction of DNA synthesis and subsequent neoplastic development by PPAR-
alpha agonists, some members of the FIFRA Science Advisory Panel (2003) [SAP Minutes No.
2003-05] noted that the interplay between PPAR-alpha agonism and Kupffer cells has not been
characterized. Thus, the results from the PPAR-alpha null mouse are not directly applicable to
the human situation in which PPAR-alpha is present and can be activated.
A different conclusion was reached by the remaining one quarter of the Panel members
who found that the weight of evidence was adequate to support a PPAR-alpha agonism mode of
action for PFOA-induced rodent liver tumors. In this view, the observation of increased liver
weights in response to PFOA but not to the prototype PPAR-alpha agonist WY-14,643 in PPAR-
alpha knock-out mice as reported in Yang et al. (2002) did not merit significance because the
study was not designed to evaluate liver toxicity, and the observation represents a single
replication without corresponding histopathology at the current time. Nor was the possible role
of Kupffer cells considered to be significant. Based on these considerations, these Panel
members believed that PPAR-alpha agonism can be considered the sole MOA for PFOA-
induced rodent liver tumors.
Issue 2: Descriptor for Carcinogenic Potential
Question 2. Please comment on the proposed descriptor for the carcinogenic potential of
PFOA.
The draft PFOA risk assessment proposes that the PFOA cancer data may be best
described as providing "suggestive evidence ofcarcinogenicity, but not sufficient to assess
human carcinogenic potentiaT' under the interim 1999 EPA Guidelines for Carcinogen Risk
Assessment (US EPA, 1999), as well as the 2003 draft EPA Guidelines for Carcinogen Risk
Assessment (US EPA, 2003). This opinion is based on the view that human studies on PFOA do
not provide adequate support of carcinogenicity, as well as on the quantitative differences
between rats and humans that OPPT believes raises uncertainties about the human relevance of
the "tumor triad" response (liver tumors, Leydig cell tumors, and pancreatic acinar cell tumors)
of PPAR-alpha agonist activation in rats.
The determination of an appropriate descriptor for the carcinogenic potential of PFOA
was discussed by the Panel in the context of the available carcinogenicity data, an evaluation of
mechanistic or MOA data, and guidance on how EPA applies various descriptors for
summarizing weight of evidence data. Panel members did not achieve full agreement on the
appropriate descriptor. Based on the above considerations, the view of about three quarters of the
Panel members was that the descriptor "likely to be carcinogenic" was more consistent with
currently available data, while the remaining one quarter of the Panel members reached the
15
-------
conclusion that the current evidence fails to exceed the descriptor "suggestive" of
carcinogenicity.
Cancer studies on PFOA
Carcinogenicity studies in Sprague-Dawley rats have shown that PFOA induces
neoplasms at multiple sites. In male rats exposed to 0 or 300 ppm ammonium perfluorooctanoate
(APFO) in the feed for 2 years, increased incidences of testicular Ley dig cell tumors (LCT) (0%
vs. 11%), pancreatic acinar cell tumors (PACT) (0% vs. 11%), and liver adenomas (3% vs. 13%)
were observed in treated animals compared to controls (Biegel et a/., 2001). In a 2-year study in
which male and female Sprague-Dawley rats were fed diets containing 0, 30 or 300 ppm APFO,
a dose-related increase in LCT was observed (0% in controls, 4% at 30 ppm, 14% at 300 ppm)
(Sibinski et al., 1987). The draft PFOA risk assessment does not address the effects in the liver
observed in the Sibinski et al. (1987) study. In that study, the incidences of hepatocellular
carcinoma in male rats were 6%, 2%, and 10%, and although no adenomas were diagnosed, the
incidences of hyperplastic nodules in the liver were 0%, 0%, and 6%. Hyperplastic nodules may
be part of the continuum of proliferative lesions in the liver carcinogenic process.
In female rats, a dose-related increase in mammary gland fibroadenomas was reported
(22% in controls, 42% at 30 ppm, and 48% at 300 ppm) by Sibinski etal. (1987). In addition, the
incidence of mammary gland adenocarcinomas was greater in the low dose group than in
controls (15% in controls, 31% at 30 ppm, and 11% at 300 ppm). The draft PFOA risk
assessment did not consider the mammary gland neoplasms to represent a compound-related
effect because of high background rates reported for fibroadenomas in Sprague-Dawley rats in
historical control data (37%) reported for female rats in the Haskell Laboratory in 1987 (Sykes).
Three-quarters of the Panel members did not believe that historical control comparisons are as
reliable as concurrent controls. A number of parameters may contribute to inter-laboratory
differences in tumor response including differences in diet, animal age at the start and
termination of studies, animal supply sources and breeding practices, environmental conditions,
vehicles and routes of administration, animal care procedures that may affect weight gain and
survival, and the use of different substrains. Thus, in their view, the concurrent control group is
the most appropriate group for evaluations of chemical-related effects. Moreover, in the
historical database of Chandra et al. (1992), the incidence in controls of mammary gland
fibroadenomas was 19.0% and the incidence of adenocarcinomas was 8.8% in female Sprague-
Dawley rats. Therefore, a neoplastic effect in the mammary gland is apparent in the Sibinski
study in comparison to Chandra et al. (1992). Those Panel members therefore believe that the
elevated tumor rates observed in female rats in the Sibinski et al. (1987) study raise concerns for
neoplastic effects induced by PFOA in the mammary gland that should not be dismissed. In
addition, while new information3 was submitted to the panel questioning the findings in the
Sibinski study, about three quarters of the Panel members urged that an independent,
appropriately-designed histopathology review of the male rat livers and the female mammary
glands from the Sibinski study be conducted to re-analyze the resulting tumor incidence data.
3 a report entitled, "Pathology Peer Review and Pathology Working Group Review of Mammary Glands from a
Chronic Feeding Study in Rats with PFOA Report" conducted by Experimental Pathology Laboratories, Inc. and
submitted to the SAB by Dr. Larry Zobel of 3M Medical Department
16
-------
The remaining Panel members believed that the comparison of the Sibinski et al. (1987)
mammary tumor data to the historical control data (Sykes, 1987) in the draft risk assessment
document was valid.
Mode-of-action analysis, uncertainties, and human relevance
The PFOA draft risk assessment proposes that there is sufficient evidence to conclude
that liver tumors induced by PFOA are due to a proposed PPAR-alpha agonist MOA (Klaunig et
al., 2003), and that this MOA is unlikely to occur in humans based on quantitative differences in
the numbers of PPAR-alpha receptors between rats and humans. In addition, the PFOA draft
risk assessment proposes that the Ley dig cell tumors (LCT) and pancreatic acinar cell tumors
(PACT) induced in the rat by PFOA probably do not represent a significant cancer hazard for
humans because of quantitative toxicodynamic differences between the rat and the human.
Thus, the panel examined issues related to our understanding of the MOA for the multiple tumor
types induced by PFOA in rats and the impact of available information on determining the
human relevance of the animal tumor responses.
Liver adenomas.
As noted under Issue 1, the Panel concurred that the collective evidence is consistent
with the hypothesis of a PPAR-alpha agonist MOA for PFOA with associated peroxisomal P-
oxidation activity, increases in absolute and relative liver weight, and liver tumors in Sprague-
Dawley rats. Issues on which the Panel members opinions diverged related to whether a PPAR-
alpha agonist MOA for liver tumor induction in rats might occur in humans and/or whether
additional MO As might be involved.
Key events in the PPAR-alpha agonist MOA.
The PFOA risk assessment did not identify dose-response data showing increases in
hepatocyte proliferation and suppression of apoptosis in rats exposed to PFOA. Many Panel
members believed this to be a critical deficiency, because these are key events in the proposed
MOA linking activation of PPAR-alpha to the liver tumor response.
Another observation that influenced most Panel members with respect to potential human
relevance of the response in rats is the observation that the same early effects actually occur in
monkeys exposed to PFOA. These effects include the induction of peroxisomal p-oxidation
activity (2.6 fold), significant increases and positive dose-response trends for absolute and
relative liver weights (1.6 fold), and the return of relative liver weight to control levels after a
13-week recovery period. Cell proliferation was evaluated in monkeys but only after 6 months
of exposure. Unfortunately, neither the rat nor the monkey studies provided data on hepatocyte
proliferation during the first 1-2 weeks of exposure, or direct measurements of apoptotic cells
during or after exposure to PFOA was stopped. The lack of data on cell proliferation and
apoptosis in animals exposed to PFOA makes it impossible to analyze dose-response
concordance between these key events and tumor induction for PFOA in relation to other PPAR-
alpha agonists. Because the available data for PFOA in rats and monkeys indicate similar
responses in the livers of rodents and primates (increased liver weight and induction of hepatic
peroxisomal enzyme activity), about three quarters of the Panel members shared the view that
human relevance for liver effects induced by PFOA by a PPAR-alpha agonism MOA cannot be
discounted.
17
-------
The remaining panel members, however, considered the increase in liver weight in rats
exposed to PFOA and the return to control levels following an 8-week recovery period
(Palazzolo, 1993) to be consistent with an increase in cell proliferation and suppression of
apoptosis by PFOA during the exposure period. In addition, the lack of an increase in hepatic
cell proliferation in rats after 1 month or more exposure to PFOA (Biegel et a/., 2001) was
considered consistent with observations of a transient increase in hepatocyte proliferation with
other peroxisome proliferators.
PPAR-alpha -independent liver effects.
As noted in response to Issue 1, about three quarters of the members of the Panel shared
the view that significant uncertainties still exist with respect to the predictability of a PPAR-
alpha agonist MOA for human cancer risk associated with exposure to PFOA. In a comparative
study of PFOA and the prototype PPAR-alpha agonist Wy-14,643, at doses of each chemical that
produced increases in liver weight and peroxisomal fatty acid acyl-CoA oxidase activity in wild-
type mice, only PFOA caused a similar 2-fold increase in liver weight (but no increase in acyl-
CoA oxidase activity) in PPAR-alpha null mice (Yang et a/., 2002). While this study was not
designed to assess liver toxicity, it confirms that PFOA is a PPAR-alpha agonist for peroxisomal
enzyme induction, and also indicates that liver changes induced by PFOA in rodents can occur
by a mechanism that is independent of PPAR-alpha activation. The lack of liver enlargement or
tumor response in PPAR-alpha null mice exposed to Wy-14,643 for 11 months has been cited
frequently as evidence that liver cancer induction by peroxisome proliferators is mediated by
PPAR-alpha activation (Peters etal, 1997). The study of Yang etal. (2002) needs to be
replicated, but appears to suggest that results with Wy-14,643 in PPAR-alpha null mice do not
predict all of the potential liver effects of PFOA.
Further, while not diminishing the conclusion that a PPAR-alpha MOA is operative in the
rodent liver carcinogenesis induced by PFOA, about three quarters of the Panel members
expressed concern over the as yet incomplete understanding of the role of Kupffer cells in the
carcinogenic process. PPAR-alpha independent stimulation of hepatocyte growth factor
production in Kupffer cells appear to be essential to the mechanism of hepatocyte replicative
DNA synthesis, suppression of apoptosis, and liver tumor development by peroxisome
proliferators. Until the interplay between PPAR-alpha agonism and Kupffer cell activation is
characterized, negative results from the PPAR-alpha null mouse may not be relevant to the
human situation in which Kupffer cells and hepatocellular PPAR-alpha are present and can be
activated.
The remaining members of the Panel believed that the finding of increased liver weights
produced by PFOA in PPAR-alpha knockout mice, as noted for Issue 1, were not significant
enough to undermine the PPAR-alpha agonism MOA, nor did they consider the absence of
information about Kupffer cell activation to be relevant to a PPAR-alpha agonism MOA for
PFOA-induced rodent liver tumors.
LCTs, PACTs, and mammary neoplasms.
Panel members did not consider the consolidation of liver tumors, LCTs, and PACTs into
a triad MOA to be justified. They believed that available evidence is inadequate to support a
18
-------
PPAR-alpha agonist MOA for the induction of LCTs and PACTs (Klaunig et al., 2003), and, at
present, available data are insufficient to characterize the MOA for PFOA-induced LCTs and
PACTs. As such, a specific MOA needs to be worked out for each tumor type. In addition,
about three quarters of the Panel members felt that the appropriate comparison for mammary
neoplasms was to concurrent not to historical controls, and in that view subscribe to the
interpretation that PFOA does increase mammary gland neoplasms, and no MOA data are
available for the mammary tumor response. As discussed in EPA's Cancer Guidelines, in the
absence of sufficient data to establish a MOA, the animal tumor responses are presumed to be
relevant to humans.
The remaining Panel members believed, in contrast to the above view, that the
comparison of PFOA-induced mammary tumor levels to historical controls was valid, and thus
deemed the evidence for mammary neoplasms to be insufficient to demonstrate such tumors in
response to PFOA. This served to support the view of these members that PPAR-alpha agonism
represented the sole MOA for PFOA-induced rodent liver tumors.
Application of cancer descriptors
The meaning of terms such as "suggestive evidence of carcinogenic potential" or "likely
to be carcinogenic to humans" may differ among some in the general public and the EPA
because of differences in perception and intent. Hence, EPA recommends a weight-of-evidence
narrative that explains the complexity of issues influencing an agent's carcinogenic potential in
humans. Descriptors are applied to provide consistency across agents that are evaluated for their
carcinogenic potential. In developing their cancer risk assessment guidelines (US EPA 1999,
2003), EPA has not provided definitive criteria for choosing a descriptor; however, examples of
the types of evidence that would be covered by a descriptor are listed. EPA also cautions that
terms such as "likely," when used as a weight-of-evidence descriptor, do not correspond to a
quantifiable probability.
About three quarters of the Panel members shared the view that while human cancer data
on PFOA are inadequate to support a definitive conclusion of the presence or absence of a causal
association, the animal data are much stronger than the examples summarized in the EPA's
Cancer Guidelines under the descriptor "suggestive evidence of carcinogenic potential." The
descriptor "suggestive" is typically applied to agents that show a marginal increase in tumors
only in a single animal study or a slight increase in a tumor response at a site with a high
background rate. The animal data for PFOA are consistent with the examples listed by EPA
under the descriptor "likely to be carcinogenic to humans" applied to agents that tested positive
in more than one species, sex, strain, site, or exposure route, with or without evidence of
carcinogenicity in humans; or a positive study that indicates a highly significant result and where
the response is assumed to be relevant to humans. These members concluded, as described
above, that data from two separate feeding studies demonstrate that PFOA is a multi-site
carcinogen in rats. Significant increases in tumor incidence and dose-response trends were
observed in male and female rats. Some of the tumor responses were observed at sites with low
background rates; the incidence of PACTs and LCTs in control rats was 0% at both sites.
Because available data are insufficient to characterize the MOA for PFOA-induced LCTS,
PACTs, or mammary tumors, the responses at these sites are presumed to be relevant to humans.
Uncertainties also still exist for the MOA(s) for liver tumors induced by PFOA.
19
-------
While opting for the descriptor "likely to be carcinogenic to humans" these Panel
members were not willing to state an associated probability value for PFOA-induced
carcinogenicity; nor do the EPA guidelines require a quantifiable probability. This group also
encouraged a cancer risk assessment for each of the PFOA-induced tumors where data permit.
The risk characterization narrative should address the state of knowledge and uncertainties in the
MO A for each tumor site and a range of approaches should be considered in the cancer dose-
response assessment.
The remaining one quarter of the Panel members did not find the weight of evidence
strong enough to exceed the descriptor "suggestive". These Panel members agreed with the
EPA's risk assessment which proposed that the PFOA cancer data may be best described as
providing "suggestive evidence ofcarcinogenicty, but not sufficient to assess human
carcinogenic potential" This view was based upon their conclusions that: 1) a mechanism for
formation of liver tumors in rats considered not relevant to humans, 2) liver cell proliferation
was absent in monkeys, 3) a strong association between PFOA exposure and tumors was not
demonstrated in human studies, 4) the belief that the testicular and pancreatic acinar tumors in
rats were probably not relevant to humans, and 5) the view that mammary tumors reported in
female rats are equivocal. Further these panel members also believed that the sole MOA for the
rodent-induced liver tumors was through PPAR-alpha agonism which they believe was not
relevant to humans.
Issue 3: Selection of Endpoints
Question 3. Please comment on the selection of these toxicity endpointsfor the risk assessment.
The Panel generally agreed with the Agency approach of considering multiple endpoints
and developing multiple margin of exposure (MOE) values at this stage in the assessment of
potential human health effects associated with PFOA. With regard to the selection of endpoints,
the initial overall philosophy should be one of inclusivity. That is, endpoints should be
considered unless evidence for an effect by PFOA is equivocal or the dose associated with the
effect is sufficiently high that other effects will clearly be of greater concern. The reason for
being inclusive is not to generate an exhaustive catalog of PFOA effects, but rather to insure that
sensitive effects (i.e., effects occurring at relatively low doses) are not overlooked or prematurely
excluded from the assessment.
The Panel agreed with inclusion of all of the endpoints in the current draft of the risk
assessment. None were recommended for deletion. However, caveats regarding the use of
organ and body weights as endpoints were offered. Organ and body weights are often among the
least sensitive endpoints for chemicals that exert specific effects on physiological or
developmental systems. Nevertheless, in the absence of information with which to select more
specific endpoints (e.g., biochemical or histological changes), body and organ weight changes
are likely to be indicative of toxicity.
Many members of the Panel also believed that additional endpoints should be included in
the risk assessment, although recognizing that this may not be possible for some endpoints
20
-------
because of the absence of sufficient information:
• Based on discussion in response to Question 2, PFOA has the potential to produce
carcinogenic effects in humans and therefore additional cancer endpoints (liver,
testicular, pancreatic acinar, and mammary) should be included in the risk assessment.
• Liver histopathology, other than cancer, should be included as an endpoint since it could
not be concluded with confidence that all liver effects are mediated through PPAR-alpha
agonism (see response to Question 1), and therefore liver histopathology from PFOA
may be relevant to humans. The Panel recognized that interpretation of some liver
changes as adverse effects may not always be apparent (e.g., liver enlargement with no
other pathology), and this should be discussed in the risk assessment.
• Immunotoxicity should be considered as an endpoint addressed quantitatively in the risk
assessment. The Panel recognizes that in order to be incorporated into the risk
assessment, immunotoxicity data will need to be derived in rats, or approaches developed
for the estimation of serum PFOA concentrations in mice.
• Consideration should be given to addition of endpoints related to lipid metabolism (see
comments under response to Issue 3, question #5).
In the view of a few Panel members who believed that data for PFOA was consistent
with the descriptor "suggestive" rather than "likely" to be carcinogenic in humans, the cancer
endpoints noted above were not recommended for inclusion in the risk assessment as they were
deemed not to be independent PFOA effects.
The Panel agreed that additional research on PFOA was needed, and encouraged studies
in the areas noted below; the Panel also encouraged exploration of methods to identify critical
targets for PFOA beyond a PPAR-alpha MOA:
Other than ataxia, no data on neurotoxicity endpoints for PFOA are available.
Neurotoxicity endpoints, including behavioral measures, should be added to the risk
assessment. PPAR-alpha receptors, as well as other PPAR receptors, are found in both
neurons and glia, and are found in multiple brain regions (frontal cortex, basal ganglia,
reticular formation). It has been proposed that, in addition to their roles common to other
tissues, these receptors in brain may have specific functions in the regulation of genes
involved in neurotransmission (Moreno etal., 2004). This would further suggest their
importance in behavioral function.
The two-generation rat study (Butenhoff et a/., 2004) involved both perinatal PFOA
exposure and direct PFOA dosing of the Fl offspring beginning at weaning. The Panel
recognized that this approach is consistent with U.S. EPA guidance regarding
developmental studies. However, consideration should be given to using developmental
endpoints in Fl generation animals prior to initiation of direct dosing so that potential
effects associated with perinatal exposure can be more clearly identified.
Current data suggest that PFOA might produce hormonal effects that would be important
to consider, but in most cases the significance of the observations are unclear. For
example, in a 26-week study of PFOA administration to cynomolgus monkeys, serum
TSH was slightly but significantly elevated in all treatment groups on the final day of the
21
-------
experiment, and serum thyroxin was slightly but significantly reduced (Butenhoff, 2002).
It is not clear whether these observations are physiologically meaningful or whether they
were strictly dependent upon treatment per se, since hormone levels appeared to change
in the control animals during the course of the experiment as well. The analysis of
Butenhoff data did not include a repeated measures ANOVA, so interactions were never
pursued. Even that, however, would not have revealed why hormone levels changed over
the course of the experiment in control animals. One study reported PFOA-induced
decreases in pituitary weight in the Fl generation female rats, but the functional
significance of this observation is unclear. Overall, the Panel thought that Margins of
Exposure (MOEs) should not be calculated for hormonal endpoints at this time, but that
additional research to clarify the hormonal effects of PFOA should be encouraged.
• Adult male rats exhibited a much slower elimination of the ammonium salt of PFOA, i.e.,
ammonium perfluorooctanoate (APFO or C8), than did females. This appears to be due
to gonadal hormones inasmuch as castration increased APFO elimination and
testosterone replacement returned the elimination rate toward normal levels. Importantly,
renal elimination was blocked by probenecid, a selective antagonist of organic anion
transporters (OATPs) (Shitara, 2004). Thus, gender differences in renal OATPs may
account for the gender differences in renal clearance of APFO. Likewise, the slower
clearance of APFO in males may account for the observation that lower doses of APFO
produced adverse effects in males compared to females. For example, the NOAEL for
APFO in a 13-week study of male CD rats was 0.56 mg/kg-day whereas females
exhibited a NOAEL of 22.4 mg/kg-day. These results suggest that specific organs (e.g.,
liver, kidney, and perhaps adrenals) are targets of APFO because of the pattern of
expression of the OATPs that transport it across the cells (OATP 1-4 in rat). Research to
identify the relationship between OATP and PFOA toxicity may offer insight into the
most important targets for PFOA effects and the best endpoints for evaluation.
Question 4. Given the available data to date, please comment on the most appropriate
lifestage/gender/species for assessing human risk.
In general, there was consensus that at this stage in the risk assessment process, no
lifestage/gender/species should be excluded from consideration in predicting human risk.
Moreover, absence of information identifying a "critical period" in development during which
PFOA may exert adverse effects on development requires inclusion of all life stages, including
fetal development. Biomonitoring data indicate children and adults alike exhibit measurable
levels of PFOA in serum, and the half-life of PFOA appears to be around 4 years. Therefore,
there is no reason to exclude any developmental period from examination. Finally, the inclusion
of data on internal dose is an important element of the dataset for PFOA which should enlighten
concerns about the use of female rats, discussed below.
Two considerations arose in evaluating the current dataset for use in assessing human
risk. In general, the EPA provides a margin of safety by using exposure values that produce the
lowest margin of exposure based on observed effect levels, usually NOAELS or LOAELs,
including those in animal models. There was general agreement that the most appropriate
criterion for assessing human risk is one that produces the lowest margin of exposure (e.g., 90l ,
95l, or 99* percentile) based on a LOAEL in animal models. The second consideration related
22
-------
to the appropriateness of the non-human primate as a model, generally considered to be most
comparable to humans.
With respect to the first, the emphasis is on having data based on the internal dose
relationships (i.e., serum PFOA levels) in some of the animal studies so that interspecies
differences in metabolism and clearance are taken into account. In addition, these data also
allow using both males and females despite a dramatic difference in clearance rate. Also,
considering empirical measures of exposures in children and adults, this view emphasizes a
concern that both developmental and adult endpoints be captured, and these endpoints have not
been evaluated in non-human primates. Therefore, the findings from adult male rats including
the 13 week study by Goldenthal (1978) in which liver weight was significantly increased, and
the Fl males in the two-generation reproductive toxicity study (Butenhoff et a/., 2004) in which
body weight was reduced should be considered in further analysis of human health risk.
The second view emphasizes the biological similarities between nonhuman primates and
humans for risk assessment. This is particularly important in the case of PFOA because there are
a number of issues with a rodent model for PFOA exposure; e.g., sexual dimorphism with
respect to elimination of PFOA, and differences in sensitivity to PPAR-alpha signaling between
rat and human. However, monkeys also exhibit a different half-life of PFOA than do humans,
and information about the potential toxicity of PFOA on non-human primates are derived
primarily from adults.
Question 5. Please comment on the appropriateness of the available animal models. Please
comment on whether additional animal models should be investigated, and if so, what
information would better enable us to ascertain potential human risks.
The available animal models are useful, but all are considered uncertain matches for
humans with respect to PFOA toxicity. Thus, most Panel members supported continued use of
multiple animal models and the need for additional models. As previously noted, some
responses to PFOA may occur via modes of action not related to PPAR-alpha agonism. Without
knowing how these PPAR-alpha independent effects are mediated, the ability to identify the
specific animal models that would be most useful is limited. Some Panel members suggested the
development and use of additional animal models without PPAR-alpha, such as transgenic or
siRNA rats. Use of these animal models would be of assistance for more clearly identifying
PPAR-alpha independent effects of PFOA.
Overall, the Panel thought that results obtained in models using female rats were
informative because they currently provide the only indication of potential effects on endpoints
specific to females (e.g., reproduction and developmental effects, mammary tumors). However,
some concerns were noted regarding the difference in toxicokinetics of PFOA in female versus
male rats and monkeys, with females exhibiting more rapid excretion.
As part of a discussion of additional sources of information and animal models to help ascertain
potential human risks, the Panel considered observations from studies in humans. The following specific
observations in regard to inclusion of the epidemiologic data as informative regarding endpoints were
shared by most Panel members:
23
-------
• The PFOA Draft Risk Assessment did not use the occupational biomonitoring data
because "data are not available for specific occupational exposures." The Panel points
out that neither are data available for "specific environmental exposures." The further
claim that information on "critical factors" like gender, sampling methods and occupation
are not available for the worker populations does not seem relevant. Gender differences
are not considered in the PFOA Draft Risk Assessment document's MOE calculations
(combined male and female values are used) because, unlike in rats, there are no apparent
gender differences in PFOA elimination in humans, at least in the sparse published data
available at the time of this review.
• In the PFOA Draft Risk Assessment document, limitations of epidemiological studies are
emphasized, while some associations (cerebrovascular disease, triglycerides and cholesterol) are
deemed less convincing, based on small numbers or inconsistencies in the results. It is
undeniable that the epidemiology studies, like the toxicological ones, have some limitations, not
the least of which are uncertainties regarding exposure. However, there is little doubt that these
workers are more highly exposed than the general population. A special strength of
epidemiological studies is that no cross-species extrapolation is needed; humans are the model. It
is also true that there may be multiple exposures in the occupational studies, but this fact alone
cannot disqualify them without simultaneously disqualifying virtually all epidemiological
studies, which doesn't seem appropriate. If the question addressed by an MOE analysis is "how
far" are actual human exposures from exposures that are associated with a health effect, any
health effect in the epidemiological studies imply the answer is "zero distance," regardless of the
actual serum values.
While conceding the small numbers and short follow-up in the available epidemiological
studies make the positive results less than compelling, they are not, conversely, reassuring. The
evidence showing increases in cholesterol and triglyceride values in worker cohorts suggest a
possibility of increased risk of cerebrovascular disease mortality.
In responding to charge question #3, therefore, many Panel members shared the view that
human cancer and alterations in lipid metabolism data be included in the relevant endpoints for
consideration. This implies that the rich data base of occupational exposures be added to the
occupational biomonitoring data to be considered. They are not now included in the PFOA Draft
Risk Assessment document because the worker epidemiological studies were not considered
suitable for quantitative risk assessment.
A contrasting conclusion reached by some Panel members was that the peer-reviewed human
data on health effects of PFOA were equivocal, thus there was not consensus that endpoints suggested
by some epidemiologic studies should be used as endpoints in the risk assessment.
Issue 4: Risk Assessment Approach
Issue 4a: Pharmacokinetic Modeling and Use of AUC as a Measure of Internal Dose
Question 6. Please comment on use of the one-compartment pharmacokinetic model.
24
-------
The purpose of developing a mathematical model to fit the serum PFOA time course data
from the single dose rat oral dosing studies in the PFOA Draft Risk Assessment document was to
estimate the AUC and Cmax values during the longer term toxicology studies with daily dosing.
The internal dose metrics calculated with this model were then compared with human serum
concentrations to establish an MOE. The equations used to describe these data sets are the same
as those usually employed in one-compartment models for uptake and elimination and were
referred to throughout the draft document as a one-compartment model.
However, the Panel was concerned that using the "one-compartment" nomenclature
without caveats and qualifications will give readers of the Draft Risk Assessment Document the
impression that PFOA pharmacokinetics follow a one-compartment description when in fact
they are much more complex. In a one-compartment model, the chemical distributes evenly
throughout a volume of distribution that is itself in rapid equilibrium with blood. Elimination
kinetics are first-order and do not change with dose level or with time. However, the data
indicate that it is clearly inappropriate to describe the observed kinetics of PFOA in rats or
monkeys as following a simple one-compartment model. The relatively complex
pharmacokinetic behavior of PFOA is reflected in several of the pharmacokinetic data sets. For
example, elimination from blood after iv dosing and tissue distribution kinetics after oral dosing
are poorly characterized by the one-compartment model. In both rats and monkeys, blood levels
are related in a complex manner to dosage and the duration of treatment.
Although the one-compartment model is not appropriate, the empirical model used in the
document and referred to as a 'one compartment model' is adequate for predicting blood levels
resulting from repeated dosing. However, the document needs to make it clear that the fitting
procedure is specific to this limited data set and useful for this one application. It is strongly
recommended that the terminology 'one-compartment' model should be stricken from the
document unless carefully qualified.
Question 7. Please comment on the use of the AUC as a measure of internal dose for rats and
humans for calculation of the MOE.
Calculating the 'blood' AUC (as a measure of average daily concentration of PFOA) is
an appropriate method of estimating the internal dose, although it is not the only possible
measure. In the absence of clear understanding of modes of action (MO A), it is also possible that
the Cmax, the integrated dose above a minimum concentration, or some other quantity may be a
more plausible measure of internal dose. For example, if the MOA was receptor-based, as might
be expected for interactions of PFOA with PPAR or other receptor proteins, one of these other
measures of dose might also be appropriate. These alternatives include receptor occupancy or
the concentration above some minimum concentration (Cmin) where Cmin is the concentration
required to initiate activation of the receptor-mediated signaling pathway. In this latter case, the
MOE would be based on the integral of (Ct-Cmin) rather than just the integral of concentration
In light of these other possible internal dose measures, the EPA document would be
strengthened if a clear rationale for the choice of the AUC were included. Since the inclusion of
this explanation may involve a detailed discussion of toxicokinetic and toxicodynamic issues,
25
-------
such a discussion would best be included as an appendix. While the report does provide an
example of how the MOE differs when based on the Cmax as compared to the AUC, it would be
helpful if the impact on the magnitude of the MOE of using each of these other internal dose
measures was explored in more detail. Calculations of MOEs based on these other measures
would provide a better idea of the extent of possible variability introduced by different internal
dose measures that may reflect a variety of possible MO As.
When estimating an AUC, it is important to note the sample that is being analyzed in the
various studies. AUCs can be calculated for serum, plasma or whole blood. These are very
different biological matrices. The document should clearly specify the biological media
measured in each study in which AUCs are reported.
Another issue to be considered is that the analyses of serum time course in the document
are based on the assumption that the analyte in serum is in the same form and the proportion of
free compound in blood is constant throughout the period of observation. This assumption does
not always hold true. For example, with some siloxanes, the blood concentrations during and
after inhalation exposure are primarily free siloxanes that are available for exhalation and
metabolism. After a period of time in the body, the siloxanes in blood appear to reside in the
lipid pool within the blood and although they are easily analyzed are no longer available for
these other clearance processes (see Andersen et a/., 2001; Reddy et a/., 2003). A situation
where the PFOA in blood at much longer times after exposure is in a distinctly different
biological pool would lead to difficulties in comparing rat AUC and human AUC values to
obtain a MOE. Interspecies differences in PFOA free fraction in plasma may also complicate
the comparison of AUC values to obtain a MOE.
The direct use of internal measures of dose by US EPA in this document represents a
promising and relatively innovative approach for risk assessments of environmental compounds
compared to the more usual practice based on comparing daily dose rates by various routes of
administration. This new approach reduces the need to include uncertainties introduced by the
use of administered or ambient doses as measures of exposure. This type of risk assessment
methodology is likely to become much more widespread due to advances in analytical chemistry
and the rapid expansion of human biomonitoring activities throughout the world. Because this
risk assessment is likely to serve as a prototype for future tissue-dose based risk assessments,
some important issues raised by this tissue-dose based approach need to be more fully
considered and adequately contrasted with the more common assessments based on comparisons
of administered doses.
To address these issues, the EPA should be encouraged to develop documentation
explaining their rationale guiding the use of these tissue-dose based risk assessment approaches.
Such documentation should compare current methods based on daily intakes with these
alternative, 'tissue-based' approaches to more explicitly address the risk characterization issues
that arise in moving to this new approach. Such a document might include discussion of (1) the
choices of tissue dose measures based on serum concentrations and the risk implications of each
choice; (2) the impacts of utilizing direct measures of tissue dose on the magnitudes of
interspecies and interindividual uncertainty factors; (3) the implications of different metrics for
characterizing distributions of human tissue dose measures on estimates of MOEs; and (4) the
26
-------
importance of routine analysis of appropriate blood concentrations; e.g., serum, plasma, etc. in
providing the information for most appropriately applying the tissue dose approach.
Issue 4b: Cross Species Extrapolation
Question 8. Please comment on the need to use or modify the default value of 10 for cross
species extrapolation given the pharmacokinetic analysis.
The internal dose analysis used in this document is considered by the Panel to be a
significant step toward reducing uncertainty related to cross species extrapolation. Although
reduced, however, cross species toxicokinetic uncertainty is not eliminated. Sources of
uncertainty remain, including the lack of information about the measured internal dose that best
predicts adverse effect in human and other species, and the bias inherent in
measurement/modeling error. While it is difficult to assign a quantitative value to the magnitude
of this uncertainty reduction, it can be stated that the toxicokinetic uncertainty value for PFOA
would fall within the range of one to three, based on the customary scale of a value of 3 for each
aspect of cross species extrapolation, pharmacokinetics and pharmacodynamics.
Pharmacodynamics aspects of PFOA cross species scaling are not addressed in a sufficient
manner to alleviate the application of some type of uncertainty factor/s (addressing
toxicodynamic equivalence across species). The assumption that PFOA serum levels are at
steady state in children 2-12 years of age has not been tested and may not be valid. The
additional complexity of multiple C-8 environmental exposures in humans versus animal
experiments involving exposures to PFOA specifically, further clouds the overall uncertainty
analysis.
While the pharmacokinetic modeling that is presented in the PFOA risk assessment is
useful, a more comprehensive way to account for biological processes that determine internal
dose is with the development of a physiologically based toxicokinetic model. The Panel
encourages EPA to continue to develop toxicokinetic models as they can improve dose-response
assessment by revealing and describing nonlinear relationships between applied and internal
dose.
A discussion of confidence should always accompany the presentation of model results
and include consideration of model validation and sensitivity analysis, stressing the predictive
performance of the model. Toxicokinetic modeling results may be presented as the preferred
method of estimating equivalent human doses or in parallel with default procedures (see Section
3.1.3), depending on the confidence in the modeling.
Standard cross-species scaling procedures are available when the data are not sufficient
to support a toxicokinetic model or when the purpose of the assessment does not warrant
developing one. The aim is to define dose levels for humans and animals that are expected to
produce the same degree of effect (U.S. EPA, 1992b), taking into account differences in scale
between test animals and humans in size and in lifespan. It is useful to recognize two
components of this equivalence: toxicokinetic equivalence, which determines administered doses
in animals, and humans that yield equal tissue doses, and toxicodynamic equivalence, which
determines tissue doses in animals and humans that yield equal lifetime risks (U.S. EPA, 1992b).
27
-------
It is equally important to note that pharmacodynamics aspects of PFOA cross species
scaling are not addressed in a sufficient manner to alleviate the application of some type of
uncertainty factor/s (addressing toxicodynamic equivalence across species). These factors may
be different for each species extrapolated. By the language used in the U.S. EPA Cancer
Guidelines, it seems evident that standard default values were never intended to act as complex
scaling factors when internal doses in human serum are compared to animal internal doses across
multiple pathways, genders, steady-state serum levels with long human half-lives and/or
different life stages.
In the case of PFOA the strong reliance on LOAEL-driven MOE calculations instead of
more appropriate Bench Mark Dose methodologies, and the absence of probabilistic approaches
to assessing human exposure and risk, was considered by most Panel members as another source
of dynamic uncertainty.
The use of an uncertainty factor/s based on data variability may be an alternative to the
traditional scaling factors given the kinetics analysis strength and in light of the larger concerns
of overall uncertainties related to dynamic analysis (as reflected in the MOE approach). This
may prove more productive when comparing relatively robust toxicokinetic dose response
models involving serum concentrations and/or their surrogates.
In conclusion, whereas toxicokinetic uncertainty is possibly reduced in this analysis, care
must be exercised in the estimation of the overall cross species uncertainty, which further
dynamic analyses may show falls below or above 10.
Issue 4c: Human Biomonitoring Data
Question 9. Please comment on the adequacy of the human exposure data for use in calculating a
MOE.
Full agreement was not reached by Panel members with respect to the utility of the human
biomonitoring data for the calculation of the MOE. Most Panel members expressed the view that the
human exposure information should be utilized in these calculations while a few Panel members
believed that these data were equivocal and thus not appropriate for the MOE calculations.
Populations used for MOE calculations
In addition to the occupational biomonitoring data, the PFOA Draft Risk Assessment document
described three separate study populations from the United States with available individual serum PFOA
levels. One consists of samples from six American Red Cross blood banks, another from a study of
Streptococcal A infection in children, and a third from elderly volunteers from Seattle who participated
in a study of cognitive function. Only the first two study populations were used in calculating the MOE
for the risk assessment.
A question was raised about reliance on the female blood bank donor population for calculating
prenatal MOEs, because the influence of pregnancy on serum PFOA levels is not known. Likewise, use
28
-------
of the samples obtained from the children for the age span of 2-12 years for the postweaning period
MOE may not be appropriate because the assumption of steady state used in the MOE analyses may not
be valid for children. Half-life issues in humans, especially when considering the impact of age at
exposure (or the critical windows of exposure model), contribute to the questions about adequacy of
using these samples (Pryor et a/., 2000; Selevan et a/., 2000; Sweeney et a/., 2001). Thus, there are a
variety of possible problems with using these data to represent the general population, but the Panel
agreed that they were likely to be reasonably representative and are better than data often available for
exercises of this nature.
It was suggested that biomonitoring data in highly exposed groups (occupational and
environmental) be included in the MOE analyses. It was noted that the existence, size and levels of
exposures of populations which may differ from those studied has yet to be fully determined. Until this
has been determined, it is not clear what percent of the general population is covered by the MOEs that
have been calculated. Thus, the appropriateness of relying solely on the blood bank and pediatric
samples for MOE calculations depends strongly on the purpose of the MOE exercise, i.e., whether it is
to assess the likelihood that any people could be suffering health effects from PFOA or only the
"general population." If the latter case, the biomonitoring data that were used may be appropriate, but
the sizes of more highly exposed populations remains unknown and this should be acknowledged.
A few members of the Panel held the view that the human data were equivocal, based on the
likely multiplicity of exposures of occupational groups, and thus should not be included in the draft
PFOA risk assessment for MOE calculations for the general population.
Depiction of the biomonitoring data
The tables and summary statistics that were used in the draft PFOA risk assessment are
somewhat uninformative and unsatisfactory. It is difficult to determine the distribution of population
exposures from these given the method of data presentation. A preferable approach would be to use a
non-parametric data-driven method to display the data (including the occupational data), using, for
example, some density estimation procedure or smoother. Inclusion of the worker data in these displays
would allow a clearer understanding of the relationships. Even side-by-side box plots would have been
preferable to what was provided. This requires having access to the raw data, however. Because such a
request is easy to satisfy, the Panel recommends that EPA provide more informative displays of the
biomonitoring data.
Appropriate summary measures for MOE calculations
At least three summary statistics are mentioned in the Draft, the geometric mean, the arithmetic
mean, and the 90 percentile.
The rationale for the use of "means" should be explained, especially the use of the geometric
means which seems the least satisfactory, since it is about 25% lower than the arithmetic mean in these
data. Use of a geometric mean for population inference (to transform a lognormal to a normal
distribution, for example) might be justified, but not for the purpose of calculating an MOE. Moreover,
the distribution does not even seem to be lognormal, as judged by the Shapiro-Wilk test. The idea that a
few censored data points are responsible for failing this test seems highly unlikely, and could have been
accounted for in the test itself.
29
-------
Means of any kind don't seem appropriate for a ubiquitous exposure. Of the three choices, the
90 percentile seems the most appropriate in that case. At least one Panel member wondered why some
even higher percentile, say 95 or even a maximum value wouldn't be better. The maximum value in
any of the samples is still an underestimate of the maximum value in the population. Even the upper
99.99 percentile represents 30,000 people in the US.
In summary, the Panel finds that:
• Use of the blood donor and pediatric biomonitoring data may be acceptable if the purpose is to
assess whether there is a potential health effect to the "general" population, although there is
some question as to the size of other non-occupational populations that might be more highly
exposed and the assumption that PFOA serum levels are at steady state may not be valid for
fetuses, neonates or children;
• Most Panel members believed that occupational biomonitoring data should be included in the
MOE calculations, especially regarding additional endpoints such as alterations in lipid
metabolism;
• A few members did not favor this inclusion based on the equivocal findings;
• The biomonitoring data should be presented in a more informative manner, for example, through
side-by-side box plots or some other method that would better depict the range of values and
distributions; and
• Thought should be given to what appropriate summary statistic for the biomonitoring
datasets used in MOE calculations should be. Some panelists believe that 90l percentiles
or higher, perhaps even maximum values might be most appropriate. In any event,
justification for use of the chosen summary measure should be made and related to the
explicit objective of the MOE analysis.
30
-------
REFERENCES
Andersen ME, Sarangapani, Reitz RH, Gallavan RH, Dobrev ID and Plotzke KP. 2001.
Physiological Modeling Reveals Novel Pharmacokinetic behavior of Inhaled
Octamethylcyclotetrasiloxane. Toxicological Sciences, 60:214-231.
Biegel LB, Hurtt ME, Frame SR, O'Connor JC, and Cook JC. 2001. Mechanisms of extrahepatic
tumor induction by peroxisome proliferators in male CD rats. Toxicological Sciences, 60:44-55.
Butenhoff J, Costa G, Elcombe C, Farrar D, Hansen K, Iwai H, Jung R, Kennedy G Jr., Lieder P,
Olsen G and Thomford P. 2002. Toxicity of ammonium perfluorooctanoate in male cynomolgus
monkeys after oral dosing for 6 months. Toxicological Sciences, 69: 244-257.
Butenhoff T, Kennedy GL Jr., Frame SR, O'Connor JC and York RG. 2004. The reproductive
toxicology of ammonium perfluorooctanoate (APFO) in the rat. Toxicology, 196:95-116.
Chandra M, Riley MG, and Johnson DE. 1992. Spontaneous neoplasms in aged Sprague-Dawley
rats. Archives of Toxicology, 66:496-502.
Goldenthal EL Final report, ninety day subacute rat toxicity study on Flourad® Fluorochemical
FC-143, International Research and Development Corporation, Study No. 137-089, 3M
Reference No. T-3141, November 6, 1978. US EPA AR226-0441.
Hasmall S, James N, Hedley K, Olsen K, Roberts R. 2001. Mouse hepatocyte response to
peroxisome proliferators: dependency on hepatic nonparenchymal cells and peroxisome
proliferator activated receptor alpha. Archives of Toxicology, 75:357-61,
Klaunig JE, Babich MA, Baetcke KP, Cook JC, Corton JC, David RM, DeLuca JG, Lai DY,
McKee RH, Peters JM, Roberts RA, and Fenner-Crisp PA. 2003. PPAR-alpha agonist-induced
rodent tumors: modes of action and human relevance. Critical Reviews in Toxicology, 33:655-
780,
Moreno, S., Farioli-Vecchioli, and Ceru, M.P. 2004. Immunolocalization of peroxisome
proliferators activated receptors and retinoid X receptors in the adult rat CNS. Neuroscience,
123:131-145.
Palazzo MJ. 1993. 13-Week dietary toxicity study with T-5180 ammonium perfluorooctanoate
(CAS No. 3825-26-1) in male rats. Laboratory Project Identification HWI 6329-100. Hazelton
Wisconsin, Inc. US EPA AR226-0449.
Parzefall W, Berger W, Kainzbauer E, Teufelhofer O, Schulte-Hermann R, and Thurman RG.
2001. Peroxisome proliferators do not increase DNA synthesis in purified rat hepatocytes.
Carcinogenesis, 22:519-23,
31
-------
Peters JM, Cattley RC, and Gonzalez FJ. 1997. Role of PPAR-alpha in the mechanism of action
of the nongenotoxic carcinogen and peroxisome proliferatorWy-14,643. Carcinogenesis,
18:2029-2033.
Peters JM, Rusyn I, Rose ML, Gonzalez FJ, Thurman RG. 2000. Peroxisome proliferator-
activated receptor alpha is restricted to hepatic parenchymal cells, not Kupffer cells: implications
for the mechanism of action of peroxisome proliferators in hepatocarcinogenesis.
Carcinogenesis, 21:823-826.
Pry or JL, Hughes C, Foster W, Hales BF and Robaire B. 2000. Critical windows of exposure for
children's health: the reproductive system in animals and humans. Environmental Health
Perspectives, 108 (Suppl 3):491-503.
Reddy MB, Dobrev ID, Plotzke KP, Andersen ME, Reitz RH, Morrow P, and Utell M. 2003. A
Physiologically Based Pharmacokinetic model for inhalation of octamethylcyclotetrasiloxane
(D4) in humans during rest and exercise. lexicological.Sciences, 72:3-18.
Rose ML, Rusyn I, Bojes HK, Germolec DR, Luster M, and Thurman RG. 1999. Role of
Kupffer cells in peroxisome proliferator-induced hepatocyte proliferation. Drug Metabolism
Reviews, 31:87-116.
SAP minutes 2003-05, USEPA, December 9, 2003, A Set of Scientific Issues Being Considered
by the Environmental Protection Agency Regarding: Proposed Science Policy: PPAR-alpha
Agonist-Mediated Hepatocarcinogenesis in Rodents and Relevance to Human Health Risk
Assessment.
Selevan SG, Kimmel CA and Mendola P. 2000. Identifying critical windows of exposure for
children's health. Environmental Health Perspectives, 108 (Suppl 3):451-455.
Sibinski LJ. 1987. Two year oral (diet) toxicity/carcinogenicity study of fluorochemical FC-143
in rats. Vol. 1-4, 3M Company/Riker Exp. No.0281CR0012.
Sweeney AM, Symanski E, Burau KD, Kim YJ, Humphrey HE and Smithci MA. 2001.
Changes in serum PBB and PCB levels over time among women of varying ages at exposure.
Environmental Research, 86:128-139.
US EPA. 1999. Draft Revised Guidelines for Carcinogen Risk Assessment (External Draft, July
1999). U.S. Environmental Protection Agency, Risk Assessment Forum, Washington, DC.
US EPA. 2003. Draft final guidelines for carcinogen risk assessment. EPA/630/P-03/001A.
NCEA-F-0644A. U.S. Environmental Protection Agency, Risk Assessment Forum, Washington,
DC.
US EPA. 2005. Guidelines for Carcinogen Risk Assessment. EPA/630/P-03/001F. U.S.
Environmental Protection Agency, Risk Assessment Forum, Washington, DC.
32
-------
Yang Q, Xie Y, Alexson SE, Nelson BD, and DePierre JW. 2002. Involvement of the
peroxisome proliferator-activated receptor alpha in the immunomodulation caused by
peroxisome proliferators in mice. Biochemical Pharmacology, 63:1893-1900.
33
------- |