UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON D.C. 20460
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
December 19, 2008
EPA-SAB-09-009
The Honorable Stephen L. Johnson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, N.W.
Washington, DC 20460
Subject: Review of EPA' s draft entitled "lexicological Review of Acrylamide"
Dear Administrator Johnson:
In response to a request from EPA's Office of Research and Development (ORD), the
Science Advisory Board (SAB) convened an expert panel to conduct a peer review of EPA's
draft Integrated Risk Information System (IRIS) assessment entitled, "Toxicologic Review of
Acrylamide" (December 2007). This draft document updates EPA's current evaluation of the
potential health effects of acrylamide.
The SAB was asked to comment on the hazard characterization and dose-response
assessment of acrylamide, including the Agency's selection of the most sensitive non-cancer
health endpoint, the use of a physiologically -based toxicokinetic (PBTK) model, the derivation
of a proposed oral reference dose (RfD), an inhalation reference concentration (RfC) for non-
cancer endpoints, as well as the cancer descriptor, oral slope factor, and inhalation unit risk for
acrylamide. The SAB Panel's report contains a number of recommendations that are aimed at
making the assessment more transparent and improving the scientific bases for the conclusions
presented. While a more detailed description of the technical recommendations is contained in
the body of the report, the key points and recommendations are highlighted below:
• The Panel agreed with the EPA's conclusion that based on the existing toxicity data base for
acrylamide, neurotoxicity appears to be the most sensitive non-cancer endpoint, and
therefore, the most appropriate for developing the RfD and RfC for non-cancer health effects.
The Panel was concerned, however, that the RfD/RfC was derived from studies which were
primarily designed as cancer bioassays and therefore did not include the most sensitive
measures of neurotoxicity.
• The Panel believed that the use of the benchmark dose methodology in this assessment was
scientifically supportable, given the nature and robustness of the data sets available on the
endpoint of concern.
-------
• The Panel supported the Agency's conclusions that exposure to acrylamide in animals leads
to heritable gene mutations and that these results indicate that it may also pose a hazard to
humans. The Panel further supported the Agency's conclusions that the available data on
heritable gene mutations are not adequate to conduct a robust assessment of this endpoint at
this time. The Panel urges further research on acrylamide-induced heritable germ cell
mutations, given the serious nature of such effects.
• The Panel concluded that the rationale and justification for acrylamide being a "likely human
carcinogen" via a mutagenic mechanism was well described and the conclusion was
scientifically supportable, although it should be further elaborated. While the Panel did
consider available information regarding non-mutagenic MOAs (e.g., hormonal) as presented
by public commenters, they did not find it to be compelling.
• The Panel encouraged the Agency to use the two main chronic bioassays in rats for deriving
the oral cancer slope factor and include an in depth discussion of the strengths and limitations
of both studies.
• The Panel commends EPA for using the PBTK model for developing the RfD, RfC and
cancer slope factor for acrylamide. The Panel did however provide some recommendations
to the Agency for improving the model as they revise their draft document. The Panel notes
that the use of internal dose metrics combined with a fairly robust understanding of the
mechanism of action may replace the use of the default interspecies factor for toxicokinetic
differences. Internal dose may be derived using the PBTK model or through application of
other pharmacokinetic approaches indicated in the Panel report.
• The Panel agreed with the use of PBTK modeling to conduct dose-route extrapolation and
commended the EPA for using the PBTK model to fill the gap created due to the absence of
robust animal toxicology studies that would support the development of an RfC. In
estimating the cancer slope factor and unit risk, human-rodent differences in toxicokinetics
were taken into account with the PBTK model, whereas toxicodynamic differences were not,
but should be, through the application of a standard factor.
• Finally, the Panel agreed that the use of the age-dependent adjustment factors to adjust the
unit risk for early life exposure is well justified and transparently and objectively described.
The SAB appreciates the opportunity to provide EPA with advice on this important
subject. Although cognizant of additional acrylamide studies currently underway, the SAB urges
EPA to move expeditiously to finalize the IRIS document on acrylamide as the Agency considers
relevant data which has been published since the release of the draft assessment. We look
forward to receiving the Agency's response.
Sincerely,
/Signed/
Dr. Deborah L. Swackhamer, Chair
EPA Science Advisory Board
/Signed/
Dr. Deborah Cory-Slechta, Chair
SAB Acrylamide Review Panel
-------
NOTICE
This report has been written as part of the activities of the EPA Science Advisory Board,
a public advisory committee providing extramural scientific information and advice to the
Administrator and other officials of the Environmental Protection Agency. The Board 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 or commercial products constitute a recommendation for use.
Reports of the EPA Science Advisory Board are posted on the EPA Web site at:
http://www.epa.gov/sab.
-------
U.S. Environmental Protection Agency
Science Advisory Board
Acrylamide Review Panel
CHAIR
Dr. Deborah Cory-Slechta, Professor, Department of Environmental Medicine, School of
Medicine and Dentistry, University of Rochester, Rochester, NY
PANEL MEMBERS
Dr. Alfred Branen, Associate Vice President, University of Idaho, Coeur d'Alene, ID
Dr. Daniel R. Doerge, Research Chemist, National Center for Toxicological Research, Food and
Drug Administration, Jefferson, AR
Dr. James S. Felton, Senior Biomedical Scientist, University of California, Lawrence
Livermore National Laboratory, Livermore, CA
Dr Timothy Fennell, Senior Research Chemist, Drug Metabolism and Pharmacokinetics, RTI
International, Research Triangle Park, NC
Dr. Penelope Fenner-Crisp, Independent Consultant, North Garden, VA
Dr. Jeffrey Fisher, Professor, Department Environmental Health Science, University of
Georgia, Athens, GA
Mr. Sean Hays, President, Summit Toxicology, Allenspark, CO
Dr. Steven Heeringa, Director, Division of Surveys and Technologies, Institute for Social
Research, University of Michigan, Ann Arbor, MI
Dr. Richard M. LoPachin, Professor of Anesthesiology, Department of Anesthesiology,
Montefiore Medical Center, Albert Einstein College of Medicine, Bronx, NY
Dr. Lorelei Mucci, Assistant Professor, Harvard Medical School, Channing Laboratory, Boston,
MA
Dr. Jerry M. Rice, Distinguished Professor, Department of Oncology, Lombardi Cancer Center,
Box 571465, Georgetown University Medical Center, Washington, DC
Dr. Dale Sickles, Professor and Vice-Chair, Department of Cellular Biology and Anatomy,
Medical College of Georgia, Augusta, GA
-------
Dr. Gina Solomon, Senior Scientist, Health and Environment Program, Natural Resources
Defense Council, San Francisco, CA
Dr. Anne Sweeney, Professor of Epidemiology, Commonwealth Medical Education, The
Commonwealth Medical College, Scranton, PA
Dr. Lauren Zeise, Chief, Reproductive and Cancer Hazard Assessment Branch, Office of
Environmental Health Hazard Assessment, California Environmental Protection Agency,
Oakland, CA
SCIENCE ADVISORY BOARD STAFF
Dr. Suhair Shallal, Designated Federal Officer, EPA Science Advisory Board, Science
Advisory Board Staff Office, Washington, DC
-------
U.S. Environmental Protection Agency
Science Advisory Board
CHAIR
Dr. Deborah L. Swackhamer, Professor, Co-Director of the Water Resources Center,
University of Minnesota School of Public, St. Paul, MN
SAB MEMBERS
Dr. Gregory Biddinger, Coordinator, Natural Land Management Programs, Toxicology
and Environmental Sciences, ExxonMobil Biomedical Sciences, Inc., Houston, TX
Dr. Thomas Burke, Professor, Department of Health Policy and Management, Johns
Hopkins Bloomberg School of Public Health, Johns Hopkins University, Baltimore, MD
Dr. James Bus, Director of External Technology, Toxicology and Environmental
Research and Consulting, The Dow Chemical Company, Midland, MI
Dr. Deborah Cory-Slechta, Professor, Department of Environmental Medicine, School of
Medicine and Dentistry, University of Rochester, Rochester, NY
Dr. Maureen L. Cropper, Professor, Department of Economics, University of Maryland,
College Park, MD
Dr. Virginia Dale, Corporate Fellow, Environmental Sciences Division, Oak Ridge
National Laboratory, Oak Ridge, TN
Dr. Kenneth Dickson, Regents Professor, Department of Biological Sciences, University
of North Texas, Aubrey, TX
Dr. David A. Dzombak, Walter J. Blenko Sr. Professor of Environmental Engineering,
Department of Civil and Environmental Engineering, College of Engineering, Carnegie
Mellon University, Pittsburgh, PA
Dr. Baruch Fischhoff, Howard Heinz University Professor, Department of Social and
Decision Sciences, Department of Engineering and Public Policy, Carnegie Mellon
University, Pittsburgh, PA
Dr. James Galloway, Professor, Department of Environmental Sciences, University of
Virginia, Charlottesville, VA
Dr. James K. Hammitt, Professor, Center for Risk Analysis, Harvard University, Boston,
MA
-------
Dr. Rogene Henderson, Senior Scientist Emeritus, Lovelace Respiratory Research
Institute, Albuquerque, NM
Dr. James H. Johnson, Professor and Dean, College of Engineering, Architecture &
Computer Sciences, Howard University, Washington, DC
Dr. Bernd Kahn, Professor Emeritus and Director, Environmental Radiation Center,
Nuclear and Radiological Engineering Program, Georgia Institute of Technology, Atlanta,
GA
Dr. Agnes Kane, Professor and Chair, Department of Pathology and Laboratory Medicine,
Brown University, Providence, RI
Dr. Meryl Karol, Professor Emerita, Graduate School of Public Health, University of
Pittsburgh, Pittsburgh, PA
Dr. Catherine Kling, Professor, Department of Economics, Iowa State University, Ames,
IA
Dr. George Lambert, Associate Professor of Pediatrics, Director, Center for Childhood
Neurotoxicology, Robert Wood Johnson Medical School-UMDNJ, Belle Mead, NJ
Dr. Jill Lipoti, Director, Division of Environmental Safety and Health, New Jersey
Department of Environmental Protection, Trenton, NJ
Dr. Michael J. McFarland, Associate Professor, Department of Civil and Environmental
Engineering, Utah State University, Logan, UT
Dr. Judith L. Meyer, Distinguished Research Professor Emeritus, University of Georgia,
Lopez Island, WA
Dr. Jana Milford, Associate Professor, Department of Mechanical Engineering,
University of Colorado, Boulder, CO
Dr. M. Granger Morgan, Lord Chair Professor in Engineering, Department of
Engineering and Public Policy, Carnegie Mellon University, Pittsburgh, PA
Dr. Rebecca Parkin, Professor and Associate Dean, Environmental and Occupational
Health, School of Public Health and Health Services, The George Washington University
Medical Center, Washington, DC
Mr. David Rejeski, Director, Foresight and Governance Project, Woodrow Wilson
International Center for Scholars, Washington, DC
Dr. Stephen M. Roberts, Professor, Department of Physiological Sciences, Director,
Center for Environmental and Human Toxicology, University of Florida, Gainesville, FL
-------
Dr. Joan B. Rose, Professor and Homer Nowlin Chair for Water Research, Department of
Fisheries and Wildlife, Michigan State University, East Lansing, MI
Dr. James Sanders, Director and Professor, Skidaway Institute of Oceanography,
Savannah, GA
Dr. Jerald Schnoor, Allen S. Henry Chair Professor, Department of Civil and
Environmental Engineering, Co-Director, Center for Global and Regional Environmental
Research, University of Iowa, Iowa City, IA
Dr. Kathleen Segerson, Professor, Department of Economics, University of Connecticut,
Storrs, CT
Dr. Kristin Shrader-Frechette, O'Neil Professor of Philosophy, Department of Biological
Sciences and Philosophy Department, University of Notre Dame, Notre Dame, IN
Dr. V. Kerry Smith, W.P. Carey Professor of Economics, Department of Economics,
W.P Carey School of Business, Arizona State University, Tempe, AZ
Dr. Thomas L. Theis, Director, Institute for Environmental Science and Policy,
University of Illinois at Chicago, Chicago, IL
Dr. Valerie Thomas, Anderson Interface Associate Professor, School of Industrial and
Systems Engineering, Georgia Institute of Technology, Atlanta, GA
Dr. Barton H. (Buzz) Thompson, Jr., Robert E. Paradise Professor of Natural Resources
Law at the Stanford Law School and Director, Woods Institute for the Environment
Director, Stanford University, Stanford, CA
Dr. Robert Twiss, Professor Emeritus, University of California-Berkeley, Ross, C A
Dr. Lauren Zeise, Chief, Reproductive and Cancer Hazard Assessment Branch, Office of
Environmental Health Hazard Assessment, California Environmental Protection Agency,
Oakland, CA
LIAISON MEMBERS
Dr. Steven Heeringa, (FIFRA SAP), Research Scientist and Director, Statistical Design Group,
Institute for Social Research (ISR), University of Michigan, Ann Arbor, MI
Dr. Melanie Marty, (CHPAC Chair), Chief, Air Toxicology and Epidemiology Branch, Office of
Environmental Health Hazard Assessment, California EPA, Oakland, CA
SCIENCE ADVISORY BOARD STAFF
Mr. Thomas Miller, Designated Federal Officer, EPA SAB, Washington, DC
-------
TABLE OF CONTENTS
EXECUTIVE SUMMARY. 10
INTRODUCTION. 16
RESPONSES TO THE CHARGE QUESTIONS. 17
Question 1. 17
Question 2. 19
Question 3. 21
Question 4. 22
Question 5. 24
Question 6. 25
Question 7. 27
Question 8 28
Question 9. 34
Question 10. 34
Question 11. 35
Question 12. 36
Question 13. 36
Question 14. 38
Question 15. 39
Question 16. 40
Question 17. 41
Question 18. 42
Question 19. 44
Question 20. 46
Question 21. 47
Question 22. 50
Question 23. 50
Question 24. 51
Question 25. 52
Question 26. 53
ABREVIATIONS 54
REFERENCES 55
Appendix A Memorandum and Charge Questions 68
Appendix B Proposed Modes of Action (MOAs) for Acrylamide Neurotoxicity 76
-------
EXECUTIVE SUMMARY
This report was prepared by the Science Advisory Board (SAB) Acrylamide Review
Panel (the "Panel") in response to a request by EPA's Office of Research and Development
(ORD) to review the Draft IRIS Toxicological Review of Acrylamide (hereafter referred to as
the draft document). The Panel deliberated on the charge questions (see Appendix A) during a
March 10-11, 2008 face-to-face meeting and discussed its draft report in a subsequent conference
call on July 16, 2008. The final draft of the panel's report was reviewed and approved during a
meeting of the chartered SAB on October 28, 2008. There were 26 charge questions that focused
on the selection of the most sensitive non-cancer health endpoint, the use of a PBTK model for
the derivation of a proposed oral reference dose (RfD), an inhalation reference concentration
(RfC) for non-cancer endpoints, as well as the cancer descriptor, oral slope factor, and inhalation
unit risk for acrylamide. The Panel encourages the Agency to review relevant data which has
been published since their draft assessment was completed as they revise and finalize the IRIS
document.
This Executive Summary highlights the Panel's major findings and recommendations as
a result of their deliberations. The responses that follow represent the views of the Panel.
Selection of Endpoint
In the draft document, EPA identified neurotoxicity as the most sensitive non-cancer
effect from exposure to acrylamide. This endpoint was based on an extensive database of animal
and human studies. Other endpoints were also considered, such as reproductive toxicity and
heritable germ cell effects. The Panel agreed that based on the existing toxicity data base for
acrylamide, neurotoxicity does appear to be the most sensitive non-cancer endpoint, and
therefore, the most appropriate for developing the RfD and RfC for non-cancer effects from
exposure to acrylamide.
Mechanism of Action
The Panel discussed two hypotheses regarding the mechanism of acrylamide
neurotoxicity. The Panel did not attempt to resolve the debate over a definitive or single MOA
for neurotoxicity; however, there was agreement that the discussion of MOA is important for
10
-------
inclusion in the draft document. The Panel found the separation of the discussion of MOA(s) for
neurotoxicity in two different sections of the document confusing and recommended their
incorporation into a single section. A more complete presentation by the Panel of these MO As
has been appended (see Appendix B) to this report for EPA's consideration as they revise their
draft document.
Derivation ofRfD
EPA's proposed RfD (0.003 mg/kg-day) for acrylamide is based on a benchmark dose
analysis of the dose-response relationship for neurotoxicity in two chronic drinking water
exposure bioassays using Fischer 344 rats. Uncertainty factors and a PBTK model were used to
extrapolate the animal dose-response to a human equivalent dose-response in the derivation of
the RfD. The Panel afforded considerable discussion to the question of whether the Friedman et
al. (1995) and Johnson et al. (1986) studies were the best choices for derivation of the
quantitative RfD (and RfC). The main concerns with these studies are that they were primarily
designed as cancer bioassays and therefore did not include the most sensitive measures of
neurotoxicity. Nevertheless, the Panel agreed that the selected studies did have some important
strengths, including reasonable statistical power due to the relatively large number of animals,
chronic dosing, and the fact that the NOAELs for the endpoint in the two studies were similar,
implying some precision in the effect estimate measured. Several Panel members noted that the
lack of sensitive functiona^ehavioral assessments is a significant data gap that should be
considered in the context of setting a database uncertainty factor. Use of the benchmark dose
methodology in this assessment was deemed scientifically supported, given the nature and
robustness of the data sets available on the endpoint of interest. The calculations and choices
made were described clearly and at an appropriate level of detail. The Panel suggested that EPA
undergo the exercise of generating an RfD from the Calleman study for purposes of comparison
with the RfD derived based on the animal data. This comparison can serve as a type of
sensitivity analysis, to help determine whether the RfD based on the Johnson study appears to be
adequately health-protective despite the insensitive endpoint used in that study.
11
-------
Heritable Germ Mutations
EPA's draft document concluded that data exist that reveal acrylamide's capacity to
induce heritable germ cell effects at doses somewhat above those at which neurotoxicity has
been observed, but that there are as yet no studies providing an in-depth examination of dose-
response or identification of credible no-effect levels. The Panel supports the Agency's
conclusions that exposure to acrylamide in animals leads to heritable gene mutations and that
these results indicate that it may also pose a hazard to humans. In addition, the Panel supports
the Agency's conclusions that the available data are not adequate to conduct a robust assessment
of this endpoint at this time. There is still uncertainty about the mode of action of acrylamide
and its metabolite, glycidamide, in the induction of heritable genetic effects. The potential for
DNA adducts of glycidamide to play a role is an attractive hypothesis for the mode of action.
The Panel found the discussion in the document on heritable germ cell effects useful and
presented in a clear, transparent manner reflective of the current science. However, the Panel
suggested that, given the serious consequences of heritable germ cell effects, the considerable
deficiencies of the database should be identified and the significance of this endpoint
emphasized.
Physiologically-Based Toxicokinetic (PBTK) modeling
A physiologically-based toxicokinetic (PBTK) model originally developed by Kirman et
al. (2003), and recalibrated by EPA with more recent kinetic and hemoglobin binding data in
rats, mice, and humans, was used in the derivation of the RfD. The PBTK model was used to
extrapolate from the animal dose-response relationship to derive a human equivalent
concentration. The Panel commends EPA for their efforts to adapt the PBTK model of Kirman
et al. (2003) for acrylamide and glycidamide, recognizing that this was a complex and
challenging task. The Panel believes, though, that the documentation is not adequate to
determine whether the recalibrated Kirman model is appropriate for its intended use. While the
Panel considered that the model structure was reasonable, the parameter estimates require greater
justification. The Panel was concerned about the ability of the model to adequately simulate the
kinetics of acrylamide and glycidamide. The Panel has proposed several modifications to the
PBTK model for making the estimates of internal dose in rats needed for both the non-cancer and
cancer assessments and for calculating the Human Equivalent Dose (HED).
12
-------
Uncertainty Factors
EPA has proposed to use the default 10X uncertainty factors (UF) to account for
intraspecies (i.e., human) differences. The Panel concurred with this choice, noting that there
were insufficient data on inter-individual differences, based upon lifestage, gender or genetic
characteristics, to support departing from the default. Consensus was not achieved on the issue of
the inclusion of an UF to account for deficiencies in the existing database.
EPA has suggested that the acrylamide IRIS document include a Table that lists points of
departure for various endpoints to facilitate a Margin of Exposure (MOE) evaluation by EPA's
Regional or Program offices, or by other end users of the assessment. The Panel recommends
the inclusion of such a table, to the extent possible, in all IRIS documents, which provides
information that may be used to conduct a variety of analyses. Uses may include, for example,
MOE analyses for specific endpoints of interest and/or for other than lifetime durations of
exposure and for windows of increased susceptibility early in the life cycle, in addition to the
traditional lifetime focus. Agency risk assessments would benefit from the inclusion of
transparently-developed, peer-reviewed consensus hazard values.
Carcinogenicity
The Panel believes that the rationale and justification for acrylamide being a "likely
human carcinogen" has been well described and the conclusion is scientifically supportable
based on the fact that it produces tumors in rodents of both sexes, that there are multiple tumor
sites, and tumors are induced via multiple routes of exposure. Acrylamide is also clearly and
reproducibly carcinogenic in both rats and mice. Nonetheless, the draft document can be
improved by expanding the discussion of biological plausibility and coherence beyond DNA
adducts. The weight of evidence supports a mutagenic mode of action for carcinogenesis, and
overall the rationale has been clearly and objectively presented. During the advisory process,
information was presented by several members of the public for the Panel's consideration.
While the Panel did consider the information regarding alternative, non-mutagenic MO As, they
did not find it to be compelling. Significant biological support and data on any putative alternate
MO As are not sufficient for either explaining cancer findings or quantifying dose response
13
-------
relationships. More than one MOA may operate for a given carcinogenic chemical, and the
likelihood that more than a single MOA is operative increases as levels of exposure increase.
EPA used two chronic drinking water exposure bioassays in Fischer 344 rats (Friedman
et a/., 1995; Johnson et a/., 1986) to derive the oral cancer slope factor, and to identify the
tumors of interest for the MOA discussion. The Panel agrees that the two chronic bioassays in
F344 rats are the main studies to consider in dose response analysis, but the rationale for using
only the Friedman et al. study for derivation of the oral cancer slope factor should be improved
with the strengths and limitations of both studies discussed in greater depth The use of the
Weibull-in-time multistage-in-dose analysis is a reasonable and scientifically justifiable way to
take into account the early mortality in the high dose group in the male study. The decision not
to employ this analysis, in the case of the female because mortality across treatment and control
groups did not differ and the overall survival appears to be fairly good, is also reasonable.
Derivation of the RfC andlUR
The draft document used area under the curve (AUC) in the blood for the putative
genotoxic metabolite, glycidamide, as the dose metric for the PBTK model analysis to derive the
human equivalent concentration. The Panel agreed that the AUC for glycidamide is the best
choice for estimating the human equivalent concentration to derive the oral slope factor. One
consideration in using this as the dose metric, however, comes from some of the human studies
in which variability is not accounted for adequately. Consideration of additional human data can
provide an improved basis for adjustments for cross-species differences in pharmacokinetics, as
well as human variability in glycidamide formation from acrylamide.
As with the RfC, EPA concluded that there were insufficient inhalation data to derive an
inhalation unit risk (IUR). The PBTK model was used in a route-to-route extrapolation of the
dose-response relationship from the oral data, and to estimate the human equivalent
concentration for inhalation exposure to acrylamide. The Panel commended the EPA for using
the PBTK model to fill the gap resulting from the absence of robust animal toxicology studies
investigating neurotoxicity via the inhalation route that would support the development of an
RfC. The Panel agreed that the absence of evidence for route of entry specific effects would
allow route-to-route extrapolation for deriving an RfC based on using the PBTK model to
calculate the human equivalent concentration (FIEC).
14
-------
Age-dependent adjustment factors
The Panel agreed that the recommendation to use the age-dependent adjustment factors is
well justified and transparently and objectively described. Additionally the Panel believed that
the discussion of uncertainties is adequate, but that human variability could be more completely
addressed. There is no characterization of sensitive populations, and this should be explored and
discussed to a much greater extent.
PBTK model and uncertainty
The Panel commends EPA for using the PBTK model for developing the RfD, RfC and
Cancer Slope Factors for acrylamide. The Panel notes that the use of internal dose metrics
combined with a fairly robust understanding of the mechanism of action may replace the use of
the default interspecies uncertainty factor for toxicokinetic differences (i.e., 101/2), but not the
default interspecies uncertainty factor for toxicodynamics. This uncertainty factor is still needed
in deriving the RfC and RfD. Further the Panel strongly encourages the Agency to move
forward with revising and finalizing their assessment.
Additional Comments
The Panel is aware that studies of acrylamide toxicity are ongoing in the FDA. However,
these studies will not be finalized or go through the peer-review process for some time.
Therefore, the Panel believed that the draft risk assessment document should not be held-up
awaiting those results, but that the Agency should consider relevant data which has been
published since their draft assessment was released in December 2007 as they finalize the IRIS
document on Acrylamide.
15
-------
INTRODUCTION
This report was prepared by the Science Advisory Board (SAB) Acrylamide Review
Panel (the "Panel") in response to a request by EPA's Office of Research and Development
(ORD) to review the Draft Toxicological Review of Acrylamide (hereafter referred to as the
"draft document"). The IRIS Toxicological Review of Acrylamide, released in December 2007,
is a compilation and summary of the available information on the potential for cancer and non-
cancer hazardous effects in humans from exposure to acrylamide.
The SAB was asked to comment on (1) whether the document is logical, clear and
concise, (2) if the discussion is objectively and transparently represented, and (3) if it presents an
accurate synthesis of the scientific evidence for non-cancer and cancer hazard. The SAB was
also asked to identify any additional relevant studies that should be included in the evaluation of
the non-cancer or cancer health effects of acrylamide, or in the derivation of toxicity values. In
addition, the SAB was asked to provide advice on 26 specific charge questions related to the
derivation of a proposed oral reference dose (RfD) and an inhalation reference concentration
(RfC) for non-cancer endpoints, as well as the cancer descriptor, oral slope factor, and inhalation
unit risk for acrylamide.
The Panel deliberated on the charge questions during a March 10-11, 2008, face-to-face
meeting and discussed their draft report in a subsequent conference call on July 16, 2008. The
final draft of the panel's report was reviewed and approved during a meeting of the chartered
SAB on October 28, 2008.
The responses that follow represent the views of the Panel. The charge to the SAB is
available in Appendix A.
16
-------
RESPONSES TO THE CHARGE QUESTIONS
Charge Question 1. Please comment on the selection of neurotoxicity as the most appropriate
choice for the most sensitive endpoint (in contrast to reproductive toxicity, heritable germ cell
effects, or other endpoint) based upon the available animal and human data.
Based on the existing toxicity data base for acrylamide, neurotoxicity does appear to be
the most sensitive endpoint, and therefore, the most appropriate for developing the (non-cancer)
RfD and RfC. Animal studies report microscopically-detected degeneration in peripheral nerve
cells at doses of 1-2 mg/kg day, as compared to levels of 3-13 mg/kg day to detect impaired male
reproductive performance. Animal studies provide a clear mechanistic understanding whereby
low-dose, subchronic exposure leads to toxicity with concomitant nerve damage. Acrylamide
has a direct or indirect effect on the motor protein kinesin or nerve terminals, producing damage
in the peripheral and central nervous systems, which leads to sensory and motor disease.
Correspondingly, reports of central-peripheral neuropathy, ataxia and muscle weakness in
exposed human cohorts have been documented since the early 1950's. Acute occupational
exposure to acrylamide can lead to an immediate neurologic response, e.g., sweating, nausea,
myalgia, numbness, paresthesia, and weakened legs and hands. Following termination of short
term exposure, these acute effects disappear.
There were issues of concern that should be noted:
1) As detailed in the response to Question 4, the determination of accurate benchmark doses
(e.g., LOAELs, NOAELs, RfDs) from the Friedman et al. (1995) and Johnson et al. (1986)
studies may be compromised by their lack of functional testing of neurotoxicity and the use
of a relatively insensitive measure, peripheral axonopathy, as the primary index
neurotoxicity.
2) There was concern that axonal degeneration observed under light microscopy was the
endpoint chosen from the Friedman et al. (1995) and Johnson et al. (1986) studies for
derivation of the RfD and RfC. Animal studies indicate that nerve terminal degeneration can
17
-------
occur prior to axonal degeneration at some doses. This would suggest that all of the cited
studies, including the subchronic Burek study and the 2 year bioassay studies of sciatic nerve
(Friedman etal., 1995) and tibial nerve (Johnson etal., 1986) axons, in looking at axonal
degeneration, may have missed a preceding terminal degeneration at a lower dose,
particularly as no specific mention of terminal degeneration is provided and
functional/behavioral measures of neurotoxicity were not included.
3) It should be noted that future studies may demonstrate effects of acrylamide exposure on
male reproductive function, as currently evidenced in animal studies by increased pre- and
post-implantation losses and decreased litter sizes, at even lower doses than those currently
associated with neurotoxicity after acrylamide dosing in animal studies. The draft document
states that "associations between human exposure to acrylamide and reproductive effects
have not been reported" (p. 187 and p. 224); rather, these associations have not been
adequately studied. The lack of human data is a major limitation in this regard. As noted in
the draft document, data also exists that reveal acrylamide's capacity to induce heritable
germ cell effects at doses somewhat above those at which neurotoxicity has been observed,
but there are as yet no studies providing an in-depth examination of dose response or
identification of credible no-effect levels. The heritable germ cell effects are very worrisome
and deserve even more consideration, including perhaps the use of this endpoint to generate
an independent RfD.
4) Although still controversial and recognizing that cigarette smoke is a complex mixture made
up of hundreds of compounds, there is growing evidence that supports an association
between cigarette smoking, a known source of acrylamide exposure, and altered semen
parameters, including concentration, morphology, motility, and DNA fragmentation
(Richthoff etal., 2008; Sepaniake^a/., 2006; Marinelli etal, 2004). The lack of data
regarding potential interactions between acrylamide and other exposures, including cigarette
smoke, alcohol use, and cosmetics (another source of acrylamide exposure) has been cited as
a major limitation in studies of human acrylamide exposure and adverse health effects (Rice
2005; draft document p. 194; p. 224). The investigation of altered semen parameters among
18
-------
occupationally exposed males, controlling for smoking and alcohol consumption, should be a
high priority.
New References
Richthoff J, Elzanaty S, Rylander L, Hagmar L, Giwercman A. Association between
tobacco exposure and reproductive parameters in adolescent males. Int J Androl 2008; 31:31-9.
Sepaniak S, Forges T, Foliguet B, Bene MC, Monnier-Barbarino P. The influence of
cigarette smoking on human sperm quality and DNA fragmentation. Toxicol 2006; 223:54-60.
Marinelli D, Gaspari L, Pedotti P, Taioli E. Mini-review of studies on the effect of
smoking and drinking habits on semen parameters. Toxicol 2004; 207:185-92.
Charge Question 2. Please comment on the discussion of mode of action for acrylamide-
induced neurotoxicity.
The Panel found the separation of the discussion of MOA(s) for neurotoxicity in two
different sections of the document (Section 4.6.1, pages 123-124; and Section 4.7.3, pages 134-
136) confusing and recommends their incorporation into a single section.
Acrylamide is a member of the type-2 alkene chemical class, which includes acrolein,
methylvinyl ketone and methyl acrylate. A weight of evidence evaluation of the current body of
data now suggests that the type-2 alkenes produce toxicity via a common molecular mechanism:
i.e., formation of adducts with essential sulfhydryl thiolate groups on proteins that play
regulatory roles in cellular processes (LoPachin etal., 2007a,b, 2008a; reviewed in LoPachin and
Barber, 2006b; LoPachin etal, 2008b).
Currently, there are two hypotheses regarding the mechanism of acrylamide
neurotoxicity: 1) Acrylamide/glycidamide inhibits fast axonal transport by forming adducts with
kinesin, the transport motor (reviewed in Sickles et a/., 2002). 2) Acrylamide disrupts nerve
nitric oxide (NO) signaling at the nerve terminal (reviewed in LoPachin et a/., 2006a). The
Panel did not attempt to resolve the debate over the MOA of neurotoxicity. It is also possible
that both MO As may be pertinent, and studies directly comparing the time course of the two
proposed MO As in a single model have not been carried out. However, the Panel agreed that the
further delineation of MO As will improve acrylamide risk assessment. Both of the proposed
19
-------
MO As suggest that visible axonal degeneration seen with light microscopy is not likely to be the
low-dose effect in the causal pathway. Regardless, it should also be evident that substantial,
detailed molecular information is available regarding mechanisms of acrylamide neurotoxicity
and that these data should be included.
Thus, the following deficiencies in the draft document were identified by the Panel:
1) As drafted, the document's coverage of research findings is incomplete and does not
adequately reflect the current molecular understanding of the mechanisms of acrylamide
neurotoxicity. Moreover, information in the document regarding the hypothesized MO As is
not presented in a sufficiently transparent manner consistent with the Agency's guidance on
identification of the key events leading to the effect of concern, i.e., use of the modified
Bradford Hill criteria with respect to dose-response concordance, temporal relationship(s),
strength, consistency, specificity of association and biological plausibility and coherence, as
is done for carcinogenicity.
2) There was insufficient discussion of acrylamide adduct chemistry and corresponding
neuronal targets pertinent to understanding the MO As.
3) There was lack of a discussion of residual questions surrounding the respective roles of the
parent toxicant, acrylamide, and its epoxide metabolite, glycidamide, in the production of
neurotoxicity.
The Panel recommends that the Agency expand its discussion of the two MO As. Panel
members provided more specific text that describes the two proposed MO As, and the Panel
offers this text to EPA for consideration in revising the acrylamide assessment. The text is given
in Appendix B of this report.
20
-------
Charge Question 3. Please comment on the qualitative discussion of acrylamide's heritable
germ cell effects and whether the discussion is clear, transparently and objectively described,
and reflective of the current science.
Discussion in the document of heritable germ cell effects, consisting of five (5) heritable
translocation studies, the two (2) specific locus studies, two (2) studies on acrylamide
transformation to glycidamide and the importance of this metabolism to toxicity, is relevant and
useful, and is presented in a clear, transparent manner reflective of the current science. However,
the discussion is a linear description of germ cell toxicity with little synthesis, analysis and
scrutiny. While some SAB members considered the presentation objective, some expressed
concerns over the lack of inclusion of all potential MO As. Given the serious consequences of
heritable germ cell effects, the considerable deficiencies of the database should be identified and
the significance of this endpoint emphasized.
The entire section is prefaced and summarized with the perspective that DNA adduct
formation and mutagenicity is the only operative mechanism for heritable germ cell effects of
acrylamide. While adducts can certainly lead to these observations, there are alternative
mechanisms for discussion. Clastogenic mechanisms, as well as, mitotic spindle defects are
viable candidates for dominant lethal effects. There is a wealth of acrylamide studies reporting
these alternative mechanisms that should be included in this discussion as well. They were
briefly outlined in the carcinogen!city section, but should also be identified here. In regards to
spindle defects, the effects of acrylamide on kinesin motors involved in cell division should be
added to the document (Sickles et al., 2007).
Adequate response data are lacking in the existing heritable germ cell studies such that
the shape of the dose response relationship cannot be ascertained. However, in Tyl et al. (2000)
dose responses are identified - a NOAEL of 2 mg/kg/d and a LOAEL of 5 mg/kg/d for a 13 week
exposure. All of the dominant lethal studies were conducted at a dose of 50 mg/kg or higher and
most with multiple exposures. The specific locus studies were conducted at 50 mg/kg/d for 5
days (Russel et al., 1991) or with a single 100-125 mg/kg exposure (Ehling and Neuhauser-
Klaus, 1992). The discrepancy between the negative results of Russel et al. (1991) and the
positive results of Ehling and Neuhauser-Klaus (1992) may be dose-related or due to other
factors. The fact that heritable translocations appeared at high frequency at the lowest doses
21
-------
tested implies that even lower doses may produce such effects.
However, in the absence of these data, the uncertainty should be identified. As a
consequence of these limitations in the database, there is some uncertainty related to the RfD.
The Panel unanimously agreed that this is an extremely serious data gap that should be a top
priority for further study. Additional studies to address the aforementioned database deficiencies
in mechanisms and dose-responses would be desirable.
The document requires correction in that the NTP/CERHR report was published in
February 2005, not 2004. Also, there appears to be a discrepancy in the text (Pg 117 indicates the
historical controls were 6%, yet on pg 116 in the discussion of the Adler et al. (1994) study, the
historical controls are listed as 5/9890 which is 0.05%).
Charge Question 4. Please comment on whether the selection of the Friedman et al, 1995 and
Johnson et al, 1986 studies as co-principal studies has been scientifically justified. Although
EPA considers Friedman et al and Johnson et al to be co-principal studies, the final
quantitative RfD value is derived only from the Johnson study. Please comment on this aspect
of the EPA's approach. Please comment on whether this choice is transparently and
objectively described in the document Please identify and provide the rationale for any other
studies that should be selected as the principal studies.
The Panel afforded considerable discussion to the question of whether the Friedman et al.
(1995) and Johnson et al. (1986) studies were the best choices for derivation of the quantitative
RfD (and RfC). The main concerns with these studies included the fact that they were primarily
designed as cancer bioassays rather than for evaluation of neurotoxicity. Specifically, the Panel
contended that the endpoint of axonal degeneration visible under light microscopy is an
insensitive measure of neurotoxicity. Alterations visible under electron microscopy or
functional/behavioral alterations would have provided more sensitive endpoints.
Nevertheless, the Panel agreed that the selected studies did have some important
strengths, including reasonable statistical power due to the relatively large number of animals,
chronic dosing, and the fact that the NOAELs for the endpoint in the two studies were similar,
implying some precision in the effect estimate measured. The Panel also noted that there are no
studies yet available which include the sensitive functional/behavioral assessments that would be
22
-------
most desirable. Several Panel members noted that this issue is a significant data gap that should
be considered in the context of setting a database uncertainty factor.
With respect to the Burek et al. (1980) study, the Panel notes that while the endpoint in
this study (axolemmal invaginations under electron microscopy) is a highly sensitive one for use
in risk assessment, the study was subchronic. One Panel member proposed that EPA consider
generating an RfD based on the data in Burek et al. (1980), but not use a subchronic-to-chronic
uncertainty factor given the existence of the two chronic studies, to compare the resulting RfD to
that based on the less sensitive endpoint of axonal degeneration. Such a comparison might begin
to quantify the degree of potential under-estimate of risk due to the less satisfactory choice of
endpoint in the Johnson and Friedman studies.
There was a brief discussion of the report of foot splay at 0.5 mg/kg in F0 males in the
Tyl et al. (2000a) two-generation reproductive toxicity/dominant lethal mutation study. The use
of this gross functional endpoint could also serve as a point of departure, although it was
considered questionable because: it was only observed in the FO generation, was found in control
animals to some degree (raising questions about the methodology used in the lab), and did not
follow a clear dose-response relationship. Overall, the Panel decided that the Tyl study was not a
good choice for derivation of the RfD.
The Panel also considered the option of deriving an RfD based on human data. Both the
Calleman et al. (1994) and the Hagmar et al. (2001) studies contain sufficient data to allow the
Agency to calculate an RfC or potentially an RfD. In this regard, the Panel made the following
observations: (1) in general, it is preferable to use human data when available; (2) the Calleman
study included a measure of internal dose (adduct levels) and a fairly sensitive measure of effect,
thereby making it appealing for risk assessment; (3) PBTK modeling could allow dose
extrapolation based on adduct levels, such that an ingested or inhaled dose could be estimated for
purposes of setting either an RfC or an RfD from the data.
However, the Panel also cautioned that there are a number of drawbacks to using the
human studies, including the following: (1) the sample sizes are small; (2) the samples mostly
include young adult males; (3) the healthy worker effect would tend to bias these studies
(especially the Calleman study) toward the null, since workers with significant neurological
symptoms would leave the workplace, thus selecting for individuals with lower genetic
susceptibilities; (4) the workers in each study were exposed to other confounding neurotoxicants
23
-------
(acrylonitrile and TV-methylolacrylamide (NMA)), but this would tend to generate a more
conservative risk estimate because these other exposures would tend to result in an over-estimate
of the effect; and (5) the exposure duration was relatively short and variable (1 month to 11.5
years in the Calleman study with an average of 3 years, and 55 days in the Hagmar study). In the
end, the Panel suggested that EPA undergo the exercise of generating an RfD from the Calleman
study for purposes of comparison with the RfD derived based on the animal data. The Panel
stopped short of recommending that the human RfD be used in place of the one in the draft
document, but instead saw this as a type of sensitivity analysis, to help determine whether the
RfD based on the Johnson study appears to be adequately health-protective despite the
insensitive endpoint used in that study.
Charge Question 5. Please comment on the benchmark dose methods and the choice of
response level used in the derivation of the RfD, and whether this approach is accurately and
clearly presented. Do these choices represent the most scientifically justifiable approach for
modeling the slope of the dose-response for neurotoxicity? Are there other response levels or
methodologies that EPA should consider? Please provide a rationale for alternative
approaches that should be considered or preferred to the approach presented in the document.
Use of the benchmark dose methodology has become the preferred approach and an
acknowledged improvement over the historically traditional NOAEL + UF procedure for the
derivation of RfDs. Its application in this instance is scientifically supported, given the nature
and robustness of the data sets available for the endpoint of interest. The calculations and
choices made were described clearly at an appropriate level of detail.
EPA's Benchmark Dose guidance provides default criteria to be used for selecting the
benchmark response (BMR). For quantal data, an excess risk of 10% is the default BMR, since
the 10% response is at or near the limit of sensitivity in most studies. In this case, even though
the BMR at 10% extra risk also was within the range of observation, the BMRs was selected for
the point of departure. The choice of a BMR5 makes sense and is well-justified: (1) the 95%
lower bound of the benchmark dose (BMD), BMDL , remained near the range of observation;
(2) the 5% extra risk level is supportable given the relatively large number of animals used in the
critical studies; and (3) the use of BMDL is consistent with the Agency's technical guidance for
24
-------
BMD analysis which allows flexibility in making such a choice. One of the strengths of the
Johnson study is that it is sufficiently large (i.e., numbers of animals/group) to allow the lower
5% bound to be identified with sufficient stability that it is usable for risk assessment purposes.
Therefore, it is reasonable to use that strength in the underlying data set and choose this number.
Such a choice is appropriately conservative (i.e., public health protective).
While alternative approaches such as averaging the BMDLs from each of the four data
sets (Friedman and Johnson, male and female) rather than using just the one for males in the
Johnson study were discussed, the Panel concluded that the steps described by the Agency in the
draft document represented the preferred approach.
Charge Question 6. Please comment on the selection of the uncertainty factors (other than the
interspecies uncertainty factor) applied to the point of departure (POD) for the derivation of
the RfD. For instance, are they scientifically justified and transparently and objectively
described in the document? [Note: This question does not apply to the interspecies uncertainty
factor which is addressed in the questions on the use of the PBTK model (see PBTK model
questions below)]
The Agency has proposed to use a composite uncertainty factor (UF) of 30: 10X to
represent human variability (10H) and 3X to reflect the toxicodynamic component of the default
interspecies uncertainty factor (10A). The other half of the lOx interspecies UF, i.e., the 3X that
would otherwise account for interspecies differences in toxicokinetics, is subsumed in the PBTK
modeling.
Two points were raised about the use of 3X as a default to account for interspecies
toxicodynamic differences. First, it was noted that the rodents are less sensitive to the neurotoxic
effects of acrylamide than humans. The Panel concluded that the application of a UF for
interspecies toxicodynamics was directionally correct. Second, there is insufficient information
available to define a chemical-specific factor and the default factor of 3X UF for interspecies in
toxicodynamics is therefore appropriate. It was noted that recent International Programme for
Chemical Safety guidelines divide the default lOAinto 2.5X for toxicodynamic differences and
4. OX for toxicokinetics differences, based primarily upon a review of the literature published in
1993 -(WHO IPCS 2005. Guidance Document for the Use of Data in Development ofChemical-
25
-------
specific Adjustment Factors (CSAFs)for Inter species Differences and Human Variability in
Dose/Concentration-Response Assessment}. The use of the factor of 3 (or V10) is consistent
with current EPA practice: according to the recent EPA (2004) Staff Paper "a default UF of 10
for interspecies variability that can now be reduced to 3 when animal data are dosimetrically
adjusted to account for toxicokinetics." The Staff paper cites the EPA (2002) RfD/RfC
methodology document. That document divides UFs "into toxicokinetic and toxicodynamic
components that have assigned default values of 3.16 (101/2) each."
EPA has proposed to use the default 10X UF to account for intraspecies (i.e., human)
differences. The Panel concurred with this choice, noting that there were insufficient data on
interindividual differences, based upon lifestage, gender or genetic characteristics, to support
departing from the default.
Consensus was not achieved on the issue of the inclusion on an UF to account for
deficiencies in the existing database that would confound the derivation of the most
scientifically-defensible RfD. EPA concluded that an UFo > 1 was not necessary, arguing that
the existing database is sufficiently robust, even though they acknowledge there are some
unresolved issues that warrant further research: describing the MOA(s) for neurotoxicity, the
potential for behavioral or functional adverse effects not detected in the assays to date, and the
uncertainty that heritable germ cell effects may occur at lower than previously reported doses.
Some Panel members agreed with EPA's position. One Panel member noted that additional UFs
were implicitly, if not explicitly, incorporated into the RfD derivation. Using the output of the
log-logistic model applied to the data set for the male rats in the Johnson study resulted in the
lowest set of BMDs/BMDLs. According to one Panel member, it was perhaps conferring an
extra UF of ~2X. In addition, using the BMDL5 as the POD, rather than the default BMDLio,
also could be seen as conferring an extra UF of ~2X.
Other Panel members, however, disagreed with the Agency's position regarding the
database UF, arguing that the remaining uncertainties have major implications that could result
in effects at significantly lower doses and thus a lower RfD. Database deficiencies include the
following:
1) EPA had to rely on the observation of axonal degeneration visible by light microscopy,
an endpoint which is not likely to be the most sensitive. EPA is using studies that were
26
-------
not designed to evaluate neurotoxicity robustly, e.g., histopathology coupled with
systematic evaluation of functional or behavioral parameters at multiple time points with
robust numbers of animals/treatment and robust number of treatment groups; these
studies should be done in adult animals and in a developmental neurotoxicity study in
order to determine whether or not critical lifestage differences exist;
2) Both existing chronic studies were done in the rat, creating some remaining uncertainty
about interspecies differences that is not addressed by the interspecies UF. Based upon
the comparison of results from the Tyl et al. (2000) 2-generation study in rats and the
Chapin et al. (1995) 2-generation study in mice, the NOAEL for (adult) neurotoxicity is
essentially the same (0.5 mg/kg/day in rats vs. 0.8 mg/kg/day in mice), but the difference
could potentially be driven by the dose spacing regimen rather than a true difference in
response. The outcomes of long-term exposure in mice hold the possibility of yielding
lower NOAELs/LOAELs/BMDs than observed/calculated from the rat data. If this were
to occur, the RfD/RfC would be lower.
3) The germ cell effects have not been fully explored and have major intergenerational
implications if they do occur at dose levels lower than those for neurotoxicity. There is a
lack of adequate data to define the dose response for heritable germ cell effects. While
the existing data describe adverse effects at doses somewhat higher than those at which
neurotoxicity was observed, BMD modeling of robust dose-response data may yield
results competitive with/lower than the neurotoxicity BMDs/BMDLs.
Charge Question 7. Please provide any other comments on the derivation of the RfD and on
the discussion of uncertainties in the RfD.
Acrylamide and Cumulative Risk Assessment
While there were no additional comments on the derivation of the RfD per se, the Panel
did want to address the potential for cumulative effects from exposure to acrylamide and other
type-2 alkenes. The Food Quality Protection Act (FQPA) of 1996 mandates EPA to consider the
"cumulative effects" of pesticides and other substances that have a "common mechanism of
toxicity" when setting, modifying or revoking tolerances for food-use pesticides. Were
27
-------
acrylamide registered as a food use pesticide, its activity as a type-2 alkene would support a
cumulative risk assessment of it and other chemicals in the class. From a scientific standpoint
and particularly from a public health perspective, this class of chemicals should be subjected to a
cumulative risk assessment (e.g., see Wilkinson et a/., 2000). Evaluating the cumulative effects
of the type-2 alkenes is particularly germane since human exposure is pervasive; i.e. chemicals in
this class are used extensively in the agricultural, chemical and manufacturing industries.
Furthermore, they are well-recognized environmental pollutants (e.g., acrolein, acrylonitrile),
food contaminants (e.g., acrylamide, methyl acrylate) and endogenous mediators of cellular
damage (e.g., acrolein, 4-hydroxy-2-nonenal) (see LoPachin et a/., 2008b). Thus, the application
of standard approaches may result in RfDs and RfCs which could be associated with risks in the
population. At a minimum, a caveat in this regard should be included in the acrylamide
assessment document.
Charge Question 8
Use of the PBTK Model
A physiologically-based toxicokinetic (PBTK) model originally developed by Kir man et
al. (2003), and recalibrated by EPA with more recent kinetic and hemoglobin binding data in
rats, mice, and humans (Boettcher et al, 2005; Doerge et al, 2005a,b; Fennell et al, 2005)
was used in the derivation of the RfD to extrapolate from the animal dose-response
relationship (observed in the co-principal oral exposure studies for neurotoxicity) to derive a
human equivalent concentration (HEC). The HEC is the external acrylamide exposure level
that would produce the same internal level of parent acrylamide (in this case the area under
the curve [AUC] of acrylamide in the blood) that was estimated to occur in the rat following
an external exposure to acrylamide at the level of the proposed point of departure, and related
to a response level of 5% (i.e., the BMDLs). The model results were used in lieu of the default
interspecies uncertainty factor for toxicokinetics differences of!01/2, which left a factor of
101/2 (which is rounded to 3) for interspecies differences in toxicodynamics.
With respect to the RfC, there are presently insufficient human or animal data to
directly derive an RfC for acrylamide. The PBTK model was thus used to conduct a route-to-
route extrapolation (oral-to-inhalation) to derive an RfC based on the dose-response
relationship observed in the co-principal oral exposure studies for neurotoxicity. In this case,
28
-------
the HEC was based on a continuous inhalation exposure to acrylamide in the air that would
yield the same A UCfor the parent acrylamide in the blood as that estimated for the rat
following an external oral exposure to acrylamide at the level of the proposed point of
departure (i.e., the BMDLs).
Please comment on whether the documentation for the recalibrated Kirman et al. (2003)
PBTK model development, evaluation, and use in the assessment is sufficient to determine if
the model was adequately developed and adequate for its intended use in the assessment.
Please comment on the use of the PBTK model in the assessment, e.g., are the model structure
and parameter estimates scientifically supportable? Is the dose metric of area-under-the-
curve (AUC)for acrylamide in the blood the best choice based upon what is known about the
mode of action for neurotoxicity and the available kinetic data? Please provide a rationale for
alternative approaches that should be considered or preferred to the approach presented in the
document.
The Panel commends EPA for their efforts to adapt the PBTK model of Kirman et al.
(2003) for acrylamide and glycidamide, recognizing that this was a complex and challenging
task. The modified Kirman et al. model was produced by changing the model initially described
for the rat, and adapting it to fit updated data published since the original publication in 2003,
and to describe toxicokinetics in humans. Three major modifications were described - the
partition coefficients for glycidamide, the metabolic rate constants for oxidation and conjugation,
and the partition coefficients for acrylamide. The simulations of the modified Kirman model
were presented as tables containing comparisons of AUC data, and the extent of metabolism of
acrylamide to glycidamide, and the extent of conjugation of each with glutathione.
However, the Panel had a number of concerns about the description of the model and its
parameterization. The Panel believed that the documentation is not adequate to determine
whether the recalibrated Kirman model is appropriate for its intended use. Among the items that
the Panel would like to see to justify the performance of the model are: the model code;
graphical presentation of the data for time course simulations; and graphical presentation of dose
response simulated by the model. Side by side comparisons of the model parameters for the rat
and human could be accomplished by combining Tables E-4 and E-6.
29
-------
The Panel noted that the model with some changes has been described in a manuscript
published in 2007 by Walker et al. If life stage considerations are planned for subsequent work,
PBTK modeling is the recommended tool for dosimetry estimates across life stages. The Panel
would like to see the model used to simulate or show the degree of consistency with data
published since 2005.
The Panel also noted that there have been additional studies of aery 1 amide, its metabolites
and adducts, with varying data quality, and varying understanding of exposures. For example,
exposures in smokers are likely a composite of exposure from diet (oral) and smoke (inhalation).
There are possible ambiguities in assignment of acrylamide and glycidamide metabolites (the
acrylamide mercapturic acid sulfoxide and the glycidamide mercapturic acids are isomeric, and
need to be resolved chromatographically for appropriate quantitation). The Panel suggests that
EPA review these reports for data quality and suitability, and if appropriate use them in
evaluation/refinement of the model.
The Panel noted discrepancies between the PBTK predicted and measured critical dose
metrics for the non-cancer (acrylamide AUC) or cancer (glycidamide AUC) PODs following
drinking water exposures in rats (see table below).
EGV
RfD
oral
cancer
BMDL
(mg/kg/day)
0.27
0.3
Critical
Dose Metric
AA_AUC
GA AUC
EPA PBTK Model
Predictions
Internal dose (uM-hr)
18.1
15.1
Tareke/Doerge
Measured Data
(2005, 2006)
Internal dose (uM-hr)
4.2
4.7
The draft document notes that the data of Doerge et al. (2005 a,b) were available (page E-
5), but it is not clear if the data were actually considered in updating the model.
30
-------
While the Panel concluded that the model structure was reasonable, the parameter
estimates require greater justification. The review notes (Page E-18 last paragraph) that: "In
comparing different versions of the model, it was also noted that the model parameters were
underdetermined, that is, there is just not enough basic toxicokinetic data to derive a unique set
of optimal parameter values, given the number of "adjustable" parameters in the current model."
The Panel was concerned about the ability of the model to adequately simulate the
kinetics of acrylamide and glycidamide. There is little justification presented for the adjustment
of parameters from the original Kirman model. The method of optimization was not well
described. The comparisons provided between observed data and model simulations are largely
for AUC in tables. Thus it is difficult to determine how the model would perform under the kind
of tests usually applied to a model, including the ability to fit kinetic data. Table E-4 indicates
that while AUC for acrylamide and glycidamide can be simulated reasonably well with the
revised rat model, and AM-GSH is reasonably close, the extent of metabolism to GA-GSH is
overestimated by 3 fold by the model. Approximately 40% of the urinary metabolites were
reported as GA-GSH (Fennell et a/., 2005), but the model simulates that 70% would be derived
from GA-GSH.
Table E-9 indicates that almost 50% of acrylamide is converted to glycidamide in
humans. The data reported in Fennell et al. (2005) indicate approximately 13.5 % of the
urinary metabolites were derived from glycidamide. Some recent studies indicate a higher degree
of glycidamide formation from acrylamide, and substantial variation among individuals in this
formation (Vesper et al. 2008; Hartmann et al. 2008). The model simulations are based on the
assumption that all of the acrylamide not accounted for by excretion in urine by 24 hours is
converted to glycidamide. As noted above, there are data not modeled that could greatly
improve the model parameter estimates, using human urine kinetic data for acrylamide,
glycidamide and urinary metabolites (e.g., Fennell et al. 2006; Hartmann et al. 2008; Vesper et
al. 2006, 2008). Table E-7 cites the Ratio of GA-GSH to AA-GSH metabolite excretion at low
doses reported by Boettcher et al. (2005) as 0.206 as a data point used for calibration. Yet the
model simulation reports a value of 0.733 (Table E-9). The half-life estimated for acrylamide in
the model is approximately 5.8 hours and the half-life estimated for glycidamide is
approximately 6.1 hours. The half life calculated from urinary excretion rate for acrylamide in
humans by Fennell et al. (2006), who studied small groups of healthy infertile adult men, was
-------
approximately half this, ranging from 3.13-3.49 hours. The issue of adjusting the parameters for
partition coefficients and the rates of glutathione conjugation and oxidation is a serious one. It is
possible to simulate the same AUC in blood with different model parameters, but with wildly
different extents of metabolism and dose to the tissues for acrylamide or glycidamide, by
adjusting partition coefficients, and metabolic rate constants. In other words, there may not be
unique solutions unless the full body of reported data can be used in model verification. It is
exceedingly important to carefully consider the extent of metabolism as a key piece of
information in making parameter selections.
The description of the parameters and calibration for the human Kirman model are
generally presented clearly on pages E-17 and E-18. A possible exception is the very general
description of the "iterative process" that was used to evaluate physiologically feasible options to
best fit the Fennell et al. (2005b) and Boettcher (2005) human data on adult adduct levels and
urinary metabolites. A rough comparison of the final rat and human values suggests increased
values for a number of tissue binding and metabolic parameters in the human model. Many of
these parameters that changed from rat to human increased roughly by a factor of 2 with the
exception of the Cytochrome P-450 oxidation rate that decreased by a factor of almost 2.1. It is
not clear from the description of the iterative process used to calibrate these values whether the
process was designed to force these parameters to move as groups or exactly what logic was
employed to adjust these multiple parameters. The general logic behind the iterative testing of
permutations of values could be clarified here without going into extreme detail.
An alternative approach that should be considered is a re-evaluation of the revised PBTK
model of Kirman et al. (2003). Determining how well it simulates the more recent data and
adjusting the metabolic parameters as necessary is one approach. The Panel had an extensive
discussion as to whether the dose metric of area-under-the-curve (AUC) for acrylamide in the
blood was the best choice based upon what is known about the mode of action for neurotoxicity
and the available kinetic data. A variety of opinions were expressed, ranging from the assertion
that AUC for acrylamide in blood was a suitable dose metric, to the belief that it may not the best
choice, but may be expedient. The best choice would be to have compartments for the tissues of
interest, and to model the amount of acrylamide and/or glycidamide reaching the tissues. The
Kirman model and the modified Kirman model are both limited by the tissue descriptions: liver,
lung, blood and a single compartment for remaining tissues.
32
-------
There was extensive discussion among the Panel members about whether the
neurotoxicity of acrylamide could clearly be attributed to acrylamide alone, to glycidamide, or to
a mixed mode of action. This question was raised in the review document (Page 136, last full
paragraph). Therefore the choice of acrylamide in blood as the dose metric may need to be
revisited as this question is clarified.
Several alternatives to the PBTK model exist for making the estimates of internal dose in
rats needed for both the non-cancer and cancer assessments and for calculating the Human
Equivalent Dose (HED). The data available in Doerge et al. (2005) and Tareke et al. (2006)
provide measured serum acrylamide and glycidamide AUCs in rats exposed at drinking water
concentrations and resulting doses near the PODs. Simple linear extrapolation could be used to
calculate the critical internal dose metrics. The hemoglobin adduct and other data available in
several recent publications (Fennell etal. 2005; Vesper et al. 2006, 2008; Hartmann et al. 2008)
together provide a robust means of estimating HEDs. The Panel also discussed the alternative
approach of using pharmacokinetic principles to interpret measurements of hemoglobin adducts
of acrylamide and glycidamide and thereby model glycidamide formation.
The Panel also raised concerns about the population variability in the metabolism and
toxicokinetics of acrylamide, and how that could be incorporated in the model. It was
recognized that there are some high quality human data sets that could be used for PBTK model
development (e.g. Fennell etal., 2005, 2006). However, there are limitations with the small
number of selected subjects compared with the general population, in describing the population
variation. The Panel has identified some studies that suggest variation in the extent of
metabolism of acrylamide to glycidamide (Vesper et al. 2006, 2008; Hartmann et al. 2008), and
differences in extent of conversion of acrylamide to glycidamide in children (Heudorf et al.,
2008). There is a need for a better understanding of exposure route differences, inter-individual
variation and life stage differences in the metabolism of acrylamide to glycidamide, and their
clearance. The Panel encourages an evaluation of the available literature, and if possible,
simulation of human variability within the PBTK model.
33
-------
Charge Question 9. Is the Young et al model adequately discussed relative to structure,
parameter values and data sets used in the model?
The Young et al. paper does not provide citations or values for many of its physiological
model parameters. This is an unusual situation for a PBTK modeling paper. For chemical
specific model parameter values, the authors fitted the chemical specific model parameter values
for each administered dose, creating a model that is calibrated for each dose. This results in an
unwieldy model for use in risk assessment. The preferred approach is to use all the administered
dose groups and create a model with one set of chemical specific model parameters that
describes all the toxicokinetic data sets. The model was based on the use of linear terms to
describe chemical specific reactions (e.g., binding, DNA adducts, and metabolism). This
approach may not hold (and non-linear terms will be needed) when developing one set of
chemical specific model parameters to describe the kinetics over a range of doses.
Do you agree with the conclusion that the recalibrated Kirman et al. 2003 model is the best for
deriving toxicity values?
In the opinion of the Panel, the recalibrated Kirman model was superior to the Young et
al. PBTK model. However, the Panel noted that the recalibrated model requires updating to
include new data sets in the rat and human. The concerns described in Charge Question 8 need
to be addressed to use the recalibrated Kirman et al. 2003 model. The Panel also noted that an
approach to calculating internal doses at the non-cancer and cancer PODs is available that relies
on measured data (and minimal linear extrapolation in a dose range that has been shown to be
linear) instead of the PBTK model. This approach also affords the ability to calculate the HED
corresponding with the critical internal dose metrics associated with the PODs (see response to
question 8). If life stages are considered, the PBTK modeling or another pharmacokinetic
approach is the preferred approach for determining a HED or HEC.
Charge Question 10. According to USEPA's RfC Methodology (1994), the use of PBTK
models is assumed to account for uncertainty associated with the toxicokinetic component of
the interspecies uncertainty factor across routes of administration. Does the use of the PBTK
34
-------
model for acrylamide objectively predict internal dose differences between the F344 rat and
humans, is the use of the model scientifically justified, and does the use of the PBTK reduce
the overall uncertainty in this estimate compared to the use of the default factor? Are there
sufficient scientific data and support for use of this PBTK model to estimate interspecies
toxicokinetic differences and to replace the default interspecies factor for toxicokinetic
differences (i.e., 101/2)? Is the remaining uncertainty factor for toxicodynamic differences
scientifically justified, appropriate and correctly used?
The Panel commends EPA for using the PBTK model for developing the RfD, RfC and
Cancer Slope Factors for acrylamide. The kinetics of acrylamide are well characterized and thus
the use of internal dose metrics that are thought to represent the critical dose metrics for non-
cancer (neurotoxicity) and cancer (various tumor types) is a preferred approach for extrapolating
across species. The Panel agrees that the use of internal dose metrics (calculated using the
PBTK model or other pharmacokinetic approaches alluded to above) combined with a fairly
robust understanding of the mechanism of action and thus the critical dose metric replaces the
use of the default interspecies factor for toxicokinetic differences (i.e., 101/2).
The Panel agreed with the use of the remaining UFs representing interspecies differences
in toxicodynamics and intraspecies variability in both toxicokinetics and toxicodynamics.
Charge Question 11. Please comment on whether the PBTK model is adequate for use to
conduct a route-to-route extrapolation for acrylamide to derive an RfC in the absence of
adequate inhalation animal or human dose-response data to derive the RfC directly. Was the
extrapolation correctly performed and sufficiently well documented?
The Panel discussed the lack of inhalation toxicology and PK studies. One Panel
member who has conducted inhalation PK exposure studies noted the difficulty with conducting
controlled rodent exposure studies and the difficulty in maintaining stable exposure
concentrations because of the low volatility of acrylamide and its propensity to sublime. The
Panel agreed with the use of PBTK modeling to conduct dose-route extrapolation. Additionally,
the Panel commends the EPA for using the PBTK model to fill the gap resulting from the
absence of robust animal toxicology studies investigating neurotoxicity via the inhalation route
35
-------
that would support the development of an RfC. The Panel agreed that the absence of evidence
for route of entry specific effects would allow route-to-route extrapolation for deriving an RfC
by using the PBTK model to calculate the human equivalent concentration (HEC). This would
yield an equivalent internal dose (Acrylamide AUC) associated with those achieved at the POD
from the oral sentinel (Johnson et al.) studies. The Panel noted that few inhalation PK studies
exist to allow a robust parameterization of the inhalation component of the PBTK model for
either rats or humans. Despite this, the Panel noted that acrylamide is very water soluble and
non-volatile, and the compound has a relatively long half-life. Therefore, the absorption of
acrylamide via inhalation should be nearly complete, and first pass effects are negligible, thereby
making the pharmacokinetics of acrylamide via inhalation easy to extrapolate from the oral case,
using simple principles of pharmacokinetics. The Panel agreed that the application of
pharmacokinetic approaches (e.g,. the use of the PBTK model) reduces uncertainty associated
with animal to human extrapolation and thus warrants replacing the default UF associated with
interspecies extrapolation for pharmacokinetic differences as was done for deriving the RfD.
The Panel noted that the air concentration one would derive using the default approach
(multiply the HED by body weight [70 kg] and dividing by daily inhalation rate [20 m3/day]
yielding 0.266 |ig/m3) is very similar to the HEC derived using the PBTK model (0.25 |ig/m3).
Therefore, if the EPA also decides to provide an extrapolation based on measured data (as
described in the response to charge question 8), the default approach of extrapolating from an
absorbed oral dose to an equivalent intake from the inhalation route (multiplying by 70 kg and
dividing by 20 m3/day) can be used with confidence to calculate the RfC.
Charge Question 12. Please provide any other comments on the derivation of the RfC and on
the discussion of uncertainties in the RfC.
The Panel has no further comments beyond those already discussed above.
Charge Question 13. Would you suggest that EPA include a Table that lists points of
departure (e.g., NOAELs, BMDs, etc.) for various endpoints that could be used, in
conjunction with exposure assessments, to conduct a MOE analysis?
36
-------
To the extent permitted by the available data, the Panel supports the concept of the
inclusion of a table in the IRIS acrylamide document that provides information which could be
used to conduct a variety of MOE analyses for specific endpoints of interest and/or for other than
lifetime durations of exposure, in addition to the traditional lifetime focus. In doing so the
magnitude of the MOE that represents a negligible risk should be reported for each point of
departure tabulated.
Currently, for those environmental agents for which sufficient data exist, IRIS documents
will present the derivation of a Reference Dose (RfD) and a Reference Concentration (RfC), as
traditionally defined, to be used in the assessment of scenarios which assume that long-term or
lifetime exposures are occurring to non-carcinogenic hazards. Additionally, in those cases where
the agent of interest has been shown to have carcinogenic potential, an oral cancer slope factor
(CSF) and/or an inhalation unit risk (IUR) may be derived, in order to estimate lifetime cancer
risks. Whether or not this step is included is determined by a weight-of-evidence evaluation of
the body of evidence supporting carcinogenic potential and an understanding, or lack thereof, of
the mode(s) of action by which the carcinogenic responses are mediated. These four values (the
RfD, RfC, CSF and IUR) are applicable in situations where the assessment is focused on the
general population exposed over a lifetime, and may have more limited utility in the assessment
of specific subpopulations and/or less-than-lifetime exposure durations.
EPA Program and Regional offices and other end-users of IRIS documents often must
develop risk assessments for specific populations and/or less-than-lifetime exposure scenarios in
order to carry out their respective legislative and regulatory mandates. These risk assessments
would benefit from the inclusion of transparently-developed, peer-reviewed consensus hazard
values.
A comprehensive table would, for example, include NOAELs, LOAELs, BMDs and
BMDLs at the 1%, 5% and 10% risk levels (as the default) for those studies deemed the most
appropriate for the assessment of specific endpoints and for acute, intermediate and long-term
exposure scenarios, data permitting. It is recognized that it will typically not be possible to fill in
every cell for every endpoint and all exposure durations of interest and that a different
BMDR/BMDLRmay better reflect the study's results. Some EPA program offices have extensive
experience in the selection of study types and durations that best lend themselves to the
assessment of specific endpoints, exposure durations and subpopulations.
37
-------
For this draft acrylamide assessment, such a table would display the relevant outcomes of
a review of the reliable and well-performed studies which evaluated the potential for
neurotoxicity in the adult and developing organism, reproductive toxicity including heritable
germ effects, developmental toxicity, and general systemic toxicity following acute, intermediate
and long-term exposure, as appropriate.
Charge Question 14. Please comment on the discussion of methods to quantitate the dose-
response for heritable germ cell effects as to whether it is appropriate, clear and objective, and
reflective of the current science. Has the uncertainty in the quantitative characterization of
the heritable germ cell effects been accurately and objectively described?
[It should be noted that the section under review is 5.5 rather than 5.4. In addition, page 215
which includes figures 5-2 and 5-2a, was inadvertently omitted in the draft EPA report and thus
not available for review by the Panel. Correction of this error, however, is not expected to
impact the recommendations of the Panel on this question as outlined below.]
Although reservations were expressed about the lack of data to quantify dose-response, it
was the consensus of the Panel that the discussion of the methods should be retained in the
report. The report adequately characterizes the current science, reflects historical attempts to
estimate these risks and notes that the quantitation methods are based only on the Dearfield et al.
(1995) publication. Concerns about the validity of the data and methods are given throughout
the section and it is appropriately noted on page 217, " these uncertainties in the assumptions and
data gaps warrant further research to improve the usefulness of the following quantitative
estimates of risk of acrylamide-induced heritable effects."
Some specific observations/recommendations/concerns are outlined below:
• The parallelogram models were clearly described and the rationale for the decision to use
the modified direct and doubling dose approach appears appropriate.
• Clearly, there is considerable uncertainty regarding the validity of the underlying
assumptions for these methods, and these methods may underestimate risk since they do not take
into account all elements that may contribute to the risk.
• The extrapolation of exposure is based on animal studies using high dosages (50 to 100
mg/kg or even higher)
38
-------
• The risk extrapolation factors (REFs; pg. 217) should be explained in more detail and
information included on how each number is derived (range, etc).
• In agreement with the report, given the differences in glycidamide production in different
species, an REF of 1 for the metabolic and dose rate variability is likely incorrect. There appear
to be significant dose-rate and species-dependent variations in acrylamide metabolism to
glycidamide (e.g., see Barber etal., 2001; Fennell and Friedman, 2005).
• An REF for uncertainty in the mode of action was recommended since the doubling dose
is dramatically higher when generated using specific locus studies which are clearly point
mutations (53.1 mg/kg using Ehling and Neuhauser-Klaus, 1992) versus using heritable
translocation data that could be based on clastogenic mechanisms (1.8, 3.3, 0.39 mg/kg for
Shelby etal, 1987, Adler etal, 1994 and Adler, 1990).
• The implementation of the modified direct approach was difficult to understand when, in
the absence of the number of human loci capable of mutating to dominantly expressed disease
alleles, it was assumed to be 1000. Clarification of how this number was derived would be
helpful (i.e. how do we know the number of mutable genes?).
• In the doubling dose approach it was not clear how the four data sets, each of which used
high acrylamide dosing rates without significant dose ranges, could accurately predict the
number of new diseases in the offspring at low doses.
Lack of current research in this area is a major concern and little has been done to update the
research and data collection based on the Dearfield et al (1995) methods. The Panel is in
agreement with the report that recommends further research and data to fill the critical data gaps
and reduce uncertainties including gaps in interspecies extrapolation factors, the quantitative
relationship between genetic alterations in germ cells and heritable disease, and the shape of the
low-dose response relationship. Research might include multiple dose studies, including dose
selection comparable to that employed in the repeated dose studies which identified
neurotoxicity as a critical effect. It is also recommended that impacts on different cell types be
determined and that biomonitoring data be utilized in any models developed.
Charge Question 15. Please comment on the scientific support for the hypothesis that
heritable germ cell effects are likely to occur at doses lower than those for neurotoxicity?
What on-going or future research might help resolve this issue?
39
-------
The Panel unanimously agreed that germ cell-induced effects should be taken very
seriously, as their implications are highly significant from a public health perspective. There is
an absence of data on these effects in lower dose ranges, making it very difficult to speculate
about the relevance of this endpoint at or below the dose levels that cause neurotoxicity.
Panelists did point out that heritable translocations appeared with very high frequency at the
lowest doses tested (i.e., 5 x 40 mg/kg resulted in 24% translocation carriers, Shelby et a/.,
1987). The high frequency of germ cell effects at these doses implies that these studies were far
from identifying a LOAEL or NOAEL, and that there would likely be germ cell effects at much
lower doses. However, the combination of lack of testing at lower doses, and the narrow dose
range in which testing has been done, makes it very difficult to extrapolate down to a low dose
range. The Panel agreed that it is a high priority to extend the heritable translocation studies
down into lower dose ranges, and that this information would be very useful for risk assessment
once it is completed.
Charge Question 16. The risks of heritable germ cell effects (Le., number of induced genetic
diseases per million off spring) for some estimated exposure in workers and the population are
presented in Table 5-11, and are based on the quantitative methods and parameter estimates
discussed in Section 5.4 of the Toxicological Review. Please comment on whether or not the
quantitation of heritable germ effects should be conducted, the level of uncertainty in the
results, if Table 5-11 is useful for risk assessment purposes, and if the RfD should be included
in the Table as one of the exposure levels.
The Panel supports the Agency's conclusions that exposure to acrylamide in animals
leads to heritable gene mutations and that these results indicate that it may also pose a hazard to
humans. In addition, the Panel supports the Agency's conclusions that the available data are not
adequate to conduct a robust assessment of this endpoint at this time.
The Panel's deliberations regarding quantifying heritable germ cell mutations centered on
the importance of including data such as those presented in Table 5-14 (not Table 5-11, as noted
in the final question), the potential significance of these endpoints to human risk assessment, and
the paucity of new data developed since the Dearfield et al. (1995) review upon which this
40
-------
section relied heavily (including Table 5-14). A majority of Panel members were supportive of
the inclusion of this table in the document and for including the RfD and RfC among the
concentrations in the table as this would facilitate comparison with the neurological endpoints.
Suggestions also included adding more information into the review regarding the role of CYP
2E1 in the dominant lethal effects of acrylamide, which indicated a requirement for metabolism
to glycidamide. While the caveats from the Dearfield et al. (1995) review were recapitulated in
the document, the Panel discussed the need to further elaborate the limitations in the underlying
data and to include reference to the new relevant studies that pertain to uncertainty and dose-
response.
Charge Question 17. Do you know of any additional data or analyses that would improve the
quantitative characterization of the dose-response for acrylamide-induced heritable germ cell
effects? Would these data also support the quantitative characterization of "total" male-
mediated reproduction risks to offspring (i.e., lethality + heritable defect)? If data are not
available, do you have any recommendations for specific needed studies?
A concern raised by the Panel was that there is a lack of a suitable data set for dose
response assessment for acrylamide-induced heritable germ cell effects. The majority of the
studies reported have been conducted in mice, using relatively high doses.
Using wild type and Cyp 2E1 knockout mice, it has been demonstrated that oxidation of
acrylamide to glycidamide is required for the dominant lethal effect (Ghanayem et a/., 2005a)
and for the induction of erythrocyte micronuclei and DNA strand breaks in lymphocytes, liver
and lung using the Comet assay (Ghanayem et al., 2005b). The greater incidence of heritable
translocation carriers in mice administered glycidamide (Generoso etal., 1996) compared with
acrylamide (Adler etal., 1994) suggests that glycidamide plays a key role in the mode of action
for heritable genetic effects.
The risk equivalent factors (REFs, page 217) need to be updated. There are profound
differences between rats, mice and humans in the extent of metabolism of acrylamide to
glycidamide, and the relative internal dose of acrylamide and glycidamide differs markedly
between mice, rats and humans. The extension of the physiologically-based pharmacokinetic
modeling approach to include the mouse should be a priority. The blood-testis barrier is thought
41
-------
to contribute to the reduction of internal dose in the testis compared with other tissues for
ethylene oxide (Fennell et a/., 2001). Testis should be included as a compartment in the model.
Data permitting, including the testis as a compartment in the model could potentially improve the
dose response characterization for this endpoint.
In reviewing data needs (page 220), it is noted that "The estimates do not take into
account other potential genotoxic mechanisms such as effects in spermatogonia stem cells,
effects in female germ cells, or induction of recessive mutations that would not appear in the first
generation, but could lead to additional adverse effects in subsequent generations." Studies to
examine the dose response for heritable genetic effects, and the effect of long-term exposure to
acrylamide are needed.
There is still uncertainty about the mode of action of acrylamide and glycidamide in the
induction of heritable genetic effects. The potential for DNA adducts of glycidamide to play a
role is an attractive hypothesis for the mode of action. With respect to the possible role for
protamine modification in the generation of effects, there was extensive Panel discussion
concerning the potential of glycidamide to form adducts with cysteine in proteins and peptides.
Adducts to protamine from acrylamide have been identified in late stage spermatids and
suggested to mediate the dominant lethal effects (Sega et a/., 1989). Whether glycidamide will
form similar protamine adducts has not been determined. Kinesin motor proteins associated with
cell division are an additional site of potential action leading to heritable germ defects (Sickles et
a/., 2007) that requires future consideration. Both AA and GA inhibit two kinesin motors
associated with spindle formation and maintenance as well as separation of chromosomes. Loss
of fidelity of chromosomal separation is related to aneuploidy, micronuclei formation and
instability of the genome. The motor protein inhibitions occur at concentrations well below the
occurrence of all heritable germ cell effects. Furthermore, glycidamide is more potent than
acrylamide. Surveying populations occupationally exposed to acrylamide in manufacturing
plants was suggested as an approach for evaluation in humans.
Charge Question 18. Have the rationale and justification for the cancer designation for
acrylamide been clearly described? Is the conclusion that acrylamide is a likely human
carcinogen scientifically supportable?
42
-------
Yes, the rationale and justification have been clearly described, although it should be
further expanded (see below), and the conclusion is scientifically supportable. Acrylamide is
clearly and reproducibly carcinogenic in both rats and mice. As outlined in the draft document, it
produced tumors at multiple sites in the rat in multiple chronic studies, and was a skin tumor
initiator in mice by multiple routes. Therefore, in accordance with EPA's cancer guidelines, it is
consistent with the "likely human carcinogen" cancer descriptor.
To paraphrase the International Agency for Research on Cancer (IARC) Monographs
Preamble, in the absence of tumor data in humans it is both reasonable and prudent to regard
evidence of carcinogenicity in experimental animals as evidence for a probable cancer hazard to
humans. This conclusion is consistent with both national and international guidelines for
carcinogenic hazard identification. The U.S. National Toxicology Program (NTP) has long
emphasized that chemicals that cause tumors at multiple sites or in more than a single species are
reasonably anticipated to be human carcinogens. Both the NTP and IARC have placed
acrylamide in cancer classifications similar to that of EPA's "likely human carcinogen" (This
could be noted in the Toxicological Review).
When experimental exposure of rats or mice to known human carcinogens is via diet or
drinking water, tumor sites observed in those species do not necessarily correspond to the same
tumor sites in humans. Exposure to chemicals that cause tumors of the mammary gland or the
liver in mice or rats, for example, does not necessarily correspond to increased cancer risk
specifically for female breast or liver in humans. The essential point to be considered is that in
any given case a tumor at these or any other site(s) results from an MOA known to operate in
humans, such as somatic cell mutagenicity.
Primary CNS tumors as a group, which are discussed at considerable length in the draft
document, should be restored to the list of experimental tumors produced by acrylamide and that
are of interest for the MOA discussion. The Panel cautions that the viruses that can cause
primary CNS tumors in hamsters and other non-human species are not relevant to this
discussion.
It should be emphasized that the spectrum of tumors consistently seen in acrylamide-
exposed rats is completely consistent with a DNA-reactive MOA, based on published data about
other substances that induce or initiate the same kinds of neoplasms. The only agents known
conclusively to induce tumors of the brain and peritesticular mesothelium in rats are all DNA-
43
-------
reactive, and in fact a single exposure to a direct-acting mutagenic carcinogen has been observed
to suffice for tumor induction at either site. The concept that acrylamide acts by a mutagenic
MOA is thus supported by the spectrum of acrylamide-associated tumors that occur in exposed
rats and mice, as well as by the biotransformation pathway of acrylamide in vivo.
Tumor initiation - promotion data for mouse skin are perhaps not sufficiently emphasized
in the draft document. First, only DNA-reactive chemicals or chemicals biotransformed to
DNA-reactive metabolites are established tumor initiators. As acrylamide is an initiator, and by
multiple routes of administration, it is a permissible inference that acrylamide is also acting by a
DNA-reactive MOA in mouse skin, as do other initiators. It is most striking that, in mice,
systemic exposure to acrylamide is more effective for skin tumor initiation than direct
application to the skin. The order of efficiency, oral > ip > dermal application, for initiation of
TPA-promotable squamous cell papillomas and carcinomas on mouse skin strongly supports the
importance of systemic exposure and post-hepatic distribution of a reactive metabolite in the
MOA for carcinogenicity at this site.
Charge Question 19. Do you agree that weight of the available evidence supports a
mutagenic mode of carcinogenic action, primarily for the acrylamide epoxide metabolite,
glycidamide (GA)? Has the rationale for this MOA been clearly and objectively presented,
and is it reflective of the current science?
A sound rationale and justification already supports the mutagenic mode of action
(MOA), and this evidence is further supported by additional new data as described below. The
weight of evidence supports a mutagenic MOA, and overall the rationale for this MOA has been
clearly and objectively presented. Some improvements to the presentation are as follows. The
discussion of biological plausibility and coherence could be expanded beyond DNA adducts and
the human relevance section could be somewhat more expansive without being repetitive. The
argument on page 145 regarding the lack of relationship of cytogenetic damage to a mutagenic
MOA should be carefully re-considered, as the literature is full of these correlations. Evidence
for and against the arguments set out should be carefully evaluated, and much better referencing
included. Reports from Bonassi and Hagmar are cited as supportive, yet contradictory findings
44
-------
from the same authors supporting an alternative argument could just as easily have been cited.
The discussion includes strong generalizations that may not hold up to close scrutiny.
There has been one published study to date that has examined biomarkers of acrylamide
exposure and human cancer risk. Olesen etal. (2008) characterized hemoglobin adducts of
acrylamide and glycidamide in a case-control study of post-menopausal breast cancer. The
authors found no association between levels of glycidamide hemoglobin adducts and breast
cancer risk. Moreover, they found no overall association between acrylamide adducts and risk.
Upon adjustment for smoking status, however, they observed a 2.7-fold (1.1-6.6) increased risk
restricted to ER+ breast cancer per 10-fold increase in acrylamide-hemoglobin level. With
respect to this study design, the authors did not match or restrict the cases and controls on
smoking status, which raises concern given the very strong link between smoking and
acrylamide adducts. Interpretability of the Olesen study with respect to supporting the mode of
carcinogenic action should be taken cautiously.
For very high levels of acrylamide exposure, the animal and other experimental data do
support a mutagenic effect of acrylamide. It has been questioned whether such a mechanism
might also apply to lower doses (and indeed, at the lowest doses to which humans are exposed),
because of uncertainty about whether the compensatory mechanisms are in place to detoxify
acrylamide. But data clearly indicate that glycidamide is formed. There are the consistent
observations in humans of glycidamide-hemoglobin adducts (Bjellaas etal., 2007; Chevolleau et
al., 2007; Vesper et al., 2006, 2007) or glycidamide urinary metabolites (Urban et al., 2006) ,
including children (Heudorf etal. 2008), thus demonstrating the widespread internal exposure to
the putative mutagenic metabolite of acrylamide at ongoing low levels of exposure in the general
population.
The Panel did not consider the carcinogenicity to be hormonally-related. The existing
short-term mouse studies in SENCAR, ICR (skin) and A/J (lung) show no such selectivity of
carcinogenicity for hormonally regulated tissues. Also, the Panel discussed the fact
acrylamide/glycidamide is not unique among DNA-reactive epoxides for carcinogenic action in
thyroid, peritesticular mesothelium, and mammary tissue (e.g., glycidol, ethylene oxide). In
addition, this argument does not consider the CNS tumors observed in both chronic acrylamide
cancer bioassays, a site that was discussed by the Panel as representing strong evidence for a
DNA-damaging mechanism (cf. Rice, 2005). Finally, a recent publication considered by the
45
-------
Panel of short-term exposures to high doses of acrylamide in male F344 rats found essentially no
evidence for hormonal dysregulation in the hypothalamus-pituitary-thyroid axis based on
measurements of gene expression, neurotransmitters, hormones, and histopathology (Bowyer et
a/., 2008). Some studies of chronic low dose exposure, such as the cohort study of acrylamide
and ovarian/endometrial cancers (Hogervorst etal., 2007) and others (Khan etal., 1999) have
shown positive associations with hormones. The Panel encourages the Agency to review all
relevant new data that has been published since their completion of the current draft assessment
as they revise and finalize this IRIS document.
Charge Question 20. Are there other MO As that should be considered? Is there significant
biological support for alternative MO As for tumor formation, or for alternative MOAs to be
considered to occur in conjunction with a mutagenic MO A? Please specifically comment on
the support for hormonal pathway disruption. Are data available on alternate MOAs sufficient
to quantitate a dose-response relationship?
No, there is not significant biological support for MOA alternatives to the mutagenic
MO A, and data on any putative alternate MOAs are not sufficient to quantify dose response
relationships. It must be emphasized that more than one MOA may operate for a given
carcinogenic chemical, and the likelihood that more than a single MOA is operative increases as
levels of exposure increase. Some well-documented non-DNA reactive MOAs appear to be
high-dose phenomena. These are often important for understanding bioassay results in
experimental animals, and sometimes for high-exposure situations in human experience, but they
are usually less important because they represent negligible risks when cumulative human
exposures to these and similarly acting compounds fall considerably below bioassay dosage
levels. MOAs that can occur both in experimental rodents and in humans and that operate both
at bioassay dosage levels in experimental animals and at lower levels as well, into the human
exposure range, are most significant for humans. In general, for chemicals such as acrylamide
where there is a compelling body of data to support a DNA-reactive MOA via biotransformation
to glycidamide, the evidence for alternative or additional high-dose MOAs would have to be
convincing to explore alternative approaches to dose response and risk assessment. One caveat
that should be mentioned is that mutations induced by acrylamide are observed following high
46
-------
doses. There are similarly acting agents, such as methylmethanesulfonate (MMS) that create N7-
Guanine, the same DNA adducts, as does glycidamide yet show a threshold for mutations. These
data are consistent with robust repair mechanisms for the specific type of DNA adducts produced
by glycidamide and MMS. However, it should also be noted that low dose exposures have not
been tested in animal mutation studies and NOAELs have not yet been established. Therefore
future research should include dose response analyses to stringently test the relationship between
DNA adducts and mutations and gain a better understanding of the effects at lower doses. The
Agency should mention the finding of inhibition of kinesin motor proteins as a newly-identified
and potential site of action of AA or GA in the production of carcinogenicity (Sickles et a/.,
2007).
Occasionally high-dose or "unique rodent-specific" MO As may be invoked or postulated
to discredit bioassay results as irrelevant to humans, especially when such putative MOAs are
observed uniquely in non-human species. Such a postulated MOA needs to be very precisely
defined and its relevance thoroughly investigated and critically tested before the postulated MOA
is accepted by the biomedical and risk assessment communities. Any MOA developed for a
single substance is at best speculative until a general pattern can be rigorously demonstrated for a
family of substances that operate via the same MOA. The hormonal disruption MOAs proposed
for acrylamide as tissue-specific alternatives to a DNA-reactive MOA are highly speculative, are
supported by at most limited evidence, and do not meet this standard as noted in response to
charge question 19. The data are insufficient for characterizing dose-response relationships for
any of these proposed alternatives.
Charge Question 21. Two chronic drinking water exposure bioassays in Fischer 344 rats
(Friedman et al, 1995; Johnson et al, 1986) were used to derive the oral slope factor, and to
identify the tumors of interest for the MOA discussion. Are the choices for the studies,
tumors, and methods to quantify risk transparent, objective, and reflective of the current
science? Do you have any suggestions that would improve the presentation or further reduce
the uncertainty in the derived values?
The two chronic bioassay studies in F344 rats are the main studies to consider in dose
response analysis. Overall the document does a good job discussing these studies, but the
47
-------
rationale for using only the Friedman et al. study for derivation of the oral slope factor is
problematic. The strengths and limitations of both studies should be discussed in greater depth.
The text describes the Friedman et al. study as "superior" and "larger and better designed" but
the Panel does not agree that this is the case, and recommends that both studies be subjected to
modeling for the purposes of deriving oral slope factors. The two studies may have fairly
similar oral slope factors. At a minimum, estimates for the second study should also be presented
to clarify the impact of study selection in the uncertainty discussion.
The methods to quantify risk are transparently presented and reflective of current science,
with the exception that a factor to scale for pharmacodynamic differences in potency between
humans and animals has not been applied. The development of unit risk based on HEC accounts
for the pharmacokinetic but not pharmacodynamic differences, and in such situations EPA's
2005 Guidelines for Carcinogen Risk Assessment (p. 3-7) indicates inclusion of a
pharmacodynamic factor be considered. The potential human variability in cancer response
attributable to human pharmacokinetic variability in handling acrylamide should be discussed
qualitatively and analyzed quantitatively. Hemoglobin adduct data could provide the basis for
such an analysis. The assumption underlying the modeling, that each and every individual of
the same age exposed to the same external dose faces the same risk of cancer, is inconsistent
with these data.
With respect to study selection, one of the reasons for not using the Johnson study had to
do with the rates of CNS tumors in this study, particularly in the controls. The Friedman et al.
study was designed "to investigate whether glial tumors in the Johnson et al. study were
significant." But, as JAice (2005) points out, the histopathological examination for glial tumors
was incomplete. Only one-fifth of the 1.0 mg/kg-day dose females' spinal cords were subjected
to histopathological examination, even though one-third of the glial tumors in the Johnson et al.
study were seen in the spinal cord. The approach to the evaluation of CNS tumors in Friedman et
al. was seen by the Panel as a significant study limitation.
Another improvement over the Johnson study noted in the document for the Friedman et
al. study was different and presumably better dose intervals. The doses for males in the
Friedman et al. and Johnson et al. studies were the same, except Johnson et al. had one
additional lower dose group. The doses in Friedman for females were 1.0 and 3.0 mg/kg-day
compared to 0.01, 0.1, 0.5 and 2.0 mg/kg-day for the Johnson study. The Friedman study did
48
-------
extend the high end of the dose response range for females and did offer a more complete dose
response function for thyroid tumors, employed somewhat larger dose groups (100 per group and
two control groups). But Johnson et al. did have 60 animals per dose group, did provide a
complete histopathological evaluation, and had more dose groups than a standard bioassay.
Another limitation of the Friedman et al. study is that the degree of histopathological
examination of oral tissue is unclear. The Friedman study does not tabulate findings for certain
tumor sites seen in the Johnson study, so quantitative comparisons are not possible and the reader
is not able to consider these sites or perform independent evaluations regarding the significance
of the findings. It appears EPA may have the data needed to do the analysis since it was able to
do a time-dependent analysis for slope estimation using the Tegeris Lab report. EPA could then
look at the data and analyze as appropriate the data for these sites.
A criticism about the possible impact of a sialodacryoadenitis virus on tumor findings
had been raised and was another reason given for using the Friedman study. On the other hand,
US FDA had raised some issues in auditing the Friedman et al. study regarding environmental
controls at the lab facility and the possibility of some under-dosing of animals. Ultimately both
studies have strengths and weaknesses and on balance neither seems clearly superior. Both are
reasonably strong studies, and thus oral slope estimates should be presented for both studies.
Some comments regarding details on tumor data presentation and analysis in the EPA
draft document follow:
Tests for dose-related trends should be conducted and presented for the all tabulated sites.
By Fisher's exact test, the mammary tumors in the 0.5 mg/kg-d group in the Friedman et al.
study are significant (p<0.05). The statistics used in the draft document that correct for
intercurrent mortality should be re-checked. It appears this group has a treatment-related finding
and this should be noted and the discussion that this group is devoid of treatment-related tumors
(page 75) changed. The clitoral gland findings in the Johnson et al. study stand out because
histology was done only on clitoral tissues observed with gross masses. This is worth an
explanatory footnote. Also given the approach taken to collecting this tissue, the clitoral tumors
in the 0.5 mg/kg dose group also appear worthy of note. All four masses analyzed indicated
tumor compared to none in controls (p<0.1). In the Friedman et al. study, CNS tumors of glial
origin should be combined for analysis as was done by WHO 2006. Considering the findings of
glial tumors in females in the Johnson study, the dose related trend for both sexes in the
49
-------
Friedman study, although falling a hair short of statistical significance at the p ^0.05 level,
provide some evidence of a CNS glial cell effect in the Friedman study. This should be
discussed. Also, the extent of examination of oral tissue in the Friedman study is unclear.
Finally, the Friedman study employed two control groups for the male rats that do not differ
from one another. For the statistical treatments, there is no apparent reason why these groups
should not be combined. The Toxicological Review did this for the dose response analysis but
may not have done the same for the pair-wise comparisons.
The data choice for modeling to address the discrepancy between the Friedman et al. and
the Tegeris laboratory reporting of thyroid tumors for the male noted in Appendix D of EPA's
draft document was appropriate. A final minor point, in the discussion of the confidence in dose
response analysis in chapter 6 (page 229), issues are raised that seem better placed in the
discussion of the hazard characterization.
Charge Question 22. The cancer slope factor (CSF) derivation includes an adjustment for
early mortality (i.e., time-to-tumor analysis). Is this adjustment scientifically supported in
estimating the risk from the 2-year bioassay data for increased incidence of tumors in the
rats?
The use of the Weibull-in-time multistage-in-dose analysis is a reasonable and
scientifically justifiable way to take into account the early mortality in the high dose group in the
male study. The decision not to employ this analysis in the case of the females is also reasonable
since mortality across treatment and control groups did not differ and the overall survival appears
to be fairly good.
Charge Question 23. Please comment on whether A UCfor glycidamide is the best choice of
the dose metric in estimating human equivalent concentration to derive the oral slope factor.
The Panel agreed that using the AUC for glycidamide is the best choice for estimating the
human equivalent concentration to derive the oral slope factor. This decision was based on the
strong evidence from experimental results that the AUC was linearly correlated with adduct
levels in single/repeat dosing studies. There was agreement that glycidamide is the more
50
-------
mutagenic metabolite based on experimental studies. The Panel felt there was good
documentation in the report regarding the correlation between levels of DNA adducts and extent
of mutations in vivo. Moreover, the metabolic conversion of acrylamide to glycidamide supports
the MOA.
One consideration in using this as the dose metric, however, comes from some of the
human studies in which variability is not accounted for adequately, specifically, inter-individual
variation is not assessed and that the value used for cross-species comparisons is based on small
numbers of healthy adult male humans. This is discussed at greater length in response to
Question 8. Consideration of additional human data (e.g., Vesper etal., 2006) to evaluate the
degree humans form glycidamide from acrylamide is clearly warranted. Such data may provide
the basis for comparing human acrylamide and glycidamide AUCs, using methodology of
Calleman, Bergmark and colleagues (Bergman etal., 1991). This in turn can provide an
improved basis for adjustments for cross-species differences in pharmacokinetics, as well as
human variability in glycidamide formation from acrylamide.
Charge Question 24. As with the RfC, there were insufficient cancer inhalation data to derive
an inhalation unit risk (IUR). The PBTK model was used in a route-to-route extrapolation of
the dose-response relationship from the oral data, and to estimate the human equivalent
concentration for inhalation exposure to acrylamide. Please comment on whether this
extrapolation to derive the inhalation unit risk was correctly performed and sufficiently well
documented.
The response to this question is nearly identical to the response to charge question #11.
The Panel agreed with the use of PBTK modeling to conduct dose-route extrapolation and
commended the EPA for using the PBTK model to fill the gap resulting from the absence of
robust animal toxicology studies investigating neurotoxicity via the inhalation route that would
support the development of an IUR. The Panel agreed that the absence of evidence for route of
entry specific effects would allow route-to-route extrapolation for deriving an IUR based on
using the PBTK model to calculate the human equivalent concentration (HEC). This would yield
an equivalent internal dose (Glycidamide- AUC) associated with those achieved at the point of
departure from the oral sentinel (Johnson et al.) studies. The Panel noted that few inhalation PK
51
-------
studies exist to allow a robust parameterization of the inhalation component of the PBTK model
for either rats or humans. Despite this, the Panel noted that acrylamide is very water soluble and
non-volatile, and the compound has a relatively long half life. Therefore, the absorption of
acryalmide via inhalation should be nearly complete, and first pass effects are negligible, thereby
making the pharmacokinetics of acrylamide via inhalation easy to extrapolate from simple
principles of pharmacokinetics. The Panel agreed that the application of pharmacokinetic
approaches (e.g,, the use of the PBTK model) reduces uncertainty associated with animal to
human extrapolation and thus warrants replacing the default uncertainty factor associated with
interspecies extrapolation for pharmacokinetic differences as was done for deriving the RfD.
The use of the PBTK model however does not address cross-species differences in
pharmacodynamics, which should be considered, following the Agency's 2005 Guidelines for
Carcinogen Risk Assessment.
The Panel noted that the air concentration one would derive using the default approach
(multiply the HED by body weight [70 kg] and dividing by daily inhalation rate [20 m3/day]
yielding 0.266 |ig/m3) is very similar to the HEC derived using the PBTK model (0.25 |ig/m3).
Therefore, if the EPA decides to also provide an extrapolation based on measured data (as
described in the response to charge question 8), the default approach of extrapolating from an
absorbed oral dose to an equivalent intake from the inhalation route (multiplying by 70 kg and
dividing by 20 m3/day) can be used with confidence to calculate the IUR.
Charge Question 25. The recommendation to use the age-dependent adjustment factors
(ADAFs) is based on the determination of a mutagenic MO A for carcinogenicity. Is this
recommendation scientifically justifiable and transparently and objectively described?
The recommendation to use the age-dependent adjustment factors is well justified and
transparently and objectively described. The Panel's deliberations regarding quantitating age-
dependent adjustment factors (Section 5.4.6) followed from discussions of a mutagenic mode of
action for acrylamide and the typically enhanced sensitivity of fetal and neonatal animals from
exposure to such agents in accordance with EPA's Supplemental Guidance for Assessing
Susceptibility from Early Life Exposure to Carcinogens (2005b). The Panel also discussed the
value of using the PBTK model to evaluate the effect of lifestage on CYP 2E1 and glutathione
52
-------
levels that could modify internal exposure to glycidamide. Such modeling results could be used
to reduce the uncertainty associated with lifestage extrapolations and the derivation of age-
dependent adjustment factors. Such efforts would be directed at pharmacokinetic aspects of the
age-dependent adjustment factors. Uncertainty regarding pharmacodynamics would remain to
be addressed by the age-dependent adjustment factors.
Charge Question 26. Please provide any other comments on the CSF or IUR, and on the
discussion of uncertainties in the cancer assessment.
The discussion of uncertainties is good, but human variability could be addressed in
greater length. It is unclear why in Table 5-13 the consideration/approach is "Method used to
protect sensitive populations." There is no characterization of sensitive populations, and this
could be explored and discussed to a much greater extent.
Specifically, not enough attention was paid to consequences of individual differences in
metabolism and cancer risk. Both the CYP2E1 polymorphisms and glutathione transferase(s)
(even though rodent data suggests no role for this pathway) polymorphisms could be looked at in
human populations. The degree to which increased activity influences the risk should be
considered, including whether this might be tumor site dependent. Also, much weight is put on
the two chronic studies in the Fischer344 rat. The limitations of not having another rodent
species should be discussed in more detail with respect to other carcinogens where 2 species
were evaluated and similar or different results were found.
A factor to scale for toxicodynamic differences between humans and animals was not
included in the derivation of the CSF and IUR. The 2005 EPA Carcinogenic Risk Assessment
Guidelines (see e.g., Guidelines pp 1-13 and 3-7) discusses how toxicodynamics can be
addressed by such a factor. The development of unit risk-based on HEC accounts for the
toxicokinetic but not toxicodynamic interspecies differences.
53
-------
ABREVIATIONS
ADAF age-dependent adjustment factor
AM-GSH Acrylamide-Glutathione
AUC area under the curve
BMD benchmark dose
BMDL benchmark dose level
BMR benchmark response
CNS Central Nervous System
CSAF Chemical-specific Adjustment Factors
CSF Cancer slope factor
DNA Deoxyribonucleic Acid
EPA Environmental Protection Agency
FQPA Food Quality Protection Act
GA or Gly Glycidamide
GA-GSH Glycidamide-Glutathione
FIEC Human Equivalent Concentration
IARC International Agency for Research on Cancer ()
IRIS Integrated Risk Information System
IUR inhalation unit risk
LOAEL Lowest Adverse Effect Level
MMS Methylmethanesulfonate
MO A mode of action
MOE Margin of Exposure
NMA N-Methylol aery 1 amide
NO Nitric Oxide
NOAEL No Adverse Effect Level
NTP/CERHR National Toxicology Program
PBPK physiologically-based pharmacokinetic
PBTK physiologically-based toxicokinetic
PK Pharmacokinetic
POD point of departure
RfC reference concentration
RfD reference dose
TP tumor promoter
UF uncertainty Factor
WHO World Health Organization
54
-------
REFERENCES
Abou-Donia, MB, Ibrahim, SM, Corcoran, JJ, etal. (1993)Neurotoxicity of glycidamide, an
acrylamide metabolite, following i.p. injections in rats. J ToxicolEnviron Health 39:447-464.
Adler, I-D. (1990) Clastogenic effects of acrylamide in different germ-cell stages of male mice.
In: Biology of Mammalian Germ Cell Mutagenesis, Banbury Report Vol. 34.
Adler, I-D, Reitmer, P, Schmoller, R, etal. (1994) Dose response for heritable translocations
induced by acrylamide in spermatids of mice. Mutat Res 309:285-291.
Barber, D.S., Hunt, J.R., Ehrich, M.F., Lehning, EJ. and LoPachin, R.M. (2001). Metabolism,
toxicokinetics and hemoglobin adduct formation in rats following subacute and subchronic
acrylamide dosing. Neurotoxicology 22, 341-353.
Barber, D.S. and LoPachin, R.M. (2004). Proteomic analysis of acrylamide-protein adduct
formation in rat brain synaptosomes. Toxicol. Appl. Pharmacol. 201, 120-136.
Barber, D.S. Stevens, S. and LoPachin, R.M. (2007). Proteomic analysis of rat striatal
synaptosomes during acrylamide intoxication at a low dose rate. Toxicol. Sci. 100, 156-167.
Bergmark, E., Calleman, CJ. and Costa, L.G. (1991). Formation of hemoglobin adducts of
acrylamide and its epoxide metabolite glycidamide in the rat. Toxicol. Appl. Pharmacol. Ill,
352-363.
Besaratinia, A. and Pfeifer, G.P. (2007). A review of mechanisms of acrylamide carcinogenicity.
Carcinogenesis 28, 519-528.
Bjellaas T., Olesen P.T., Frandsen H., Haugen M., St01en L.H., Paulsen I.E., Alexander J.,
Lundanes E., Becher G. (2007). Comparison of estimated dietary intake of acrylamide with
hemoglobin adducts of acrylamide and glycidamide. Toxicol Sci. 98(1): 110-7.
55
-------
Boettcher, M. I, Schettgen, T., Kutting, B., Pischetsrieder, M. and Angerer, J. (2005).
Mercapturic acids of aery 1 amide and glycidamide as biomarkers of the internal exposure to
acrylamide in the general population. Mutat Res 580:167-76.
Bonassi, S, Lando, C, Ceppi, M, etal. (2004) No association between increased levels of high-
frequency sister chromatid exchange cells (HFCs) and the risk of cancer in healthy individuals.
Environ Mol Mutagen. 43 (2): 13 4-6.
Bowyer, J.F., Latendresse, J.R., Delongchamp, R.R., Muskhelishvili, L., Warbritton, A.R.,
Thomas, M., Tareke, E., McDaniel, L.P., Doerge, D.R. (2008). The effects of subchronic
acrylamide exposure on gene expression, neurochemistry, hormones, and histopathology in the
hypothalamus-pituitary-thyroid axis of male Fischer 344 rats. Toxicol Appl Pharmacol.
15;230(2):208-15.
Brady, S. T., Pfister, K. K., and Bloom, G. S. (1990). A monoclonal antibody against kinesin
inhibits both anterograde and retrograde fast axonal transport in squid axoplasm. Proceedings of
the National Academy of Science USA 87, 1061-1065.
Brat DJ, Brimijoin S. (1993) Acrylamide and glycidamide impair neurite outgrowth in
differentiating N1E. 115 neuroblastoma without disturbing rapid bidirectional transport of
organelles observed by video microscopy. JNeurochem. 60(6):2145-52.
Burek, J.D., Albee, R.R., Beyer, I.E., Bell, T.J., Carreon, R.M., Morden, D.C., Wade, C.E.,
Hermann, E.A. and Gorzinski, SJ. (1980). Subchronic toxicity of acrylamide administered to
rats in drinking water followed by up to 144 days of recovery. J Environ Pathol Toxicol. 4, 157-
182.
Calleman, CJ, Wu, Y, Tian G, et al. (1994) Relationships between biomarkers of exposure and
neurological effects in a group of workers exposed to acrylamide. Toxicol Appl Pharmacol
126:361-371.
56
-------
Cavanagh, J. B. (1964). The significance of the "dying-back" process in experimental and human
neurological disease. IntRevExp Pathol3, 219-267.
Chapin, RE, Fail, PA, George, JD, etal. (1995) The reproductive and neuronal toxicities of
acrylamide and three analogues in Swiss mice, evaluated using the continuous breeding protocol.
Fundam Appl Toxicol 27:9-24.
Chevolleau S., Jacques C., Canlet C., Tulliez J., Debrauwer L. (2007). Analysis of hemoglobin
adducts of acrylamide and glycidamide by liquid chromatography-electrospray ionization
tandem mass spectrometry, as exposure biomarkers in French population. J Chromatogr A.
1167(2): 125-34.
Clarke, C. H., and Sickles, D. W. (1996). Decreased GAP-43 accumulation in neurite tips of
cultured hippocampal neurons by acrylamide. Neurotoxicology 17, 397-406.
Costa, LG, Deng, H, Gregotti, C, etal. (1992) Comparative studies on the neuro- and
reproductive toxicity of acrylamide and its epoxide metabolite glycidamide in the rat.
NeuroToxicology 13:219-224.
Costa, LG, Deng, H, Calleman, CJ, etal. (1995) Evaluation of the neurotoxicity of glycidamide,
an epoxide metabolite of acrylamide: behavioral, neurochemical and morphological studies.
Toxicology. 98:151-161.
Crofton, K.M., Padilla, S., Tilson, H.A., Anthony, D.C., Raymer, J.H. and MacPhail, R.C.
(1996). The impact of dose-rate on the neurotoxicity of acrylamide: the interaction of
administered dose, target tissue concentrations, tissue damage and functional effects. Toxicol.
Appl. Pharmacol. 139, 163-176.
Dearfield, KL, Douglas, GR, Ehling, UH, etal. (1995) Acrylamide: a review of its genotoxicity
and an assessment of heritable genetic risk. Mutat Res. 330:71-99.
57
-------
DeGrandchamp, R. L., and Lowndes, H. E. (1990). Early degeneration and sprouting at the rat
neuromuscular junction following acrylamide administration. Neuropathology and Applied
Neurobiology 16, 239-254.
DeGrandchamp, R. L., Reuhl, K. R., and Lowndes, H. E. (1990). Synaptic terminal degeneration
and remodeling at the rat neuromuscular junction resulting from a single exposure to acrylamide.
Taxicol Appl Pharmacol. 105:422-43 3.
Doerge, D.R., Gamboa da Costa, G., McDaniel, L.P., Churchwell, M.I., Twaddle, N.C. and
Beland, F.A. (2005a). DNA adducts derived from administration of acrylamide and glycidamide
to mice and rats. Mut. Res. 580:131-141.
Doerge, D. R., Young, J. F., McDaniel, L. P., Twaddle, N. C., and Churchwell, M. I. (2005b).
Toxicokinetics of acrylamide and glycidamide in Fischer 344 rats. Toxicol Appl Pharmacol.
208:199-209.
Ehling UH, Neuhauser-Klaus A. (1992) Reevaluation of the induction of specific-locus
mutations in spermatogonia of the mouse by acrylamide. Mutat Res. 283(3):185-91.
Feng, Y. and Forgac, M. (1992b). Cysteine 254 of the 73-kDa A subunit is responsible for
inhibition of the coated vesicle (H+)- ATPase upon modification by sulfhydryl reagents. J. Biol.
Chem. 267, 5817-5822.
Fennell, T.R. and Brown, C.D. (2001). A physiologically-based pharmacokinetic model for
ethylene oxide in mouse, rat, and human. Toxicol Appl Pharmacol. 173:161-75.
Fennell, T.R. and Friedman, M.A. (2005). Comparison of acrylamide metabolism in humans and
rodents. In: Chemistry and Safety of Acrylamide in Food. (M. Friedman andD. Mottram, eds).
Springer Science+Business Media, Inc., NY. pp 109-116.
58
-------
Fennell, T. R., Sumner, S. C., Snyder, R. W., Burgess, J., Spicer, R., Bridson, W. E. and
Friedman, M. A. (2005). Metabolism and hemoglobin adduct formation of acrylamide in
humans. ToxicolSci. 85:447-59.
Fennell, T. R., Sumner, S. C., Snyder, R. W., Burgess, J. and Friedman, M. A. (2006). Kinetics
of elimination of urinary metabolites of acrylamide in humans. Toxicol Sci. 93:256-67.
Friedman, M.A., Dulak, L.H. and Stedham, M.A. (1995). A lifetime oncogenicity study in rats
with acrylamide. FundAppl Toxicol. 27:95-105.
Generoso, WM, Sega, GA, Lockhart, AM, etal. (1996) Dominant lethal mutations, heritable
translocations, and unscheduled DNA synthesis induced in male mouse germ cells by
glycidamide, a metabolite of acrylamide. Mutat Res. 371:175-183.
Ghanayem, BI, Witt, KL, El-Hadri, L, etal. (2005a) Comparison of germ cell mutagenicity in
male CYP2El-null and wild-type mice treated with acrylamide: evidence supporting a
glycidamide-mediated effect. BiolReprod. 72(1): 157-63.
Ghanayem, BI, Witt, K, Kissling, GE, etal. (2005b) Absence of acrylamide-induced
genotoxicity in CYP2El-null mice: Evidence consistent with a glycidamide-mediated effect.
Mutat Res. 578:284-297.
Gho, M., McDonald, K., Ganetzky, B., and Saxton, W. M. (1992). Effects of kinesin mutations
on neuronal functions. Science. 258:313-316.
Gould, R.M. and Brady, S.T. (2004) Neuropathology: many paths lead to hereditary spastic
paraplegia. Current Biology. 14:R903-R904.
Hagmar, L, Tornqvist, M, Nordander, C, etal. (2001) Health effects of occupational exposure to
acrylamide using hemoglobin adducts as biomarkers of internal dose. ScandJ Work Environ
Health. 27(4):219-226.
59
-------
Hartmann, B.C., Boettcher, M.I., Schettgen, T., Fromme, H., Drexler, H., Angerer, J. (2008).
Hemoglobin adducts and mercapturic acid excretion of acrylamide and glycidamide in one study
population. JAgric FoodChem. 56(15):6061-8.
Harris, CH, Gulati, AK, Friedman, MA, etal. (1994) Toxic neurofilamentous axonopathies and
fast axonal transport. V. Reduced bidirectional vesical transport in cultured neurons by
acrylamide and glycidamide. J Tox andEnvion Health. 42:343-356.
Heudorf, U., Hartmann, E., Angerer, J. (2008). Acrylamide in children - exposure assessment via
urinary acrylamide metabolites as biomarkers. Int JHygEnviron Health. [Jun 12 Epub ahead of
print]
Hinson JA, Roberts DW. (1992) Role of covalent and noncovalent interactions in cell toxicity:
effects on proteins. Annu Rev Pharmacol Toxicol. 32:471-510.
Hogervorst JG, Schouten LJ, Konings EJ, Goldbohm RA, van den Brandt PA. (2007). A
prospective study of dietary acrylamide intake and the risk of endometrial, ovarian, and breast
cancer. CancerEpidemiolBiomarkersPrev. 16(11):2304-13.
Kurd, D. D., and Saxton, W. M. (1996). Kinesin mutations cause motor neuron disease
phenotypes by disrupting fast axonal transport in drosophila. Genetics 144:1075-1085.
Johnson, K.A., Gorzinski, S.J., Bodner, K.M., Campbell, R.A., Wold, C.H., Friedman, M.A. and
Mast, R.W. (1986). Chronic toxicity and oncogenicity study on acrylamide incorporated in the
drinking water of Fischer 433 rats. Toxicol. Appl. Pharmacol. 185:154-168.
Khan, M.A., Davis, C.A., Foley, G.L., Friedman, M.A. and Hansen, L.G. (1999). Changes in
thyroid gland morphology after acute acrylamide exposure. Toxicological Sciences. 47: 151-157.
60
-------
Kirman, C. R., Gargas, M. L., Deskin, R., Tonner-Navarro, L. and Andersen, M. E. (2003). A
physiologically based pharmacokinetic model for acrylamide and its metabolite, glycidamide, in
the rat. J Toxicol Environ Health A 66:253-74.
Lehning, E.J., Persaud, A., Dyer, K.R., Jortner, B.S., LoPachin, R.M. (1998). Biochemical and
Morphologic characterization of acrylamide peripheral neuropathy. Toxicol Appl Pharmacol.
151:211-221.
Lehning, E.J., Balaban C.D., Ross J.F. and LoPachin R.M. (2002a). Acrylamide neuropathy. I.
Spatiotemporal characteristics of nerve cell damage in rat cerebellum. Neurotoxicology. 23:397-
414.
Lehning, E.J., Balaban C.D., Ross J.F. and LoPachin R.M. (2002b). Acrylamide neuropathy. II.
Spatiotemporal characteristics of nerve cell damage in brainstem and spinal cord.
Neurotoxicology. 23:415 -429.
Lehning, E.J., Balaban C.D., Ross J.F. and LoPachin R.M. (2003). Acrylamide neuropathy. III.
Spatiotemporal characteristics of nerve cell damage in forebrain. Neurotoxicology 24:125-136.
LoPachin R.M. and Lehning E.J. (1994). Acrylamide-induced distal axon degeneration: A
proposed mechanism of action. Neurotoxicology. 15:247-260.
LoPachin R.M., Ross J.F. and Lehning E.J. (2002a). Nerve terminals as the primary site of
acrylamide action: A hypothesis. Neurotoxicology. 23:43-60.
LoPachin R.M., Ross J.F., Reid, M.L., Das, S., Mansukhani, S. and Lehning E.J. (2002b).
Neurological evaluation of toxic axonopathies in rats: acrylamide and 2,5-hexanedione.
Neurotoxicology. 23:95-l 10.
LoPachin, R.M., Balaban, C.D. and Ross, J.F. (2003). Acrylamide axonopathy revisited.
Toxicol. Appl. Pharmacol. 188:135-153.
61
-------
LoPachin, R.M., Schwarcz, A.I., Gaughan, C.L., Mansukhani, S. and Das, S. (2004). In vivo and
in vitro effects of acrylamide on synaptosomal neurotransmitter uptake and release.
NeuroToxicology 25:349-363.
LoPachin, R.M. and DeCaprio, A.P. (2005). Protein adduct formation as a molecular mechanism
in neurotoxicity. Tox Sci. 86;214-225.
LoPachin, R.M. Barber, D.S., He, D. and Das, S. (2006a). Acrylamide inhibits dopamine uptake
in rat striatal synaptic vesicles. Tox Sci. 89:224-234.
LoPachin, R.M. and Barber, D.S. (2006b). Synaptic cysteine sulfhydryl groups as targets of
electrophilic neurotoxicants. Tox. Sci. 94;240-255.
LoPachin, R.M., Barber, D.S., Geohagen, B.C., Gavin, T., He, D. and Das, S. (2007a). Structure-
toxicity analysis of Type-2 alkenes: in vitro neurotoxicity. Tox. Sci. 95:136-146.
LoPachin, R.M., Gavin, T., Geohagen, B.C. and Das, S. (2007b). Neurotoxic mechanisms of
electrophilic type-2 alkenes: soft-soft interactions described by quantum mechanical parameters.
Tox. Sci. 98:561-570.
LoPachin, R.M., Barber, D.S. and Gavin, T. (2008a). Molecular mechanism of the conjugated
a,|3-unsaturated carbonyl derivatives: relevance to neurotoxicity and neurodegenerative diseases.
Tox. Sci. 104(2):235-49..
LoPachin, R.M., Gavin, T., Geohagen, B.C. and Das, S. (2008b). Synaptosomal toxicity and
nucleophilic targets of 4-hydroxy-2-nonenal. Tox. Sci. submitted.
Moser, V.C., Anthony, D.C., Sette, W.F. andMacPhail, R.C. (1992). Comparison of subchronic
neurotoxicity of 2-hydroxyethyl acrylate and acrylamide in rats. Fund. Appl Toxicol. 18:343-
352.
62
-------
Olesen PT, Olsen A, Frandsen H, Frederiksen K, Overvad K, Tj0nneland A. (2008). Acrylamide
exposure and incidence of breast cancer among postmenopausal women in the Danish Diet,
Cancer and Health Study. IntJ Cancer. 122(9):2094-100.
Pearson, RG and Songstad, J. (1967) Application of the principle of hard and soft acids and
bases to organic chemistry. J. Am. Chem. Soc. 89: 1827-1836.
Reid, E. (2003) Science in motion: commom moleculare pathological themes emerge in the
hereditary spastic paraplegias. JMed Genet. 40:81-86.
Reid, E., Kloos, M., Ashley-Koch, A., Hughes, L., Bevan, S., Svenson, I. K., Graham, F. L.,
Gaskell, P. C., Dearlove, A., Pericak-Vance, M. A., Rubinsztein, D. C., and Marchuk, D. A.
(2002). A Kinesin Heavy Chain (KIF5A) Mutation in Hereditary Spastic Paraplegia (SPG10).
Am. J. Hum. Genet. 71:1189-1194.
Rice, J.M. (2005). The carcinogenicity of acrylamide. Mutation Res. 580, 3-20.
Russell, LB, Hunsicker, PR, Cacheiro, NL, et al. (1991) Induction of specific-locus mutations in
male germ cells of the mouse by acrylamide monomer. Mutat Res. 262:101-107.
Sabri, M. L, and Spencer, P. S. (1990). Acrylamide impairs fast and slow axonal transport in rat
optic system. NeurochemicalResearch 15:603-608.
Sega, GA, Valdivia Alcota, RP, Tancongco, CP, etal. (1989) Acrylamide binding to the DNA
and protamine of spermiogenic stages in the mouse and its relationship to genetic damage. Mutat
Res. 216:221-230.
Shelby, MD, Cain, KT, Cornett, CV, etal. (1987) Acrylamide: Induction of heritable
translocations in male mice. EnvironMutagen. 9:363-368.
63
-------
Sickles, D. W. (1989a). Toxic neurofilamentous axonopathies and fast anterograde axonal
transport. I. The effects of single doses of acrylamide on the rate and capacity of transport.
Neurotoxicology. 10:91-102.
Sickles, D. W. (1989b). Toxic neurofilamentous axonopathies and fast anterograde axonal
transport. II. The effects of single doses of neurotoxic and non-neurotoxic diketones and B,B'-
iminodipropionitrile (IDPN) on the rate and capacity of transport. Neurotoxicolog.y 10: 103-112.
Sickles, D. W. (1991). Toxic neurofilamentous axonopathies and fast anterograde axonal
transport. III. Recovery from single injections and multiple dosing effects of acrylamide and 2,5-
hexanedione. ToxicolApplPharmacol. 108:390-396.
Sickles, D. W. (1992). Toxic neurofilamentous axonopathies and fast anterograde axonal
transport. IV. In vitro analysis of transport following acrylamide and 2,5-hexanedione.
Toxicology Letters. 61:199-204.
Sickles, D. W., Brady, S. T., Testino, A. R., Friedman, M. A., and Wrenn, R. A. (1996). Direct
effect of the neurotoxicant acrylamide on kinesin-based microtubule motility. Journal of
Neuroscience Research. 46:7-17.
Sickles, D. W., Stone, J. D., and Friedman, M. A. (2002). Fast axonal transport: a site of
acrylamide neurotoxicity? Neurotoxicology 23:223-251.
Sickles, D.W., Sperry, A.O., Testino, A., Friedman, M. (2007) Acrylamide effects on kinesin-
related proteins of the mitotic/meiotic spindle. Toxicol Appl Pharmacol. 222(1): 111-21.
Spencer, P. S., and Schaumburg, H. H. (1991). A review of acrylamide neurotoxicity. Part 2.
Experimental animal neurotoxicity and pathologic mechanisms. Canadian Journal of
Neurological Sciences pp. 152, 69.
64
-------
Stone, J. D., Peterson, A. P., Eyer, 1, Oblak, T. G., and Sickles, D. W. (1999). Axonal
neurofilaments are non-essential elements of toxicant-induced reductions in fast axonal transport:
Video-enhanced differential interference microscopy in peripheral nervous system axons.
Toxicol Appl Pharmacol. 161:50-58.
Stone, J. D., Peterson, A. P., Eyer, J., and Sickles, D. W. (2000). Neurofilaments are non-
essential elements of toxicant-induced reductions in fast axonal transport: Pulse labeling in CNS
neurons. Neurotoxicology. 21, 447-458.
Tareke, E., Twaddle, N.C., McDaniel, L.P., Churchwell, M.I., Young, J.F., and Doerge, D.R.
(2006). Relationships between biomarkers of exposure and toxicokinetics in Fischer 344 rats and
B6C3F1 mice administered single doses of acrylamide and glycidamide and multiple doses of
acrylamide. Toxicol Appl Pharmacol. 217:63-75.
Tyl, RW, Friedman, MA, Losco, PE, et al. (2000) Rat two-generation reproduction and dominant
lethal study of acrylamide in drinking water. Reprod Toxicol. 14:385-401.
Urban M., Kavvadias D., Riedel K., Scherer G., Tricker A.R. (2006). Urinary mercapturic acids
and a hemoglobin adduct for the dosimetry of acrylamide exposure in smokers and nonsmokers.
Inhal Toxicol. 18(10):831-9.
U.S. EPA. (2002) A review of the reference dose and reference concentration processes. Risk
Assessment Forum, Washington, DC; EPA/630/P-02/0002F. Available from
http://www.epa.gov/iris/backgr-d.htm.
U.S. EPA. (2004) An Examination of EPA Risk Assessment Principles and Practices: Staff Paper
Prepared for the U.S. Environmental Protection Agency by members of the Risk Assessment
Task Force, Washington, DC; EPA/100/B-04/001. Available from:
http://www.epa.gov/osa/pdfs/ratf-fmal.pdf
65
-------
U.S. EPA (2005a) Guidelines for carcinogen risk assessment. Risk Assessment Forum,
Washington, DC; EPA/630/P-03/001B. Available from:
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=l 16283.
U.S. EPA (2005b) Supplemental Guidance for Assessing Susceptibility from Early Life
Exposure to Carcinogens. Risk Assessment Forum, Washington, DC; EPA/630/R-03/003F..
Available from: http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=l 16283.
Vesper, H.W., Ospina, M., Meyers, T., Ingrahm, L., Smith, A., Gray, J.G. and Meyers, G.L.,
(2006). Automated method for measuring globin adducts of acrylamide and glycidamide at
optimized Edman reaction conditions, Rapid Commun. Mass. Spectrom. 20:959-964.
Vesper, H.W., Bernert, J.T., Ospina, M., Meyers, T., Ingham, L., Smith, A., Myers, G.L. (2007).
Assessment of the relation between biomarkers for smoking and biomarkers for acrylamide
exposure in humans. Cancer EpidemiolBiomarkers Prev. 16(ll):2471-8.
Vesper, H.W., Slimani, N., Hallmans, G., Tj0nneland, A., Agudo, A., Benetou, V., Bingham, S.,
Boeing, H., Boutron-Ruault, M.C., Bueno-de-Mesquita, H.B., Chirlaque, D., Clavel-Chapelon,
F., Crowe, F., Drogan, D., Ferrari, P., Johansson, L, Kaaks, R., Linseisen, J., Lund, E., Manjer,
J., Mattiello, A., Palli, D., Peeters, P.H., Rinaldi, S., Skeie, G., Trichopoulou, A., Vineis, P.,
Wirfalt, E., Overvad, K., Stromberg, U. (2008). Cross-sectional study on acrylamide hemoglobin
adducts in subpopulations from the European Prospective Investigation into Cancer and
Nutrition (EPIC) Study. JAgric Food Chem. 56(15):6046-53.
Walker, R.A., O'Brien, E.T., Epstein, D.L., and Sheetz, M.P. (1997). N-ethylmaleimide and
ethacrynic acid inhibit kinesin binding to microtubules in a motility assay. CellMotility and the
Cytoskeleton. 37:289-299.
Walker, K., Hattis, D., Russ, A., Sonawane, B. and Ginsberg, G. (2007). Approaches to
acrylamide physiologically based toxicokinetic modeling for exploring child-adult dosimetry
differences. J ToxicolEnviron Health A 70:2033-55.
66
-------
Wilkinson, CF, Christoph, GR, Julien, E, Kelley, JM, Kronenberg, J, McCarthy, J and Reiss, R
(2000). Assessing the risks of exposures to multiple chemicals with a common mechanism of
toxicity: How to cumulate? Reg. Toxicol. Pharmacol. 31:30-43.
World Health Organization (2006). Joint FAO/WHO Expert Committee on Food Additives
(2005 : Rome, Italy) Evaluation of certain food contaminants : Sixty-Fourth Report of the Joint
FAO/WHO Expert Committee on Food Additives. WHO Technical Report Series No. 930.
WHO, Geneva.
Young, JF, Luecke, RH, Doerge, DR. (2007) Physiologically based pharmacokinetic/
pharmacodynamic model for acrylamide and its metabolites in mice, rats, and humans. Chem Res
Toxicol. 20(3):388-99.
67
-------
Appendix A Memorandum and Charge Questions
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C. 20460
l-obruarv 1. ^(H>S
MEMOHAMU M
SI *UM" I: Kcqnest lor SAB review ol the Drult IRIS A^si-ssrncnl for AcnJnirndc
FROM: [hi t LHe, Hi |>., Aelinjd Director _
Naluitkiil t'enlct lor IMU irorirtiuntaf Assessment. KcscJirch Ifianulf Park (f!24.1-()l I
' nt'Koscarch and IX-M-'Ui
TO: SUL' Sliiillul. I'h.D.
f)L'sij2n^ici,l I ci.li.Tsl OKkx-r
t:i'A Sciuncc Achisor> Rwml S(alT ( Jftloc ( I-HW1 >
I'hii ii to iCLHiciL a CL-V it-vs b> the Suit-rice AiK i>i>r\ F^oitrd of the draft document
"1 oxiuolii!j.ical Rc\iew nf Aei\ luruiJi.1 iC'AS No 7'J-i.Wi- ] }" in Mipporl o)' sunininrv intbrniinion -un llif
IntL'jiraled Risk [nlorrriijliori S\stum (IKCs) I Ms clrtcumcnt i^ ail as<>es.si)ieiit nl'lhc p<»lci)iiaj for cancer
ami MiMK-itn.eerct1i.vis fi'.|]oxuiiji t-xposMis.- Li> ;ILTV -Liinidc. I lit I oxivologieul Review of Acrvlamidc was
fjrL-piiruiJ by ihe N;iii>in;il t enter for t\nviri>nmi:ti[dl A^sL^hiiienL (NCE'iA), which is the health risk
itt*e*snii.-iil progrutu in ihc Olll-Le ol' Resirareh iim.l I )eveh:''pniein. The tJoeuniciil I>us been made available
['or psihlic cuminem on lUe Aet-ncv ':•, NtT.A ^t-b site Jl ihe 1'uLlowinjj I.IKI.:
1 1 up : /.•' e I pub .epu . L'O\ ; ncc:i 'cl'm 'reft; /M'd i s p lay cfhV.'do ki= 1 X 7 "7 2l>
I he KiXJLulojiiuLi] HL'\ it'w ol ALTykimiile hrondk supports i>e1i\ (lies authori/cd in tilt I '>90
k'lin Air Ae( anil is ;ippLieiil>le loms;i1ioii ;imJ reiiiiUuir\ needs of nil pronnim OfHecsaiul
ejiiom ii\ LJt. LilniiliiiL! . ihe eujicer and nonianeer offeets I'lillow. in^ CX[X>>uii.' U> uerj liirtiide. I'.PA la.sl
ublished :)n assessment of the potential ha/urdous fffeei of aer> Nimiclc in I*->HK I he current asst-ss
levicvts niore ruecnl ilaui. :ind applies more recent iiiothodolouv tin deriving lo\k'i1v values.
Antitehed arc Ihe eharet- L|aesl»ons h> ihe Science Arlvisi>rj Board ihai piv*vidf liuckj.in.M,iiid
in lorn Kili i MI as v.ell as live qiieslii>n> asld issuts thut ;ire lo be ihe Jix;us ofihc SieieiKC Ai.Kpsi>r\ littar
eon>nlLalii:.|i on ill is aisessnicnl.
AllnelmienC Chnry.c lor lil'A's S^ieiKv Ad>, ison Jlmird (SAH) - IRIS I'ltxteo I Laical Kc\«evt ol"
p
68
-------
Charge Questions
Selection of Studies and Endpoints for the Oral Reference Dose (RfD)
In the draft document, the proposed most sensitive noncancer effect from exposure to acrylamide
is neurotoxicity. This endpoint is based on an extensive database of animal and human studies.
The next most sensitive effect is reproductive toxicity, which was in the 3-5 fold higher exposure
range for a no effect response in animal studies. No human data were identified for acrylamide
related reproductive effects. Heritable germ cell effects, a potentially serious noncancer effect,
have been observed in male mice, however, the lowest dose levels tested are considerably higher
(two orders of magnitude) than the doses where neurotoxicity were observed, and there is
uncertainty about the shape of the low-dose-response relationship.
1. Please comment on the selection of neurotoxicity as the most appropriate choice for the most
sensitive endpoint (in contrast to reproductive toxicity, heritable germ cell effects, or other
endpoint) based upon the available animal and human data.
2. Please comment on the discussion of mode of action for acrylamide-induced neurotoxicity.
Is the discussion clear, transparently and objectively described, and accurately reflective of
the current scientific understanding?
3. Please comment on the qualitative discussion of aery 1 amide's heritable germ cell effects and
whether the discussion is clear, transparently and objectively described, and reflective of the
current science.
Derivation of the Reference Dose (RfD)
The proposed RfD (0.003 mg/kg-day) for acrylamide is based on a benchmark dose analysis of
the dose-response relationship for neurotoxicity in two chronic drinking water exposure
bioassays using Fischer 344 rats. Uncertainty factors and a PBPK model are used to extrapolate
the animal dose-response to a human equivalent dose-response in the derivation of the RfD.
69
-------
4. Please comment on whether the selection of the Friedman et al., (1995) and Johnson et al.,
(1986) studies as co-principal studies has been scientifically justified. Although EPA
considers Friedman et al. and Johnson et al. to be co-principal studies, the final quantitative
RfD value is derived only from the Johnson study. Please comment on this aspect of EPA's
approach. Please also comment on whether this choice is transparently and objectively
described in the document. Please identify and provide the rationale for any other studies
that should be selected as the principal study(s).
5. Please comment on the benchmark dose methods and the choice of response level used in the
derivation of the RfD, and whether this approach is accurately and clearly presented. Do
these choices represent the most scientifically justifiable approach for modeling the slope of
the dose-response for neurotoxicity? Are there other response levels or methodologies that
EPA should consider? Please provide a rationale for alternative approaches that should be
considered or preferred to the approach presented in the document.
6. Please comment on the selection of the uncertainty factors (other than the interspecies
uncertainty factor) applied to the point of departure (POD) for the derivation of the RfD. For
instance, are they scientifically justified and transparently and objectively described in the
document? [Note: This question does not apply to the interspecies uncertainty factor which is
addressed in the questions on the use of the PBPK model (see PBPK model questions
below)]
7. Please provide any other comments on the derivation of the RfD and on the discussion of
uncertainties in the RfD.
Use of a PBPK Model in the Derivation of the RfD and the Inhalation Reference
Concentration (RfC)
A physiologically-based toxicokinetic (PBTK) model originally developed by Kirman et al.
(2003), and recalibrated by EPA with more recent kinetic and hemoglobin binding data in rats,
mice, and humans (Boettcher et al., 2005; Doerge et al., 2005a,b; Fennell et al., 2005) was used
in the derivation of the RfD to extrapolate from the animal dose-response relationship (observed
in the co-principal oral exposure studies for neurotoxicity) to derive a human equivalent
concentration (FIEC). The FIEC is the external acrylamide exposure level that would produce the
70
-------
same internal level of parent acrylamide (in this case the area under the curve [AUC] of
acrylamide in the blood) that was estimated to occur in the rat following an external exposure to
acrylamide at the level of the proposed point of departure, and related to a response level of 5%
(i.e., the BMDLs). The model results were used in lieu of the default interspecies uncertainty
factor for toxicokinetics differences of 101/2, which left a factor of 101/2 (which is rounded to 3)
for interspecies differences in toxicodynamics.
With respect to the RfC, there are presently insufficient human or animal data to directly derive
an RfC for acrylamide. The PBPK model was thus used to conduct a route-to-route extrapolation
(oral-to-inhalation) to derive an RfC based on the dose-response relationship observed in the co-
principal oral exposure studies for neurotoxicity. In this case, the HEC was based on a
continuous inhalation exposure to acrylamide in the air that would yield the same AUC for the
parent acrylamide in the blood as that estimated for the rat following an external oral exposure to
acrylamide at the level of the proposed point of departure (i.e., the BMDLs).
8. Please comment on whether the documentation for the recalibrated Kirman et al. (2003)
PBTK model development, evaluation, and use in the assessment is sufficient to determine if
the model was adequately developed and adequate for its intended use in the assessment.
Please comment on the use of the PBTK model in the assessment, e.g., are the model
structure and parameter estimates scientifically supportable? Is the dose metric of area-
under-the-curve (AUC) for acrylamide in the blood the best choice based upon what is
known about the mode of action for neurotoxicity and the available kinetic data? Please
provide a rationale for alternative approaches that should be considered or preferred to the
approach presented in the document.
9. Is the Young et al. (2007) PBTK model adequately discussed in the assessment with respect
to model structure, parameter values, and data sets used to develop the model? Do you agree
with the conclusion (and supporting rationale) that the recalibrated Kirman et al. (2003)
model (model structure and parameter values presented in the Toxicological Review)
currently represents the best model to use in the derivation of the toxicity values?
10. According to US EPA's RfC Methodology (1994), the use of PBTK models is assumed to
account for uncertainty associated with the toxicokinetic component of the interspecies
71
-------
uncertainty factor across routes of administration. Does the use of the PBTK model for
acrylamide objectively predict internal dose differences between the F344 rat and humans, is
the use of the model scientifically justified, and does the use of the PBTK reduce the overall
uncertainty in this estimate compared to the use of the default factor? Are there sufficient
scientific data and support for use of this PBTK model to estimate interspecies toxicokinetic
differences and to replace the default interspecies factor for toxicokinetic differences (i.e.,
101/2)? Is the remaining uncertainty factor for toxicodynamic differences scientifically
justified, appropriate and correctly used?
11. Please comment on whether the PBTK model is adequate for use to conduct a route-to-route
extrapolation for acrylamide to derive an RfC in the absence of adequate inhalation animal or
human dose-response data to derive the RfC directly. Was the extrapolation correctly
performed and sufficiently well documented?
12. Please provide any other comments on the derivation of the RfC and on the discussion of
uncertainties in the RfC.
Margin of Exposure (MOE) Analysis
IRIS documents do not include exposure assessments, which precludes the ability to conduct a
Margin of Exposure (MOE) analysis. It has been suggested, however, that the acrylamide
assessment include a Table that lists points of departure for various endpoints to facilitate a MOE
evaluation by EPA's Regional or Program offices, or by other end users of the assessment.
13. Would you suggest that EPA include a Table that lists points of departure (e.g., NOAELs,
BMDs, etc.) for various endpoints that could be used, in conjunction with exposure
assessments, to conduct a MOE analysis?
Ouantitating Heritable Germ Cell Effects
The Toxicological Review includes a discussion of methods to quantitate the risk for heritable
germ cell effects (Section 5.4). The questions below address the uncertainty and utility of the
quantitative results.
72
-------
14. Please comment on the discussion of methods to quantitate the dose-response for heritable
germ cell effects as to whether it is appropriate, clear and objective, and reflective of the
current science. Has the uncertainty in the quantitative characterization of the heritable germ
cell effects been accurately and objectively described?
15. Please comment on the scientific support for the hypothesis that heritable germ cell effects
are likely to occur at doses lower than those seen for neurotoxicity? What on-going or future
research might help resolve this issue?
16. The risks of heritable germ cell effects (i.e., number of induced genetic diseases per million
offspring) for some estimated exposure in workers and the population are presented in Table
5-11, and are based on the quantitative methods and parameter estimates discussed in Section
5.4 of the Toxicological Review. Please comment on whether or not the quantitation of
heritable germ effects should be conducted, the level of uncertainty in the results, if Table 5-
11 is useful for risk assessment purposes, and if the RfD should be included in the Table as
one of the exposure levels.
17. Do you know of any additional data or analyses that would improve the quantitative
characterization of the dose-response for acrylamide-induced heritable germ cell effects?
Would these data also support the quantitative characterization of "total" male-mediated
reproduction risks to offspring (i.e., lethality + heritable defect)? If data are not available, do
you have any recommendations for specific needed studies?
Carcinogenicity of Acrylamide
In accordance with EPA's 2005 Guidelines for Carcinogen Risk Assessment
(www.epa.gov/iris/backgr-d.htm), acrylamide is described as likely to be carcinogenic to humans
based on: (1) significant increased incidences of thyroid tumors in male and female rats, scrotal
sac mesotheliomas in male rats, and mammary gland tumors in female rats in two drinking water
bioassays; (2) initiation of skin tumors following oral, intraperitoneal, or dermal exposure to
acrylamide and the tumor promoter, TPA, in two strains of mice; and (3) increased incidence of
lung adenomas in another mouse strain following intraperitoneal injection of acrylamide.
Evidence from available human studies is judged to be limited to inadequate.
73
-------
The mechanisms by which acrylamide may cause cancer are poorly understood, but EPA has
determined that the weight of the available evidence supports a mutagenic mode of carcinogenic
action, primarily for the acrylamide epoxide metabolite, glycidamide (GA). Other mode(s) of
action (MOA) have been proposed for the carcinogenicity of acrylamide, but there is less
support.
18. Have the rationale and justification for the cancer designation for acrylamide been clearly
described? Is the conclusion that acrylamide is a likely human carcinogen scientifically
supportable?
19. Do you agree that weight of the available evidence supports a mutagenic mode of
carcinogenic action, primarily for the acrylamide epoxide metabolite, glycidamide (GA)?
Has the rationale for this MOA been clearly and objectively presented, and is it reflective of
the current science?
20. Are there other MO As that should be considered? Is there significant biological support for
alternative MOAs for tumor formation, or for alternative MO As to be considered to occur in
conjunction with a mutagenic MOA? Please specifically comment on the support for
hormonal pathway disruption. Are data available on alternate MOAs sufficient to quantitate a
dose-response relationship?
21. Two chronic drinking water exposure bioassays in Fischer 344 rats (Friedman et a/., 1995;
Johnson et a/., 1986) were used to derive the oral slope factor, and to identify the tumors of
interest for the MOA discussion. Are the choices for the studies, tumors, and methods to
quantify risk transparent, objective, and reflective of the current science? Do you have any
suggestions that would improve the presentation or further reduce the uncertainty in the
derived values?
22. The cancer slope factor (CSF) derivation includes an adjustment for early mortality (i.e.,
time-to-tumor analysis). Is this adjustment scientifically supported in estimating the risk from
the 2-year bioassay data for increased incidence of tumors in the rats?
23. The dose metric used in the PBTK model analysis to derive the human equivalent
concentration was area under the curve (AUC) in the blood for the putative genotoxic
metabolite, glycidamide. Please comment on whether AUC for glycidamide is the best
choice of the dose metric in estimating the human equivalent concentration to derive the oral
74
-------
slope factor. If other dose metrics are preferable, please provide the scientific rationale for
their selection.
24. As with the RfC, there were insufficient cancer inhalation data to derive an inhalation unit
risk (IUR). The PBTK model was used in a route-to-route extrapolation of the dose-response
relationship from the oral data, and to estimate the human equivalent concentration for
inhalation exposure to acrylamide. Please comment on whether this extrapolation to derive
the inhalation unit risk was correctly performed and sufficiently well documented.
25. The recommendation to use the age-dependent adjustment factors (ADAFs) is based on the
determination of a mutagenic MOA for carcinogenicity. Is this recommendation scientifically
justifiable and transparently and objectively described
26. Please provide any other comments on the CSF or IUR, and on the discussion of
uncertainties in the cancer assessment.
75
-------
Appendix B
Proposed Modes of Action (MOAs) for Acrylamide Neurotoxicity
The following text on the two proposed MOAs for acrylamide neurotoxicity was written
by one panel member. It is offered for the Agency's consideration in writing the revised version
of the acrylamide IRIS document:
1. Disruption of Nitric Oxide (NO) Signaling at the Nerve Terminal Hypothesis
Acrylamide is a conjugated a,|3-unsaturated carbonyl derivative in the type-2 alkene
chemical class. Because electrons in pi orbitals of a conjugated system are mobile, the a,|3-
unsaturated carbonyl structure of acrylamide is characterized as a soft electrophile according to
the hard-soft, acid-base principle (reviewed in Pearson, 1967). Characteristically, soft
electrophiles will preferentially form Michael-type adducts with soft nucleophiles, which in
biological systems are primarily sulfhydryl groups on cysteine residues (Hinson and Roberts,
1992; LoPachin and DeCaprio, 2005). Free sulfhydryl groups can exist in the reduced thiol-state
or in the anionic thiolate-state and recent research indicates that the highly nucleophilic thiolate
is the preferential adduct target for acrylamide (LoPachin et al., 2007b; see also Friedman et al.,
1995). Based on the pKa of cysteine (pH 8.5), at physiological pH (7.4) the thiolate state exists
only in unique protein motifs called catalytic triads, where proton shuttling through an acid-base
pairing of proximal amino acids (e.g., aspartic acid and lysine) regulates the protonation and
deprotonation of the cysteine sulfhydryl group. Indeed, both mass spectrometric and kinetic data
have demonstrated the selective adduction of cysteine residues on many neuronal proteins
(Barber and LoPachin, 2004; Barber et al., 2007). Furthermore, it is now recognized that the
redox state or nucleophilicity of cysteine sulfhydryl groups within catalytic triads can determine
the functionality of these proteins (reviewed in LoPachin and Barber, 2006; Stamler etal., 2001).
In contrast to acrylamide, the epoxide metabolite glycidamide (Gly), is a hard electrophile that
preferentially forms adducts with hard nucleophiles such as nitrogen, carbon and oxygen.
Nucleotide residues of DNA contain abundant hard nucleophilic targets, which is consistent with
the formation of glycidamide adducts on adenine and guanine bases in acrylamide-intoxicated
animals (Doerge etal., 2005; reviewed in Besaratinia and Pfeifer, 2007).
Based on the observation that the processes affected (e.g., neurotransmitter release and
storage) and corresponding kinetics (Km, Vmax) were similar in synaptosomes exposed in vitro to
76
-------
acrylamide and those isolated from aery 1 ami de-intoxicated rats (Barber and LoPachin, 2004;
LoPachin et al, 2004, 2006), LoPachin and colleagues have reasoned that the parent compound,
acrylamide, is responsible for neurotoxicity. Moreover, cysteine thiolate groups have clear
regulatory functions in many critical neuronal processes (LoPachin and Barber, 2006), whereas
protein valine, lysine and histidine residues, which are the likely hard nucleophilic targets for a
hard electrophile such as Gly, have unclear functional and therefore toxicological relevance.
Quantitative morphometric and silver stain analyses of PNS and CNS of acrylamide-intoxicated
animals have shown that axon degeneration was an epiphenomenon related to dose-rate; i.e.,
degeneration occurred at lower but not higher dose-rates. In contrast, nerve terminal
degeneration occurred regardless of dose-rate and in correspondence with the onset and
development of neurological deficits (Crofton etal., 1996; Lehning etal., 1998, 2002a,b, 2003;
reviewed in LoPachin et al., 1994, 2002, 2003), suggesting the nerve terminals as a primary site
of action. Subsequent neurochemical studies showed that both in vitro and in vivo acrylamide
exposure produced early disruptions of neurotransmitter release, reuptake and vesicular storage
(Barber and LoPachin, 2004; LoPachin etal., 2004, 2006, 2007a). Further, proteomic analyses
indicated that the inhibition of presynaptic function was due to the formation of cysteine adducts
on proteins that regulate neurotransmitter handling; e.g., Cys 264 of 7V-ethylmaleimide sensitive
factor, Cys 254 of v-ATPase (see Barber and LoPachin, 2004; Barber etal., 2007; Feng and
Forgac, 1992; LoPachin etal., 2007a,b, 2008b; reviewed in LoPachin and Barber, 2006). The
anionic sulfhydryl state, which is only found in the catalytic triads of regulatory proteins, is an
acceptor for nitric oxide (NO) and, therefore, has lead to the proposal that acrylamide-induced
neurotoxicity results from disruption of neuronal NO signaling (LoPachin and Barber, 2006;
LoPachin et al., 2008a).
2. Fast Axonal Transport Disruption Hypothesis
Another proposed MOA is that both acrylamide and Gly inhibit the movement of
materials in fast axonal transport (Sickles etal, 2002). According to the "kinesin/axonal
transport" hypothesis, toxicant inhibition of kinesin could lead to reductions in the axonal
delivery of macromolecules that would eventually produce a deficiency of essential proteins
required to maintain axon structure and/or function. Distal axons and nerve terminals are
particularly vulnerable to transport defects based upon an exceptionally large axonal volume (as
77
-------
much as 1000 times the volume of the neuron cell body) and the dependence of these distal
regions on long distance transport (100 fold longer length than diameter of the cell body). This
regional sensitivity is consistent with the previously identified distal spatial distribution of
toxicant-induced damage (Cavanagh, 1964).
Microtubule motility assays using purified kinesin from bovine brain identified a dose-
dependent inhibition of kinesin as well as a less sensitive effect on microtubules (Sickles etal.,
1996). Preincubation of either kinesin or taxol-stabilized microtubules produced a reduction in
the affinity between kinesin and microtubules, recognized as a reduced number of microtubules
bound or locomoting on an absorbed bed of kinesin. Microtubules that were locomoting did so in
a less directed or staggering type of progression. The inhibitions were due to covalent adduction,
presumably through sulfhydryl alkylation, although adduction of other amino acid residues such
as valine was possible. The non-neurotoxic analogue, propionamide had no effect. Other
investigators have identified kinesin inhibition by sulfhydryl reagents such as N-ethylmaleimide
and ethacrynic acid (Walker etal., 1997). As with acrylamide, inhibition by these sulfhydryl
reagents produced the characteristic staggering movement of microtubules. The reaction was
slow and temperature dependent suggesting a sterically hindered cysteine residue as an important
adduct target. Additional studies have demonstrated a comparable effect of glycidamide on
kinesin (Sickles, unpublished data). The predicted outcome of such an effect would be reduced
quantity of flow, precisely the outcome from several experiments where rate of transport versus
quantity could be discriminated (Sickles, 1989a; Sickles, 1989b; Stone etal., 1999).
Fast axonal transport has been studied in a variety of model systems using diverse
techniques. A comprehensive survey of acrylamide effects on fast anterograde and retrograde
axonal transport (Sickles et al. ,2002) revealed that all studies measuring fast transport within 24
hours of acrylamide exposure demonstrated significant reductions, whereas longer postexposure
delay was not associated with changes in transport. Furthermore, a reduction in transport
quantity (but not rate) has been reported within 20 minutes of exposure. The duration of this
effect was 16 hours, with full recovery at 24 hours (Sickles, 1991). Quantitation of transport
after multiple dosings (i.e. 4, 7 or 10 doses) had a similar effect on transport in the proximal
sciatic nerve (Sickles, 1991). The changes in transport were not due to an effect on protein
synthesis and exposure of only the axons confirmed that the target was axonal (Sickles, 1989a;
Sickles, 1992). Collectively, these results suggested action on a target that is replaced via the fast
78
-------
transport system, consistent with kinesin. The actions of acrylamide on fast axonal transport
were independent of effects on axonal neurofilaments, as similar reductions were observed in
wild-type and transgenic mice lacking axonal neurofilaments (Stone et a/., 1999; Stone et a/.,
2000). The same results were observed using radiolabelling of proteins in mouse optic nerves
and differential interference microscopy of isolated sciatic nerve axons. Other recent studies
have identified a parallel inhibition of retrograde axonal transport by acrylamide (Sabri and
Spencer, 1990), although it is unclear whether this effect is due to inhibition of cytoplasmic
dynein, the retrograde axonal transport motor, or whether this is a result of indirect effects of
kinesin motor inhibition (Brady etal., 1990).
The predicted outcome from axonal transport compromise is a reduction in vital
macromolecules in the distal axons and an accumulation of transported material within the axon.
Morphological studies have consistently identified accumulations of tubulovesicular profiles and
neurofilaments in axons of acrylamide-intoxicated animals (Spencer and Schaumburg, 1991),
which are morphological elements transported via kinesin along microtubules. Other studies
have identified reduced synaptic vesicles in neuromuscular junctions (DeGrandchamp and
Lowndes, 1990; DeGrandchamp etal., 1990). A reduction in GAP-43 in the terminal neurites of
cultured primary spinal cord neurons following acrylamide exposure has been observed (Clarke
and Sickles, 1996). Future studies are required to quantitate reductions in specific axonal
compartments using a variety of neurotoxic and non-neurotoxic dosing regimens in vivo to
confirm the loss of physiologically or structurally important macromolecules.
Additional supportive data for the axonal transport hypothesis comes from several studies
of kinesin knockouts as well as similarity to human diseases. While most knockouts are lethal,
low level mutations of kinesin motors in Drosophila have identified an identical spatial pattern of
dysfunction and morphological similarity in axonal pathology (Gho et a/., 1992; Kurd and
Saxton, 1996) as with acrylamide intoxication. The group of neurological disorders classified as
hereditary spastic paraplegias has a spatial pattern of ataxia, spasticity and muscle weakness as
observed with acrylamide intoxication. Some of these types have been associated with mutations
in kinesin motors (Reid etal., 2002), while others are the result of either axonal or glial protein
mutations. However, the common theme is alteration in axonal transport (Reid, 2003; Gould and
Brady, 2004).
79
-------
Role of Acrylamide vs. Glycidamide
The respective adduct chemistries of acrylamide and glycidamide are well understood
and could have fundamental implications for neurotoxicity regardless of the proposed
mechanism; i.e., kinesin inhibition (Sickles etal., 2002) or blockade of NO signaling (LoPachin
and Barber, 2006; LoPachin et a/., 2008). Accordingly, an obvious data gap in the current
mechanistic understanding of acrylamide neurotoxicity, is the relative roles of the parent
compound and Gly. Thus, although early research suggested that Gly produced neurotoxicity
both in whole animal (Abou-Donia et a/., 1993) and in vitro (Harris et a/., 1994) model systems,
other studies using similar models failed to find neurotoxic effects associated with this
metabolite (Brat and Brimijoin, 1993; Costa et a/., 1992, 1995; Moser et a/., 1992). Clearly,
resolving the relative roles of acrylamide vs. glycidamide is an important issue that will require
more research. Although the adduct chemistry of these toxicants has been reasonably defined,
the precise molecular mechanisms and sites of neurotoxicity are unknown.
80
------- |