UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON D.C. 20460
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
August 26, 2011
EPA-SAB-011-014
The Honorable Lisa P. Jackson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, N.W.
Washington, D.C. 20460
Subj ect: SAB Review of EPA's Reanalysis of Key Issues Related to
Dioxin Toxicity and Response to NAS Comments (May 2010)
Dear Administrator Jackson:
EPA's Office of Research and Development (ORD) requested that the Science Advisory Board (SAB)
review the Agency's draft report entitled EPA's Reanalysis of Key Issues Related to Dioxin Toxicity and
Response to NAS Comments (May 2010) ("Report"). The Report contains EPA's technical response to
key comments in the 2006 National Academy of Sciences (NAS) report, Health Risks from Dioxin and
Related Compounds: Evaluation of the EPA Reassessment. The NAS reviewed EPA's 2003 exposure
and human health reassessment of dioxin and recommended that the Agency should: more thoroughly
justify and communicate its approaches to dose-response modeling for the health effects of dioxin,
taking into consideration both nonlinear and linear methods for characterizing cancer risk; improve the
transparency and clarity of the selection of key data sets for the dioxin dose-response analysis;
reevaluate its cancer weight-of-evidence determination for dioxin based on the Agency's 2005
Guidelines for Carcinogen Risk Assessment, consider using physiologically-based pharmacokinetic
(PBPK) modeling in the dioxin risk assessment; and improve transparency, thoroughness and clarity in
quantitative uncertainty analysis. The NAS also encouraged EPA to calculate a reference dose (RfD),
which had not been derived in the 2003 reassessment.
In response to EPA's request, the SAB convened an expert panel to review the Agency's Report. The
SAB Panel was asked to comment on the scientific soundness of EPA's responses to the NAS
recommendations. The enclosed SAB report provides the consensus advice and recommendations of the
Panel, with the exception of one member who offered a dissenting opinion mainly on 2,3,7,8
tetrachlorodibenzo-p-dioxin (TCDD) carcinogenicity.
The SAB finds that EPA's Report is generally clear, logical, and responsive to many but not all of the
recommendations of the NAS. We have provided recommendations to further enhance the transparency,
clarity, and scientific integrity of the Report. The SAB has identified deficiencies in EPA's Report with
respect to the completeness of its consideration of two critical elements of the TCDD assessment: 1)
^£0S7^
%
\
53S
^ PBO^°
-/
-------
nonlinear dose-response for TCDD carcinogenicity, and 2) uncertainty analysis of TCDD toxicity. Our
major comments and recommendations are provided below:
• The SAB commends EPA for the comprehensive and rigorous process that was used to identify,
review, and evaluate the TCDD literature. The criteria for study selection have been clearly
articulated, well justified, and applied in a scientifically sound manner. To further improve
clarity and transparency of the Report, we recommend that EPA include a better means of
tracking and describing which studies did not satisfy inclusion criteria. Similarly, we recommend
that EPA strengthen its justification for excluding studies of dioxin-like compounds. The Report
can be enhanced by incorporating information from studies with dioxin-like compounds into a
qualitative discussion of the weight-of-evidence for cancer and noncancer endpoints.
• EPA used the Emond physiologically-based pharmacokinetic model to evaluate the internal dose
of TCDD in human and rodent tissue, and to estimate the continuous daily TCDD intake over the
relevant period of exposure. The SAB agrees with EPA that this model provides the best
available basis for the dose metric calculations. We also support EPA's use of blood TCDD
concentrations as the relevant dose metric. However, we recommend that EPA expand the
discussion of other published models, evaluate the impact of model selection on dose metric
prediction, provide a more quantitative uncertainty analysis, and conduct an external peer review
of the mouse model because it has not been published in the peer-reviewed literature.
• The SAB agrees with EPA's characterization of TCDD as carcinogenic to humans in accordance
with EPA's 2005 Guidelines for Carcinogen Risk Assessment. The SAB recommends that in the
weight-of-evidence characterization EPA build upon all available data to support its decision and
clearly indicate how different types of data support each other. One dissenting Panel member,
however, expressed the view that at best, there is equivocal evidence for TCDD classification as
a human carcinogen.
• The SAB agrees with EPA's selection of the Cheng et al. (2006) study, which analyzed the
National Institute for Occupational Safety and Health (NIOSH) occupational cohort, as the
critical study for the quantitative cancer assessment. The SAB also agrees that it is appropriate to
use all-cancer mortality as the basis of the oral slope factor because of the extensive dose-
response information and because in the case of TCDD there appear to be multiple targets for
carcinogenic action.
• The SAB finds that the Report did not respond adequately to the NAS recommendation to adopt
both linear and nonlinear methods of extrapolation in order to account for the uncertainty of the
dose-response curve for TCDD. The Report states that only a linear approach could be justified.
We recommend that EPA revise the Report to provide a discussion of evidence of possible
modes of action that include both linear and nonlinear alternatives for the cancer endpoint. In the
absence of a definitive nonlinear mode of action, estimates based on the linear option can serve
as the baseline for comparison with other estimates.
• The SAB supports EPA's selection and use of two co-critical epidemiologic studies for the
derivation of the RfD for TCDD. These studies evaluated the effects of human exposure to
TCDD following accidental release at a chemical plant near Seveso, Italy. The SAB finds that
the study endpoints used by EPA to determine the RfD (decrease in sperm count and motility and
increased thyroid stimulating hormone in blood) are relevant to public health. The selection of
-------
these endpoints also resolves the critical issue of differing windows of susceptibility to
environmental toxic agents over the course of the life cycle, with pre- and periconceptional
exposures comprising the window of greatest susceptibility. We recommend, however, that EPA
provide a discussion of the strengths and weaknesses of the studies and an indication of whether
these weaknesses affect the RfD determination. The SAB agrees with the benchmark dose
modeling approaches used by EPA in the Report and the decision to use human data as preferred
to animal data for the RfD determination.
• EPA's Report discusses a broad range of philosophical and methodological issues to be
considered in conducting an uncertainty analysis for TCDD toxicity. The SAB acknowledges the
challenges of a unified quantitative uncertainty analysis. However, we do not agree with the
position taken in the Report that such an analysis is unfeasible and we have suggested a number
of methods that could be used for this purpose.
• Finally, EPA's Report could be improved by editing and restructuring to better integrate the
material presented in various sections, eliminate redundancies, and move some material into
appendices to provide more succinct responses to NAS concerns. In addition, we recommend
including a glossary in the Report to help minimize confusion and misinterpretation among
diverse users.
The SAB appreciates the opportunity to provide EPA with advice on this important subject. EPA should
move in a proficient and expeditious manner to finalize the IRIS document for dioxin. We look forward
to receiving the Agency's response.
Sincerely,
/Signed/
/Signed/
Dr. Deborah L. Swackhamer, Chair
EPA Science Advisory Board
Dr. Timothy J. Buckley, Chair
SAB Dioxin Review Panel
Enclosure
-------
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
1
-------
U.S. Environmental Protection Agency
Science Advisory Board
Dioxin Review Panel
CHAIR
Dr. Timothy J. Buckley, Associate Professor and Chair, Division of Environmental Health Sciences,
College of Public Health, The Ohio State University, Columbus, OH
MEMBERS
Dr. Harvey Clewell, Director of the Center for Human Health Assessment, The Hamner Institutes for
Health Sciences, Research Triangle Park, NC
Dr. Louis Anthony (Tony) Cox, Jr., President, Cox Associates, Denver, CO
Dr. Elaine Faustman, Professor and Director, Institute for Risk Analysis and Risk Communication,
School of Public Health, University of Washington, Seattle, WA
Dr. Scott Ferson, Senior Scientist, Applied Biomathematics, Setauket, NY
Dr. Jeffrey Fisher, Research Toxicologist, National Center for Toxicological Research, U.S. Food and
Drug Administration, Jefferson, AR
Dr. Helen Hakansson, Professor of Toxicology, Unit of Environmental Health Risk Assessment,
Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden
Dr. Russ Hauser, Frederick Lee Hisaw Professor, Department of Environmental Health, Harvard
School of Public Health, Boston, MA
Dr. B. Paige Lawrence, Associate Professor, Departments of Environmental Medicine and
Microbiology and Immunology, School of Medicine and Dentistry, University of Rochester School of
Medicine and Dentistry, Rochester, NY
Dr. Michael I. Luster, Professor, Department of Community Medicine, West Virginia University
Health Sciences Center, Morgantown, WV
Dr. Paolo Mocarelli, Professor of Clinical Biochemistry, Department of Clinical Laboratory, Hospital
of Desio-Nuovo Monoblous, University of Milano Bicocca, Desio-Milano, Italy
Dr. Victoria Persky, Professor, Epidemiology and Biostatistics Program, School of Public Health,
University of Illinois at Chicago, Chicago, IL
Dr. Sandra L. Petersen, Professor, Associate Graduate Dean, Department of Veterinary and Animal
Sciences, College of Natural Sciences, University of Massachusetts-Amherst, Amherst, MA
11
-------
Dr. Karl Rozman, Professor, Pharmacology, Toxicology and Therapeutics, The University of Kansas
Medical Center, Kansas City, KS
Dr. Arnold Schecter, Professor, Environmental and Occupational Health Sciences, School of Public
Health-Dallas Campus, University of Texas, Dallas, TX
Dr. Allen E. Silverstone, Professor, Department of Microbiology and Immunology, Health Science
Center, SUNY Upstate Medical University, Syracuse, NY and Adjunct Professor of Environmental
Medicine, University of Rochester School of Medicine and Dentistry, Rochester, NY.
Dr. Mitchell J. Small, The H. John Heinz III Professor of Environmental Engineering, Department of
Civil and Environmental Engineering and Engineering and Public Policy, Carnegie Mellon University,
Pittsburgh, PA
Dr. Anne Sweeney, Professor of Epidemiology, Department of Epidemiology and Biostatistics, School
of Rural Public Health, Texas A&M Health Science Center, College Station, TX
Dr. Mary K. Walker, Professor, Division of Pharmaceutical Sciences, College of Pharmacy, University
of New Mexico, Albuquerque, NM
SCIENCE ADVISORY BOARD STAFF
Dr. Thomas Armitage, Designated Federal Officer, U.S. Environmental Protection Agency,
Washington, DC
Dr. Diana Wong, Designated Federal Officer, U.S. Environmental Protection Agency, Washington, DC
Did not participate in the discussion of the charge questions specifically pertaining to use of the
Mocarelli et al. (2008) study in deriving the reference dose.
Dissenting opinion in Appendix B.
111
-------
U.S. Environmental Protection Agency
Science Advisory Board
BOARD
CHAIR
Dr. Deborah L. Swackhamer, Professor and Charles M. Denny, Jr. Chair in Science, Technology and
Public Policy - Hubert H. Humphrey School of Public Affairs and Co-Director of the Water Resources
Center, University of Minnesota, St. Paul, MN
SAB MEMBERS
Dr. David T. Allen, Professor, Department of Chemical Engineering, University of Texas, Austin, TX
Dr. Claudia Benitez-Nelson, Full Professor and Director of the Marine Science Program, Department
of Earth and Ocean Sciences, University of South Carolina, Columbia, SC
Dr. Timothy J. Buckley, Associate Professor and Chair, Division of Environmental Health Sciences,
College of Public Health, The Ohio State University, Columbus, OH
Dr. Patricia Buffler, Professor of Epidemiology and Dean Emerita, Department of Epidemiology,
School of Public Health, University of California, Berkeley, CA
Dr. Ingrid Burke, Director, Haub School and Ruckelshaus Institute of Environment and Natural
Resources, University of Wyoming, Laramie, WY
Dr. Thomas Burke, Professor, Department of Health Policy and Management, Johns Hopkins
Bloomberg School of Public Health, Johns Hopkins University, Baltimore, MD
Dr. Terry Daniel, Professor of Psychology and Natural Resources, Department of Psychology, School
of Natural Resources, University of Arizona, Tucson, AZ
Dr. George Daston, Victor Mills Society Research Fellow, Product Safety and Regulatory Affairs,
Procter & Gamble, Cincinnati, OH
Dr. Costel Denson, Managing Member, Costech Technologies, LLC, Newark, DE
Dr. Otto C. Doering III, Professor, Department of Agricultural Economics, Purdue University, W.
Lafayette, IN
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. T. Taylor Eighmy, Vice President for Research, Office of the Vice President for Research, Texas
Tech University, Lubbock, TX
iv
-------
Dr. Elaine Faustman, Professor and Director, Institute for Risk Analysis and Risk Communication,
School of Public Health, University of Washington, Seattle, WA
Dr. John P. Giesy, Professor and Canada Research Chair, Veterinary Biomedical Sciences and
Toxicology Centre, University of Saskatchewan, Saskatoon, Saskatchewan, Canada
Dr. Jeffrey K. Griffiths, Professor, Department of Public Health and Community Medicine, School of
Medicine, Tufts University, Boston, MA
Dr. James K. Hammitt, Professor, Center for Risk Analysis, Harvard University, Boston, MA
Dr. Bernd Kahn, Professor Emeritus and Associate Director, Environmental Radiation Center, Georgia
Institute of Technology, Atlanta, GA
Dr. Agnes Kane, Professor and Chair, Department of Pathology and Laboratory Medicine, Brown
University, Providence, RI
Dr. Madhu Khanna, Professor, Department of Agricultural and Consumer Economics, University of
Illinois at Urbana-Champaign, Urbana, IL
Dr. Nancy K. Kim, Senior Executive, Health Research, Inc., Troy, NY
Dr. Kai Lee, Program Officer, Conservation and Science Program, David & Lucile Packard Foundation,
Los Altos, CA (affiliation listed for identification purposes only)
Dr. Cecil Lue-Hing, President, Cecil Lue-Hing & Assoc. Inc., Burr Ridge, IL
Dr. Floyd Malveaux, Executive Director, Merck Childhood Asthma Network, Inc., Washington, DC
Dr. Lee D. McMullen, Water Resources Practice Leader, Snyder & Associates, Inc., Ankeny, IA
Dr. Judith L. Meyer, Professor Emeritus, Odum School of Ecology, University of Georgia, Lopez
Island, WA
Dr. James R. Mihelcic, Professor, Civil and Environmental Engineering, University of South Florida,
Tampa, FL
Dr. Jana Milford, Professor, Department of Mechanical Engineering, University of Colorado, Boulder,
CO
Dr. Christine Moe, Eugene J. Gangarosa Professor, Hubert Department of Global Health, Rollins
School of Public Health, Emory University, Atlanta, GA
Dr. Horace Moo-Young, Dean and Professor, College of Engineering, Computer Science, and
Technology, California State University, Los Angeles, CA
v
-------
Dr. Eileen Murphy, Grants Facilitator, Ernest Mario School of Pharmacy, Rutgers University,
Piscataway, NJ
Dr. Duncan Patten, Research Professor, Hydroecology Research Program, Department of Land
Resources and Environmental Sciences, Montana State University, Bozeman, MT
Dr. Stephen Polasky, Fesler-Lampert Professor of Ecological/Environmental Economics, Department
of Applied Economics, University of Minnesota, St. Paul, MN
Dr. C. Arden Pope III, Professor, Department of Economics, Brigham Young University, Provo, UT
Dr. Stephen M. Roberts, Professor, Department of Physiological Sciences, Director, Center for
Environmental and Human Toxicology, University of Florida, Gainesville, FL
Dr. Amanda Rodewald, Professor of Wildlife Ecology, School of Environment and Natural Resources,
The Ohio State University, Columbus, OH
Dr. Jonathan M. Samet, Professor and Flora L. Thornton Chair, Department of Preventive Medicine,
University of Southern California, Los Angeles, CA
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, Philip E. Austin Professor of Economics, Department of Economics,
University of Connecticut, Storrs, CT
Dr. Herman Taylor, Director, Principal Investigator, Jackson Heart Study, University of Mississippi
Medical Center, Jackson, MS
Dr. Barton H. (Buzz) Thompson, Jr., Robert E. Paradise Professor of Natural Resources Law at the
Stanford Law School and Perry L. McCarty Director, Woods Institute for the Environment, Stanford
University, Stanford, CA
Dr. Paige Tolbert, Professor and Chair, Department of Environmental Health, Rollins School of Public
Health, Emory University, Atlanta, GA
Dr. John Vena, Professor and Department Head, Department of Epidemiology and Biostatistics,
College of Public Health, University of Georgia, Athens, GA
Dr. Thomas S. Wallsten, Professor and Chair, Department of Psychology, University of Maryland,
College Park, MD
Dr. Robert Watts, Professor of Mechanical Engineering Emeritus, Tulane University, Annapolis, MD
vi
-------
Dr. R. Thomas Zoeller, Professor, Department of Biology, University of Massachusetts, Amherst, MA
SCIENCE ADVISORY BOARD STAFF
Dr. Angela Nugent, Designated Federal Officer, U.S. Environmental Protection Agency, Washington,
DC
Ms. Stephanie Sanzone, Designated Federal Officer, U.S. Environmental Protection Agency,
Washington, DC
vii
-------
TABLE OF CONTENTS
ABBREVIATIONS AND ACRONYMS ix
1. EXECUTIVE SUMMARY 1
2. INTRODUCTION 8
3. RESPONSES TO EPA'S CHARGE QUESTIONS 9
3.1. General Charge Questions 9
3.2. Transparency and Clarity in the Selection of Key Data Sets for Dose-Response Analysis 11
3.3. The Use of Toxicokinetics in the Dose-Response Modeling for Cancer and Noncancer
Endpoints 15
3.4. Reference Dose 21
3.5. Cancer Assessment 28
3.6. Feasibility of Quantitative Uncertainty Analysis 35
REFERENCES 43
APPENDIX A: EPA'S CHARGE QUESTIONS A-l
APPENDIX B: DISSENTING OPINION B-1
APPENDIX C: VALUE OF INFORMATION C-l
APPENDIX D: EDITORIAL COMMENTS AND CORRECTIONS D-l
viii
-------
ABBREVIATIONS AND ACRONYMS
AhR
aryl hydrocarbon receptor
BMD
benchmark dose
BMDL
benchmark dose lower bound
BMR
benchmark response level
CYP
cytochrome P450
DLC
dioxin-like compound
ED
effective dose
EPA
U.S. Environmental Protection Agency
HED
human equivalent dose
IRIS
integrated risk information system
LASC
lipid-adjusted serum concentrations
LOAEL
lowest-observed-adverse-effect level
MOA
mode of action
NAS
National Academy of Sciences
NHEERL
National Health and Environmental Effects Research Laboratory
NIOSH
National Institute for Occupational Safety and Health
NOAEL
no-ob served-adverse-effect level
NRC
National Research Council
OSF
oral slope factor
PBPK
physiologically-based pharmacokinetic
PCDDs
polychlorinated dibenzo-p-dioxin
PCDFs
polychlorinated dibenzofuran
POD
point of departure
RfD
reference dose
RR
relative risk
SAB
Science Advisory Board
T3
triiodothyronine
14
thyroxine
TCDD
2,3,7,8 -tetrachl orodib enzo-p-di oxin
TEF
toxicity equivalence factor
TEQ
toxicity equivalence
TSH
thyroid stimulating hormone
UF
uncertainty factor
WHO
World Health Organization
IX
-------
1. EXECUTIVE SUMMARY
In 2003, EPA, along with other federal agencies, asked the National Academy of Sciences (NAS) to
review aspects of the science in EPA's draft dioxin reassessment entitled, Exposure and Human Health
Reassessment of 2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) and Related Compounds and, in 2004,
EPA sent the 2003 draft dioxin reassessment to the NAS for review. In 2006, the NAS released the
report of its review entitled, Health Risks from Dioxin and Related Compounds: Evaluation of the EPA
Reassessment. The NAS recommended that EPA should: more thoroughly justify and communicate its
approaches to dose-response modeling for the health effects of dioxin, taking into consideration both
nonlinear and linear methods for characterizing cancer risk; improve the transparency and clarity of the
selection of key data sets for the dioxin dose-response analysis; reevaluate its cancer weight-of-evidence
determination for dioxin based on the Agency's 2005 Guidelines for Carcinogen Risk Assessment,
consider using physiologically-based pharmacokinetic (PBPK) modeling in the dioxin risk assessment;
and improve transparency, thoroughness and clarity in quantitative uncertainty analysis. The NAS also
encouraged EPA to calculate a reference dose (RfD), which had not been derived in the 2003
reassessment.
EPA's Office of Research and Development (ORD) prepared the draft report, entitled EPA 's Reanalysis
of Key Issues Related to Dioxin Toxicity and Response to NAS Comments (May 2010) (EPA, 2010)
(hereafter referred to as the Report), and requested that the EPA Science Advisory Board (SAB) conduct
an independent external peer review of the Report. This Executive Summary highlights the findings and
recommendations of the SAB Dioxin Review Panel (the "Panel") in response to charge questions
concerning each of the six sections of th e Report. EPA's charge questions are provided in Appendix A.
General Charge (Charge Questions 1.1 - 1.2)
EPA asked the SAB to comment on whether the Report is clear and logical and whether there are other
critical studies that would make a significant impact on the conclusions of the hazard characterization or
dose-response assessment of the chronic noncancer and cancer health effects of TCDD.
As further discussed in the responses to the charge questions in Section 3.1 of this report, the SAB finds
that the Report is generally clear, logical, and responsive to many but not all of the recommendations of
the NAS. The SAB has, however, provided many recommendations to further improve the clarity,
organization, and responsiveness of various parts of the Report. The SAB was impressed with the
process that EPA used to identify, review, and evaluate the relevant literature. The SAB finds that
EPA's process was comprehensive and rigorous and included public participation. However, the SAB
recommends that the Report be improved by: incorporating text to better integrate the material presented
in the individual chapters, providing greater clarity and transparency in indicating which studies do not
satisfy criteria for inclusion in EPA's assessment of TCDD, and editing the Report to provide greater
clarity in writing and make it more concise by moving some material into appendices.
During its deliberation, the SAB did not identify any additional studies that would make a significant
impact on the conclusions of the hazard characterization and dose-response assessment. The SAB
recommends that EPA provide more discussion and clarity concerning the exclusion of null
epidemiologic studies. In addition, as further discussed in the responses to the relevant charge questions,
the SAB has identified deficiencies in the Report with respect to the completeness of its consideration of
1
-------
two critical elements: 1) nonlinear dose-response for TCDD carcinogenicity, and 2) uncertainty analysis.
As discussed below, the SAB has provided recommendations to improve the Report in these areas.
Transparency and Clarity in the Selection of Key Data Sets for Dose-Response Analyses (Charge
Questions 2.1-2.3)
The NAS recommended that EPA develop a clear and readily understandable methodology for
evaluating and including epidemiologic and animal bioassay data sets in dose-response evaluations. The
SAB Panel was asked to comment on whether EPA had been responsive to NAS concerns and had
applied epidemiology and animal bioassay study criteria considerations in a scientifically sound manner.
Section 2 of the Report contains a clear presentation of the process EPA used to select key data sets for
dose-response analysis and is thus responsive to NAS recommendations in this area. The Report also
clearly identifies the studies that were used for dose-response analysis. However, the SAB has provided
recommendations to further enhance the overall clarity and transparency of Section 2 of the Report. The
SAB recommends careful and extensive editing to revise and consolidate Section 2.
The SAB also finds that EPA's epidemiology and animal bioassay study criteria and considerations are
scientifically justified, clearly described, and applied in a scientifically sound manner. However, the
SAB has provided recommendations to improve and strengthen the scientific justification and clarity of
description of EPA's study criteria and considerations. EPA should provide better justification of the
rationale for using only studies where the exposure was primarily to TCDD for derivation of the
reference dose. This justification should include both scientific and practical reasons. EPA should
incorporate information from studies with dioxin-like compounds (DLCs) into a qualitative discussion
of the weight-of-evidence for cancer and noncancer endpoints. There is a wealth of literature on these
compounds that can be used to support mode of action determination as well as the plausibility of
individual biological effects that are used as the basis for risk assessment. In addition, the SAB has
provided a number of specific recommendations to further clarify the justifications for some of the study
inclusion and exclusion criteria.
Use of Toxicokinetics in Dose-Response Modeling for Cancer and Noncancer Endpoints (Charge
Questions 3.1 3.5)
Section 3 of EPA's Report discusses the use of a physiologically-based pharmacokinetic (PBPK) model
(Emond et al., 2004, 2005, 2006) with blood concentration as the dose metric rather than first-order
body burden. EPA asked the SAB to comment on the scientific justification for application of this
model. EPA also developed a PBPK model to estimate TCDD concentration in mouse tissues. The SAB
was asked to comment on the scientific rationale for development of the mouse model and the
performance of the model.
The SAB agrees with EPA's use of blood TCDD concentration as a surrogate for tissue exposure to
TCDD. Blood TCDD concentration is a better choice than using body burden (as in the 2003
Reassessment) because it is more closely related to the biologically relevant dose metric: the free
concentration of dioxin in the target tissues. It is important to recognize, however, that TCDD
distribution within tissues such as the liver can be non-uniform. The SAB further agrees that the PBPK
model developed by Emond et al. (2004, 2005, 2006) provides the best available basis for the dose
metric calculations in the assessment. However, EPA should clarify how the model was used in studies
2
-------
that reported the concentrations of dioxin in plasma, serum, blood, or blood fat:blood measurements.
The SAB also recommends additional discussion of: other published models, the intended use of the
Emond model in the assessment, and the basis for selection of the Emond model. The EPA
modifications to the published Emond model are minor and appropriate. However, the use of 0.6 as the
Hill coefficient in the Emond model for CYPla2 induction is well outside the confidence interval of
0.78 and 1.14 reported by Walker et al. (1999). The use of a Hill coefficient value well below unity
would lead to a nonlinear model behavior that is biologically implausible. As a result, when the human
model was used for extrapolation to lower doses (in the calculation of risk-specific doses), the model
would tend to estimate a lower exposure level for a given blood concentration. The SAB recommends
repeating the human Emond model calculations with multiple values for the Hill coefficient to
characterize the resulting uncertainty in the exposure estimates.
A more quantitative uncertainty analysis needs to be conducted for the PBPK model. Methods that could
be useful and informative for such analysis are suggested in the response to Charge Question 6.2. The
sensitivity analysis in the Report left out the Hill coefficient, which is one of the most important
parameters in the model for low-dose extrapolation. Model sensitivities are species, dose, and dose-
scenario dependent, so they need to be determined under the same exposure conditions that dose metrics
are calculated.
The mouse model developed by EPA based on the published rat model (Emond et al., 2004, 2005, 2006)
is appropriate. However, EPA should conduct an external peer review of the mouse model. The SAB
agrees with the average daily dose calculation approaches described in the Report. However, the EPA
should carefully explain how the early life stage internal doses were calculated because serum thyroid
stimulating hormone (TSH) levels in newborns are used as a critical effect.
Reference Dose (Charge Questions 4.1 4.8)
In Section 4 of the Report EPA discusses the use of two co-critical studies (Mocarelli et al., 2008 and
Baccarelli et al., 2008) for development of the reference dose for TCDD. EPA asked the SAB to
comment on the scientific justification for selection and use of these studies to develop the reference
dose.
Selection of Critical Studies and Effects
The SAB supports EPA's selection of the Mocarelli et al. (2008) and Baccarelli et al. (2008) studies for
identifying "co-critical" effects for the derivation of the reference dose (RfD). These two human
epidemiological studies are well designed and provide sufficient exposure information, including
biological concentrations that could be used to establish acceptable lifetime daily exposure levels. The
rationale for selecting these two studies over numerous other available studies for determining the RfD
is clearly described but study weaknesses are not clearly delineated. The SAB recommends that EPA
provide a discussion of the strengths and weaknesses of these studies with an indication of whether the
weaknesses affect determination of the RfD. In addition, EPA should make use of the comprehensive
data base of both animal and human epidemiological studies to demonstrate a consistent and integrative
signal of toxicity across species and endpoints for TCDD. The coherence of evidence provided by the
studies should be made stronger in the Report by including discussion of both human and experimental
animal studies that have examined the effects of dioxin and DLCs on other reproductive and endocrine
endpoints. In this regard, dose-response relationships as well as comparisons of no-observed-adverse-
3
-------
effect levels (NOAELs) and lowest-observed-adverse-effect levels (LOAELs) should be discussed. With
respect to TCDD, the SAB agrees with EPA's assertion that traditional (e.g., immune, endocrine,
reproductive) endpoints are more appropriate than biochemical endpoints for establishing points of
departure (PODs). The associations of traditional endpoints with health outcomes have been well studied
and they are more tightly associated with adverse outcomes than biochemical endpoints. However,
because of the wealth of data on P450s and their importance in chemical response to exogenous agents,
EPA should discuss biochemical endpoints relevant to establishing and strengthening the proposed
reference dose.
Estimation of Continuous Exposure for Mocarelli etal. (2008)
Mocarelli et al. (2008) reported male reproductive effects (decrease in sperm count and motility)
observed later in life for boys with high acute exposure to TCDD between the ages of 1 and 9 (average
age 5 years), followed by low level background dietary exposure. EPA identified a 10 year critical
exposure window and estimated the continuous TCDD intake as the average of the high acute exposure
and the 5 year average exposure during the critical exposure window. The SAB finds that the pattern of
exposure in the study at the city of Seveso (high acute exposure) posed some extrapolation issues for the
EPA, particularly whether the same endpoints and or dose-response from high acute exposures would be
expected when extrapolating to low-dose chronic exposures. It would be useful for EPA to provide a
discussion of published examples in which dioxin studies were conducted using both high-dose acute
and low-dose chronic exposures in animals for the same endpoint, and how the outcomes compare both
qualitatively and quantitatively. The life stage-specific approach to hazard and dose-response
characterization for children's health risk assessment in EPA's Framework for Assessing Health Risks of
Environmental Exposures to Children (EPA, 2006) is also relevant to addressing this issue and should
be discussed.
Designation of a 20% Decrease in Sperm Count as a LOAEL for Mocarelli et al. (2008)
The SAB supports the use of the designated change from normal sperm counts and sperm motility for
determining an RfD. While the shifts observed in sperm counts may or may not pose a significant health
effect in a single individual, such shifts on a population basis could presumably lead to an increased
incidence of adverse health outcomes. The SAB recommends that World Health Organization (WHO)
reference values for male reproductive parameters and life stage differences in sperm counts in humans
be discussed in the Report.
Determination of Effective Exposure Estimate for the Baccarelli et al.(2008) Study
EPA determined the maternal intake at the LOAEL from the maternal serum-TCDD vs. neonatal TSH
regression model by finding the maternal TCDD lipid-adjusted serum concentrations (LASC) at which
neonatal TSH exceeded 5 |i-units/ml (|iU/ml). EPA then used the Emond PBPK model under the human
gestational scenario to estimate the continuous daily oral TCDD intake that would result in a TCDD
LASC corresponding to a neonatal TSH of 5 |iU/ml at the end of gestation. EPA estimated the effective
maternal intake as 0.024 ng/kg-day. The SAB supports EPA's decision to use the Baccarelli et al. (2008)
estimates of the relevant effective doses. However, EPA should clarify how these measurements relate
to ranges and variations in exposure in utero.
4
-------
Designation of 5ju-U TSHper ml blood as a LOAEL for Baccarelli et al. (2008)
EPA selected a LOAEL of 5|iU TSH per ml blood in neonates. The SAB supports EPA's designation of
the TSH endpoint within the context of the broader dioxin literature. While the shift observed in TSH
levels may or may not pose a significant health effect in a single individual, such a shift on a population
basis could presumably lead to an increased incidence of adverse health outcomes. There is a need to
better describe the potential adverse health outcomes related to altered neonatal TSH levels. For
example, in addition to effects on growth, both cognitive and motor deficits have been found in young
adults with congenital hypothyroidism (Oerbeck et al., 2003; Oerbeck et al., 2007). The Report could
better describe the consequences of transient hypothyroidism on reproductive outcomes (Anbalagan et
al., 2010).
Selection of Uncertainty Factors
A composite uncertainty factor of 30 (an uncertainty factor of 10 for the lack of a NOAEL, and an
uncertainty factor of 3 for human interindividual variability) was applied to the LOAEL of 0.020 ng/kg-
day from Mocarelli et al. (2008) to obtain the RfD. The SAB agrees that EPA has used the appropriate
uncertainty factors for the derivation of the RfD. However, a discussion of the decision not to include an
uncertainty factor for data quality is needed.
Benchmark Dose (BMD) Modeling of animal bioassay data and EPA 's Choice of POD from These
Studies
The SAB supports the use of the BMD modeling approaches and agrees that the animal data have
sufficient limitations that preclude their use to establish a RfD.
Cancer Assessment (Charge Questions 5.1 5.8)
In Section 5 of the Report, EPA has provided: a weight-of-evidence characterization of TCDD as a
known human carcinogen, conclusions regarding the mode of carcinogenic action for TCDD, EPA's
selection of data sets for cancer dose-response modeling, and consideration of approaches for
assessment of TCDD carcinogenicity. The SAB was asked to comment on the scientific soundness of
these aspects of EPA's cancer assessment.
Weight-of-Evidence Cancer Descriptor
The SAB agrees with EPA's conclusion that TCDD is "Carcinogenic to Humans." However, one Panel
member expressed a dissenting view that at best, there is equivocal evidence for the carcinogenicity of
TCDD in the occupational setting where body burdens were much higher than current or previous
background levels (see dissenting opinion in Appendix B). The SAB recommends that the Agency
provide more discussion of the power of the studies used and the difficulties involved when assessing
rare tumors. EPA should consider including studies with substantial DLC exposure where toxicity
equivalence factors (TEFs) can be calculated in the weight-of-evidence discussion. EPA should also
attempt to characterize the uncertainty regarding the carcinogenicity of TCDD at low human exposures,
since the minimum dose at which carcinogenic effects would be expected to occur cannot be clearly
5
-------
delineated from the current epidemiological human data. EPA has concluded that AhR activation is a
necessary but not sufficient precursor event in the carcinogenic activity of TCDD. Therefore, it would
be beneficial if the Agency could evaluate available data on AhR activation and related effects in human
cells and animal models to help inform the doses at which these precursor events are observed for
comparison to the epidemiological data.
Mode of Action
The SAB asserts that the mode of action for TCDD toxicity should be "reasonably well known" rather
than "largely unknown," although the exact mechanism of action has not been fully delineated for any
distinct TCDD toxicity endpoint. EPA should provide a discussion of the evidence for possible modes of
action that include both linear and nonlinear alternatives, and the description of the nature of a receptor
mediated dose-response should be expanded by including more evidence regarding the nonlinearity of
the receptor mediated dose-response for dioxin.
Selection of Critical Study for Cancer Endpoint
The SAB supports the use of the Cheng et al. (2006) study for quantitative cancer assessment. This
study incorporated information on gradation of exposure. However, expanded discussion of several
other studies would support the weight-of-evidence for carcinogenicity in less common cancers such as
lymphomas and soft tissue sarcoma. It is appropriate to use all-cancer mortality for the assessment
because TCDD appears to have multiple targets for carcinogenic action and because of the extensive
dose-response information. The SAB agrees that the use of the Emond model to estimate risk-specific
doses from Cheng et al. (2006) dose-response modeling results is scientifically justified and clearly
described. However the value of the Hill coefficient used in the model is problematic. The Cheng et al.
(2006) study did not provide completely clear information regarding risks below current background
exposure levels. The SAB therefore suggests that EPA expand the discussion to consider the possibility
that mode of action considerations could help indicate whether linear extrapolation of the Cheng et al.
(2006) data is appropriate to obtain risk estimates in this range of exposures.
Nonlinear Approach for Assessment of TCDD Carcinogenicity
The SAB finds that the Report did not respond adequately to the NAS recommendation to adopt "both
linear and nonlinear methods of risk characterization to account for the uncertainty of dose-response
relationship shape below the ED01 " EPA should present both linear and nonlinear risk assessment
approaches. In the absence of a definitive nonlinear mode of action, the linear option results can serve as
the baseline for comparison with other estimates. The examples in the current document should be
formalized and extended to allow for such a comparison.
Quantitative Uncertainty Analysis (Charge Questions 6.1 6.3)
Section 6 of the Report discusses a broad range of philosophical and methodological issues to be
considered in conducting an uncertainty analysis for TCDD toxicity. The SAB was asked to comment on
the scientific soundness of this section of the Report.
Section 6 of EPA's Report provides many useful insights for EPA's dioxin reassessment, but it is not
scientifically justified. As further discussed in the responses to the charge questions in Section 3.6 of this
6
-------
report, the SAB does not agree with EPA's argument that conducting a unified quantitative uncertainty
analysis for TCDD toxicity is unfeasible. EPA's decision to not conduct an integrated quantitative
uncertainty analysis may be based primarily on grounds of practicality or timeliness. In particular, EPA
argues that a complete quantitative uncertainty analysis would require data and resources not available.
The SAB disagrees with this logic. While EPA may lack an adequate empirical basis for full Monte-
Carlo propagation of input distributions, there are other options available. More limited evaluations can,
and should, be implemented to inform critical issues in the dioxin reassessment. In the response to
Charge Question 6.2 the SAB offers a number of methods that could be used. The SAB recommends
that EPA revise its argument that quantitative uncertainty analysis for dioxin toxicity is unfeasible.
EPA's document contrasted volitional uncertainty with cognitive uncertainty. The SAB recommends
that the term "volitional uncertainty," which might also have been called "decisional uncertainty," be
dropped from the Agency's document. EPA should display different modeling choices and the
consequences of making them. It is recommended that EPA apply standard tools and techniques for
analysis of model uncertainty.
In addition, the SAB finds that the sensitivity studies EPA has already completed (e.g., sensitivity
analysis and uncertainty analysis of dose metrics derived for the risk assessment) are useful. While there
is a need to minimize further delay of the finalization of EPA's dioxin assessment, the SAB
recommends that these completed sensitivity studies be integrated into whatever overall uncertainty
analysis the Agency elects to undertake.
7
-------
2. INTRODUCTION
EPA has been preparing an assessment of the potential health impacts of 2,3,7,8-tetrachlorodibenzo-p-
dioxin (TCDD) for many years. In 2003, EPA released an external review draft report entitled, Exposure
and Human Health Reassessment of 2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) and Related
Compounds (U.S. EPA, 2003) (hereafter referred to as the 2003 Reassessment). The EPA Science
Advisory Board (SAB) had previously reviewed the Agency's draft Reassessment in 1995 and 2000
(EPA SAB, 1995, 2001), and the 2003 Reassessment was reviewed by the National Academy of
Sciences (NAS). In 2006, the National Research Council (NRC) of the National Academies published
their evaluation of EPA's reassessment, Health Risks from Dioxin and Related Compounds: Evaluation
of the EPA Reassessment (NRC, 2006).
The NAS identified key recommendations that they believed would result in substantial improvement to
the 2003 Reassessment and thus support a scientifically robust characterization of human responses to
exposures to TCDD. The NAS recommended that EPA should: more thoroughly justify and
communicate its approaches to dose-response modeling for the health effects of dioxin, taking into
consideration both nonlinear and linear methods for characterizing cancer risk; improve the transparency
and clarity of the selection of key data sets for the dioxin dose-response analysis; reevaluate its cancer
weight-of-evidence determination for dioxin based on the Agency's 2005 Guidelines for Carcinogen
Risk Assessment, consider using physiologically-based pharmacokinetic (PBPK) modeling in the dioxin
risk assessment; and improve transparency, thoroughness and clarity in quantitative uncertainty analysis.
The NAS also encouraged EPA to calculate a reference dose (RfD), which had not been derived in the
2003 Reassessment.
In 2010, EPA's Office of Research and Development (ORD) prepared the draft report, EPA 's Reanalysis
of Key Issues Related to Dioxin Toxicity and Responses to NAS Comments (EPA, 2010) (hereafter
referred to as the Report) and requested that the EPA Science Advisory Board (SAB) conduct an
independent external peer review of the draft document. In its review, the SAB was asked to consider
the accuracy, objectivity, and transparency of EPA's reanalysis and responses to the NAS.
In response to ORD's request, the SAB convened an expert panel to conduct the review. The Panel held
an initial public teleconference on June 24, 2010 to receive an orientation to EPA's Report. The Panel
then held two public face-to-face meetings (July 13 - 15, 2010 and October 27 - 29, 2010) to deliberate
on the charge questions (see Appendix A) and two public teleconferences (March 1 and 2, 2011) to
discuss its draft report. The SAB Panel's draft report was considered and approved by the chartered
SAB on a public teleconference call on June 6, 2011. Oral and written public comments were considered
throughout the advisory process.
EPA asked the SAB to specifically focus its review on six sections of the dioxin assessment. The charge
questions (provided in Appendix A) focused on: transparency and clarity in the selection of key data sets
for dose-response analysis, the use of physiologically-based pharmacokinetic (PBPK) modeling in dose-
response modeling for cancer and noncancer endpoints, derivation of a proposed oral reference dose
(RfD) for noncancer endpoints, cancer weight-of-evidence classification, mode of action of dioxin
carcinogenicity, derivation of oral slope factor (OSF) for dioxin, and quantitative uncertainty analysis.
This SAB report provides the consensus advice and recommendations of the Panel, with the exception of
one member who offered a dissenting opinion mainly on the TCDD carcinogenicity (see Appendix B).
8
-------
3. RESPONSES TO EPA'S CHARGE QUESTIONS
3.1. General Charge Questions
The Panel was asked to comment on: whether the EPA's Report was clear and logical, whether the
Agency had objectively and clearly presented the key National Academy of Sciences (NAS)
recommendations, and whether there were other critical studies that would make a significant impact on
the conclusions of the hazard characterization or dose-response assessment of the chronic noncancer and
cancer health effects of TCDD.
Charge Question 1.1. Is the draft Response to Comments clear and logical? Has EPA
objectively and clearly presented the three key NRC recommendations?
Response:
Although the Panel has provided some editorial suggestions to improve the clarity of EPA's Report, in
general the Panel finds that the Agency has developed a report that is clear, logical and responsive to
many but not all of the recommendations of the NAS. While we provide a general assessment of
opportunities for improvement in the context of this overview charge question, most of the issues related
to clarity, organization, and responsiveness are addressed more completely and specifically in the
context of the subsequent, more specific charge questions.
With respect to the first question, the Panel finds that EPA was effective in developing a clear,
transparent, and logical response to the NAS. The Panel is particularly impressed with the process that
EPA used for identifying, reviewing, and evaluating the relevant literature. EPA's process was
comprehensive, rigorous, and included public participation. The Agency's report, EPA's Reanalysis of
Key Issues Related to Dioxin Toxicity and Response to NAS Comments (May 2010), consists of two
volumes. The first volume contains the main text of the Report and is 690 pages long (including 37
pages of references). The second volume contains appendices and is 1,159 pages long. Because of the
size and complexity of EPA's Report, the Panel finds the Executive Summary to be particularly
important and valuable in providing a concise and accurate summary. As described in detail in the
responses to the charge questions in Section 3.2 of this report, the Panel has identified the need for better
integration across chapters and greater clarity and transparency in indicating which studies did not
satisfy inclusion criteria, and therefore were not carried forward for further consideration in subsequent
chapters of the Report to meet particular needs. In contrast, the Panel is satisfied that the inclusion
criteria and the associated retained studies are well described and transparent. Given the enormity of the
dioxin published literature, the Panel recognizes that it is not a trivial matter to characterize the studies
that were not considered. Therefore the Panel suggests that the Report be revised to generally indicate
how this issue was considered. Information about studies that were excluded could be provided in an
appendix.
The Report is long and dense with a considerable amount of jargon and in some places it is quite
repetitive. These features, while a necessity of this type of document, at times detract from clarity or
make the EPA's logic difficult to discern. The Panel finds some instances where the Report would
benefit from greater clarity in writing. For example, topic sentences are sometimes not easily connected
to paragraph content. A specific example of this is in the second paragraph on page xxvii of Volume 1
9
-------
where the text does not clearly identify separate EPA activities to address NAS comments. Another
example of a Report section that should be edited to improve clarity is the qualitative discussion of the
uncertainty in the RfD (Section 4.4 of Volume 1). The clarity of this section could be improved by
including bullet points to highlight and separate key points and/or provide links to information in other
sections of the document (e.g., Section 6 - Feasibility of Quantitative Uncertainty Analysis). A careful
review by a qualified technical editor is needed. Similarly, the clarity and accessibility of the Report
should be enhanced by the inclusion of a glossary to help minimize confusion and misinterpretation
among the diverse users of the document. At 690 pages, Volume 1 is a formidable report. The Panel
appreciates the dilemma of preparing a report that is both complete and rigorous and at the same time
succinct and efficient. EPA should make greater use of appendices and eliminate redundancies in order
to provide a more approachable document.
With respect to the second part of Charge Question 1.1 (i.e., objectivity and clarity of presentation of the
three key NAS recommendations), the Panel finds that EPA has been successful. EPA's Report clearly
presents the key NAS recommendations. However, as described more fully in responses to the relevant
specific charge questions below, the Panel has identified deficiencies in the Report with respect to the
completeness of its consideration of two critical elements: 1) nonlinear dose-response for TCDD
carcinogenicity and 2) uncertainty analysis.
Recommendations
• As further discussed in the responses to the charge questions in Section 3.2 of this report, the
Panel recommends that the Report be revised to provide greater clarity and transparency in the
discussion of studies that did not satisfy inclusion criteria for use in the dioxin assessment. The
Panel recommends that the Report be revised to provide an overview of reasons why studies
were excluded.
• The Report would benefit from greater clarity in writing. The Panel therefore recommends that
the Report be carefully reviewed by a qualified technical editor and revised to incorporate such
improvements as better integration across chapters, better connection between topic sentences
and paragraph content, and elimination of repetition.
• The Panel recommends that the clarity and accessibility of EPA's Report be enhanced by the
inclusion of a glossary to help minimize confusion and misinterpretation among the diverse users
of the document.
• The Panel recommends that EPA find additional efficiencies (e.g., greater use of appendices and
elimination of redundancies) to yield more succinct and approachable document. In this regard,
we recommend that EPA develop a summary document, longer than the current executive
summary but shorter than the full Report, that captures the highlights of the response to the NAS.
• As discussed in the responses to other charge questions in this report, the Panel has identified
deficiencies in EPA's Report with respect to the completeness of its consideration of two critical
elements: 1) nonlinear dose-response for TCDD carcinogenicity, and 2) uncertainty analysis. In
the relevant charge question responses below, the Panel provides recommendations to improve
the Report in these areas.
10
-------
Charge Question 1.2. Are there other critical studies that would make a significant impact on the
conclusions of the hazard characterization and the dose-response assessment of the chronic
noncancer and cancer health effects of TCDD?
Response:
During its deliberation, the Panel did not identify any additional studies that would impact the hazard
characterization or the dose-response assessment. However, EPA's Report should provide more
discussion of null epidemiologic studies and a clearer indication of the reasons for excluding these
studies.
Recommendations
• The Panel recommends that EPA's Report provide more discussion of null epidemiologic studies
and a clearer indication of the reasons for excluding these studies.
3.2. Transparency and Clarity in the Selection of Key Data Sets for Dose-Response Analysis
The NAS committee recommended that EPA develop a clear and readily understandable methodology
for evaluating and including epidemiologic and animal bioassay data sets in dose-response evaluations.
Section 2 of EPA's Report describes the Agency's approach to ensuring transparency and clarity in the
selection of the studies for dose-response analyses. The Panel was asked to comment on: whether EPA
had been responsive to NAS concerns about transparency and clarity in data set selection, whether the
epidemiology and animal bioassay study criteria and considerations had been scientifically justified and
clearly described, and whether EPA had applied the epidemiology and animal bioassay study criteria
considerations in a scientifically sound manner.
EPA developed and applied two sets of criteria for the animal bioassays and epidemiologic data. The
Agency collected and evaluated these studies, including studies from the 2003 Reassessment and newer
studies found through literature searches and through public submissions. The Panel views with favor all
of the efforts made by EPA to develop this section of the document. The Panel compliments the Agency
for its efforts to present the nuanced differences and complicating issues surrounding this subject in a
comprehensive and logical manner. The intention of the comments and recommendations provided
below is to assist the EPA in further improvement of Section 2.
Charge Question 2.1. Is this section responsive to the NAS concerns about transparency and
clarity in data set selection for dose-response analysis?
Response:
Members of the Panel find that Section 2 of the Report is responsive to NAS concerns about
transparency and clarity. Moreover, it is perceived and appreciated that, in addressing these concerns,
EPA has improved the approach in the original 2003 Reassessment. The EPA's collaboration with
Argonne National Laboratory, and invitation to the public to engage in updating the literature search to
identify all appropriate studies for evaluation, as well as the conduct of the Dioxin Workshop in
February of 2009, were instrumental in enhancing the transparency and clarity regarding the process of
11
-------
selection of studies for the dose-response analysis. The development of clear criteria for study
evaluation and inclusion was crucial in addressing the concerns raised by the NAS.
EPA's Report presents a clear identification of the study selection process and the studies that were used
for dose-response analysis. For example, the process and criteria used to select key data sets for dose-
response analyses is described in Section 2.3 of the Report and in the Executive Summary. Flow
diagrams (e.g., ES-1 and ES-2) clearly demonstrate how studies were chosen for inclusion. Likewise,
Appendix B, which includes a point-by-point evaluation of which epidemiological studies were included
and excluded, is useful and provides a detailed rationale explaining why the EPA used the particular
studies selected in the Report. However, more discussion is needed to clearly explain exclusion of null
epidemiologic studies. In addition, the results of the literature search performed by EPA are available
online. Clarity could be improved by providing search words used for the MedLine searches. A clear
case for including high-quality human studies over animal studies is also made.
While Section 2 of the Report is deemed responsive to NAS concerns, the Panel finds that overall clarity
and transparency regarding dataset selection would be further and markedly enhanced if EPA were to
make Section 2 (and the document as a whole) more concise. In its present form, Section 2 is viewed by
the Panel as overly verbose, to the detriment of overall clarity and we provide the following
recommendations to improve the Report.
Recommendations
• The Panel strongly recommends careful and extensive editing to revise and consolidate Section 2
and the Report as a whole. Specifically, editing should include aspects of English grammar and
syntax, minimizing redundancies, and efforts to provide more succinct responses to NAS
concerns.
• The Panel recommends restructuring Section 2 to make it easier to follow a study used by EPA
from one section of the Report to another. In other words, EPA should improve overall document
integration using Section 2 as the foundation for this integration.
Charge Questions 2.2 and 2.3
Charge Question 2.1. Are the epidemiology and animal bioassay study criteria/considerations
scientifically justified and clearly described?
Charge Question 2.3. Has EPA applied the epidemiology and animal bioassay study
criteria/considerations in a scientifically sound manner? If not, please identify and provide a
rational for alternative approaches.
Response:
The Panel's discussion of Charge Questions 2.2 and 2.3 was highly integrated. Therefore, comments and
specific recommendations that stem from these two questions are presented together.
The Panel finds that EPA's study criteria and considerations are scientifically justified and clearly
described, and that these were generally applied in a scientifically sound manner. Thus, Section 2 is
deemed responsive to NAS concerns regarding the scientific justification and clarity of description for
12
-------
epidemiology and animal bioassay study criteria/considerations. However, several concerns were
discussed by the Panel, and are summarized here.
The Panel's major concern pertains to improving clarity with regard to the decision to include or exclude
particular studies and groups of studies from the data sets to be used. The rationale for distinct criteria
for epidemiological and animal studies should be made stronger, and data set selection for noncancer
and cancer endpoints has room for further clarification and justification. There was discussion, with
differences of opinion among members of the Panel, regarding EPA's scientific justification and clarity
of description concerning the Agency's decision to exclude dioxin-like compounds. There is consensus
among Panel members that the following recommended improvements would strengthen this section,
and thus the document as a whole.
Recommendations
Rationale for excluding dioxin-like compounds
• EPA should better justify the rationale for using studies where the exposure is primarily to
TCDD (or for animal studies only to TCDD) to calculate the reference dose. This justification
should include scientific and practical reasons.
• There is a wealth of literature on dioxin-like compounds that can be used to support mode of
action determination as well as the plausibility of individual biological effects that are used for
risk assessment. EPA should incorporate information from studies with dioxin-like chemicals
into a qualitative discussion of the weight-of-evidence for cancer and noncancer endpoints.
Study inclusion and exclusion criteria and considerations
• While the Panel generally finds that EPA's study criteria and considerations are scientifically
justified and clearly described, the Agency should further clarify the justifications for some
specific study inclusion and exclusion criteria/considerations. To be clear, this recommendation
does not indicate that the Panel suggests that a different approach to data set selection is needed.
However, the approach used should be explained more effectively and clearly. In this regard, the
following specific recommendations are provided to address points of concern raised by Panel
members about the study inclusion and exclusion criteria:
o EPA should remove the criterion that studies must contain an explicit statement of TCDD
purity. For research purposes, TCDD is available from a limited set of vendors, and all sell it
as a highly purified compound. Thus, for the animal studies, it is highly unlikely that any
study would be conducted using impure TCDD. Therefore, excluding a study simply due to
absence of statements regarding TCDD purity runs the risk of excluding high quality studies
because the author or journal editorial staff did not elect to include this piece of information.
o EPA should revise the explanation of the in vivo mammalian bioassay evaluation indicating
that the "study design is consistent with standard toxicological practices." This is too vague
as it likely has different meaning to readers from different backgrounds. In addition to
defining this more clearly, it is recommended that, if possible, a reference should be provided
to an EPA document in which these practices are described in detail.
13
-------
o EPA should define the phrase "common practices," and if possible cite appropriate Agency
documents to which the reader can refer for further detail. To provide further context, this
recommendation refers specifically to statements such as the following one on page 2-5:
"The study criteria shown below and in Figure 2-3 for animal bioassay data reflect EPA's
preferences for TCDD-specific study inclusion, some of which are based on common
practices and guidance for POD selection and RfD and OSF derivation."
o EPA should provide a more thorough (albeit concise) discussion of data set limitations to
educate the reader regarding Agency decisions about study inclusion/exclusion criteria. For
instance, consider adding an expanded discussion on suitability of studies of immunological
effects and/or thyroid and diabetes (e.g., Calvert,1999; Steenland, 2001; Baccarelli et al.,
2002, 2004).
Considerations concerning selection of epidemiology studies
• The Panel recommends that EPA better justify and explain considerations relating to the
selection of epidemiology studies. The following specific recommendations are provided. Many
of these specifically address the use of more standard epidemiology vocabulary and descriptors.
o EPA evaluated the available epidemiologic cohorts and studies based on five considerations
presented on pages 2-6 and 2-7 of the Report. The Panel finds that Consideration #2 (page 2-
6) is worded awkwardly and that epidemiologic terms are misspecified. The Panel therefore
recommends that EPA revise Consideration #2 as follows:
¦ Define "susceptible to important biases." This is a non-specific term and the biases
should be explained.
¦ Clarify what is meant by "control for potential confounding exposures." Does this refer to
only exposure to dioxin-like compound exposures or was it meant to more broadly refer
to other exposures as well (NIOSH cohort studies)? Does the text "bias arising from
study design" refer to selection bias or is this phrase used more broadly to describe how
exposure and outcome are measured and covariate data collected?
¦ Define what is meant by the phrase "bias arising from statistical analyses." It is unclear if
bias is the correct term, rather this may refer to model misspecification.
o With regard to scientific justification and application of Consideration #3 (listed on page 2-
7), the Panel recommends that EPA provide more discussion and clarity on the exclusion of
null epidemiologic studies.
o In Exclusion Criterion #3 (listed on page 2-7) EPA should define "reported dose."
o The Panel recommends that the discussion in Section 2 of the consideration of "confounding
and other potential sources of bias" be clarified. The differences between males and females
with regard to TCDD half-life are discussed, but the description of the number of males and
females in each study population were often missing or very difficult to determine. Also, in
the occupational cohort studies, the possibility of men and women performing different job
tasks also increased the possibility that the men and women were exposed at different levels.
However, when the job categories with assigned TCDD exposure levels were presented,
there was often no discussion of the numbers by gender in the categories. For example, the
14
-------
Manz et al. study (1991) of the Hamburg cohort (1,583 men and 399 women) does not
describe the TCDD categories by gender. In addition, the validity of the TCDD exposure
levels assigned to the categories was examined "in a group of 48 workers who provided
adipose tissue samples" (Page 2-41, lines 18-19). How were these workers selected? How
many were approached but refused to provide a sample? Assessment of selection bias in this
and other similar circumstances was lacking in some of the studies. This is particularly
notable in the lack of overall response rates reported for several of these studies. Inclusion of
these factors in the study review would be very helpful.
o The Panel recommends that discussion of the consideration that "statistical precision, power,
and study follow-up are sufficient" be clarified. These metrics can be difficult to determine
with the smaller sample size populations, but there are studies that can be very useful even
given the small samples. For example, the relative risks calculated for increasing TCDD
exposure and risk of breast cancer in the Seveso study were greatly increased in the 3rd and
4th highest exposure categories, but the relative risks were not statistically significant (page
2-56, lines 1-8 of the Report).
3.3. The Use of Toxicokinetics in the Dose-Response Modeling for Cancer and Noncancer
Endpoints
EPA used a physiologically-based pharmacokinetic (PBPK) model (Emond et al., 2004, 2005, 2006)
with blood concentration as the dose metric rather than first-order body burden. The Panel was asked to
comment on the scientific justification for EPA's application of this model, the model modifications that
EPA implemented, and EPA's characterization of uncertainty in the model. EPA also developed a PBPK
model to estimate TCDD concentration in mouse tissues. The Panel was asked to comment on the
scientific rationale for development of the mouse model, the performance of the mouse model, and
whether model uncertainty had been adequately characterized. In addition, the Panel was asked to
comment on the use of the Emond PBPK model to estimate human intake based on internal exposure
measures, EPA's sensitivity analysis of the kinetic modeling, and EPA's estimate of lifetime average
daily dose
Charge Question 3.1. The 2003 Reassessment utilizedfirst-order body burden as the dose metric.
In the draft Response to Comments document, EPA used a physiologically-based
pharmacokinetic (PBPK) model (Emond et al., 2004, 2005, 2006) with whole blood
concentration as the dose metric rather than first-order body burden. This PBPK model was
chosen, in part, because it includes a biological description of the dose-dependent elimination
rate of TCDD. EPA made specific modifications to the published model based on more recent
data. Although lipid-adjusted serum concentrations (LASC) for TCDD are commonly used as a
dose metric in the literature, EPA chose whole blood TCDD concentrations as the relevant dose
metric because serum and serum lipid are not true compartments in the Emond PBPK models
(LASC is a side calculation proportional to blood concentration).
Please comment on the following:
Charge Question 3. l.a. Please comment on the justification of applying a PBPK model with
whole blood TCDD concentration as a surrogate for tissue TCDD exposure in lieu of using first-
order body burden for the dose-response assessment of TCDD.
15
-------
Response:
The use of body burden in the 2003 Reassessment represents an improvement over the usual default
metric of administered dose (mg/kg/d) because the default metric would not properly reflect the
accumulation of dioxin in the tissues over time. However, because the accumulation of dioxin in liver is
dose-dependent, body burden would not serve as a direct surrogate for tissue exposure. The use of blood
concentration is a better choice than body burden because it is more closely related to the biologically
relevant dose metric: the free concentration of dioxin in the target tissues (liver, fetus, etc.). It is
important to recognize, however, that TCDD distribution within tissues such as the liver can be non-
uniform. Blood concentrations are routinely used to estimate biologically effective exposures for
pharmaceuticals.
The rationale for the use of blood concentration rather than lipid adjusted serum concentration (LASC)
should not be based on the Emond model structure. It would be trivial to change the model so that LASC
could be predicted. Indeed, the model is apparently used to estimate LASCs in the RfD calculations
(e.g., page xli, line 21 in the Executive Summary of the Report). The question that should be addressed
is only whether blood concentrations or LASCs provide better surrogates for cross-species and cross-
study comparisons of free dioxin concentration in the target tissues. LASC is the preferred measure for
reporting dioxin biomonitoring data, and is the measurement reported in most of the human
epidemiological studies. A metric that considers blood lipid content is also more likely to reflect free
dioxin concentration in the plasma, and hence free concentration in the target tissue. The EPA pointed
out (page xxxiv in the Executive Summary of the Report) that the LASC was related to the blood
concentration by a scalar; however, EPA incorrectly concluded that the metrics are equivalent and later
(page 3-511, line 6 of the Report) discussed the fact that the relationship between them was subject to
inter-individual and inter-species variation. If the LASC were used to drive the distribution of TCDD to
tissues, the pharmacokinetic outcome would be different from using blood as the driver because the
tissue:blood ratio would differ. If the blood fat:blood and tissue:blood values were accounted for in the
model, the use of blood and LASC would be similar. It is not clear at this point how this issue was
addressed in the dose metric calculations. Consideration of this issue is unlikely to drastically affect the
outcome of the risk calculations, but it would be important for a quantitative uncertainty analysis.
Recommendations
• The use of the blood metric is acceptable for the PBPK model. EPA should clarify how the
model deals with studies that report the concentration of dioxin in plasma, serum, blood or blood
fat:blood measurements.
Charge Question 3.1. b. Please comment on the scientific justification for using the Emond et al.
model as opposed to other available TCDD kinetic models.
Response:
The Emond model provides the best available basis for the dose metric calculations in the assessment. It
is the product of a high-caliber, multi-year research effort at EPA's National Health and Environmental
Effects Research Laboratory, and represents a significant effort in terms of data collection. This model
builds on prior PBPK modeling efforts conducted by Andersen et al. (1997). However, additional
16
-------
discussion of other published models and quantitative evaluation of the impact of model selection on
dose metric predictions should also be provided.
Recommendations
• The Report should discuss how the model was intended to be used in the assessment, which
would then dictate why a particular model was selected. That is, for the intended purposes, was
the Emond model more robust and/or simpler than other models (such as Andersen et al., 1993,
Andersen et al., 1991, and Simon et al. 2009), and did it contain sufficient details for biological
determinants deemed important by the Agency?
Charge Question 3. I.e. Please comment on the modifications implemented by EPA to the
published Emond et al. model.
Response:
The EPA modifications to the published Emond model (modifications described on page 3-44 of
the Report account for volume of plasma and describe urinary clearance using blood concentration and
not a lumped compartment) are minor and appropriate. The model changes are scientifically appropriate
and well supported.
Charge Question 3.1.d. Please comment on whether EPA adequately characterized the
uncertainty in the kinetic models.
Response:
The Report presents a reasonably thorough qualitative characterization of the uncertainty in the kinetic
models that is sufficient to support their use in the assessment. A more quantitative uncertainty analysis
is needed. Methods that could be useful and informative for such an analysis are suggested in the
response to Charge Question 6.2. It is critical to demonstrate the dependence of human equivalent dose
(HED) and risk predictions on uncertainty and variability in the model parameters, particularly those
with high sensitivity (Evans and Andersen, 2000). Moreover, dose metric uncertainty needs to be
determined under the same exposure conditions that dose metrics are calculated: both for the various
studies that serve as the basis for the dose-response assessments and for human exposures at the
corresponding HEDs and risk specific doses.
The Hill coefficients for CYPlal and CYPla2 induction used in the Emond model were 1.0 and 0.6,
respectively, based on fitting of kinetic data from single doses of dioxin (Wang et al., 1997;
Santostefano et al., 1998). However, Walker et al. (1999) subsequently estimated a Hill coefficient of
0.94 for both CYPlal and CYPla2 induction using chronic exposures which were more relevant to the
use of the Emond model in the dioxin risk assessment. The value of 0.6 used in the Emond model is well
outside the confidence interval of 0.78 to 1.14 reported by Walker et al. (1999). The use of a Hill
coefficient value well below unity would lead to a nonlinear model behavior that is biologically
implausible (hypersensitivity to induction at doses near zero). As a result, when the human model was
used for extrapolation to lower doses (as in the calculation of risk-specific doses) the model would tend
to estimate a lower exposure level for a given blood concentration. This effect can be seen in Table ES-1
of the Report, where a 5 order-of-magnitude change in risk is associated with a 6 order-of-magnitude
17
-------
change in risk specific dose. That is, the model-estimated risk specific doses in the vicinity of 10"6 risk
are about a factor of 10 lower (more conservative) than linear extrapolation. The evidence for this
parameter needs to be carefully reviewed and the reasonable range of values determined. At the least,
the human Emond model calculations will need to be repeated with multiple values of the Hill
coefficient to characterize the resulting uncertainty in the estimates. When this is done, the Agency
should also consider increasing the fat:blood partition in the human model from 100 to 200 to be more
consistent with the human data (Patterson and Mackay, 1987; Schecter and Ryan, 1989; Schecter et al.,
1989; Iida et al., 1999; Maruyama et al., 2002). The Hill coefficient is not likely to have as significant an
effect on calculations with the animal models, since low-dose extrapolation was not performed in the
animals, but this should also be verified by sensitivity/uncertainty analysis of the animal models. Public
comments were submitted to the Panel recommending consideration of a Hill coefficient value of 1.0
and pointing out why lower values are inappropriate.
Recommendations
• The Panel recommends additional efforts to fully characterize the uncertainty in the models with
special consideration of the Hill coefficient value.
Charge Question 3.2. Several of the critical studies for both noncancer and cancer dose-
response assessment were conducted in mice. A mouse PBPK model was developedfrom an
existing rat model in order to estimate TCDD concentrations in mouse tissues, including whole
blood.
Please comment on the following:
Charge Question 3.2. a. Please comment on the scientific rationale for the development of EPA 's
mouse model based on the published rat model (Emond et al., 2004, 2005, 2006).
Response:
The Panel agrees that an appropriate approach was used to develop the mouse model on the basis of the
published rat model and the available mouse kinetic data. It should be noted that the NAS
recommendation to use human data for dose metric could be accomplished because dose-dependent
elimination of TCDD has been described in humans, albeit in just a few cases. Dose-dependent
elimination has been reported repeatedly in animals and the PBPK model reflected this dose-
dependence. Using CYP1A2 data from humans (caffeine metabolism) and mice would offer an
opportunity to validate and/or adjust the mouse model.
Recommendations
• An external peer review of the mouse model should be conducted because this model has not
been published in the peer-reviewed literature. This is typically a requirement for models to be
used by the Agency.
Charge Question 3.2.b. Please comment on the performance of the mouse model in reference to
the available data.
18
-------
Response:
The Panel finds that the mouse model performed reasonably well, apart from under-prediction of urinary
excretion data. The urinary excretion data can be improved by taking into account the fact that urine
contains metabolites only, which partition differently from the parent compound. The model appears to
be adequate for use in estimating dose metrics for the assessment, but with greater uncertainty than the
rat and human models. This is considered a reasonable approach to solve a deficiency in published
PPBK models to meet the needs of this assessment.
The EPA's suggestion in the RfD chapter that the clustering of mouse points of departure (PODs) at the
lowest doses was due to mouse model failure is inappropriate and should be rewritten.
Recommendations
• EPA should use the mouse model. The scientific credibility of the model will be enhanced by its
publication in an appropriate peer reviewed journal.
Charge Question 3.2.c. Please comment on whether EPA adequately characterized the
uncertainty in the mouse and rat kinetic models. Please comment specifically on the scientific
justification of the kinetic extrapolation factor from rodents to humans.
Response:
EPA provides an adequate characterization of the qualitative uncertainty in the mouse and rat kinetic
models sufficient to justify their use, together with the human model, to estimate rodent-to-human
extrapolation factors. On the other hand, formal recalibration of the PBPK model parameters using a
Hierarchical Bayesian approach such as Markov chain Monte Carlo analysis is not considered necessary
or particularly useful. However, a more quantitative uncertainty analysis is needed.
Recommendations
• A more quantitative uncertainty analysis is recommended. Methods that could be useful and
informative for such an analysis are suggested in the response to Charge Question 6.2 in this
report.
Charge Question 3.3. Please comment on the use of Emond et al. PBPK model to estimate
human intakes based on internal exposure measures.
Response:
The modified Emond model is the best available approach for estimating exposures on the basis of
internal exposure measurements. Nevertheless, there is considerable uncertainty associated with
attempting to reconstruct prior exposures in a human population (e.g., Seveso).
19
-------
Recommendations
• Model simulations of the Cheng et al. (2006), Moccarelli et al. (2008), and Bacarelli et al. (2008)
studies need to be described in more detail. The impact of model parameter and exposure
uncertainties on model predicted dose metrics should be evaluated quantitatively and reported.
Charge Question 3.4. Please comment on the sensitivity analysis of the kinetic modeling (see
Section 3.3.5).
Response:
The Report only presents the sensitivity analysis published by Emond et al. (2006), which is not entirely
adequate for the purposes of this assessment. The analysis left out the Hill coefficient, which is one of
the most important parameters in the model for low dose extrapolation (Evans and Andersen, 2000).
Moreover, model sensitivities are species, dose, and dose-scenario dependent, so they need to be
determined under the same exposure conditions as those for which dose metrics are calculated: both for
the various studies that serve as the basis for the dose-response assessments and for human exposures at
the corresponding HEDs and risk specific doses. This represents the most pragmatic path forward for an
evaluation of model sensitivity as it relates to potential environmental regulation.
Recommendations
• In order to authenticate the model for its intended purpose, EPA should provide a sensitivity
analysis of the model under the same exposure conditions for which dose metrics were
calculated.
Charge Question 3.5. Both EPA 's noncancer and cancer dose-response assessments are based
on a lifetime average daily dose. Did EPA appropriately estimate lifetime average daily dose? If
not, please suggest alternative approaches that could be readily developed based on existing
data.
Response:
The Panel agrees with the average daily dose calculation approaches described in the Report. It was not
clear to some Panel members how the computational estimates of internal dose for newborns were
carried out since a lactation model was not used. This is important because of the use of TSH in
newborns as a critical effect. EPA, and Baccarelli et al. (2008), developed an empirical description of
the relationship between maternal TCDD levels (lipid adjusted) in serum at birth of neonate and the
measured serum TSH in the newborns up to 3 days of age. The Emond et al. model was run in an
iterative fashion by adjusting chronic daily intake (ng/kg/day) in the human gestation model to predict
maternal serum level of TCDD at term that was associated with infant serum thyroid stimulating
hormone (TSH) concentration of 5 |i-units/ml (|iU/ml) (by using the regression equation). The result
was 0.024 ng/kg bw/day.
Recommendations
• EPA should carefully explain how the early life stage internal doses are calculated.
20
-------
3.4. Reference Dose
EPA selected two co-critical studies (Mocarelli et al., 2008 and Baccarelli et al., 2008) for development
of the reference dose for TCDD. The Panel was asked to comment on the scientific justification for
selection and use of these studies to develop the reference dose.
Charge Question 4.1. The Mocarelli et al. (2008) and Baccarelli et al. (2008) studies were
selected as co-critical studies for the derivation of the RfD. Is the rationale for the choice of
Mocarelli and Baccarelli scientifically justified and clearly described? Please identify and
provide the rationale for any other studies that should be selected, including the rationale for
why the study would be considered a superior candidate for the derivation of the RfD. In
addition, male reproductive effects and changes in neonatal thyroid hormone levels, respectively,
were selected as the co-critical effects for the RfD. Please comment on whether the selection of
these critical effects is scientifically justified and clearly described. Please identify and provide
the rationale for any other endpoints that should be selected as the critical effect.
Response:
The Panel finds that use of the Mocarelli et al. (2008) and Baccarelli et al. (2008) studies is appropriate
for identifying "co-critical" effects for the RfD calculation. These are human epidemiological studies
that were well designed and executed. The studies provide sufficient exposure information, including
biological concentrations that could be used to help establish acceptable life-time daily exposure levels.
Some of the strengths of the human studies include the use of a well-characterized human cohort,
conducted by dioxin epidemiology experts, and the fact that similar PODs were found across a broad
spectrum of other reported dioxin toxicities in multiple species. The rationale for selecting these two
studies over numerous other available studies is clearly described and, overall, EPA has provided a well-
considered and rational discussion of why these two human studies were selected for determining the
RfD. However, one issue discussed by the Panel is that, while the strengths of the two human studies are
well-described, the study weaknesses are not always clearly delineated. For example, in the Baccarelli
(2008) study there is limited discussion of how the presence of polychlorinated dibenzo-p-dioxins
(PCDDs), polychlorinated dibenzofurans (PCDFs), and coplanar polychlorinated biphenyls (PCBs) that
were also found in the blood might confound the interpretation of TCDD association with elevated TSH
levels. In addition there is no discussion of the potential impact of residential histories (e.g., individuals
who may have moved in and out of Zone A after the accident). The Panel believes that it is important to
discuss both the strengths and weaknesses of these two studies. More discussion of the weaknesses of
the studies is needed.
As indicated above, the Panel agrees that the major strengths of the human studies are the use of a well-
characterized dioxin-exposed human cohort, conducted by dioxin epidemiology experts, and the fact
that similar PODs were found across a broad spectrum of other reported dioxin toxicities in multiple
species. However, in isolation from each other, and lacking a description of supportive animal and
epidemiological studies, the studies are less useful for setting the RfD. The Panel emphasizes the need to
consider these other supportive studies within the context of the weight of the dioxin and dioxin-like
compound (DLC) database. The strength of the RfD should not be based solely on these two human
epidemiology studies, but rather should be supported by integration with other similar supporting dioxin
and DLC studies. The Panel strongly supports looking at the comprehensive data base of both animal
21
-------
and human epidemiological studies together to demonstrate a consistent and integrative signal of
toxicity across species and endpoints for TCDD. It is suggested that similar studies with DLCs should
also be included as these would be supportive, at least for a semi-quantitative comparative analysis. This
"collective" impact of the studies is stated in the Report but the coherence of evidence needs to be made
stronger as it represents the contextual framing for understanding dioxin health impacts. This response
would include discussions of both human and experimental animal studies that have examined the
effects of dioxin or DLCs on other reproductive and endocrine endpoints and should, for example,
include discussion of dose-response relationships as well as comparisons of no-observed-adverse effect
levels (NOAELs) and lowest-observed-adverse effect levels (LOAELs).
The Panel notes that Figures 4.3 and 4.4 in the Report show quantitative comparisons across the RfDs
and benchmark dose lower bounds (BMDLs) calculated from the animal and epidemiological studies.
These figures are useful in understanding the quantitative similarities (to the PODs in the chosen
studies) in these calculations. The Panel also notes that since the figures do not have an indication of
endpoints being measured, just the reference to the publications, the consistency in signal (i.e., the
similarities in PODs determined) is not as readily apparent as it could be.
Although it has been addressed in the Report, the Panel recommends expanding the discussion of the
known human age-specific variability in endpoints such as sperm counts, though the data from
Moccarelli et al. (2008) do show ranges and variance (in Figure 3/Table 2), and neonatal TSH levels.
Recommendations
• EPA should provide a discussion of the strengths and weaknesses of the Mocarelli et al. (2008) and
Baccarelli et al. (2008) studies with an indication of whether the weaknesses affect determination of
the RfD.
• EPA should label the endpoints for studies included in Figures 4.3 and 4.4.
• The comprehensive data base of both animal and human epidemiological studies, including studies
with DLCs (e.g., studies cited in Goodman et al., 2010), should be discussed together to demonstrate
a consistent and integrative signal and coherence of evidence of toxicity across species and
endpoints for TCDD.
Charge Question 4.2. In the Seveso cohort, the pattern of exposure to TCDD is different from the
average daily exposure experienced by the general population. The explosion in Seveso created
a high dose pulse of TCDD followed by low level background dietary exposure in the exposed
population. In the population, this high dose pulse of TCDD was slowly eliminated from body
tissues over time. There is uncertainty regarding the influence of the high-dose pulse exposure
on the effects observed later in life.
Charge Question 4.2. a. Mocarelli et al. (2008) reported male reproductive effects observed later
in life for boys exposed to the high dose pulse of TCDD between the ages of 1 and 10. EPA
identified a 10 year critical exposure window. In the development of the candidate RfD, EPA
used an exposure averaging approach that differs from the typical approach utilizedfor animal
bioassays. EPA determined that the relevant exposure should be calculated as the mean of the
22
-------
pulse exposure and the 10-year critical exposure window average. Please comment on the
following:
Charge Question 4.2.a.i. Please comment on EPA 's approach for identifying the exposure
window and calculating average exposure for this study.
Response:
The Panel discussed extensively, both as part of the deliberations on Section 4 of the Report and also as
part of the discussion on Section 3, extrapolation issues posed by the pattern of exposure found in the
study at the city of Seveso. Issues raised included the question of whether the same endpoints and or
dose-response would be expected from such exposure scenarios with high acute exposures when
extrapolating to low-dose chronic exposures. It would be useful for EPA to provide a discussion of
published examples in which dioxin studies were conducted using both high-dose acute and low-dose
chronic exposures in animals for the same endpoint and how the outcomes compare both qualitatively
and quantitatively (see further discussion and recommendations in the response to Charge Question 4.8).
It would be important to determine whether similar results were observed for similar endpoints. Several
Panel members indicated that there were sufficient data in the immunological or reproductive areas that
may allow such a comparison. Several chronic dioxin animal studies may be useful in this regard
(Yoshizawa et al., 2009; Sand et al., 2010; Yoshizawa et al., 2010). The life stage-specific approach to
hazard and dose-response characterization for children's health risk assessment found in EPA's
Framework for Assessing Health Risks of Environmental Exposures to Children (EPA, 2006), is also
relevant to addressing this issue and should be discussed. The Panel also recommends that the
publication of Bell et al., (2010), which summarized and presented data on some differences about
chronic vs. acute exposure in maternal transfer, be considered in this discussion.
Charge Question 4.2.a.ii. Please comment on EPA 's designation of a 20% decrease in sperm
count (and an 11% decrease in sperm motility) as aLOAEL for Mocarelli et al. (2008).
Response:
The Panel finds that the designated changes from normal sperm counts and sperm motility are of public
health relevance and therefore of interest for determining an RfD. Collectively, there is support for
these endpoints within the context of the broader dioxin literature. The Panel discussed whether the
magnitude of these changes would represent an adverse health effect. While the shifts observed in
sperm counts may or may not pose a significant health effect in a single individual, such shifts on a
population basis could presumably lead to an increased incidence of adverse health outcomes.
Designation of the decreases in sperm count and motility as a LOAEL for Mocarelli et al. (2008) is
appropriate. Although there was concern expressed about the sample size used for sperm number and
known variability in the biological endpoint, the Panel finds that sample collection was conducted
consistently across subjects and the differences in groups were apparent.
The Panel recommends that further discussion of World Health Organization (WHO) reference values
for male reproductive parameters be included in the Report, as was done for the relevant TSH levels.
Several references are available which provide background information and current values
recommended by WHO regarding sperm counts (e.g., Skakkebaek, 2010). The Panel recommends that
23
-------
the standard deviations or range of changes from Mocarelli et al. (2008) be discussed in the Report
because this provides a better understanding of the potential magnitude of effect.
Life stage differences in sperm counts were discussed by the Panel. Members of the public also provided
comments on this issue. It would be appropriate to indicate in the Report that life stage differences
clearly exist in sperm counts in humans and to cite and discuss the EPA life stage document (EPA,
2006).
Recommendations
• Discussion on WHO reference values for male reproductive parameters should be included in the
Report (e.g., Skakkebaek, 2010).
• EPA should indicate and discuss within the Report that life stage differences clearly exist in sperm
counts in humans as documented (for example) in EPA, 2006.
• The standard deviations or range of changes from the Mocarelli (2008) study should be discussed in
the Report to provide a better understanding of the potential magnitude of effect.
Charge Question 4.2.b. For Baccarelli et al. (2008), the critical exposure window occurs long
after the high-dose pulse exposure. Therefore, the variability in the exposure over the critical
exposure window is likely to be less than the variability in the Mocarelli et al. subjects. EPA
concluded that the reported maternal exposures from the regression model developed by
Baccarelli et al. provide an appropriate estimate of the relevant effective dose as opposed to
extrapolating from the measured infant TCDD concentrations to maternal exposure.
Additionally, EPA selected a LOAEL of 5 piU TSH per ml blood in neonates; as this was
established by World Health Organization (WHO) as a level above which there was concern
about abnormal thyroid development later in life. Please comment on the following:
Charge Question 4.2.b.i. Please comment on EPA 's decision to use the reported maternal levels
and the appropriateness of this exposure estimate for the Baccarelli et al. study.
Response:
The Panel supports EPA's decision to use the Baccarelli et al. (2008) estimates of the relevant effective
doses. However, it should be made clearer how these measurements relate to ranges and variations in
exposure in utero.
Recommendations
• EPA should clarify how the Baccarelli et al. (2008) exposure measurements relate to ranges and
variations in exposure in utero.
Charge Question 4.2.b.ii. Please comment on EPA 's designation of 5 /ilJ TSH per ml blood as a
LOAEL for Baccarelli et al., ( 2008).
24
-------
Response:
The change in TSH levels reported by Baccarelli et al. (2008) is of public health relevance and therefore
of interest for determining an RfD. EPA's designation of 5|iU TSH/ml blood as a LOAEL for Baccarelli
et al. is appropriate. As discussed above, collectively, there is support for this endpoint within the
context of the broader dioxin literature. There was discussion on whether the magnitude of these
changes would represent an adverse health effect. The Panel notes that the shift observed in TSH levels
may or may not pose a significant health effect in a single individual, but such a shift on a population
basis could presumably lead to an increased incidence of adverse health outcomes. The Panel also
discussed the variability in neonatal TSH levels but concerns about this issue were minimized by the fact
that samples were all collected on the same postnatal day. The Panel suggests that if any follow-up data
on thyroid hormone levels, such as T3, T4 or TSH levels, are available from the population studied, then
these results should be discussed in the Report. The Panel discussed several studies describing health
effects associated with elevated neonatal TSH levels not always recognized as associated with
congenital hyperthyroidism (CH). There is a need to better describe the potential adverse health
outcomes related to altered neonatal TSH levels. For example, in addition to effects on growth, both
cognitive and motor deficits have been found in young adults with congenital hypothyroidism (Oerbeck
et al., 2003; Oerbeck et al., 2007). The Report could better describe the consequences of transient
hypothyroidism on reproductive outcomes e.g., see Anbalagan et al. (2010). Other references that relate
to this question include: Chevrier et al. (2007), Dimitropoulos et al. (2009), and Yr (2008).
Recommendations
• EPA should better describe the potential adverse health outcomes related to altered neonatal TSH
levels (e.g., effects on both cognitive and motor deficits).
Charge Question 4.3. Please comment on the rationale for the selection of the uncertainty
factors (UFs) for the RfD. If changes to the selected UFs are proposed, please identify and
provide a rationale.
Response:
A composite uncertainty factor of 30 (an uncertainty factor of 10 for the lack of a NOAEL, and an
uncertainty factor of 3 for human interindividual variability) was applied to the LOAEL of 0.020 ng/kg-
day from Mocarelli et al. (2008) to obtain the RfD. The Panel agrees that the appropriate uncertainty
factors (UFs) were included. The exclusion or inclusion of the UFs in the Report is obvious, clearly
discussed, and adequately rationalized. The Report would be more transparent if EPA included a short
discussion of the basis for the decision not to include a UF for data quality.
Recommendations
• EPA should include in the Report a short discussion of the basis for the decision not to include a UF
for data quality.
Charge Question 4.4. EPA did not consider biochemical endpoints (such as CYP induction,
oxidative stress, etc.) as potential critical effects for derivation of the RfD for TCDD due to the
uncertainties in the qualitative determination of adversity associated with such endpoints and
25
-------
quantitative determination of appropriate response levels for these types of endpoints in relation
to TCDD exposure. Please comment on whether the decision not to consider biochemical
endpoints is scientifically justified and clearly described.
Response:
Biochemical endpoints such as P450 activation, increased oxidative stress, etc. may be acceptable
endpoints to establish PODs, particularly when the quantitative relationship between the biochemical
endpoint and an adverse health outcome is clearly evident. However, with respect to TCDD, the Panel
agrees that more traditional endpoints (e.g., immune, endocrine, reproductive) are more appropriate
because associations of these endpoints with health outcomes are well studied and provide a stronger
association to an adverse outcome than biochemical endpoints. However, because of the wealth of data
on P450s and their importance in disease development, normal development, and chemical response to
exogenous agents, EPA should discuss biochemical endpoints, particularly P450s, relevant to
strengthening the proposed reference dose.
Recommendations
• Because of the wealth of data on P450s and their importance in chemical response to exogenous
agents, EPA should discuss biochemical endpoints (particularly P450s) relevant to strengthening the
proposed reference dose.
Charge Question 4.5. In using the animal bioassays, EPA averaged internal blood TCDD
concentrations over the entire dosing period, including 24 hours following the last exposure.
Please comment on EPA 's approach for averaging exposures including intermittent and one-day
gestation exposure protocols.
Response:
For animal studies it has been shown that for some effects acute exposure could give different results
than chronic exposure. For TCDD, however, its persistence might suggest that such differences may be
less pronounced. In Baccarelli et al. (2008) there is extensive discussion regarding the use of the
exposure average time for the TCDD concentrations. This is of biological significance as several papers
have indicated the unique aspects of high peak exposure of TCDD as occurred in Seveso and in several
of the animal studies. The endpoints affected as a result of these peaks do not always translate to impacts
from lower chronic exposures. As previously stated, it would be helpful to discuss any available animal
studies comparing high-dose acute vs. low-dose chronic effects on similar endpoints for dioxin or DLCs.
By returning to the broader animal literature and using time and dose-response studies from the dioxin
and DLC studies, biological support for the two critical endpoints might be found.
Charge Question 4.6. Please comment on the benchmark dose (BMD) modeling conducted by
EPA to analyze the animal bioassay data and EPA 's choice ofpoints of departure (PODs) from
these studies.
26
-------
Response:
The Panel agrees with the BMD modeling approaches used in Section 4 of the Report. EPA conclusions
that the animal data had sufficient limitations that precluded their use to establish a RfD are adequately
justified. The reasons provided, however, are quite diverse, (e.g., no NOAEL, not considered an adverse
effect, the effect at the LOAEL is too divergent from the control group, insufficient dose groups at the
low-end of the dose-response curve, monotonic responses) and there is no way for the reader to
determine which study has particular deficiencies without going back to the original paper. To help
address this gap, the Panel suggests that several of the best animal studies be discussed in some detail so
these limitations are more apparent to the reader. As indicated previously, the EPA authors need to
better cite the endpoint guidance that is present within EPA documents for defending these approaches
and application of BMD models for the critical effects. This is especially necessary given public
comments that EPA was not following its own guidelines.
Recommendations
• EPA conclusions that animal data had limitations that precluded their use in establishing the RfD are
adequately justified, but there is no way for the reader to determine which study has particular
deficiencies without going back to the original paper. To help address this gap, several of the best
animal studies should be discussed in some detail so the limitations are more apparent.
Charge Question 4.7. For the animal bioassay modeling, EPA applied the kinetic extrapolation
at the level of the POD prior to applying the uncertainty factors because EPA has less
confidence in the kinetic model output at lower doses reflective of the RfD. Please comment on
whether this approach was scientifically justified and clearly described.
Response:
The EPA approach of applying the kinetics on the actual data present at the POD is preferred in
this assessment (see additional discussion in the responses to the charge questions in Section 3.3 of this
report - The use of toxicokinetics in the dose-response modeling for cancer and noncancer endpoints).
Charge Question 4.8. Please comment as to whether EPA 's qualitative discussion of uncertainty
in the RfD is justified and clearly described.
Response:
The Panel agrees that EPA's discussion of the uncertainties in deriving the RfD using the Seveso cohort
is justified and clearly described. Section 4 of the Report discusses study limitations regarding the need
to adjust from acute exposure to average daily dose, the issue of critical windows, co-exposure to DLCs,
and the strength/weaknesses of the animal data. The Panel agrees with EPA that the major limitation of
the Seveso cohort is the uncertainty arising from how well the effects resulting from high-dose acute
exposure translate to low-dose daily exposures. Again, it would be useful to re-review the animal studies
to identify whether there are any studies where dioxin or DLCs were administered by acute as well as
chronic (or even subchronic) exposure and comparable endpoints were examined. If so, the information
can be used to help confirm or refute the accuracy of the "average daily dose" adjustment. This is of
particular concern in the Mocarelli study as "time periods of susceptibility" appear in male reproductive
27
-------
development and these periods (windows) may be very short. Again, animal studies, particularly those
involving male reproduction, may be helpful.
It would also be useful to include a discussion of potential uncertainty in the exposure estimates from
the Baccarelli study. Serum dioxin levels were only established in a subset of the cohort (approximately
51) at the time of the study while dioxin levels from the main cohort were estimated from data collected
from zone of residence (A or B) at a much earlier time.
The discussion in the Report of whether the background DLC exposure may have a significant impact,
particularly at the lower TCDD exposure levels, is important. While the Panel agrees that the true DLC
impact cannot be determined, it would be helpful to provide some general estimates of the variability
that may occur at the proposed RfD.
Recommendations
• EPA should re-review the animal studies to identify whether there are any studies where dioxin or
DLCs were administered by acute as well as chronic (or even subchronic) exposure and comparable
endpoints were examined. If so, the information can be used to help confirm or refute the accuracy
of the "average daily dose" adjustment.
• EPA should include in the Report a discussion of potential uncertainty in the exposure estimates
from the Baccarelli study. Serum dioxin levels were only established in a subset of the cohort
(approximately 51) at the time of the study while dioxin levels from the main cohort were estimated
from data collected from zone of residence (A or B) at a much earlier time.
• Background DLC exposure may have a significant impact, particularly at the lower TCDD exposure
levels. While the Panel agrees that the true DLC impact cannot be determined, we recommend that
EPA provide some general estimates of the variability that may occur at the proposed RfD.
3.5. Cancer Assessment
In the Report EPA has provided: a weight-of-evidence characterization of TCDD as a known human
carcinogen, conclusions regarding the mode of carcinogenic action for TCDD, EPA's selection of data
sets for cancer dose-response modeling, and consideration of approaches for assessment of TCDD
carcinogenicity. The Panel was asked to comment on the scientific soundness of these aspects of EPA's
cancer assessment.
Charge Question 5.1. Weight-of-Evidence Cancer Descriptor: The 2003 Reassessment
concluded that TCDD is a "known human carcinogen. " In the current draft Response to
Comments document, EPA concluded that under the 2005 Guidelines for Carcinogen Risk
Assessment (U.S. EPA, 2005) TCDD is "carcinogenic to humans. "Is the weight-of-evidence
characterization scientifically justified and clearly described?
Response:
The Panel agrees with the characterization that "TCDD is carcinogenic to humans" under EPA's 2005
Guidelines for Carcinogen Risk Assessment. Available occupational epidemiologic studies (e.g., studies
28
-------
of chemical manufacturing workers in the National Institute for Occupational Safety and Health,
NIOSH, cohort) provide convincing evidence of an association between TCDD and human cancer that
cannot be reasonably attributed to chance or confounding and other types of bias, and with a
demonstration of temporality, strength of association, consistency, biological plausibility, and a
biological gradient. The findings of these studies have been previously described in EPA's 2003
Reassessment and in the Report. Additional evidence from animal studies and from mechanistic studies
provides support for the characterization of TCDD as carcinogenic to humans. One Panel member
expressed a dissenting opinion (see Appendix B of this report) that at best, there is equivocal evidence
for the carcinogenicity of TCDD in the occupational setting where body burdens were much higher than
current or previous background levels. The Panel provides the following recommendations to strengthen
the Report.
Recommendations
• EPA should provide more discussion of the power of studies used and the difficulties involved
when assessing rare tumors. Thoroughly addressing these aspects will make the weight-of-
evidence characterization in Section 5 of the Report more clear and transparent.
• In the weight-of-evidence characterization, EPA should build on all the available data to support
the decision. It needs to be made clear how different types of data (in vitro, in vivo, human)
support each other; or not.
• EPA should consider including studies with substantial DLC exposure where TEFs can be
calculated. Specific experimental studies include Li and Rozman (1995), Rozman et al. (1993,
2005), and Viluksela et al. (1994, 1997a,b, 1998a,b).
• EPA should attempt to characterize the uncertainty regarding the carcinogenicity of TCDD at
low human exposures, since the minimum dose at which carcinogenic effects would be expected
to occur cannot be clearly delineated from the current epidemiological human data. The Agency
has concluded that AhR activation is a necessary but not sufficient precursor event in the
carcinogenic activity of TCDD. Therefore, it would be beneficial if the Agency could evaluate
available data on AhR activation and related effects in human cells and animal models to help
inform the doses at which these precursor events are observed for comparison with the
epidemiological data.
Charge Question 5.2. Mode of Action: The mode of action of a carcinogen can inform
identification of hazards and approaches usedfor a dose-response assessment. The mode of
carcinogenic action for TCDD has not been elucidated for any tumor type. EPA concluded that,
while interaction with the Ah receptor is likely to be a necessary early event in TCDD
carcinogenicity in experimental animals, the downstream events involved are unknown.
Charge Question 5.2.a. Are the available data related to mode(s) of action for the
carcinogenicity of TCDD appropriately characterized and clearly presented?
29
-------
Response:
The Panel appreciates the attempts by the Agency to further develop cancer mode of action concepts
based on available dioxin liver, lung, and thyroid toxicity data. Such innovative and explorative work is
clearly fundamental to the continued need to further develop risk assessment sciences and to make more
detailed and integrated use of already existing and published data.
The Panel compliments the Agency for providing an up-to-date dioxin cancer mode of action section in
its response to NAS comments. It could, however, be improved by incorporating additional data on
linear and nonlinear modes of action in different target tissues and life stages. A large amount of data
related to the mode of action for the carcinogenicity of TCDD is described in the Report, but the focus
appears to be on presenting evidence that supports the use of a default linear approach rather than
providing a balanced evaluation of alternative mode of action hypotheses.
The discussion of the likely dose-response for receptor mediated processes focuses only on the first step,
binding of the agonist to the receptor, which is ultimately linear at low concentrations. However, no
discussion is given to the nature of the dose-response for the down-stream sequelae of receptor
activation, for which there is evidence of nonlinearity. It is, in fact, the fundamentally nonlinear nature
of the dose-response for receptor mediated processes that underlies the conviction of a large segment of
the scientific community that a nonlinear approach should be preferred for the risk assessment for
dioxin.
Recommendations
• EPA should further expand the discussion of mode of action data available to delineate linear
versus nonlinear modes of action and effects in different target tissues at different life stages.
Charge Question 5.2.b. Do the available data support EPA 's conclusion that the overall mode(s)
of action for TCDD-induced carcinogenesis is largely unknown? Please comment on whether
this evaluation is clearly described.
Response:
The Panel notes that much is known about TCDD toxicity and mode of action. Some Panel members
indicated that the characterization of the mode of action should be "reasonably well known" rather than
"largely unknown." Nevertheless, the Panel agrees that the exact mechanism-of-action has not been
fully delineated for any distinct TCDD-toxicity end-point. For example, it was pointed out that most
TCDD toxicities are mediated by activation of the AhR. Many studies have demonstrated that TCDD
can activate or interfere with the activity of estrogen receptors, as well as other steroid receptors. Such
interference can disrupt the regulation of cell proliferation, cell death and tissue differentiation. By
disrupting these cell functions, TCDD can have profound and lasting effects as demonstrated by studies
showing that TCDD exposure during development produces adult neural dysfunctions.
Recommendations
• EPA should provide a discussion of the evidence for possible modes of action that include both
linear and nonlinear alternatives.
30
-------
• EPA should describe the receptor mediated nonlinear mode of action for dioxin (e.g., Van den
Heuvel et al., 1994; Li and Rozman 1995; Andersen et al., 1997; Bhattacharya et al., 2010; Gim
et al., 2010) and DLCs (e.g., Rozman et al., 1993, 2005; Stahl et al., 1994; Viluksela et al., 1994,
1997a,b, 1998a,b), as well as evidence regarding the fundamentally nonlinear nature of receptor
mediated cellular responses (e.g., Andersen et al., 1999; Louis and Becskei, 2002; Zhang et al.,
2010).
Charge Question 5.3. Is EPA's approach for selecting data sets from the key epidemiologic
studies and animal bioassays identifiedfor cancer dose response modeling scientifically justified
and clearly described?
Response:
EPA derived candidate oral slope factors (OSFs) for all cancer mortality from human epidemiological
studies as well as for individual and combined tumor incidence from rodent cancer bioassays. The study
inclusion criteria outlined in the Report were used to select the studies for TCDD dose-response
modeling. EPA chose to give higher consideration to the human epidemiological data in developing the
OSF for TCDD and decided to use the results of the Cheng et al. (2006) study based on total cancer
mortality. The Panel agrees with this approach because the Cheng et al. (2006) study incorporated
information on gradation of exposure and because in the case of TCDD there appear to be multiple
targets for carcinogenic action. Expanded discussion of several other studies would support the weight-
of-evidence for carcinogenicity in less common cancers such as lymphomas and soft tissue sarcoma.
The Panel discussed the possible value of including studies with DLCs in the evaluation of the weight-
of-evidence, in light of the small number of studies involving primarily exposure to TCDD. There are a
numerous studies in the literature involving the health effects of DLCs. These include rice oil poisoning
incidents in Japan and Taiwan. These incidents have been described, and additional references have
been provided, in Schecter and Gasiewicz (2003).
Recommendations
• EPA should present in a clear and visible format, for example in a table, an indication of which
studies were carried forward or not, and the reasons for the decisions made. The weight-of-
evidence discussion should be expanded to include evidence from studies of individual cancers
for which precise gradation of exposure data is lacking.
Charge Question 5.4. For the animal bioassay data, potential cancer oral slope factors (OSFs)
were calculated by linear extrapolation (using a linear, non threshold cancer approach) from the
point of departure (POD). EPA also estimated the composite risk of the occurrence of several
tumor types from the animal cancer bioassay data.
Please comment on the following:
Charge Question 5.4.a. Please comment on whether the approach for estimating cancer risk,
including the use of tumor modeling of the TCDD animal cancer bioassay data, is scientifically
justified and clearly described.
31
-------
Response:
The Panel agrees that the approach for estimating cancer risk from animal studies was scientifically
justified and clearly described.
Charge Question 5.4.b. Please comment on the choice of using a BMDLoi as the POD for the
development of candidate oral slope factors derived from the TCDD animal cancer bioassays.
Response:
The Panel notes the consistency of the selection of the BMDLoi as the POD with Agency guidelines and
has no further comments.
Charge Question 5.5. EPA selected Cheng et al. (2006) - an analysis of the NIOSH occupational
cohort - as the critical study for oral slope factor (OSF) development. This study was chosen
because it considers dose-dependent elimination of TCDD rather than first-order kinetics.
Charge Question 5.5. a. Please comment on whether the rationale for this selection is
scientifically justified and clearly described. Please identify and provide the rationale for any
other studies that should be considered and provide a critical evaluation of the study and of its
suitability for meeting the goals of a quantitative cancer assessment.
Response:
The Panel agrees that Cheng et al (2006) is the appropriate study for OSF development. The selection of
this study is well described.
Charge Question 5.5.b. Cheng et al. (2006) analyzed all-cancer mortality. Please comment on
the use of all-cancer mortality as the basis of the OSF.
Response:
The Panel agrees that in this case, it is appropriate to use the Cheng et al. (2006) study, which analyzed
all-cancer mortality, to derive the OSF because of the extensive dose-response information and because
in the case of TCDD there appear to be multiple targets for carcinogenic action.
Charge Question 5.5.c. Please comment on whether the use of the Emond PBPK model in the
estimation of risk-specific doses from the Cheng et al. dose-response modeling results is
scientifically justified and clearly described.
Response:
The Panel agrees that the use of the Emond model to estimate risk-specific doses from the Cheng et al.
(2006) dose-response modeling results is scientifically justified and clearly described. This is because
the "concentration-and-age-dependent elimination model" (CADM) used in Cheng et al. (2006) was
considered to be less useful for predicting reliable metrics for sensitive human subpopulations. Also, the
dose conversions were consistent with those used in the derivation of the RfD. However, as discussed in
32
-------
the response to Charge Question 3.1.d, the Panel is concerned about the value of the Hill coefficient
used.
Charge Question 5.5.d. EPA elected to use the log linear relationship of fat concentration and
rate ratio to estimate risk-specific doses at all risk levels. EPA could have estimated a POD for
cancer risk itself at a single risk level (BMR) for extrapolation to the origin. Please comment on
EPA 's choice of extrapolation approach.
Response:
Since the fat concentrations generated by CADM were not linear with the oral exposure at higher doses,
a single oral slope factor to be used for all risk levels could not be obtained. EPA used the upper 95%
bound on the slope (from Cheng et al., 2006) of the linear relationship between the natural logarithm of
the rate ratio and the cumulative fat TCDD concentration (fat-AUC) to estimate risk-specific doses for
TCDD at all risk levels. The Panel agrees that the Agency has chosen the appropriate extrapolation
approach.
Charge Question 5.5. e. The slope factor derivedfrom Cheng et al. (2006) was extrapolated
below the background TCDD exposure levels experienced by the NIOSH cohort. Please
comment on this extrapolation.
Response:
The ability of the Cheng study to be informative regarding risks below current background exposure
levels is not completely clear. Consideration of both linear and nonlinear mode of action may clarify the
extrapolation of risks below background exposure levels.
Recommendations
• EPA should expand the discussion in the Report to consider the possibility that mode of action
considerations could help to inform whether linear extrapolation of the Cheng data to obtain risk
estimates in this range of exposures is appropriate.
Charge Question 5.6. Please comment on whether EPA has clearly described the major
qualitative uncertainties in the derivation of the OSF.
Response:
The Panel finds the description of qualitative uncertainties in the derivation of the OSF to be clear and
adequate.
Charge Question 5.7. EPA did not consider dioxin-like compounds (DLCs) in the cancer dose-
response modeling because the occupational exposures in the available cohorts were primarily
to TCDD. BackgroundDLC exposures were not incorporated in the dose-response modeling
because EPA judged that it was not possible to disaggregate the responses from background
exposure to DLCs and occupational exposure to TCDD. Please comment on whether this
approach is scientifically justified and clearly described.
33
-------
Response:
While the Panel finds that it is important to include DLC studies in the weight-of-evidence analysis, we
are conflicted on their use as a source of dose-response estimates for TCDD. The Panel notes the
scientific importance and regulatory relevance of including a coordinated TEQ/DLC discussion in the
Report. Including TEQ/DLC aspects in the evaluation would allow for the use of additional studies with
dose-response information that more closely mirror environmental exposures. On the other hand, the
Panel recognizes the complications associated with developing a TCDD risk estimate that is dependent
on current TEF values.
Recommendations
• DLC studies should be considered in the weight-of-evidence discussion. There is a wealth of
literature on these compounds that can be used to support mode of action determination as well
as the plausibility of individual biological effects that are used as the basis for risk assessment.
Charge Question 5.8. The NRC suggested that EPA consider nonlinear approaches for the
assessment of TCDD carcinogenicity. In the Response to Comments, EPA presents two
illustrative nonlinear approaches for cancer, but considers both inappropriate to use because
lack of MO A information.
Charge Question 5.8.a. Please comment on these two illustrative nonlinear approaches
including EPA 's conclusions regarding the limitations of these approaches.
Response:
EPA's Report does not respond adequately to the NAS recommendation to adopt "both linear and
nonlinear methods of risk characterization to account for the uncertainty of dose-response relationship
shape below the EDoi." Instead of adopting both linear and nonlinear methods, the EPA argues that
only a linear approach can be justified, and derives two examples of RfD development using a nonlinear
approach that is characterized as an illustrative exercise only. The choice not to include both linear and
nonlinear risk assessment approaches for TCDD is inconsistent with the Agency's 2005 Guidelines for
Carcinogen Risk Assessment (EPA, 2005) (page 3-23/24):
"Nonlinear extrapolation having a significant biological support may be presented in addition to
a linear approach when the available data and a weight-of-evidence evaluation support a
nonlinear approach, but the data are not strong enough to ascertain the mode of action applying
the Agency's mode of action framework."
"In the absence of data supporting a biologically based model for extrapolation outside of the
observed range, the choice of approach is based on the view of mode of action of the agent
arrived at in the hazard assessment. If more than one approach (e.g., both a nonlinear and linear
approach) are supported by the data, they should be used and presented to the decision maker."
34
-------
Recommendations
• EPA should present both linear and nonlinear risk assessment approaches. In the absence of a
definitive nonlinear mode of action, the linear option results can serve as the baseline for
comparison with other estimates. The examples in the current document should be formalized
and extended to allow for such a comparison.
Charge Question 5.8.b. Are there other nonlinear approaches that could be readily developed
based on existing data for the assessment of TCDD carcinogenicity? If so, please suggest
alternative approaches and describe their utility and suitability for meeting the goals of a
quantitative cancer assessment.
Recommendations
• Since the EPA nonlinear analysis only used studies in S-D rats that were identified in Section 2
of the Report for potential noncancer dose-response modeling, additional alternative PODs
should be added. For example, Simon et al. (2009), which was cited in EPA's Report, provided a
number of alternative PODs for a nonlinear approach that should be included in the EPA risk
assessment.
3.6. Feasibility of Quantitative Uncertainty Analysis
In its evaluation of EPA's 2003 Reassessment, the NAS committee recommended that EPA improve the
transparency, thoroughness, and clarity in quantitative uncertainty analysis (QUA). Section 6 of EPA's
Response to NAS Comments document addresses NAS comments regarding QUA. The Panel was asked
to comment on: whether Section 6 of EPA's Report was clearly presented and scientifically justified;
EPA's conclusion that a QUA is not feasible; the discussion of volitional uncertainty, and the utility of
the limited sensitivity studies presented by EPA.
Charge Question 6.1. Please comment on the discussion in this Section. Is the response clearly
presented and scientifically justified?
Response:
As discussed below, the Panel finds that Section 6 of EPA's Report is clearly presented, but it is not
scientifically justified. In particular, the Panel disagrees with EPA's argument that, since the most
detailed and complete available methods for conducting a QUA are unfeasible, no analysis can be
conducted at all. There are a number of approaches for conducting a QUA, some of which are feasible
given current knowledge and data. Specific methods that can be implemented in a timely manner using
available data and knowledge are suggested.
Clarity of the EPA response to the NAS presented in Section 6 of the Report
The EPA response is clearly presented and provides many useful insights for the Agency's dioxin
reassessment. The Report addresses a broad range of philosophical and methodological issues in
conducting an uncertainty analysis for TCDD toxicity, specifically for estimates of cancer oral slope
factors and noncancer reference doses. Section 6 is successful in identifying the challenges involved in
35
-------
assessing uncertainty in toxicity estimates based on a small set of available models for toxicokinetics,
dose-response relationships, and low dose extrapolation, with limited application, testing, and
verification; and a small set of animal bioassay, epidemiological or clinical/case studies, many with
differing endpoints, dose metrics, and (in the case of the human studies) uncertain exposure and subject
data. In its discussion of available methods, the Report is somewhat biased in its treatment of certain
statistical methods (discussed below) which could address some of these issues (though the Report does
note the potential contribution of the methods at the end of Section 6 as part of ongoing or future
studies) and overly pessimistic regarding our ability to provide improved quantitative estimates for
certain portions of the toxicity assessment.
Some Panel members note that the whole section should be rewritten to make it more understandable to
non-statisticians. As further discussed in the editorial comments on Section 6 in Appendix D of this
report, some phrasing and word choices in the text should be reconsidered, in particular "exotic
methods," "volitional uncertainty," and "epistemic uncertainty." The Panel finds that the definition of
"quantitative uncertainty analysis" is overly narrow and should be expanded to embrace other common
and useful methods discussed below. In a few other places, the Report's wording in Section 6 is strongly
at variance with the literature on uncertainty analysis (see editorial comments in Appendix D of this
report).
Scientific justification of the arguments presented in Section 6
The Panel finds that the arguments in Section 6 are not scientifically justified. In Section 6, EPA's
decision to not do an integrated quantitative uncertainty analysis is presented and a variety of theoretical
issues are discussed, but EPA's decision may be based primarily on grounds of practicality or
timeliness. EPA indicates that a complete quantitative uncertainty analysis would require data and
resources not available. We disagree with this logic. More limited evaluations can, and should, still be
implemented to inform critical issues in the dioxin reassessment. EPA should be methodical in
considering what variables and components of the assessment would be included in the analysis. The
Panel finds that the uncertainty narratives and sensitivity analyses already in the document are an
excellent beginning and may constitute the lion's share of the work necessary to implement quantitative
uncertainty analysis based on simple bounding.
The Panel does not concur with the specific argument EPA used to justify not doing a unified QUA. If
the answer to the question of why EPA did not undertake one is that it was not possible to specify
precise marginal distributions and dependence functions from existing data, then the conclusion would
be that EPA has not been responsive to the NAS criticism, because there are many possible approaches
that could be used that do not depend on such specifications. If the argument is that EPA guidance does
not require a QUA, then one might agree that the NAS criticism is perhaps itself unreasonable. If EPA
had asserted that it actually had done an uncertainty analysis in the form of uncertainty factors (UFs) and
the limited sensitivity studies that were performed, then that might be understandable, though not
consistent with the current state-of-the-art in risk and uncertainty analysis. Even if the argument had
been that mounting a QUA is a significant and controversial undertaking itself and that doing one should
not delay the finalization of the Report, then such a practicality argument would be understandable given
the protracted delay in completing the dioxin reassessment.
Instead, EPA asserts that "Data are the ultimate arbiter of whether quantitative uncertainty analysis ...
has sufficient evidentiary support." This flies in the face of how uncertainty analyses are normally
36
-------
conceived. Of course, the absence of data is never a substantive reason not to conduct an uncertainty
analysis; it is the reason to do one.
In its Report, EPA indicates that it needs an "underlying distribution from which to sample" in order to
conduct a quantitative uncertainty analysis. The Panel notes that this is not necessarily true, and it is
facile to shrug off a call to characterize and account for important uncertainties in the assessment
process on these grounds alone. If one can estimate the value of a quantity, then one should be able to
express the uncertainty about the value, otherwise one does not really have a scientific measurement in
the first place. One is not forced to identify precise probability distributions and dependence functions
for everything that is to be characterized as uncertain. Even when the uncertainty is volitional (or
decisional or just model uncertainty), there can be relevant ranges that are interesting to decision makers
and stakeholders. In some cases, the analysis may be formally closer to a sensitivity analysis, but some
appropriate response is usually possible, if not always practicable. To its credit, EPA has acknowledged
the legitimacy of the call for QUA by NAS and undertaken some efforts in this direction.
In the Report, EPA calls uncertainty analysis an "emerging area in science" and this is inarguably true,
but it does not seem reasonable to hold that methodological research is necessary for EPA to do
anything more comprehensive to respond to NAS's criticism, even if we disallow the use of expert
elicitation. Under a commitment to the idea that analyses be data-driven, it is possible to do something
that is useful, even if it is not predicated on precise distributions. There are a variety of ways to conduct
a quantitative uncertainty analysis, even an entirely probabilistic one that obeys the Kolmogorov axioms
(Gillies, 2000) that require neither extensive data nor expert elicitation. The response to Charge
Question 6.2 below provides a list of various ways (with references) to accomplish this. The list includes
probability trees or model choice trees that articulate the structure of the model and dependencies,
sensitivity analyses, simple interval analysis that just propagates the plausible ranges, and the
supervaluation approach that uses nested inner and outer intervals (with the inner range representing the
values that most everyone considers to be plausible values and the outer range representing
conservatively broad ranges). There is also a continuous and unbounded version of nesting intervals in
an approach known as info-gap analysis that would be useful if one cannot develop finite bounds on
some of the inputs. One can also propagate bounds on distribution functions, so whatever imperfect
information about each input variable's distribution is available, one can fashion bounds on distribution
functions and propagate them through the calculations, with or without assumptions or information
about the dependencies among variables.
The Panel notes that the approaches mentioned above require EPA to make certain modeling judgments,
in the same way that developing any analysis requires judgments. However, this does not mean that
analysts would be required to make up numbers or elicit any expert opinion. Such an analysis does not
necessarily require a lot of extra work by EPA. These methods can be simple to develop, and they are
mostly computationally trivial. Of course, the more comprehensive the analysis is, the harder it is to
complete. But the analysis does not have to be fully comprehensive to provide useful insights.
We note that there was not perfect consensus among Panel members about the value of a quantitative
uncertainty analysis. Some on the Panel agree that an uncertainty analysis is not an absolute good. For
instance, if the final answer is already clear, an uncertainty analysis can be a waste of time and
resources. It would not be reasonable to insist on another analysis which would merely waste time and
resources. Likewise, if the analysis is done poorly, or without appeal to available evidence from the real
world, it can be misleading. For instance, the idea, mentioned in footnote 66 on page 6-20 of EPA's
37
-------
Report, of arbitrarily converting uncertainty factors to independent lognormal random variables in a
scattered attempt to mount a QUA would entail a suite of unjustified and probably untenable
assumptions rendering the exercise nearly pointless. Finally, if the analysis is used strategically to avoid
rendering or finalizing a decision that is proper, it can be counterproductive. However, most members of
the Panel indicate that quantitative uncertainty analysis is an integral part of any good assessment, and
that one is essential to address the many empirically unresolved questions and issues that have arisen in
this assessment which beg for explicit consideration in the context of an uncertainty analysis. In its
discussion of the other charge questions, the Panel has identified a number of important issues that
should be addressed in an eventual uncertainty analysis.
Other methods to be considered
The Panel finds that relevant Bayesian methods have been inadequately addressed and improperly
dismissed in Section 6. In particular, methods that should be given a more extensive and balanced
discussion with more citations to the literature include: 1) Bayesian hierarchical modeling (Coull et al.,
2003; Axelrad et al., 2007; Ryan, 2008; Choi et al., 2010;) which is used for combining information
from multiple studies, and 2) Bayesian model averaging (Morales et al., 2006; Viallefont et al., 2001;
Wheeler and Bailer 2007, 2009) which would be useful for considering more than one dose-response
equation, while allowing the data to weight their relative likelihood and contribution to the estimate.
These Bayesian methods should not be referred to as "exotic." For example, in agreeing with the
Section 6 authors that these methods should be pursued in ongoing and future case studies, White et al.
(2009) refer to them as "advanced," rather than exotic. Specifically, they recommend that health
scientists should explore statistical approaches to model selection and suggest that "improvements to
statistical approaches for model selection, such as model averaging, should be pursued. Case study
applications of these advanced statistical approaches will identify potential strengths and weaknesses of
the approaches and their significance for risk characterization" (White et al., 2009).
Recommendations
• The Panel recommends that EPA revise Section 6 of the Report because, as discussed above, the
arguments in this section are not scientifically justified. In particular, EPA should consider
revising its argument that quantitative uncertainty analysis is unfeasible for the dioxin
assessment. Specific suggestions regarding feasible methods for quantitative uncertainty analysis
are provided herein.
Charge Question 6.2. Please comment on EPA 's overall conclusion that a comprehensive
quantitative uncertainty analysis is not feasible.
Response:
As discussed above, the Panel rejects EPA's argument that a quantitative uncertainty analysis is
unfeasible. Although a quantitative uncertainty analysis is challenging, the Panel does not agree that it is
impossible or even impractical to undertake one. While it may well be true that we lack an adequate
empirical basis for full Monte-Carlo propagation of input distributions, there are many other options
available. Many on the Panel indicated that the present circumstances warrant a compromise approach
that would be simple and achievable with modest effort by the Agency. Various bounding approaches,
sensitivity studies, uncertainty set analyses, and event trees (probability trees without the probabilities)
38
-------
are suggested as possible approaches that could be used. With such methods, legitimate and
comprehensive uncertainty analyses (including even fully probabilistic analyses) are possible. They
would be useful and sufficient to respond to NAS' criticism.
The Panel agrees with EPA's assertion that expert elicitation would be problematic and should be "off
the table." However, many on the Panel further suggest that value-of-information methods would also
be very useful, although feedback from EPA included reservations about this idea. A discussion of value
of information methods is provided in Appendix C of this report.
The Panel considered the use of bounding approaches for quantitative uncertainty analysis and asked
EPA to provide information about the limitations of bounding approaches. In response, EPA asked Dr.
Roger Cooke (an author of EPA's Report) to send the Panel a document on bounding analysis. The short
bounding analysis document provided to the Panel by Dr. Cooke focused on the features of interval
analysis, although this is not by any means the only approach that might be useful in the context of the
dioxin assessment. The bounding analysis document mentions one issue that could be construed as a
disadvantage of this simplest bounding approach. It is the idea that the ranges are supposed to be
absolute bounds on the possible values of each input variable. So, for instance, the only thing one can
say about a percentage is that it is between zero and 100%, or the only thing one can say about a
dispersal distance is that it is between zero and the circumference of the Earth (these are Dr. Cooke's
examples). But the Panel finds that this criticism seems to represent a misunderstanding of the word
"absolute." Vacuous (e.g., physically limiting) bounds are not the only bounds that can be used in
interval analysis. In fact, they are meant to be informed by observed study results. Furthermore, one is
not necessarily limited to interval ranges and interval analysis.
The Panel suggests that there are in fact a variety of methods that, with proper application, could be
useful and informative, including:
• Sensitivity analysis studies (even if not completely comprehensive) (Saltelli et al., 2000a,b;
Frey and Patil, 2002),
• Interval analysis (Moore 1966; Neumaier, 1990) which has been widely used for decades and
can be applied to complex models and even blackbox models (Trejo and Kreinovich, 2001),
• Nesting of intervals, e.g., two levels, wide and narrow can give conservative and optimistic
characterizations of overall uncertainty (van Fraassen, 1966, 1980),
• Probability bounds analysis (Ferson and Long, 1995; Ferson et al., 2003) including Bayesian p-
boxes (Montgomery, 2009), which has been used in a variety of applications (Regan et al.,
2002a,b; Aughenbaugh and Paredis, 2007; Dixon, 2007; Karanki et al., 2009; Minnery et al.,
2009), including assessments at two Superfund sites (EPA, 2002-2005, 2007),
• Info-gap decision theory (Ben-Haim, 2006) which has been used in several applications,
(Regan et al., 2005; Davidovitch et al., 2009; Hall and Harvey, 2009; Rout et al., 2009;
Yokomizo, 2009),
• Robust optimization (Bertsimas and Brown, 2009; Bertsimas et al., 2009; Bertsimas et al.,
2010; Ben-Tal et al., 2010), and
• Probability trees, which are distributional methods for considering alternative assumptions and
models at various stages of the toxicity assessment. Small (2008) explains that the distributional
approach for characterizing uncertainty in cancer risk assessment was developed by Evans,
Sielken, and co-workers beginning in the 1990s (Sielken, 1990; Holland and Sielken, 1993;
Evans et al., 1994a,b, 1995; Sielken et al., 1995; Sielken and Valdez-Flores, 1996, 1999) and has
39
-------
also been referred to as information analysis, weight-of-evidence analysis, the comprehensive
methodology, and comprehensive realism (Sielken, 1990; Sielken et al.,1995). The method has
since been acknowledged in a number of reviews of cancer risk assessment practice and research
needs (Boyce, 1998; Moschandreas and Karuchit, 2002; Zeise et al., 2002), and applied in
various forms for risk assessment of different chemical compounds (Crump, 1994; Humphreys et
al., 2001; Rai et al., 2002; Kirman et al., 2004; Starr et al., 2006; David et al., 2006). The
distributional approach enables consideration of a "portfolio-of-mechanisms" that may
contribute to carcinogenesis (Cox, 2006).
These methods are nontrivial and potentially valuable alternatives to traditional probabilistic uncertainty
analysis, and they are able to provide insights on critical uncertainties in the assessment endpoints and
the ongoing and future research needed to achieve their resolution. The motivation for all of these
approaches is the recognition that the use of a single set of assumptions for the components of a cancer
risk assessment, whether default, conservative, or otherwise, fails to capture the full range of plausible
or likely relationships, how these relationships depend upon our current state of knowledge, the
implications for computed values of potency or unit risk, and the opportunities for improved estimates.
The methods require modeling judgment as any analysis does, but they can provide a basis for ongoing
integration and value of information assessment as new studies and knowledge accumulate over time
(Brusick et al., 2008). These methods can at least provide useful bounds on the plausible risks and on the
value of information (VOI) of reducing uncertainties further (especially, perhaps, on whether the dose-
response relation has a threshold).
There are, of course, many significant benefits to undertaking a quantitative uncertainty analysis.
Although a completely comprehensive analysis might indeed be too much to expect, it is possible and
practical to provide readers with much more useful information about uncertainty. A policy maker might
reasonably expect the Report to provide insight into major uncertainties and questions such as the
following:
• How likely is it that TCDD is not a human carcinogen at current exposure levels? Full
discussion of this uncertainty may help to overcome probability neglect and action bias (Patt and
Zeckhauser, 2000).
• How likely is it that TCDD at current exposure levels has health effects that have not yet been
identified in the toxicological or epidemiological literature (Diamanti-Kandarakis et al., 2009;
Soto and Sonnenschein, 2010)?
• What is the probability that reducing TCDD exposures would not reduce cancer risk at all, or
only by amounts that would not be measurable, based on recent epidemiological studies and
updates such as Pesatori et al. (2009)?
• What is the probability that reducing TCDD exposures would reduce cancer risk in the whole
U.S. population, or targeted subpopulations, by amounts significantly greater than a prediction
derived from the cancer slope factor estimated by EPA?
• What is the probability that reducing TCDD exposures would increase cancer risk (e.g., if the
dose-response relation is J-shaped or U-shaped)?
• What is the decision-analytic value of information (Vol) from collecting more information on
AhR kinetics and dose-response before making risk management decisions?
• What is the probability that TCDD interacts with other compounds to which U.S. or targeted
subpopulations are exposed, increasing cumulative risk for cancer or other health effects
(Carpenter et al., 2002)?
40
-------
Although many members of the public believe that it is imprudent or even morally wrong to delay
tighter regulation of TCDD exposures (perhaps reflecting beliefs that TCDD is a potent carcinogen,
developmental toxin, etc.) many on the Panel expressed the view that EPA should provide a thorough
quantitative decision analysis that makes explicit the current uncertainties and trade-offs and that shows
the conditions under which acting now or postponing action are the optimal actions. Without such
quantitative analysis, risk management decisions for TCDD will not be adequately informed, and
principles other than those of rational decision making (e.g., the biases discussed in Sunstein and
Zeckhauser, 2010) may dominate risk management decisions for TCDD. EPA's uncertainty analysis
should provide the scientific basis for improved decision making. The current decision, in effect, to
"punt" on quantitative uncertainty analysis is not adequate for informing responsible risk management
decision and policy-making, and is not justified.
Recommendations
• The Panel recommends that EPA reconsider the argument for not doing a quantitative
uncertainty analysis, or undertake one. EPA could follow the recommendation of the NAS on
this point by using one or more of the techniques suggested above.
Charge Question 6.2. a. Please comment on the discussion in Section 6 regarding volitional
uncertainty and how this type of uncertainty limits the ability to conduct a quantitative
uncertainty analysis.
Response:
In the Report, EPA contrasts volitional uncertainty with cognitive uncertainty. The Panel recommends
that the term "volitional uncertainty," which might also have been called "decisional uncertainty,"
should be dropped from the Report. EPA should instead display the different modeling choices and the
consequences of making them. The decisions mentioned in the discussion in Section 6 of volitional
uncertainty are modeling choices, and they should be dealt with using techniques for model uncertainty.
Standard tools and techniques for analysis of model uncertainty can be applied.
Recommendations
• The Panel recommends that EPA delete from the Report the notion of "volitional uncertainty."
EPA should display the different modeling choices and the consequences of making them.
Charge Question 6.3. Throughout the document (including the Appendices), EPA presents a
number of limited sensitivity analyses (e.g., toxicokinetic modeling, RfD ranges, cancer OSF
ranges, cancer RfD development). Please comment on the approaches used, and the utility of
these sensitivity analyses in clarifying potential significant uncertainties.
Response:
The Panel congratulates EPA on the sensitivity studies that it has already done and considers them to be
very useful. These studies should be integrated and unified in an overall uncertainty analysis. The Panel
emphasizes that EPA has already done the lion's share of the effort needed in their considerations
41
-------
described in the uncertainty narratives. EPA should take credit for this hard work and extend the
sensitivity studies to respond fully to the NAS criticism.
Recommendations
• The Panel recommends that sensitivity studies that have already been completed be integrated
into whatever additional uncertainty analysis EPA elects to undertake.
42
-------
REFERENCES
Anbalagan, J, A.M. Sashi, G. Vengatesh, J. A. Stanley, R. Neelamohan, and M.M. Aruldhas. 2010.
Mechanism underlying transient gestational-onset hypothyroidism-induced impairment of posttesticular
sperm maturation in adult rats. Fertility and Sterility 93(8): 2491-2497.
Andersen, M. E., and W.F. Greenlee. 1991. Biological determinants of TCDD pharmacokinetics and
their relationship to a biological-based risk assessment. In Biological Basis for Risk Assessment of
Dioxins and Related Compounds, M. Gallo, R. J. Scheuplein, and K. A. Van Der Heijden (eds.), 291-
307, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.
Andersen, M.E., and H.A. Barton. 1999. Biological regulation of receptor-hormone complex
concentrations in relation to dose-response assessments for endocrine-active compounds. Toxicological
Sciences 48(l):38-50.
Andersen, M. E., J.J. Mills, M L. Gargas, L. Kedderis, L.S. Birnbaum, D. Neubert, and W.F. Greenlee
1993. Modeling receptor-mediated processes with dioxin: Implications for pharmacokinetics and risk
assessment. Risk Analysis 13:25-36.
Andersen, M.E., L.S. Birnbaum, H.A. Barton, and C.R. Eklund. 1997. Regional hepatic CYP1A1 and
CYP1A2 induction with 2,3,7,8-tetrachlorodibenzo-p-dioxin evaluated with a multicompartment
geometric model of hepatic zonation. Toxicology and Applied Pharmacology 144(1): 145-155.
Aughenbaugh, J. M., and C.J.J. Paredis. 2007. Probability bounds analysis as a general approach to
sensitivity analysis in decision making under uncertainty. SAE 2007 Transactions Journal of Passenger
Cars: Mechanical Systems, (Section 6) 116:1325-1339, SAE International, Warrendale, Pennsylvania.
[Available online at: http://www.srl.gatech.edu/Members/jaughenbaugh/papers_presentations/2007-01-
1480.pdf]
Axelrad D.A., D C. Bellinger, L.M. Ryan, and T.J. Woodruff. 2007. Dose-response relationship of
prenatal mercury exposure and IQ: an integrative analysis of epidemiologic data. Environmental Health
Perspectives 115:609-615.
Baccarelli, A., P. Mocarelli, D.G. Patterson, Jr., M. Bonzini, A.C. Pesatori, N. Caporaso, and M.T.
Landi. 2002. Immunologic effects of dioxin: new results from Seveso and comparison with other
studies. Environmental Health Perspectives 110:1169-1173.
Baccarelli, A., A.C. Pesatori, S.A. Masten, D.G. Patterson, Jr., L.L. Needham, P. Mocarelli, N.E.
Caporaso, D. Consonni, J. A. Grassman, P. A. Bertazzi, and M.T, Landi. 2004. Aryl-hydrocarbon
receptor-dependent pathway and toxic effects of TCDD in humans: a population-based study in Seveso,
Italy. Toxicology Letters 149:287-292.
Baccarelli, A., S.M. Giacomini, C. Corbetta,, M.T. Landi, M. Bonzini, D. Consonni, P Grillo, D.G.
Pattterson, Jr., A.C. Pesatori, and P. A. Bertazzi. 2008. Neonatal thyroid function in Seveso 25 years
after maternal exposure to dioxin. PLoSMed 5: el61.
43
-------
Bell, D.R., S. Clode, M.Q. Fan, A. Fernandez, P.M. Foster, T. Jiang, G. Loizou, A. MacNicoll, B.G.
Miller, M. Rose, L. and S. White. 2010. Interpretation of studies on the developmental reproductive
toxicology of 2,3,7,8-tetrachlorodibenzo-p-dioxin in male offspring. Food and Chemical Toxicology
48:1439-1447.
Ben-Haim, Y. 2006. Info-Gap Decision Theory: Decisions Under Severe Uncertainty. 2nd edition,
Academic Press, London.
Ben-Tal, A., W. Davidson, D. Bertsimas, and D.B. Brown. 2010. A soft robust model for optimization
under ambiguity. Operations Research 58 (4): 1220-1234. [Available online at:
http://www.mit.edu/~dbertsim/papers/Robust%200ptimization/A%20Soft%20Robust%20Model%20for
%200ptimization%20Under%20Ambiguity.pdf]
Bertsimas, D., and D.B. Brown. 2009. Constructing uncertainty sets for robust linear optimization.
Operations Research 57(6): 1483-1495.
Bertsimas, D., O. Nohadani, and K.M. Teo. 2009. Nonconvex robust optimization for problems with
constraints. INFORMS Journal on Computing. Articles in Advance, 1-15. [Available online at:
http://www.mit.edu/~dbertsim/papers/Robust%200ptimization/Nonconvex%20Robust%200ptimization
%20for%20Problems%20with%20Constraints.pdf]
Bertsimas, D., D.A. Iancu, and P. A. Parrilo. 2010. Optimality of affine policies in multistage robust
optimization. Mathematics of Operations Research 35(2):363-394. [Available online at:
http://www.rnit.edu/~dbertsirn/papers/Robust%200ptimization/Optimality%20of%20Affine%20Policies
%20in%20Multistage%20Robust%200ptimization.pdf
Bhattacharya S., R.B. Conolly, N.E. Kaminski, R.S. Thomas, M.E. Andersen, and Q. Zhang. 2010. A
bistable switch underlying B-cell differentiation and its disruption by the environmental contaminant
2,3,7,8-tetrachlorodibenzo-p-dioxin. ToxicologicalSciences 115(1):51 -65.
Boyce, C.P. 1998. Comparison of approaches for developing distributions for carcinogenic slope factors.
Human and Ecological Risk Assessment 4(2):527-577.
Brusick, D., M.J. Small, E.L. Cavalieri, D. Chakravarti, X. Ding, D.G. Longfellow, J. Nakamura, E.C.
Rogan, and J.S. Swenberg. 2008. Possible genotoxic modes of action for naphthalene. Regulatory
Toxicology and Pharmacology 51(2):43-50.
Calvert, G.M., M.H. Sweney, J. Deddens, and D.K. Wall. 1999. Evaluation of diabetes mellitus, serum
flucose, and thyroid function among United States workers exposed to 2,3,7,8-tetrachlorodibenzo-p-
dioxin. Occupational and Environmental Medicine 56:270-276.
Carpenter, D.O., K. Arcaro, and D.C. Spink. 2002. Understanding the human health effects of chemical
mixtures. Environmental Health Perspectives 110(suppl. l):25-42.
Cheng, H, L. Aylward, C. Beall, T.B. Starr, R.C. Brunet, G. Carrier, and E. Delzell. 2006. TCDD
exposure-response analysis and risk assessment. Risk Analysis 26: 1059-1071.
44
-------
Chevrier, J., B. Eskenazi, A. Bradman, L. Fenster, and D.B. Barr. 2007. Associations between prenatal
exposure to polychlorinated biphenyls and neonatal thyroid-stimulating hormone levels in a Mexican-
American population, Salinas Valley, California. Environmental Health Perspectives 115(10): 1490-
1496.
Choi, T., M.J. Schervish, K.A. Schmitt, and M.J. Small. 2010. Bayesian hierarchical analysis for
multiple health endpoints in a toxicity study. Journal of Agricultural, Biological, and Environmental
Statistics 15(3):290-307 [Available online at:
http ://www. springerlink. com/content/2h416p25 81210773/fulltext.pdf]
Coull, B., M. Menzetti, and L. Ryan. 2003. A Bayesian hierarchical model for risk assessment of
methylmercury. Journal of Agricultural, Biological and Environmental Statistics 8(3):253-270.
Cox, L.A. 2006. Quantifying potential health impacts of cadmium in cigarettes on smoker risk of lung
cancer: A portfolio-of-mechanisms approach. Risk Analysis 26(6): 1581-1599.
Crump, K.S. 1994. Risk of benzene-induced leukemia: A sensitivity analysis of the Pliofilm cohort with
additional follow-up and new exposure estimates. Journal of Toxicology and Environmental Health
42:219-242.
David, R.M., H.J. Clewell, P.R. Gentry, T.R. Covington, D.A. Morgott, and D.J. Marino. 2006. Revised
assessment of cancer risk to dichloromethane II. Application of probabilistic methods to cancer risk
determinations. Regulatory Toxicology and Pharmacology 45(1):55—65.
Davidovitch, L., R. Stoklosa, J. Majer, A. Nietrzeba, P. Whittle, K. Mengersen, and Y. Ben-Haim. 2009.
Info-gap theory and robust design of surveillance for invasive species: The case study of Barrow Island.
Journal of Environmental Management 90(8): 278 5-2793.
Diamante-Kandarakis, E., J-P. Bourguignon, L.C. Giudice, R. Hauser, G.S. Prins, A.M. Soto, R.T.
Zoeller, and A.C. Gore. 2009. Endocrine disrupting chemicals: an Endocrine Society scientific
statement. Endocrine Reviews 30:293-342.
Dimitropoulos, A., L. Molinari, K. Etter, M. Lang-Muritano, O.G. Jenni, R.H. Largo, and B. Latal.
2009. Children with congenital hypothyroidism: long-term intellectual outcome after early high-dose
treatment. Pediatric Research 65(2):242-248.
Dixon, W.J. 2007. The Use of Probability Bounds Analysis for Characterizing and Propagating
Uncertainty in Species Sensitivity Distributions. Arthur Rylah Institute for Environmental Research,
Technical Report Series No. 163, Department of Sustainability and Environment. Victoria, Australia
[Available online at:
http://www.land.vic.gov.au/CA256F310024B628/0/F11A9B711262E4C3CA2573840014660E/$File/A
RI+Technical+Report+163+-
+The+use+of+probability+bounds+analysis+for+characterising+and+propagating+uncertainty+in+speci
es+sensitivity+distributions.pdf]
Emond, C., L.S. Birnbaum, and M.J. DeVito 2004. Physiologically based pharmacokinetic model for
developmental exposures to TCDD in the rat. Toxicological Sciences 80:115-133.
45
-------
Emond, C., J.E. Michalek, and L.S. Birnbaum. 2005. Comparison of the use of a physiologically based
pharmacokinetic model and a classical pharmacokinetic model for dioxin exposure assessments.
Environmental Health Perspectives 113:1666-1668.
Emond, C., L.S. Birnbaum, and M.J. DeVito. 2006. Use of a physiologically based pharmacokinetic
model for rats to study the influence of body fat mass and induction of CYP1A2 on the
pharmacokinetics of TCDD. Environmental Health Perspectives 114:1394-1400.
EPA (U.S. Environmental Protection Agency) 2002-2005. GE/Housatonic River Site in New England.
Several documents available online at http://www.epa.gov/regionl/ge/thesite/restofriver-reports.html
describing the U.S. EPA's use of probability bounds analysis in the "rest of river" component of the
assessment for the GE/Housatonic River Site in New England, including both the Ecological Risk
Assessment and the Human Health Risk Assessment.
EPA (U.S. Environmental Protection Agency). 2005. Guidelines for Carcinogen Risk Assessment.
EPA/630/P-03/001F. U.S. Environmental Protection Agency, Washington, DC.
EPA (U.S. Environmental Protection Agency). 2006. A Framework for Assessing Health Risks of
Environmental Exposures to Children. EPA/600/R-05/093F, National Center for Environmental
Assessment, Office of Research and Development, Washington, DC.
EPA (U. S. Environmental Protection Agency). 2007. Region 6 SuperfundProgram: Calcasieu Estuary
Initiative, http://www.epa.gov/earthlr6/6sf/sfsites/calcinit.htm, [accessed July 2009]
EPA (U.S. Environmental Protection Agency). 2010. EPA's Reanalysis of Key Issues Related to Dioxin
Toxicity and Response to NAS Comments (May 2010 draft). EPA/600/R-10/038A.
EPA SAB (U.S. Environmental Protection Agency Science Advisory Board). 1995. A Second Look at
Dioxin. EPA-SAB-EC-95-021. U.S. EPA Science Advisory Board, Washington, DC.
EPA SAB (U.S. Environmental Protection Agency Science Advisory Board). 2001. Dioxin
Reassessment - An SAB Review of the Office of Research and Development's Reassessment of Dioxin.
EPA-SAB-EC-01-006. U.S. EPA Science Advisory Board, Washington, DC.
Evans, M.V. and Andersen, M.E. 2000. Sensitivity analysis of a physiological model for 2,3,7,8,-
tetrachlorodibenzo-p-dioxin (TCDD): assessing the impact of specific model parameters on
sequestration in liver and fat in the rat. Toxicological Sciences 54:71-80.
Evans, J.S., J.D. Graham, G.M. Gray, and R.L. Sielken Jr. 1994a. A distributional approach to
characterizing low-dose cancer risk. Risk Analysis 14(l):25-34.
Evans, J.S., G.M. Gray, R.L. Sielken Jr., A.E. Smith, C. Valdez-Flores, and J.D. Graham. 1994b. Use of
probabilistic expert judgment in distributional analysis of carcinogenic potency. Regulatory Toxicology
and Pharmacology 20(1): 15-36.
46
-------
Evans, J.S., J.D. Graham, G.M. Gray, and R.L. Sielken, Jr. 1995. A distributional approach to
characterizing low-dose cancer risk. In Low-Dose Extrapolation of Cancer Risks, S. Olin, W. Farland,
C. Park, L. Rhomberg, R. Scheuplein, T. Starr, J. Wilson (eds.), 253-274, ILSI Press Washington, DC.
Ferson, S., and T.F. Long. 1995. Conservative uncertainty propagation in environmental risk
assessments. Environmental Toxicology and Risk Assessment, Third Volume, ASTM STP 1218, J.S.
Hughes, G.R. Biddinger and E. Mones (eds.), 97-110, American Society for Testing and Materials,
Philadelphia, .
Ferson, S., V. Kreinovich, L. Ginzburg, D.S. Myers, and K. Sentz. 2003. Constructing Probability
Boxes andDempster-Shafer Structures. SAND2002-4015. Sandia National Laboratories, Albuquerque,
NM.
Frey, H.C., and S.R. Patil. 2002. Identification and review of sensitivity analysis methods. Risk Analysis
22(3):553-578.
Gillies, D. 2000. Philosophical Theories of Probability. Routledge, New York, NY.
Gim, J, H.S. Kim, J. Kim, M. Choi, J.R. Kim, Y.J. Chung, and K.H. Cho. 2010. A system-level
investigation into the cellular toxic response mechanism mediated by AhR signal transduction pathway.
Bioinformatics 26(17):2169-2175.
Goodman, J.E., L.E., Kerper, C.P. Boyce, R.L. Prueitt, and L.R. Rhomberg. 2010. Weight-of-evidence
analysis of human exposures to dioxins and dioxin-like compounds and associations with thyroid
hormone levels during early development. Regulatory Toxicology and Pharmacology 58(l):79-99.
Hall, J.W., and H. Harvey 2009. Decision making under severe uncertainty for flood risk management: a
case study of info-gap robustness analysis. Proceedings of the Eighth International Conference on
Hydroinformatics, January 12-16, 2009, Concepcion, Chile. [Available online at:
http://www.floodsite.net/html/partner_area/project_docs/T20-08-05-hic-infogap.pdf]
Holland, C.D., and R.L Sielken, Jr. 1993. Quantitative Cancer Modeling and Risk Assessment (Chapter
7). Prentice Hall, Englewood Cliffs, NJ.
Humphreys, S.H., C. Carrington, and M. Bolger. 2001. A quantitative risk assessment for fumonisins B1
and B2 in U.S. corn. Food Additives and Contaminants 18(3):211-220.
Iida, T., H. Hirakawa, T. Matsueda, J. Magayama, and T. Nagata. 1999. Polychlorinated dibenzo-p-
dioxins and related compounds: correlations of levels in human tissues and in blood. Chemosphere
38:2767-2774.
Karanki, D.R., H.S. Kushwaha, A.K. Verma, and S. Ajit. 2009. Uncertainty analysis based on
probability bounds (p-box) approach in probabilistic safety assessment. Risk Analysis 29(5):662-75.
Kirman, C. R., L.M. Sweeney, M.J. Teta, R.L. Sielken, C. Valdez-Flores, R.J. Albertini, and M.L.
Gargas. 2004. Addressing nonlinearity in the exposure-response relationship for a genotoxic carcinogen:
Cancer potency estimates for ethylene oxide. Risk Analysis 24(5): 1165-1183.
47
-------
Li, X., and K.K. Rozman. 1995. Subchronic effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and
their reversibility in male Sprague-Dawley rats. Toxicology 97:133-140.
Louis M., and A. Becskei. 2002. Binary and Graded Responses in Gene Networks. Sciences Signal
Transduction Knowledge Environment (143):pe33.
Manz, A., J. Berger, J.H. Dwyer, D. Flesch-Janys, S. Nagel, and H. Waltsgott. 1991. Cancer mortality
among workers in chemical plant contaminated with dioxin. Lancet 338: 959-964.
Maruyama, W., K. Yoshida, T. Tanaka, and J. Nakanishi. 2002. Determination of tissue-blood partition
coefficients for a physiological model for humans, and estimation of dioxin concentration in tissues.
Chemosphere 46:975-985.
Minnery, J.G., J.G. Jacangelo, L.I. Boden, D.J. Vorhees and W. Heiger-Bernays. 2009. Sensitivity
analysis of the pressure-based direct integrity test for membranes used in drinking water treatment.
Environmental Science and Technology 43 (24): 9419-9424.
Mocarelli, P., P.M. Gerthoux, D.G. Patterson, Jr., S. Milani, G. Limonta, M. Bertona, S. Signorini, P.
Tramacere, L. Colombo, C, Crespi, P. Brambilla, C. Sarto, V. Carreri, E. Sampson, W.E. Turner, and L.
Needham. 2008. Dioxin exposure, from infancy through puberty, produces endocrine disruption and
affects human semen quality. Environmental Health Perspectives 1 16:70-77.
Montgomery, V. 2009. New Statistical Methods in Risk Assessment by Probability Bounds. Ph.D.
dissertation, Durham University, UK. [Available online at: http://maths.dur.ac.uk/stats/people/fc/thesis-
VM.pdf ]
Moore, R.E. 1966. Interval Analysis. Prentice Hall, Englewood Cliffs, NJ.
Morales, K.H., J.G. Ibrahim, C.-J. Chen, and L.M. Ryan. 2006. Bayesian model averaging with
applications to benchmark dose estimation for arsenic in drinking water. Journal of the American
Statistical Association 101(473):9-17.
Moschandreas, D.J., and S. Karuchit, 2002. Scenario-model-parameter - A new method of cumulative
risk uncertainty analysis. Environment International 28 (4):247-261.
NRC (National Research Council). 2006. Health Risks from Dioxin and Related Compounds: Evaluation
of the EPA Reassessment. National Academies Press, Washington, D.C. [Available online at:
http ://www. nap. edu/catal og. php?record_id= 11688]
Neumaier, A. 1990. Interval Methods for Systems of Equations. Cambridge University Press,
Cambridge, UK.
Oerbeck, B., K. Sundet, B.F. Kase, and S. Heyerdahl. 2003. Congenital hypothyroidism: influence of
disease severity and L-thyroxine treatment on intellectual, motor, and school-associated outcomes in
young adults. Pediatrics 112:923-930.
48
-------
Oerbeck, B., I. Reinvang, K. Sundet, and S. Heyerdahl. 2007. Young adults with severe congenital
hypothyroidism: cognitive event related potentials (ERPs) and the significance of an early start of
thyroxine treatment. Scandinavian Journal of Psychology 48:61-67.
Paterson, S, and D. Mackay. 1987. A steady-state fugacity-based pharmacokinetic model with
simultaneous multiple exposure routes. Chemosphere 6:395-408.
Patt, A., and R. Zeckhauser. 2000. Action bias and environmental decisions. Journal of Risk and
Uncertainty 21(l):45-72. [Available online at:
http://www.springerlink.com/content/k47064873365w720/]
Pesatori, A.C., D. Consonni, M. Rubagotti, P. Grillo, and P. A. Bertazzi. 2009. Cancer incidence in the
population exposed to dioxin after the "Seveso accident": twenty years of follow-up. Environmental
Health 8:39. [Available online at:
http://www.ncbi. nlm.nih.gov/pmc/articles/PMC2754980/?tool=pmcentrez&report=abstract]
Rai, S.N., S. Bartlett, D. Krewski, and J. Paterson. 2002. The use of probabilistic risk assessment in
establishing drinking water quality objectives. Human and Ecological Risk Assessment 8(3):493-509.
Regan, H.M., B.K. Hope, and S. Ferson. 2002a. Analysis and portrayal of uncertainty in a food-web
exposure model. Human and Ecological Risk Assessment 8:1757-1777.
Regan, H.M., B.E. Sample, and S. Ferson. 2002b. Comparison of deterministic and probabilistic
calculation of ecological soil screening levels. Environmental Toxicology and Chemistry 21:882-890.
Regan, H.M., Y. Ben-Haim, B. Langford, W.G. Wilson, P. Lundberg, S.J. Andelman, and M.A.
Burgman. 2005. Robust decision-making under severe uncertainty for conservation management.
Ecological Applications 15(4):1471-1477.
Rout, T.M., C.J. Thompson, and M.A. McCarthy. 2009. Robust decisions for declaring eradication of
invasive species. Journal of Applied Ecology 46:782-786.
Rozman, K.K., B.U. Stahl, M. Viluksela, and L.S. Birnbaum. 1993. Multiple dose (subchronic) toxicity
of heptachlorodibenzo-p-dioxin. In Dioxin '93, H. Fiedler, H. Frank, O. Hutzinger, W. Parzefall, A. Riss
and S. Safe (eds.), 13:133-136.
Rozman, K.K., M. Lebofsky, and D.M. Pinson. 2005. Chronic toxicity and carcinogenicity of
1,2,3,4,6,7,8-heptachlorodibenzo-p-dioxin displays a distinct dose/time threshold (c x t = k) and a
profound subthreshold hormetic effect: prolongation of life. Food and Chemical Toxicology 43:729-40.
Ryan, L. 2008. Combining data from multiple sources, with applications to environmental risk
assessment. Statistics inMedicine 27(5):698-710.
Saltelli, A., and S. Campolongo. 2000. Sensitivity analysis as an ingredient of modeling. Statistical
Science 15(4):377-395.
Saltelli, A., K. Chan, and M. Scott (eds.). 2000. Sensitivity Analysis. Wiley and Sons, New York, NY.
49
-------
Santostefano, M.J., X. Wang, V.M. Richardson, D.G. Ross, M.J. DeVito, and L.S. Birnbaum . 1998. A
pharmacodynamic analysis of TCDD-induced cytochrome P450 gene expression in multiple tissues:
dose- and time-dependent effects. Toxicology and Applied Pharmacology 151:294-310.
Sand, S., N. Fletcher, D. von Rosen, F. Kalantari, M. Viluksela, J.T. Tuomisto, J. Tuomisto, A. Falk-
Filipsson, and H. Hakansson. 2010. Quantitative and statistical analysis of differences in sensitivity
between Long-Evans and Han/Wistar rats following long-term exposure to 2,3,7,8-tetrachlorodibenzo-p-
dioxin. Regulatory Toxicology and Pharmacology 57(2-3): 136-145.
Schecter, A.J., and J.J. Ryan. 1989. Blood and adipose tissue levels of PCDDs/PCDFs over three years
in a patient after exposure to polychlorinated dioxins and dibenzofurans. Chemosphere 18(l-6):635-642.
Schecter, A., and T. Gasiewicz (eds)., 2003. Dioxins and Health, 2nd Edition. Wiley Interscience,
Hoboken, NJ.
Schecter, A.J., J. Mes, and D. Davies. 1989. Polychlorinated biphenyl (PCB), DDT, DDE and
hexachlorobenzene (HCB) and PCDD/F isomer levels in various organs in autopsy tissue from North
American patients. Chemosphere 18(1 -6):811-818.
Sielken, R.L., Jr. 1990. A weight-of-evidence approach to quantitative cancer risk assessment:
Information analysis. In Risk Assessment in Chemical Carcinogenesis. G. Schettler, D. Schmahl, and T.
Klenner (eds.), Springer-Verlag, New York. Proceedings of the Satellite Symposium on Risk
Assessment in Chemical Carcinogenesis, Heidelberg, Germany August 24-25, 1990.
Sielken, R.L., Jr., and C. Valdez-Flores. 1996. Comprehensive realism's weight-of-evidence based
distributional dose-response characterization. Special Issue of the Human and Ecological Risk
Assessment on Theoretical, Toxicological and Biostatistical Foundations for Deriving Probability
Distribution Functions for Reference Doses and Benchmark Doses with Application to Carcinogens and
Noncarcinogens 2(1): 175-193.
Sielken, R. L., Jr., and C. Valdez-Flores. 1999. Probabilistic risk assessment's use of trees and
distributions to reflect uncertainty and variability and to overcome the limitations of default
assumptions. Special Issue of Environmental International on Modeling and Simulation 25:755-772.
Sielken, R.L., Jr., R.S. Bretzlaff, and D.E. Stevenson. 1995. Challenges to default assumptions stimulate
comprehensive realism as a new tier in quantitative cancer risk assessment. Regulatory Toxicology and
Pharmacology 21:270-280.
Simon, T., L.L. Aykward, C.R. Kirman, J.C. Rowlands, and R.A. Budinsky. 2009. Estimates of cancer
potency of 2,3,7,8-tetrachlorodibenzo(p)dioxin using linear and nonlinear dose-response modeling and
toxicokinetics. Toxicological Sciences 112(2):490-506.
Skakkeback, N.E. 2010. Normal reference ranges for semen quality and their relations to fecundity.
Asian Journal of Andrology 12(l):95-98.
50
-------
Soto, A.M. and C. Sonnenschein. 2010. Environmental causes of cancer: endocrine disrupters as
carcinogens. Nature Reviews Endocrinology 6:364-371.
Stahl, B.U., M. Viluksela, J. Deliberto, L.S. Birnbaum, andK.K. Rozman. 1994. Subchronic (13 week)
toxicity of a mixture of four chlorinated dibenzo-p-dioxins in Sprague-Dawley rats. In Dioxin '94 (H.
Fiedler, O. Hutzinger, L. Birnbaum, G. Lambert, L. Needham and S. Safe, (eds.), 21:341-345.
Starr, T.B., G. Matanoski, M.W. Anders, and M.E. Andersen. 2006. Workshop overview: Reassessment
of the cancer risk of dichloromethane in humans. Toxicological Sciences 91(1): 20-28.
Steenland, K., G. Calvert, N. Ketchum, and J. Michalek. 2001. Dioxin and diabetes mellitus: an analysis
of the combined NIOSH and Ranch Hand data. Occupational and Environmental Medicine 58:641-648.
Sunstein, C.R., andR. Zeckhauser. 2010. Dreadful possibilities, neglected probabilities. The Irrational
Economist: Making Decisions in a Dangerous World, E. Michel-Keijan and P. Slovic (eds.), Public
Affairs Books, NY: New York. [Available online at: http://www.hks.harvard.edu/fs/rzeckhau/Sunstein4-
6-09.pdf]
Trejo, R., and V. Kreinovich. 2001. Error estimations for indirect measurements: randomized vs.
deterministic algorithms for 'black-box' programs. In Handbook of Randomized Computing, S.
Rajasekaran, P. Pardalos, J. Reif, and J. Rolim, (eds.), 673-729, Springer, New York, NY
van Fraassen B. 1966. Singular terms, truth value gaps, and free logic. Journal of Philosophy 63(17):48-
495.
van Fraassen, B., 1980, The Scientific Image. Oxford: Oxford University Press.
Van den Heuvel, J.P., G.C. Clark, M.C. Kohn, A.M. Tritscher, W.F. Greenlee, G.W. Lucier, and D.A.
Bell. 1994. Dioxin-responsive genes: examination of dose-response relationships using quantitative
reverse transcriptase-polymerase chain reaction. Cancer Research 54(l):62-68.
Viallefont, V., A.E. Raftery, and S. Richardson. 2001. Variable selection and Bayesian model averaging
in case-control studies. Statistics inMedicine 20:3215-3230.
Viluksela, M., B.U. Stahl, and K.K. Rozman. 1994. Subchronic (13-week) toxicity of
heptachlorodibenzo-p-dioxin in male Sprague-Dawley rats. Chemosphere 29:2381-2393.
Viluksela, M., B.U. Stahl, L.S. Birnbaum, K.W. Schramm, A. Kettrup, and K.K. Rozman. 1997a.
Subchronic/chronic toxicity of heptachlorodibenzo-p-dioxin (HpCDD) in rats. Part 1. Design, general
observations, hematology and liver concentrations. Toxicology and Applied Pharmacology 46:207-216.
Viluksela, M., B.U. Stahl, L. Birnbaum, and K.K. Rozman. 1997b. Subchronic/chronic toxicity of
heptachlorodibenzo-p-dioxin (HpCDD) in rats. Part 2. Biochemical effects. Toxicology and Applied
Pharmacology 46:217-226.
Viluksela, M., B.U. Stahl, L. Birnbaum, K.W. Schramm, A. Kettrup, and K.K. Rozman. 1998a.
Subchronic/chronic toxicity of a mixture of four chlorinated dibenzo-p-dioxins (CDDs) in rats. Part I.
51
-------
Design, general observations, hematology and liver concentrations. Toxicology and Applied
Pharmacology 151:57-69.
Viluksela, M., BU.. Stahl, L. Birnbaum, and K.K. Rozman. 1998b. Subchronic/chronic toxicity of a
mixture of four chlorinated dibenzo-p-dioxin (CDDs) in rats. Part II. Biochemical effects. Toxicology
Applied Pharmacology 151:70-78.
Walker, N.J., C.J. Portier, and S.F. Lax 1999. Characterization of the dose-response of CYP1B1,
CYP1 Al, and CYP1A2 in the liver of female Sprague-Dawley rats following chronic exposure to
2,3,7,8-tetrachl orodi benzo-/>di ox i n. Toxicology and Applied Pharmacology 154:279-286.
Wang, X., M.J. Santostefano, M.V. Evans, V.M. Richardson, J.J. Diliberto, and L.S. Birnbaum. 1997.
Determination of parameters responsible for pharmacokinetic behavior of TCDD in female Sprague-
Dawley rats. Toxicology and Applied Pharmacology 147:151-168.
Wheeler, M.W., and A.J. Bailer. 2007. Properties of model-averaged BMDLs: a study of model
averaging in dichotomous risk estimation. Risk Analysis 27:659-670.
Wheeler, M.W., and A.J. Bailer. 2009. Comparing model averaging with other model selection
strategies for benchmark dose estimation. Environmental and Ecological Statistics 16(1):37-51.
White, R.H., I. Cote, L. Zeise, M. Fox, F. Dominici, T.A. Burke, P.D. White, D. Hattis, and J.M. Samet.
2009. State-of-the-science workshop report: issues and approaches in low dose-response extrapolation
for environmental health risk assessment. Environmental Health Perspectives 117:283-287.
Yokomizo, H., H.P. Possingham, M.B. Thomas, and Y.M. Buckley. 2009. Managing the impact of
invasive species: the value of knowing the density-impact curve. Ecological Applications 19(2):376-
386.
Yoshizawa, K., A.E. Brix, D.M. Sells, M.P. Jokinen, M. Wyde, D.P. Orzech, G.E. Kissling, N.J.
Walker, and A. Nyska. 2009. Reproductive lesions in female Harlan Sprague-Dawley rats following
two-year oral treatment with dioxin and dioxin-like compounds. Toxicologic Pathology 37(7):921-937.
Yoshizawa, K., N.J. Walker, A. Nyska, G.E. Kissling, M.P. Jokinen, A.E. Brix, D.M. Sells, and M.E.
Wyde. 2010. Thyroid follicular lesions induced by oral treatment for 2 years with 2,3,7,8-
tetrachlorodibenzo-p-dioxin and dioxin-like compounds in female Harlan Sprague-Dawley rats.
Toxicologic Pathology 38(7): 1037-1050.
Yr, J. 2008. Neuro-developmental deficits in early-treated congenital hypothyroidism. Annals of the
Academy of Medicine, Singapore 37,Supp. 12:42-43.
Zeise, L., D. Hattis, M. Andersen, A.J. Bailer, S. Bayard, C. Chen, H. Clewell, R. Conolly, K. Crump,
D. Dunson, A. Finkel, L. Haber, A.M. Jarabek, R. Kodell, D. Krewski, D. Thomas, T. Thorslund, and J.
Wassell. 2002. Improving risk assessment: Research opportunities in dose response modeling to
improve risk assessment. Human and Ecological Risk Assessment 8(6): 1421-1444.
52
-------
Zhang Q., S. Bhattacharya, M.E. Andersen, and R.B. Conolly. 2010. Computational systems biology
and dose-response modeling in relation to new directions in toxicity testing. Journal of Toxicology and
Environmental Health B 13:253-276.
53
-------
APPENDIX A: EPA'S CHARGE QUESTIONS
UNITED states environmental protection agency
NATIONAL CENTER FOR ENVIRONMENTAL ASSESSMENT
WASHINGTON, DC 20460
May 27,2010
OFFICE OF
RESEARCH AND DEVELOPMENT
Memorandum
SUBJECT; Request for Science Advisory Board Review of the Draft Report, "EPA's
Reanalysis of Key Issues Related to Dioxin Toxicity and Response to MAS
Comments"
FROM: Becki Clark, Deputy Director
National Center for Environmental Assessment {860IP)
Office of Research and Development
TO: Vanessa T, Vu, Ph.D., Director
EPA Science Advisory Board (1400F)
This is to request a review by the Science Advisory Board of the draft report entitled "EPA's
Reanalysis of Key Issues Related to Dioxin Toxicity and Response to MAS Comments"
(EPA/600/R-10/038A). This draft report details the Environmental Protection Agency's (EPA)
response to key comments and recommendations included in the 2006 NAS report ("Health
Risks from Dioxin and Related Compounds: Evaluation of the EPA Reassessment") on their
review of the (EPA) 2003 draft dioxin reassessment. This draft report also includes significant
new analyses on both the potential cancer and noncancer human health effects that may result
from chronic exposures to dioxins.
Attached is the Charge that provides background information as well as questions that are to be
the focus of the Science Advisory Board review of this draft report.
Please let me know if you have any questions. Thank you.
Attachment: Charge for EPA's Science Advisory Board - Review of the Draft Report, "EPA's
Reanalysis of Key Issues Related to Dioxin Toxicity and Response to NAS Comments"
cc; Peter W. Prcuss
Annette Gatchett
Glenn Rice
Cheryl Itkin
A-l
-------
Charge to the Science Advisory Board for Peer Review of Draft Report
"EPA's Reanalysis of Key Issues Related to Dioxin Toxicity
and Response to NAS Comments"
May, 2010
EPA has been preparing an assessment of the potential health impacts of 2,3,7,8-Tetrachlorodibenzo-p-
Dioxin (TCDD) for many years. In 2003, EPA released an external review draft report entitled,
Exposure and Human Health Reassessment of 2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) and
Related Compounds (U.S. EPA, 2003) (herein referred to as "2003 Reassessment") that was reviewed
by the EPA Science Advisory Board (SAB), and then by the National Academy of Sciences (NAS). In
2006, the National Research Council (NRC) of the National Academies published their report of EPA's
reassessment, Health Risks from Dioxin and Related Compounds: Evaluation of the EPA Reassessment
(NRC, 2006).
The current Report EPA 's Reanalysis of Key Issues Related to Dioxin Toxicity and Response to NAS
Comments {"Response to Comments") before the SAB is a response to the review by the NRC, and
includes new analyses completed in response to the NRC recommendations and recently published
literature, as well as a discussion of topics where our views differed. The draft Response to Comments
document is not an assessment per se; it is designed to supplement the information provided in the 2003
Reassessment. However, the draft Response to Comments provides a noncancer reference dose and
updated cancer values. Detailed discussions of many of the issues addressed in the draft Response to
Comments are available in the 2003 Reassessment and have not been reproduced in the current Report -
whenever appropriate; the reader is directed to the pertinent chapters of the 2003 Reassessment.
The NRC identified three key recommendations that they believed would result in substantial
improvement to the EPA 2003 Reassessment and thus support a scientifically robust characterization of
human responses to exposures to TCDD. These three key areas are (1) improved transparency and
clarity in the selection of key data sets for dose-response analysis, (2) further justification of approaches
to dose-response modeling for cancer and noncancer endpoints, and (3) improved transparency,
thoroughness, and clarity in quantitative uncertainty analysis. The NRC Report also encouraged EPA to
calculate a reference dose (RfD), which had not been derived in the 2003 Reassessment. The draft
Response to Comments document addresses each of these issues. Please consider the accuracy,
objectivity, and transparency of EPA's reanalysis and responses in your review.
General Charge Questions
1.1 Is the draft Response to Comments clear and logical? Has EPA objectively and clearly presented
the three key NRC recommendations?
1.2 Are there other critical studies that would make a significant impact on the conclusions of the
hazard characterization or dose-response assessment of the chronic noncancer and cancer health
effects of TCDD?
Specific Charge Questions
Section 2. Transparency and Clarity in the Selection of Key Data Sets for Dose-Response
Analysis
A-2
-------
2.1. Is this Section responsive to the NAS concern about transparency and clarity in data-set selection
for dose-response analysis?
2.2. Are the epidemiology and animal bioassay study criteria/considerations scientifically justified and
clearly described?
2.3. Has EPA applied the epidemiology and animal bioassay study criteria/considerations in a
scientifically sound manner? If not, please identify and provide a rationale for alternative
approaches.
Section 3. The Use of Toxicokinetics in Dose-Response Modeling for Cancer and Noncancer
Endpoints
3.1 The 2003 Reassessment utilized first-order body burden as the dose metric. In the draft Response to
Comments document, EPA used a physiologically-based pharmacokinetic (PBPK) model (Emond
et al., 2004, 2005, 2006) with whole blood concentration as the dose metric rather than first-order
body burden. This PBPK model was chosen, in part, because it includes a biological description of
the dose-dependent elimination rate of TCDD. EPA made specific modifications to the published
model based on more recent data. Although lipid-adjusted serum concentrations (LASC) for
TCDD are commonly used as a dose metric in the literature, EPA chose whole blood TCDD
concentrations as the relevant dose metric because serum and serum lipid are not true
compartments in the Emond PBPK models (LASC is a side calculation proportional to blood
concentration).
Please comment on:
3.1.a. The justification of applying a PBPK model with whole blood TCDD concentration as a
surrogate for tissue TCDD exposure in lieu of using first-order body burden for the dose-
response assessment of TCDD.
3.1.b. The scientific justification for using the Emond et al. model as opposed to other available
TCDD kinetic models.
3.1.e. The modifications implemented by EPA to the published Emond et al. model.
3.1 .d. Whether EPA adequately characterized the uncertainty in the kinetic models.
3.2. Several of the critical studies for both noncancer and cancer dose-response assessment were
conducted in mice. A mouse PBPK model was developed from an existing rat model in order to
estimate TCDD concentrations in mouse tissues, including whole blood.
Please comment on:
3.2.a. The scientific rationale for the development of EPA's mouse model based on the published
rat model (Emond et al., 2004, 2005, 2006).
3.2.b. The performance of the mouse model in reference to the available data.
A-3
-------
3.2.c. Whether EPA adequately characterized the uncertainty in the mouse and rat kinetic
models. Please comment specifically on the scientific justification of the kinetic
extrapolation factor from rodents to humans.
3.3 Please comment on the use of the Emond et al. PBPK model to estimate human intakes based on
internal exposure measures.
3.4. Please comment on the sensitivity analysis of the kinetic modeling (see Section 3.3.5).
3.5. Both EPA's noncancer and cancer dose-response assessments are based on a lifetime average daily
dose. Did EPA appropriately estimate lifetime average daily dose? If not, please suggest
alternative approaches that could be readily developed based on existing data.
Section 4. Reference Dose
4.1. The Mocarelli et al. (2008) and Baccarelli et al. (2008) studies were selected as co-critical studies
for the derivation of the RfD. Is the rationale for this selection scientifically justified and clearly
described? Please identify and provide the rationale for any other studies that should be selected,
including the rationale for why the study would be considered a superior candidate for the
derivation of the RfD. In addition, male reproductive effects and changes in neonatal thyroid
hormone levels, respectively, were selected as the co-critical effects for the RfD. Please comment
on whether the selection of these critical effects is scientifically justified and clearly described.
Please identify and provide the rationale for any other endpoints that should be selected as the
critical effect.
4.2. In the Seveso cohort, the pattern of exposure to TCDD is different from the average daily exposure
experienced by the general population. The explosion in Seveso created a high dose pulse of
TCDD followed by low level background dietary exposure in the exposed population. In the
population, this high dose pulse of TCDD was slowly eliminated from body tissues over time.
There is uncertainty regarding the influence of the high-dose pulse exposure on the effects
observed later in life.
4.2.a. Mocarelli et al. (2008), reported male reproductive effects observed later in life for boys
exposed to the high dose pulse of TCDD between the ages of 1 and 10. EPA identified a
10 year critical exposure window. In the development of the candidate RfD, EPA used an
exposure averaging approach that differs from the typical approach utilized for animal
bioassays. EPA determined that the relevant exposure should be calculated as the mean of
the pulse exposure and the 10-year critical exposure window average. Please comment on
the following:
4.2.a.i. EPA's approach for identifying the exposure window and calculating average
exposure for this study.
4.2.a.ii. EPA's designation of a 20% decrease in sperm count (and an 11% decrease in
sperm motility) as a LOAEL for Mocarelli et al. (2008).
4.2.b. For Baccarelli et al. (2008), the critical exposure window occurs long after the high-dose
A-4
-------
pulse exposure. Therefore, the variability in the exposure over the critical exposure
window is likely to be less than the variability in the Mocarelli et al. subjects. EPA
concluded that the reported maternal exposures from the regression model developed by
Baccarelli et al. provide an appropriate estimate of the relevant effective dose as
opposed to extrapolating from the measured infant TCDD concentrations to maternal
exposure. Additionally, EPA selected a LOAEL of 5 |i-units TSH per ml blood in
neonates; as this was established by World Health Organization (WHO) as a level above
which there was concern about abnormal thyroid development later in life. Please
comment on the following:
4.2.b.i. EPA's decision to use the reported maternal levels and the appropriateness of this
exposure estimate for the Baccarelli et al. study.
4.2.b.ii. EPA's designation of 5 |i-units TSH per ml blood as a LOAEL for Baccarelli et
al. (2008).
4.3. Please comment on the rationale for the selection of the uncertainty factors (UFs) for the RfD. If
changes to the selected UFs are proposed, please identify and provide a rationale.
4.4. EPA did not consider biochemical endpoints (such as CYP induction, oxidative stress, etc.) as
potential critical effects for derivation of the RfD for TCDD due to the uncertainties in the
qualitative determination of adversity associated with such endpoints and quantitative
determination of appropriate response levels for these types of endpoints in relation to TCDD
exposure. Please comment on whether this decision is scientifically justified and clearly described.
4.5. In using the animal bioassays, EPA averaged internal blood TCDD concentrations over the entire
dosing period, including 24 hours following the last exposure. Please comment on EPA's
approach for averaging exposures including intermittent and one day gestation exposure
protocols.
4.6. Please comment on the benchmark dose (BMD) modeling conducted by EPA to analyze the animal
bioassay data and EPA's choice of points of departure (PODs) from these studies.
4.7. For the animal bioassay modeling, EPA applied the kinetic extrapolation at the level of the POD
prior to applying the uncertainty factors because EPA has less confidence in the kinetic model
output at lower doses reflective of the RfD. Please comment on whether this approach was
scientifically justified and clearly described.
4.8. Please comment as to whether EPA's qualitative discussion of uncertainty in the RfD is justified
and clearly described.
Section 5. Cancer Assessment
5.1. Weight of Evidence Cancer Descriptor: The 2003 Reassessment concluded that TCDD is a
"known human carcinogen." In the current draft Response to Comments document, EPA
concluded that under the 2005 Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005)
TCDD is "carcinogenic to humans." Is the weight-of-evidence characterization scientifically
justified and clearly described?
A-5
-------
5.2. Mode of Action: The mode of action of a carcinogen can inform identification of hazards and
approaches used for a dose-response assessment. The mode of carcinogenic action for TCDD has
not been elucidated for any tumor type. EPA concluded that, while interaction with the Ah receptor
is likely to be a necessary early event in TCDD carcinogenicity in experimental animals, the
downstream events involved are unknown.
5.2.a. Are the available data related to mode(s) of action for the carcinogenicity of TCDD
appropriately characterized and clearly presented?
5.2.b. Do the available data support EPA's conclusion that the overall mode(s) of action for
TCDD-induced carcinogenesis is largely unknown? Please comment on whether this
evaluation is clearly described.
5.3. Is EPA's approach for selecting data sets from the key epidemiologic studies and animal bioassays
identified for cancer dose response modeling scientifically justified and clearly described?
5.4. For the animal bioassay data, potential cancer oral slope factors (OSFs) were calculated by linear
extrapolation (using a linear, nonthreshold cancer approach) from the point of departure (POD).
EPA also estimated the composite risk of the occurrence of several tumor types from the animal
cancer bioassay data.
5.4.a. Please comment on whether the approach for estimating cancer risk, including the use of
tumor modeling of the TCDD animal cancer bioassay data, is scientifically justified and
clearly described.
5.4.b. Please comment on the choice of using a BMDLoi as the POD for the development of
candidate oral slope factors derived from the TCDD animal cancer bioassays.
5.5. EPA selected Cheng et al. (2006) - an analysis of the NIOSH occupational cohort - as the critical
study for oral slope factor (OSF) development. This study was chosen because it considers dose-
dependent elimination of TCDD rather than first-order kinetics.
5.5.a. Please comment on whether the rationale for this selection is scientifically justified and
clearly described. Please identify and provide the rationale for any other studies that
should be considered and provide a critical evaluation of the study and of its suitability
for meeting the goals of a quantitatively cancer assessment.
5.5.b. Cheng et al. (2006) analyzed all-cancer mortality. Please comment on the use of all-
cancer mortality as the basis of the OSF.
5.5.c. Please comment on whether the use of the Emond PBPK model in the estimation of risk-
specific doses from the Cheng et al. dose-response modeling results is scientifically
justified and clearly described.
5.5.d. EPA elected to use the log linear relationship of fat concentration and rate ratio to
estimate risk-specific doses at all risk levels. EPA could have estimated a POD for
cancer risk itself at a single risk level (BMR) for extrapolation to the origin. Please
A-6
-------
comment on EPA's choice of extrapolation approach.
5.5.e. The slope factor derived from Cheng et al. (2006) was extrapolated below the
background TCDD exposure levels experienced by the NIOSH cohort. Please comment
on this extrapolation.
5.6. Please comment on whether EPA has clearly described the major qualitative uncertainties in the
derivation of the OSF.
5.7. EPA did not consider dioxin-like compounds (DLCs) in the cancer dose-response modeling
because the occupational exposures in the available cohorts were primarily to TCDD. Background
DLC exposures were not incorporated in the dose-response modeling because EPA judged that it
was not possible to disaggregate the responses from background exposure to DLCs and
occupational exposure to TCDD. Please comment on whether this approach is scientifically
justified and clearly described.
5.8. The NRC suggested that EPA consider nonlinear approaches for the assessment of TCDD
carcinogenicity. In the Response to Comments, EPA presents two illustrative nonlinear approaches
for cancer, but considers both inappropriate to use because of the lack of MO A information.
5.8.a. Please comment on these two illustrative nonlinear approaches including EPA's
conclusions regarding the limitations of these approaches.
5.8.b. Are there other nonlinear approaches that could be readily developed based on existing
data for the assessment of TCDD carcinogenicity? If so, please suggest alternative
approaches and describe their utility and suitability for meeting the goals of a
quantitative cancer assessment.
Section 6. Feasibility of Quantitative Uncertainty Analysis from NAS Evaluation of the 2003
Reassessment
6.1. Please comment on the discussion in this Section. Is the response clearly presented and
scientifically justified?
6.2. Please comment on EPA's overall conclusion that a comprehensive quantitative uncertainty
analysis is not feasible.
6.2.a. Please comment on the discussion in Section 6 regarding volitional uncertainty and how
this type of uncertainty limits the ability to conduct a quantitative uncertainty analysis.
6.3. Throughout the document (including the Appendices), EPA presents a number of limited
sensitivity analyses (e.g., toxicokinetic modeling, RfD ranges, cancer OSF ranges, cancer RfD
development). Please comment on the approaches used, and the utility of these sensitivity analyses
in clarifying potential significant uncertainties.
A-7
-------
APPENDIX B: DISSENTING OPINION FROM KARL ROZMAN, PH.D.
The University of Kansas Medical Center
Kail K. Rozman, Ph.D.
Professor
School of Medicine
Department of Pharmacology,
Toxicology and Therapeutics
December 9, 2010
Thomas Armitage, Ph.D.
Designated Federal Officer
USEPA Science Advisory Board (MOOR)
1200 Pennsylvania Ave., N.W.
Washington, D.C. 20460
RE: A Dissenting Opinion
Dear Tom,
As 1 have indicated in my previous written and oral opinions to this panel, I disagree with
the panel conclusions regarding the carcinogenicity of TCDD and the adequacy of the BPA
response to the criticisms of the NAS report.
There is at best equivocal evidence (statistically not significant) for the carcinogenicity of
"l'CDD (or DLCs) in the occupational setting where the body burdens were at least 100 or 1000
times higher than the current or previous background levels. Therefore, the consideration of a
practical threshold for any defined population requires acceptance of the compelling scientific
conclusion that there is negligible (essentially zero) carcinogenic risk at current background
levels which are much lower than past levels. Any other conclusion is incompatible with sound
science and no amount of modeling or data manipulation will transform a non-existing effect at
occupational exposure levels into a risk at current background levels other than the non-
scientific, policy-driven, non-threshold extrapolation by EPA,
Further, it is my opinion that the EPA document (2010 Reanalysis of Key Issues Related
to Dioxin Toxicity and Response to NAS Comments, 600/-10/038A) is deliberately non-
responsive to the recommendations of the NAS report.
Respectfully,
Karl K.Rozman, Ph.D., D.A.B.T.
Professor
3901 Rainbow Blvd., MS1018, Kansas City, KS 66160 * (913) 588-7717 ¦ FAX (913} 588-7501
krozmari@kumc.sdu
B-l
-------
Additional Information Provided by Dr. Karl Rozman to Support Dissenting Opinion
The gist of my conclusion based on the evidence presented is that TCDD and its congeners is/are not
known human carcinogens at occupational and high environmental exposure levels; indeed not even at
levels that induce clinical chloracne (ATSDR, 1998; ATSDR, 2011). There is no question that chronic
TCDD exposures are carcinogenic for rodents at levels that are comparable to or somewhat higher than
those that cause chloracne in humans. Therefore, I have never taken issue with TCDD as a rodent
carcinogen or even as a possible human carcinogen. It is my conclusion (after more than 100 peer-
reviewed publications of mine dealing with this class of chemicals) that given sufficiently high (greater
than the chloracnegenic dose) exposures, dioxins may very well cause cancer in humans just as they do
in rodents. Rather, my dissent with the Panel is based on the unassailable logic that if even very high
exposures (reviewed in ATSDR, 2011) do not cause cancer in human beings, then past exposures (about
7ppt) and far lower current exposure levels (2 ppt or less) do not produce carcinogenic risk to humans.
As you can see from the ATSDR (1998, 2011) summaries, the un-manipulated Fingerhut Study failed to
detect an increase in total cancer, although there were many cases of chloracne. Analyses of several
hundred, perhaps more than a thousand blood samples revealed the presence of some very high levels of
TCDD. Manipulation of those data by U.S. EPA following the lead of IARC (1997) and NTP (2001)
resulted in U.S. EPA declaring TCDD as a known human carcinogen. From that point on, there was no
limit to extrapolation between species and from high-to-low doses. But modeling of an effect non-
existent in humans is nothing more than fiction.
There are numerous major scientific objections to the application of U.S. EPA's policy forcing default
linear treatment of all carcinogenicity dose-responses. In the case of TCDD and congeners, linear
treatment of the dose-response cannot be justified given the abject failure of these materials to elicit
genotoxicity in a wide variety of test systems (e.g., ATSDR, 1998). As such, the U.S. EPA default
assumption of linearity in the case of TCDD and congeners is policy masquerading as science.
I for one cannot accept a "policy" pretending to be rigorous science when the intent is to hide non-
scientific assumptions for the purpose of short-term political, perceived gain.
References
Agency for Toxic Substances and Disease Registry (ATSDR). 1998. Toxicological Profile for
Chlorinated Dibenzo-p-Dioxins. U.S. Department of Health and Human Services, Public Health
Services, Atlanta, GA.
Agency for Toxic Substances and Disease Registry (ATSDR). 2001. Addendum to the Toxicological
Profile for Chlorinated Dibenzo-p-Dioxins. U.S. Department of Health and Human Services, Atlanta,
GA.
International Agency for Research on Cancer (IARC). 1997. LARC Monographs on the Evaluation of
Carcinogenic Risks to Humans. International Agency for Research on Cancer, Lyon, France
National Toxicology Program (NTP) 2001. Report on Carcinogens: Dioxins (9th ed). U.S. Department
of Health and Human Services, Public Health Service, Atlanta, GA.
B-2
-------
APPENDIX C: VALUE OF INFORMATION
When human health risk assessments include an explicit representation of uncertainty, the potential
value of new information (VOI) can be estimated by its ability to reduce uncertainties that matter most
to the assessment target. While methods for determining VOI are most commonly associated with the
decision analysis literature in the context of informing management or regulatory decisions (Raiffa,
1968; Keeney, 1982; Winkler and Murphy, 1985; Finkel and Evans, 1987; Taylor et al., 1993; Clemen,
1996; Chao and Hobbs, 1997), there are many steps in a scientific assessment well before (or even
without subsequent) decision support and decision making where VOI evaluations can be of benefit in
characterizing current scientific knowledge and the potential for its improvement. EPA should integrate
these methods into their current and ongoing assessments of dioxin toxicity.
When uncertainty in a scientific assessment is measured by the variance of model predictions, a first
measure of VOI is the extent to which this variance might be reduced by new or additional data (e.g.,
Patwardhan and Small, 1992; Brand and Small, 1995; Abbaspour et al., 1996; Chao and Hobbs, 1997;
Sohn et al., 2000; Bosgra et al., 2005; Cooke, 2009). The relative contribution of different model
assumptions and parameter uncertainties to the variance of the estimated effect (e.g., the BMD), or the
cancer slope factor) provides an indication of which of these uncertainties would be most beneficial to
address. In addition, a VOI assessment considers the potential for the component uncertainties to be
reduced, based on the feasibility, resource requirements (time and funding), and likelihood of success of
the studies that would be needed to achieve the necessary improvement in scientific knowledge.
A scientific VOI study may also target a key classification inference that results from a risk assessment,
for example, whether a compound is genotoxic. Assuming the current assessment leads one to assign an
inconclusive probability to this outcome (e.g., between 10% and 90%, so that neither inference can be
rejected with a high degree of confidence), then potentially valuable studies are those able to shift
subsequent probabilities to high values (e.g., above 90, 95, or 99%) with a positive result (e.g., providing
support for genotoxicity) and/or to low values (below 10, 5, or 1%) with a negative result.
To illustrate, Small (2008) presents a simple probability tree model (a "distributional approach") for
assessing genotoxicity based on studies of DNA damage response caused by naphthalene and its
metabolites. In the proposed studies a series of isogenic cell lines deficient in various DNA metabolism
pathways are used to characterize the DNA damage responses caused by the targeted compounds.
Following results from the cultured cells, mice deficient in the specific DNA damage responses would
be exposed to naphthalene. Possible inferences are identified based on the assessed sensitivity and
selectivity of study results to the genotoxicity of naphthalene. Study outcomes considered include: i)
DNA damage responses in the isogenic cells; ii) increased numbers of stable DNA adducts in the DNA
repair deficient mouse lung; and iii) heightened Clara cell toxicity in the DNA repair deficient mouse
lung. Illustrative results using Netica are presented as follows:
C-l
-------
<( (a) Prior probabilities )>
Naphthalene Genotoxic?
true 50.0
talse 50.0
Cell DNA Damage Response?
yes 42.5
no 57.5
< Tumors in DNA R-D Mouse?
yes 50.0
no 50.0
Fig. 5. Illustrative consideration of
selected experimental results for
naphthalene Study IV (outcomes
assumed independent, prior probability
of genotoxicity set to 0.5, and sensitivities
and selectivities chosen by author solely
for illustration of methodology): (A)
Prior probability before study; (B)
Positive outcomes for both study results;
and (C) Positive results for cell DNA
damage, but negative results for
increased tumors in DNA
repair-deficient mouse.
(b) Both results positi\e )>
Naphthalene Genotoxic?
true 94.8
false 5.15
I
Cell DNA Damage Response'
yes Tool
no 01
esponse?
)|-4-hJ
>11111
< Tumors in DNA R-D
yes
no
100
z
j (c) Mixed results (cell damage, but no
( increase In tumors In DNA repair deficient
\ mouse)
Naphthalene Genotoxic?
true 36.4
false 63.6
4+
Cell DNA Damage Response?
yes
no
100
0
'p^onsc^^
< Tumors In DNA R-D Mouse?
yes
no
0
100
As rioted, the results shown above are intended solely to demonstrate the way in which study results can
be combined to support or refute targeted inferences.
Even when the uncertainty tree method is only used to delineate the set of possible outcomes and
relationships among steps and assumptions in the risk assessment (i.e., mode of action; dosimetry
measures for exposure; the mathematical form of the dose-response relationship; the experimental data
set(s) used to fit the relationship; and the procedure used for interspecies extrapolation) without the
assignment of probabilities to the tree branches, key assumptions and the experiments needed to support
or refute them can still be identified. These will typically involve elements of the assessment that,
depending on their resolution, effectively restrict the set of possible outcomes to either a positive or a
negative inference regarding the endpoint of the risk assessment. Establishing a procedure of this type
will allow the Agency to put in place a more formal mechanism for identifying, conducting, and
integrating the results of key studies for future assessments.
References
Abbaspour, K. C., R. Schulin., E. Schlappi and H. Fluhler 1996. A Bayesian approach for incorporating
uncertainty and data worth in environmental projects. Environmental Modeling and Assessment 1:151-
158.
Bosgra, S., P.M. Bos, T.G. Vermeire, R.J. Luit, and W. Slob. 2005. Probabilistic risk characterization:
an example with di(2-ethylhexyl) phthalate. Regulatory Toxicology and Pharmacology 43:104-113.
C-2
-------
Brand, K.P., and M.J. Small. 1995. Updating uncertainty in an integrated risk assessment: Conceptual
framework and methods. Risk Analysis 15(6):719-731.
Chao, P.T., and B.F. Hobbs. 1997. Decision analysis of shoreline protection under climate change
uncertainty. Water Resources Research 33(4):817-830.
Clemen, R.T. 1996. Making Hard Decisions: An Introduction to Decision Analysis (2nd ed.), Duxbury
Press, Belmont, CA.
Cooke, R.M. 2009. (ed) Uncertainty Modeling in Dose Response, Bench-testing Environmental Models.
Wiley, New York, NY
Finkel, A.M., and Evans, J.S. 1987. Evaluating the benefits of uncertainty reduction in environmental
health risk management. Journal Air Pollution Control Association, 37:1164-1171
Keeney, R.L. 1982. Decision analysis: an overview. Operations Research 39: 803-838.
Patwardhan, A., and M.J. Small. 1992. Bayesian methods for model uncertainty analysis with
application to future sea level rise. Risk Analysis 4:513-523.
Raiffa, H. 1968. Decision Analysis: Introductory Lectures on Choice under Uncertainty. Addison-
Wesley, Reading, MA.
Small, M.J. 2008. Methods for assessing uncertainty in fundamental assumptions and associated models
for cancer risk assessment. Risk Analysis 28(5): 1289-1307.
Sohn, M.D., M.J. Small, and M. Pantazidou. 2000. Reducing uncertainty in groundwater site
characterization using Bayes Monte Carlo methods. Journal of Environmental Engineering 126(10):
893-902.
Taylor, A.C., J. Evans, and T. McKone. 1993. The value of animal test information in environmental
control decisions. Risk Analysis 13:403-412.
Winkler, R.L., and A.H. Murphy. 1985. Decision analysis. In Probability, Statistics, and Decision
Making in the Atmospheric Sciences, A.H. Murphy and R.W. Katz (eds.), 493-524, Westview Press,
Boulder, CO,
C-3
-------
APPENDIX D: EDITORIAL COMMENTS AND CORRECTIONS
The Panel's advice and responses to the charge questions are contained in the body of this report.
However, in the course of the review, the following editorial comments and corrections have been
provided. This is not intended to be an exhaustive list.
a. Section 2
Minor suggestions to further improve clarity regarding data set inclusion/exclusion criteria.
Page xxxvii, Lines 16-19. The sentence needs clarification. It currently gives the impression that studies
that were eliminated for further analysis would have NOAELs available.
Pages 2-234 - 2-247. EPA should consider adding information to the appendices and/or tables to
provide readers with clarification regarding the exclusion of particular studies. For example, an extra
column in Table 2-7 listing, by numbered reference, the criteria that were or were not met for each study
would be helpful.
b. Section 6
Page 6-2. Add NRC (1996).
Page 6-3, bottom: The word "margins" should be "marginals."
Page 6-3, line 26: If EPA wants to use the adverb "always", the phrase "as a joint distribution" should
be "as some characterization of a joint distribution" to be correct.
Page 6-4, lines 9-12: This text is strange and off-putting. A reader might ask who wrote this and why. It
seems opinionated and unnecessary.
Page 6-4, line 9: The tone is too pedagogical ("This is not the place . . .").
Footnote 54: The discussion of alternatives to strict, single-measure probability theory is ham-handed.
Neither interval probabilities nor imprecise probabilities (sensu Walley, 1991) depart from probability
theory; they follow the Kolmogorov axioms. They are motivationally and essentially equivalent to
sensitivity analyses, except they do not make use of sampling strategies and can be more
comprehensive.
Lines 29-30: It is simply untrue that sensitivity analyses have to be systematic. The word "systematic"
might better be "comprehensive" and the word "essential" should be weakened, perhaps to
"advantageous".
Page 6-5, lines 4-7 and footnote 55: There seem to be only two axioms mentioned in the text, but
Kolmogorov needs three to make probability theory.
Page 6-5: The meaning of the phrase "epistemic uncertainty" given on this page is plainly incorrect.
Epistemic uncertainty is the uncertainty that arises from imperfect knowledge such as from limitations
on the amount or quality of data available or deficiencies in our causal understanding about a system. It
D-l
-------
is not true that a quantity about which there is epistemic uncertainty is necessarily fixed. Although it is
perhaps clear how one might come to this mistaken impression, no researchers use the phrase to imply
that the underlying quantity has no variability (although all would admit that this could be the case given
our ignorance about it). Indeed, a variable can have both forms of uncertainty. For example, when body
weight varies across a population, but with a distribution that is unknown, the variable has both aleatory
and epistemic uncertainty. This mistake echoes in a couple of other places throughout this section.
It is not clear what the authors take to be the difference between epistemic uncertainty and what they call
"cognitive uncertainty." It seems that the latter phrase was introduced because the meaning of
"epistemic uncertainty" had been misunderstood. Normally, the phrase "cognitive uncertainty" would
refer to an individual person's uncertainty about the validity of the results of his or her own information
processing. The assertion that cognitive uncertainty may be represented by probability (i.e., by precise
probability measures) is unnecessary and may be misleading. In fact, researchers in human cognition
and neuroscience have shown that humans process this kind of uncertainty (which they often call
"ambiguity") separately and differently from what we think of as probability or frequentist risk
(Glimcher, 2003; Hsu et al., 2005). The section can omit the phrase "cognitive uncertainty" altogether
and use in its place "epistemic uncertainty." There are slight differences between the two ideas (e.g.,
epistemic uncertainty could be shared by members in a group, whereas cognitive uncertainty is always
personal), but these appear to be unimportant in this context.
Page 6-5: The words "aleatoric" and "aleatory" are both used on this page as (synonymous) adjectives
of uncertainty. Actually, in the engineering literature, only "aleatory" is preferred for this use. In any
case, please pick one to use.
Page 6-5, line 10: The assertion that the frequentist and Bayesian interpretations are not mutually
exclusive may be misleading. They are mutually exclusive in the sense that it would be improper to mix
and match components of each into an analysis. It would be appropriate to omit the clause with the
phrase "mutually exclusive," although it is surely fair to say that subjective probabilities can and do
track relative frequencies.
Page 6-5, lines 30-32: The text on the subject of dependence is strange here, and also in section 6.1.3.3.
It is incorrect that the "[ijssues involving.. .epistemic and aleatory uncertainty translate into issues of
dependence." This is just wrong (even under their unusual definition of "epistemic").
Page 6-6: Section 6.1.3.2 starting on this page discusses a way to address uncertainty for sample data.
This Spartan treatment does not mention that sampling uncertainty is not the only kind of uncertainty
that can be associated with data, nor that it may not even be the largest kind of uncertainty.
Mensurational uncertainty (including the plus-minus part of a measurement, and censoring) may be
more important. In some cases, the family or shape of the marginal distribution may be unknown, which
is a kind of model uncertainty. As suggested on page 6-35, such uncertainties can be significant. The
section suggests only a resampling approach to expressing the uncertainty, but fails to mention the often
severe limitations of such approaches, and says nothing about what one might do if there is no relevant
sample data.
Page 6-6, line 20: Maybe the last word of the header should be plural.
Line 21: Modern practice has replaced "error" with "uncertainty" in this context.
D-2
-------
Footnote 56: EPA could add "or subtracting" after "adding."
Page 6-7, line 14: "The role of dependence modeling" should be replaced with "Dependence among
variables."
Page 6-7. More examples of use of expert judgment for health assessment are available and should be
cited.
Page 6-7, last paragraph: This paragraph extending onto the next page should be rewritten. The example
is reasonable and important, but the discussion about it is confused. The first sentence is incorrect. The
uncertainty mentioned in the second sentence may be epistemic, but the sentence is erroneous in its
claim. In the following sentences, the words "variable" and "fixed" (or "constant") should be used rather
than "aleatoric" and "epistemic." It is nonsense to say that a kinetic constant is "completely correlated
across individuals." It's not correlated; it is invariant. This case is not an example of a dependence issue.
There is no correlation between a distribution and a fixed quantity (even if it's uncertain). Correlation is
defined between varying quantities. If the number is fixed, whether or not we know what it is, then one
cannot say it is correlated with anything. The authors may have come to this twisted language because
they're thinking of the uncertainties in terms of how they might plan to quantitatively characterize them
in a Monte Carlo simulation (repeatedly selecting a random deviate for the kinetic constant but assigning
it to every individual). Of course, variables such as body fat, age, and smoking, on the other hand, can
and do exhibit correlations that definitely should be accounted for in the quantitative assessments.
Likewise, it is also important to keep track of the constancy of particular quantities about which we may
not know the precise value. These two issues should be untangled and discussed in a less confusing way.
Page 6-8, line 12: The first paragraph of section 6.1.3.4 seems to be saying that one can sometimes
express model uncertainty as parametric uncertainty, which simplifies its handling. This could be said
more plainly. It would be helpful to mention that this trick cannot always be used (as when the possible
models cannot be listed). It might also be especially helpful to mention that this trick is not so much a
way to propagate model uncertainty as a way to sweep it under the rug. Model averaging, including
Bayesian model averaging (which is mentioned in several places, including 6-36, lines 3ff), erases
model uncertainty in the same way that averaging variable quantities erases their variation.
Page 6-8. line 13: Omit the unnecessary fancy after the semicolon.
Lines 15-17: This sentence is nonsense, if we understand what a linear low-dose model is. Parsing the
sentence, it seems to say "uncertainty over a... slope.. .may be quantified, but uncertainty.. .in
slope.. .cannot be captured" which is self-contradictory. Perhaps what the authors mean to say is that the
linearity assumption is not itself subject to uncertainty quantification.
Page 6-9, line 1: The mathematical symbol x should be italicized, as should all Roman letters
throughout the document that represent unknown quantities, i.e., are symbols representing something
else rather than names like "e" the base of the natural logarithms.
Lines 14 and 16: The prefixes "pseudo" and "quasi" are not words. Hyphens are needed.
Page 6-9, line 18: Provide citations for dependence modeling.
D-3
-------
Page 6-9: Section 6.1.3.6 might also mention graphs, and other traditional communication tools other
than correlation indices.
Page 6-10, line 4: Add mention of methods that identify uncertain assumptions or parameters that are
important for determining whether the model is consistent with observed data (Hornberger and Spear,
1983) and for affecting a decision that is made as a result of the model (Merz et al., 2009).
Page 6-10, lines 29-30: Do the authors mean "this probabilistic language," referring to the word "likely"
in the quoted text?
Page 6-11, line 19: Of course there is no guarantee that linear will be protective.
Page 6-13, line 18: Of course it isn't really apodictic knowledge at all, but rather only an opinion or an
assumption. We see the authors' point and agree with it entirely, but perhaps they should use a word
other than "apodictic" here since it's not technically correct.
Page 6-14, lines 33-34: The parenthetical phrase "volitional uncertainty" should be expanded into a
sentence that says what the authors mean to express. The phrase "cognitive uncertainty" does not mean
anything in this context. Perhaps if the authors expanded it into a sentence too, maybe making it
"epistemic uncertainty" along the way, it would be possible to understand what they are trying to say
here.
Footnote 62: "Effective" is misspelled, as is "cancer."
Page 6-16, line 5: We note that it's not really a guarantee of course.
Line 8: The word "common" should be "predominant".
Page 6-16, line 20: Perhaps we can say that variability (and uncertainty) in the factors that are used to
determine a particular UF can be considered in choosing the particular value of the UF.
Page 6-17, lines 3-14: This problem can be addressed using a Bayesian analysis with a beta conjugate
for the uncertain response probability, p, with uniformative (uniform) prior for p. The probability that
"an experiment with a null response might have yielded a positive response" can be estimated from the
predictive distribution (which will depend on the number of test animals in the original study that
yielded zero responses) for the next experiment (with any number of exposed animals).
Page 6-17, line 28: The word "band" should be "limit".
Page 6-20, footnote 66: The text starting "each have an error factor" should be followed by "of rather
than "or".
Page 6-21, line 6: It would be helpful to say something about what the concerns are.
Page 6-21, lines 12-14: NAS was not suggesting that EPA use the uncertainty factors approach to mount
an uncertainty analysis, but rather a more modern approach.
D-4
-------
Page 6-22, line 19: Would it be appropriate to note "and establishes a concomitant reduction in some
UFs?"
Line 29: The word "invokes" should perhaps be "would require".
Page 6-23, line 33 and passim: The word "exotic" is a poor choice that is unnecessarily and
transparently loaded.
Page 6-25, line 29: This sentence is ungrammatical.
Page 6-26, line 24 and Figure 6-1: Would it be helpful to draw the 45-degree line on the graph?
Page 6-27, line 10: The word "epistemic" here is acceptable.
Line 14: The word "epistemic" here should be replaced by "fixed across individuals," and " is estimated
from" should be replace by "varies with." How does half life's estimability from data imply that it is
variable?
Page 6-28, lines 1-2: One would need the dependence between the variables to proceed.
Line 9: We suggest that "and" should be "although."
Page 6-29, line 1-2: There are bounding techniques based on the classical Frechet inequality that do not
require any knowledge of or any assumptions about dependencies.
Line 32: Omit "to."
Page 6-31, line 24: The pessimistic conclusion is a bit strong. Any estimate made from data is amenable
to a quantitative uncertainty analysis so, if one is measuring anything, one can propagate uncertainties
such as mensurational uncertainty, sampling uncertainty, and perhaps even surrogacy uncertainty. It's
not quite as hard to get quantitative models as the text here seems to suggest.
Page 6-32, lines 13-14: The dour conclusion is confusing. One could do a sensitivity analysis in this
case, couldn't one? If so, it seems that some kind of uncertainty analysis is clearly possible.
Page 6-33: The example in the text box is great, but the second table seems to say the log-likelihood for
LLD is 2.46 and for Hill is 2.16, which would make LLD's larger than Hill's, which contradicts what's
said in the text.
Page 6-34, line 4: Shouldn't "Delivered dose" be a new bullet?
Line 8: We don't think this statement is true. Perhaps "statistically more powerful" should be "typically
yield more sensitive".
Lines 24-25: We don't think it is necessary or helpful to persist with Box's platitude. Model uncertainty
is the uncertainty about a model's predictions that arises from doubt about the relevance of that model
for making such predictions.
D-5
-------
Page 6-37, line 29: The caveat is overwrought. Exploring relevant alternative values in a sensitivity
analysis could constitute a quantitative uncertainty analysis, even if the exploration is limited.
Page 6-37, line 30: This sentence is false. Analytical methods of propagation (convolution) don't
"sample" anything, and analyses based on intervals or imprecise probabilities don't depend on
uncertainty "distributions" (i.e., precise probability distributions).
It is important to keep in mind that, in general, we are not necessarily limited to identifying precise
probability distributions for everything that is to be characterized as uncertain (as seems to be suggested
here). Simple intervals about uncertain quantities can support a straightforward, albeit crude, interval
analysis that propagates uncertainty about parameters and other model choices to statements about the
range of possible results. Similarly, an approach based on interval probabilities, probability boxes, or
general imprecise probabilities (Walley, 1991) can combine such intervals with precise distributions if
they are known for some other inputs, and with structures that are intermediate between coarse intervals
and delicate probability distributions when some but incomplete knowledge is available. If the inputs are
profoundly uncertain, the results from such analyses are likely to be wide in reflection of these
uncertainties. In almost all cases, it is possible to be entirely rigorous without necessarily being precise
and without completely specifying each probability distribution.
Page 6-37, line 31: There does not need to be a specified "underlying distribution from which to
sample" in order to conduct a quantitative uncertainty analysis. It is facile to shrug off a call to
characterize and account for important uncertainties in the assessment process on these grounds alone.
Even when the uncertainty is volitional, there can be relevant ranges that are interesting to decision
makers and stakeholders. In such cases, the analysis may be formally closer to a sensitivity analysis, but
some appropriate response is usually possible, if not always practicable. To their credit, EPA has
acknowledged the legitimacy of the call and undertaken some efforts in this direction, notably Tables 5-
18 and 5-19 (although some kind of graphical summary of the results might have been better).
Page 6-38, line 30 and passim: The adjective "data driven" needs a hyphen, as it has elsewhere in the
document.
Line 23-24: We think this sentence is true, but, again, sampling from a distribution is not the only way
to conduct a quantitative uncertainty analysis.
Line 26: What is "(2.a)?"
Page 6-41, line 23: Omitting the word "extra" would make the sentence more easily understandable.
Line 31: What does "How Forward?" mean? Is this idiomatic?
Section 6.5.2: The assertions in this section are rather surprising and questionable. EPA says that
uncertainty quantification is an "emerging area in science" and that it is "an area where research could
be focused" because "the requisite knowledge does not yet exist" to apply quantitative uncertainty
analysis in assessments such as this one for dioxin. The document peremptorily dismisses the utility of
"convening a blue-ribbon panel" to identify the proper approach and suggests instead that "multiple
approaches should be encouraged." Is the inference that the present review panel shouldn't try to say
what the proper approaches to uncertainty quantification are, even if we think the area is more mature
than emerging? It is hard to understand what these statements are suggesting. Will the Agency support
D-6
-------
intramural and extramural research efforts in this direction? If not, what do the statements mean? Is it
impossible that EPA could benefit from some tech transfer efforts as well as basic research on
uncertainty quantification? The paragraph beginning on page 6-42 (line 3) mentions a European idea of
bench-test exercises to compare different approaches. It may be worth mentioning that this idea has been
implemented in the United States as well (Ferson et al., 2004; Oberkampf et al., 2004).
The document's reference list is alphabetically arranged, but seems to go from Z back to A again on
page R-33.
References
Ferson, S., C.A. Joslyn, J.C. Helton, W.L. Oberkampf, and K. Sentz. 2004. Summary from the epistemic
uncertainty workshop: consensus amid diversity. Reliability and Engineering and System Safety 85:355-
370.
Glimcher, P.W. 2003. Decisions, Uncertainty, and the Brain: The Science of Neuroeconomics. The MIT
Press, Cambridge, MA.
Hornberger, G.M., and R.C. Spear. 1983. An approach to the analysis of behavior and sensitivity in
environmental systems. In Uncertainty and Forecasting of Water Quality, M.B. Beck and G. van
Stratten (eds.), 101-116, Springer-Verlag.
Hsu, M., M. Bhatt, R. Adolphs, D. Tranel, and C.F. Camerer. 2005. Neural systems responding to
degrees of uncertainty in human decision-making. Science 310(5754): 1680-1683. [Available online at:
http://www.sciencemag.org/content/310/5754/1680.abstract]
Merz, R., J. Parajka, and G. Bloschl. 2009. Scale effects in conceptual hydrological modeling. Water
Resources Research 45:W09405.
Oberkampf, W.L., J.C. Helton, C.A. Joslyn, S.F. Wojtkiewicz, and S. Ferson. 2004. Challenge
problems: uncertainty in system response given uncertain parameters. Reliability and Engineering and
System Safety 85:11 -20.
Walley, P. 1991. Statistical Reasoning with Imprecise Probabilities. Chapman and Hall/CRC Press
D-7
------- |