Comparative Dietary Risks:
Balancing the Risks and Benefits
of Fish Consumption
Results of a Cooperative Agreement between
The U.S. Environmental Protection Agency
and
Toxicology Excellence for Risk Assessment (TERA)
Final
August 1999
Work performed in part undrr U S EPA Cooperative Agreement CX825499-01-0. Although the information in this
docurent has been funded in part by the United States Environmental
j'rotection Agency, this text may not necessarily reflect its
views and no official endorsement should be inferred
-------
Cooperative Agreement with U S EPA on Comparative Dietary. Risk
Table of Contents Page
List of Tables and Figures vi
List of Abbreviations ix
Foreword xi
Authors and Contributors xii
Acknowledgements xiv
Executive Summary xv
1. INTRODUCTION 1-1
1.1 References 1-5
2. HEALTH BENEFITS FROM EATING FISH 2-1
2.1 Introduction 2-1
2.2 Health Benefits Associated with Fish Consumption 2-2
2.2.1 Coronary Heart Disease (CHD) and Myocardial Infarction (MI) 2-2
2.2.1.1 Kromhout, Bosschieter and Coulander, 1985. “The Zutphen Study” (1) 2-9
2.2.1,2 Kromhout, Feskens and Bowles, 1995. General practice patients in
Rotterdam, the Netherlands. (2) 2-9
2.2.1.3 Norell, Ahibom, Feychting and, Pedersen, 1986. “Swedish Twins” (3) 2-9
2.2.1.4 Vollset, Heuch and, Bjelke, 1985. “Norway Postal Dietary Survey (4) 2-10
2.2.1.5 Curb and Reed, 1985. “Honolulu Heart Program” (5) 2-10
2.2.1.6 Fraser, Sabate, Beeson and, Strahan, 1992. “The Adventist Health Study” (6) ... 2-11
2.2.1.7 Moms, Manson, Rosner, Buring, Willett and Hennekens, 19g5. “The
US Physicians’ Health Study: 4 years” (7) 2-11
2.2.1.8 Albert, Hennekens, O’Donnell, Ajani, Carey, Willett, Ruskin and Manson,
1998. “The US Physicians’ Health Study: 12 years” (8) 2-12
2.2.1.9 Ascherio, Rimm, Stampfer, Biovannucci and Willett, 1995. “The Health
Professionals Follow-Up Survey” (9) 2-12
2.2.1.l0Gramenzi, Gentile, Fasoli, Negri, Parazzine and La Vecchia, 1990 (10) 2-13
2.2.1.11 Siscovick etal., 1995. Case-control Study in Seattle and King County,
Washington. (11) 2-13
2.2.1.12 Burr, Gilbert, Holliday, Elwood, Fehily, Rogers, Sweetnam and Deadman,
1989. “The Diet and Reinfarction Trial (DART)” (12) 2-13
2.2.1.l3Daviglus eta]., 1997. “The Western Electric Study” (13) 2-14
2.2.1.14 Conclusions and Weight of Evidence for an Association between Coronary
Heart Disease and Fish Consumption 2-14
2.2.2 Studies of Other Possible Health Effects of Fish Consumption 2-16
2.2.2.1 Smoking-Related Chronic Obstructive Pulmonary Disease (COPD) 2-19
2.2.2.2 Lung Damage from Smoking 2-19
2.2.2.3 Rheumatoid Arthritis 2-19
2.2.2.4 Childhood Asthma 2-20
2.2.2.5 Plaque Psoriasis 2-20
2.2.2.6 Colon Cancer 2-20
2.2.2.7 Gastrointestinal Disease 2-20
Toxicology Excellence for Risk Assessment iI 8/06/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
2.2.2.8 Dyslipideniia in Non-Insulin-Dependent Diabetes Mellitus 2-21
2.3 Antioxidant Levels 2-21
2.4 Health Benefits During Pregnancy, Lactation and Infancy 2-21
2.5 Health Benefits for Children Consuming Fish 2-22
2.6 Conclusions and Research Needs 2-23
2.7 References 2-24
3 NUTRITIONAL ASPECTS OF FISH COMPARED WITH OTHER PROTEIN
SOURCES 3-1
3.1 Introduction 3-1
3.2 Per Capita Consumption of Fish (Finfish and Shellfish) 3-1
3.3 Nutritional Content and Contamination Levels for Fish and Other Protein Sources 3-3
3.3.1 Selection of Nutrients, Foods and Contaminants for Tables 3-1 and 3-2 3-3
3.3.2 Substituting Other Foods for Fish: Effects on Macronutrient Profiles 3-4
3.4 Fish as a Protein Source 3-5
3.4.1 Protein Quality 3-5
3.4.2 Fish Protein vs Other Dietary Protein Sources 3-5
3.5 Fish as a Source of Essential Fatty Acids 3-6
3.6 Cholesterol 3-7
3.7 Vitamins 3-7
3.7.1 Vitamins B3, B6, and B12 3-7
3.7.2 Vitamin A 3-8
3.7.3 Vitamin D 3-9
3.8 Minerals 3-9
3.8.1 Calcium 3-9
38.2 Iron 3-10
3.8.3 Zinc 3-10
3.8.4 Selenium 3-11
3.9 Effects of Food Preparation Methods on Nutritional Benefits 3-11
3.10 Effects of Food Preparation Methods on Contaminant Levels 3-12
3.11 Conclusions and Research Needs 3-12
3.12 References 3-22
4 HEALTH RISKS FROM EATING CONTAMINATED FISH 4-1
4.1 Introduction 4-1
4.2 Calculating Risk above the Reference Dose for Noncancer Endpoints 4-2
4.2.1 EPA/ChemRisk Model 4-3
4.2.2 Other Approaches to Calculate Risk above the RID 4-5
4.3 Dose Response Information for the Six Selected Target Substances 4-6
4.3.1 DDT and Metabolites (DDE and DDD) 4-6
4.3.2 Methylmercury 4-7
4.3.3 Dioxin 4-10
4.3.4 Polychiorinated Biphenyls (PCBs) 4-11
4.3.5 Chlordane 4-13
4.3.6 Chlorpyrifos 4-16
4.4 Multigenerational Study of Great Lakes Salmon Fed to Rats 4-17
Toxicology Excellence for Risk Assessment 8/06/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary4Risk
4.5 Breast Milk as a Source of Contaminants 4-18
4.6 Conclusions and Research Needs 4-19
4.7 References 4-19
5 SOCIO-CULTURAL CONSIDERATIONS OF FISH CONSUMPTION 5-1
5.1 Introduction 5-1
5.2 Asian-Americans and Fish 5-2
5.3 Native Americans and Fish 5-4
5.4 Subsistence Fishing 5-5
5.5 Low-Income, Urban Anglers 5-5
5.6 Conclusion and Research Needs 5-7
5.7 References 5-8
6 FRAMEWORK AND CASE STUDIES 6-1
6.1 Introduction 6-1
6.2 Goals of the Comparative Dietary Risk Framework 6-2
6.3 Inputs for the Comparative Dietary Risk Framework 6-3
6.3.1 Potential Health Benefits of Fish Consumption 6-3
6.3.2 Measuring Severity of Health Outcomes and Magnitude of Health Benefits 6-4
6.3.2.1 Introduction 6-4
6.3.2.2 Incorporation of Severity into the Framework 6-6
6.3.3 Estimates of Human Health Risk 6-8
6.3.4 Dietary Considerations 6-8
6.3.5 Cultural Considerations 6-12
6.4 The BenefitfRisk Framework 6-13
6.4.1 Algorithm for Health Benefits 6-17
6.4.2 Algorithm for Health Risk 6-19
6.4.3 Algorithm for the Fish Consumption Index (FCI) 6-20
6.5 Demonstrating the Framework 6-20
6.5.1 Quantitative Example of the Framework 6-22
6.5.1.1 Calculations for Estimating Benefits 6-24
6.5.1.2 Calculations for Estimating Risks 6-25
6.5.1.2.1 Excess Lifetime Cancer Risk 6-25
6.5.1.2.2 Excess Lifetime Non-Cancer Risk 6-26
6.5.1.3 Estimating the FCI 6-28
6.5.2 Impacts from Changes in Contaminant Concentrations 6-28
6.5.3 Evaluation of Different Subgroups 6-32
6.5.4 Mixtures of Chemicals and Multiple Endpoints 6-34
6.5.5 Cultural Benefits 6-41
6.5.6 Personal Perception of Severity 6-43
6.6 Case Studies 6-44
6.6.1 Case Study: The Florida Everglades 6-44
6.6.1.1 Background 6-44
6.6.1.2 Summary of Existing Data 6-44
6.6.1.3 Exposure Assessment 6-45
6.6.1.4 Calculation of FCI 6-46
Toxicology Excellence for Risk Assessment iv 8/06/99
-------
Coo rative Agreement with U.S. EPA on Comparative Dietary Risk
6.6.1.5 Discussion 6-48
6.6.1.6 A Method for Verifying Fish Consumption Estimates 6-51
6.6.2 Vietnamese Immigrant Women Consuming Lake Ontario SportfIsh 6-52
6.6.2.1 Background 6-52
6.6.2.2 Summary of Existing Data 6-53
6.6.2.2.1 Descriptive Data 6-53
6.6.2.2.2 Biochemical Data 6-54
6.6.2,3 Exposure Assessment 6-55
6.6.2.4 Calculation of FCI 6-56
6.6.2.4.1 Salmon from Credit River 6-56
6.6.2.4.2 Rockbass and Smailmouth Bass from the Niagara River 6-60
6.6.2.5 Discussion 6-60
6.7 Overall Conclusions and Research Needs 6-61
6.8 References 6-69
7 USING AND COMMUNICATING THE COMPARATIVE DIETARY RISK
FRAMEWORK 7-1
7.1 Overview of Risk Communication as a Process 7-1
7.2 Designing, Implementing, and Evaluating a Communication Program for the
Comparative Dietary Risk Framework 7-2
7.2.1 Problem Analysis 7-5
7.2.2 Audience IdentifIcation and Needs Assessment 7-6
7.2.3 Communication Program Strategy Design and Implementation 7-7
7 2 4 Evaluation 7-10
7.3 Research Needs and Further Work 7-10
7.4 References 7-11
8 CONCLUSIONS AND RESEARCH NEEDS 8-1
8.1 Overall Conclusijns and Research Needs 8-1
8.2 Chapter 2 8-1
8.3 Chapter 3 8-2
8.4 Chapter 4 8-3
8.5 Chapter 5 8-4
8.6 Chapter 6 8-5
8.7 Chapter 7 8-7
8.8 Final Comment 8-8
8.9 References 8-8
Toxicology Excellence for Risk Assessment v 8/06/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
List of Tables and Figures
List of Tables
Table 2-1. Studies of Fish Consumption and Coronary Heart Disease (CHD) 2-3
Table 2-2. Studies of Fish Consumption and Other Endpoints 2-17
Table 3-3. Percent of energy (calories) from macronutrients based upon one
day’s diet which included a 150 gram serving of fish, chicken or
hotdogs. 1 (calculations based on Candat, 1994) 3-4
Table 3-1 Nutrition Values and Contaminant Levels in Fish - values for
100 g edible portion 3-14
Table 3-2 Nutrition Values and Contaminant Levels in Other Protein
Sources - values for 100 g edible portion 3-18
Table 4-1. Frequency of residue presence in fish, and the number of states
that have issued advisories for the chosen chemicals 4-2
Table 4-2: Methylmercury Responses at Multiples of the Reference Dose 4-10
Table 4-3: Chlordane Responses at Multiples of the Reference Dose 4-16
Table 6-1. Relative Risks for Various Endpoints listed in Table 2-1 6-4
Table 6-2. Severity Ranking of Effects and Benefits and Resulting Multipliers
for the Framework 6-7
Table 6-3. Caveats with the Use of Severity Schemes Shown in Table 6-2 for
Adjusting Quantitative Information on Risks and Benefits 6-8
Table 6-4. Input Parameters To Estimate Benefits 6-22
Table 6-5. Inputs Parameters To Estimate Risks 6-23
Table 6-6. Calculation of estimated daily doses using total hair Hg data from
Fleming etal. (1995) 6-45
Table 6-7. Dose-response estimates for methylmercury (Price eta!. 1997) 6-46
Table 6-8. Dose (mg/kg-day) as a Function of Fish Consumption Rate and
RfDs for Contaminants from EPA (1999) 6-57
Table 6-9. Hazard Indices Assuming Additive Toxicity for Salmon taken from the
Credit River. Calculations for Individual, all Compounds, and by Target
Organ or Critical Effect. HI> 1 Indicates Possibility of Toxic Effect 6-57
Table 6-10. Cancer Incidence, Cancer Risk (Including Severity Factor), Benefit
(Including Magnitude) and FCI for Salmon Taken from the Credit River 6-58
Table 6-11. Dose (mg/kg-day) of chemicals detected in smallmouth bass taken
from the Niagara River as a function of fish consumption (g/day) 6-59
Table 6-12 Dose (mg/kg-day) of chemicals detected in rockbass taken from the
Niagara River as a function of fish consumption (g/day) 6-59
Table 6-13. FCI at 38 glday for Salmon Rockbass and Smalimouth Bass 6-61
Table 6-14. Hazard Index for PCBs at 38 g/day 6-61
Toxicology Excellence for Risk Assessment vi - 8/06/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
List of Figures
Figure 4-1. Dose-Response Curves for Methylmercury 4-9
Figure 4-2. Dose-Response Curves for Chiordane 4-15
Figure 6-1. Disability as a Function of Protein Intake 6-10
Figure 6-2. Disability as a Function of Target Organ Impairment 6-11
Figure 6-3. Relative risk as a function of intake rate and source of protein 6-12
Figure 6-4. Relative risk of benefits and toxicity as a function of different
amounts of fish consumed assuming contamination with 2.1 ppm
methylmercury and 12 ppm chlordane 6-14
Figure 6-5. Health Scale as a Function of Fish Consumption Rate 6-21
Figure 6-6. Low Concentration Carcinogen 6-29
Figure 6-7. High Concentration Carcinogen 6-29
Figure 6-8. Low Concentration Non-Cancer 6-30
Figure 6-9. High Concentration Non-Cancer 6-30
Figure 6-10. Low Concentration Cancer and Non-Cancer 6-31
Figure 6-11. High Concentration Cancer and Non-Cancer 6-31
Figure 6-12. Low Concentration Non-Bioaccumulative 6-35
Figure 6-13. High Concentration Non-Bioaccumulative 6-35
Figure 6-14. Low Concentration Bioaccumulative 6-35
Figure 6-15. High Concentration Bioaccumulative 6-35
Figure 6-16. Change in FCI as more chemicals are evaluated for health
risk in fish Figule 6-16 6-36
Figure 6-17. Non-critical effects begin to manifest themselves at doses
much greater than the critical effect 6-37
Figure 6-18. Non-critical effects manifested at doses similar to critical
effect but dose response curves are shallower 6-38
Figure 6-19. Non-critical effects begin at doses similar to the critical
effect and their dose response curves are - similar 6-39
Figure 6-20. FCI changes when cultural benefits of fish consumption are added 6-42
Figure 6-21a Estimated Risk, Benefit, and FCI for Mercury Contaminated Fish
from the Everglades for the General Population 6-49
Figure 6.2 lb Estimated Risk to the Fetus As a Function of Everglades Fish
Consumption 6-50
Figure 6-22. Risk, Benefit, and FCI as a Function of Consumption of Salmon
from the Credit river 6-62
Figure 6-23. Total Cancer Risk and Individual Components for Salmon
Taken from the Credit River 6-63
Figure 6-24 Hypothetical Risk, Benefit, and FCI Assuming that the Shape
of the Noncancer Dose-Response Curve for PCBs is the Same as
that for Methylmercury for Salmon from the Credit River 6-64
Figure 6-25. Risk, Benefit, and FCI as a Function of Niagara River Rockbass
Consumption 6-65
Figure 6-26. Risk, Benefit, and FCI as a Function of Niagara River Smailmouth
Bass Consumption 6-66
Toxicology Excellence for Risk Assessment vii 8/06/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Figure 7-1. The risk communication process, adapted from Velicer and
Knuth (1994) 7-3
Toxicology Excellence for Risk Assessment viii 8/06/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
List of Abbreviations
AA Arachidonic Acid
ADI Allowable Daily Intake
AFS American Fisheries Society
AOC Areas of Concern
ARIC Arteriosclerosis Risk in Communities
AR Attributable Risk
ATSDR Agency for Toxic Substances and Disease Registry
ChE Cholinesterase
CSF Cancer Slope Factor
B 1 Background incidence of health endpoint i
BMD Benchmark Dose
BMDL Lower confidence limit on a benchmark dose
BMDL 1 O Lower b und on dose corresponding to 10% risk (used to be explicit that the
lower bound and not the maximum likelihood estimate is being used)
BMI Body Mass Index
BMR Benchmark Response
BW Body Weight
CHD Coronary Heart Disease
CNS Central Nervous System
COPD Chronic Obstructive Pulmonary Disease
CSF Cancer Slope Factor
CSFII Continuing Survey of Food Intakes by Individuals
DHA Docosahexanoic Acid
DL Detection Limit
ECG Electrocardiogram
ED Effective Dose
EPA Eicosapentanoic Acid
EPA Environmental Protection Agency
FA Fatty Acid
Trans-FA Trans-fatty Acid
FCI Fish Consumption Index
FDA Food and Drug Administration
FEL Frank Effect Level
FEV Forced Expiratory Volume
FVC Forced Vital Capacity
HDL High Density Lipoprotein
HHP Honolulu Heart Program
HI Hazard I idex
IRIS Integrated Risk Information System
LDL Low Density Lipoprotein
LOAEL Lowest Observed Adverse Effect Level
MeHg Methylmercury
MI Myocardial Infarction
MOE Margin of Exposure
Toxicology Excellence for Risk Assessment ix 8/06/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
MRL Minimal Risk Level
NEJM New England Journal of Medicine
NIDDM Non-insulin-dependent diabetes mellitus
NTP National Toxicology Program
NOAEL No Observed Adverse Effect Level
NOEL No Observed Effect Level
PAH (s) Polyaromatic Hydrocarbon(s)
PCB (s) Polychlorinated Biphenyl (s)
PCDD (s) Polychlorinated dibenzodioxin(s)
PCDF(s) Polychlorinated dibenzofurans(s)
ppm Parts Per Million
PUFA Polyunsaturated Fatty Acid
QALY Quality Adjusted Life Years
R Risk
RQ Reportable Quantity
RR Relative Risk
RR 1 Relative Risk of health endpoint i at a given consumption rate
RfC Reference Concentration
RD Reference Dose
RSD Risk Specific Dose
S Severity
S 1 Severity of health endpoint i
SFA Saturated Fatty Acid
UF Uncertainty factor
WHO World Health Organization
Toxicology Excellence for Risk Assessment x 8/06/99
-------
Coop irative Agreement with U.S. EPA on Comparative Dietary Risk
Foreword
This document is the result of a cooperative agreement between Toxicology Excellence for Risk
Assessment (TER4) and the U.S. Environmental Protection Agency (U.S. EPA), Office of
Water. TERA formed a Research Team of scientists to collectively develop knowledge of
problems regarding assessing health risks and benefits posed by consumption of chemically
contaminated fish and determine a method to evaluate both risks and benefits together. The final
outcome of this cooperative agreement is this report, which summarizes what is known about
health risks from consumption of contaminated fish, health benefits from consuming fish, and
general problems associated with comparisons of these risks and benefits. Moreover, this report
proposes a framework for comparing the health benefits and health risks in a quantitative
fashion.
The results of this research are intended to lead to a better understanding of the relative health
risks and benefits of consumption of contaminated fish. The authors of this report anticipate that
the proposed framework will be used by local risk managers and fish consumers to further
evaluate health benefits, health risks and other dietary information on contaminated fish.
Furthermore, states and tribes may use the results of this or subsequent work in assessing local
conditions and developing policies towards site-specific fish consumption advisories. An
Advisory Committee of state, local, tribal, industry and environmental scientists provided input
during the course of this research on the design and use of the framework. This Advisory
Committee reviewed a draft of this document and suggested improvements.
Funding for this work was provided by the U.S. EPA under Cooperative Agreement number
CX825499-O1-O and by TERA. Mr. Jeffrey Bigler of the U.S. EPA Office of Water was the
Project Officer. Although the information in this document has been funded in large part by the
United States Environmental Protection Agency, it does not necessarily reflect the views of the
Agency and no official endorsement should be inferred.
We would welcome your comments on this document. Please contact Toxicology Excellence for
Risk Assessment (TERA) at 513-542-7475 (RISK), or tera@tera.org (e-mail).
Toxicology Excellence for Risk Assessment xi 8/06/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Authors and Contributors
To conduct this research and write this document, Toxicology Excellence for Risk Assessment
(TERA) formed a Research Team of scientists from a number of key disciplines, including risk
assessment, nutrition science, environmental anthropology, medicine and public health, risk
communication and toxicology. The Research Team members each contributed knowledge and
inspiration from their respective fields to write or contribute to specific chapters, as well as
collaborate on the quantitative framework outline.
Authors and Research Team
Paul D. Anderson, Ph.D. (Chapter 6)
Ogden Environmental and Energy Services, Westford, MA
Daniel Cartledge, Ph.D. (Chapter 5)
Monmouth College, Department of Sociology and Anthropology, Monmouth, IL
Martha Daviglus, M.D. (Chapter 2)
Northwestern University Medical School, Department of Preventive Medicine, Chicago, IL
Michael Dourson, Ph.D. (Chapters 4, 6 and 8)
Toxicology Excellence for Risk Assessment, Cincinnati, OH
Barbara A. Knuth, Ph.D. (Chapter 7)
Cornell University, Department of Natural Resources, Ithaca, NY
Elaine Murkin, M.Sc., (Chapters 2, 3 and 6)
University of Guelph, Division of Applied Human Nutrition, Guelph, Ontario, Canada
Jacqueline Patterson, M.En. (Chapters 1 and 8)
Toxicology Excellence for Risk Assessment, Cincinnati, OH
Paul Price, M.S. (Chapter 4)
Ogden Environmental and Energy Services, Portland, ME
Judy Sheeshka, Ph.D. (Chapters 2, 3 and 6)
University of Guelph, Division of Applied Human Nutrition, Guelph, Ontario, Canada
Jerry Stober, Ph.D.
U.S. Environmental Protection Agency EPA, Athens, GA
Jason Unrine, B.S. (Chapter 4 and 6)
Toxicology Excellence for Risk Assessment, Cincinnati, OH
Toxicology Excellence for Risk Assessment xii 8/06/99
-------
Coop raUve Agreement with U S EPA on Comparative Dietary Risk
Contributors (Advisory Committee )
An Advisory Committee was formed in 1997 at the beginning of this project to provide advice
and assistance to TERA and the Research Team by identifying target and countervailing risks,
suggesting case study ideas and providing comments on the practicality and usefulness of the
framework. The Advisory Committee met in February 1999 to review a draft of this document.
The Committee members provided many helpful and constructive suggestions for revisions;
many of which are reflected in the final document. TERA and the Research Team greatly
appreciated the input aiid suggestions of the Advisory Committee. Their comments have
significantly strengthened this document.
Henry Anderson, M.D.
Bureau of Public Health, State of Wisconsin, Madison, Wisconsin
Michael Bolger, Ph.D.
Food and Drug Administration, Washington, D.C.
J. Milton Clark, Ph.D.
U.S. Environmental Protection Agency, Region 5, Chicago, Illinois
John Festa, Ph.D.
American Forestry Products Association, Washington, D.C.
Kory Groetsch, M.S.
Great Lakes Indian Fish & Wildlife Commission, Odanah, Wisconsin
Neil Kmiecik, M.S.
Great Lakes Indian Fish & Wildlife Commission, Odanah, Wisconsin
Amy Kyle, Ph.D.
School of Public Health, University of California
& Natural Resources Defense Council, San Francisco, California
Randall Manning, Ph.D.
Department of Natural Resources, State of Georgia, Athens, Georgia
Gerald Pollock, Ph.D.
California Environmental Protection Agency, Sacramento, California
Edward Ohanian, Ph.D.
U.S. Environmental Protection Agency, Washington, D.C.
Andy Smith, Ph.D.
Bureau of Health, State of Maine, Augusta, Maine
Toxicology Excellence for Risk Assessment xiii 8/06/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Acknowledgements
It was necessary to understand the perspectives of multiple disciplines in order to create this
framework and document. We appreciate the many contributions of our colleagues. In
particular, our Advisory Committee (listed elsewhere) was invaluable in providing advice and
suggestions. We also appreciate the scientific contributions of Dr. Barbara Harper, Mr. Stuart
Harris, and Dr. Rafael Ponce who were interested in this project and shared their ideas with us.
A number of TERA staff assisted in this endeavor. Ms. Joan Dollarhide provided initial thinking
and scoping of the project and Dr. Lynne Haber provided scientific review and input of the final
document. We thank both of them. We also appreciate the patience and perseverance of Ms.
Meg Poehlmann and Ms. Caitlin McArleton in finalizing the text and references.
Finally, we thank our EPA Project Officer, Mr. Jeffrey Bigler. His vision for the project,
especially as it fits with other EPA work, was most helpful in motivating us beyond our
individual disciplines towards an integrated and interdisciplinary product.
Toxicology Excellence for Risk Assessment xiv 8/06/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Executive Summary
A comparative dietary risk framework (hereafter referred to as the framework) has been
developed under this Cooperative Agreement for comparing the possible health risks of
consuming contaminated fish, while considering the potential health benefits lost by not eating
fish. The result of using the framework is a crude quantitative representation of the risk and
benefit associated with eating contaminated fish. The output of the framework is referred to as
the fish consumption index (FCI).
The FCI is an estimate of relative risk. It is not an estimate of absolute risk. In other words, it
does not provide users of the framework with an estimate of their increased or decreased
incidence of a particular health outcome. It simply provides a mechanism by which users can
weigh the possible health risks versus the possible health benefits of eating contaminated fish.
Cultural benefits of catching and eating fish (or detriments of not being able to fish or consume
fish) may also be considered, however the current version of the framework does not attempt to
quantify these benefits.
Before considering risks and benefits, a determination should be made that alternatives to
contaminated fish are not available. Perhaps lower contaminated fish sources are available
sufficient to maintain the individual’s desired level of fish consumption. Situations where the
weighing of benefits and risks may be necessary may include subsistence populations where
alternatives to contaminated locally caught fish are limited.
The framework is designed to provide information for a range of fish consumption rates,
allowing a user to roughly estimate the range of consumption rates at which people may have a
net benefit, a net risk, and the consumption rate at which no net change in the health index would
be likely. However, the suggested framework has a number of significant data gaps. These gaps
are sufficiently large so as to prevent any definitive conclusions. Moreover, these gaps prevent
making any overall recommendations on the existing fish consumption advisory programs of the
U.S. or other countries. Further study is needed to confirm and extend the preliminary findings
discussed in this document.
Use of the framework and FCI does not imply the proper choice is simply achieving a situation
in which the net risks and benefits are zero. Nor is it a justifIcation for accepting fish
consumption risks as long as there is a net benefit. Rather, the framework helps make the risks
and benefits transparent. Decisions about acceptable risks and distribution of risks and benefits
throughout society should be made collectively by the communities affected, and are not a focus
of this text. That the FCI may demonstrate cases in which fish consumption benefits may
outweigh the risks is not a license to pollute. Rather, society must determine policy about long-
term goals for minimizing environmental pollution based on a range of ethical, economic, social,
and other criteria. Again, the purpose of this text is to discuss the underlying scientific issues
associated with comparing the risks and benefits of fish consumption. It does not address the
social, economic or ethical considerations.
There is some evidence for an association between decreased risk of coronary heart disease
(CHD) or myocardial infarction (MI) and consumption of small amounts of fish, including
Toxicology Excellence for Risk Assessment xv 8/06/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
mainly lean (non-fatty) fish. In addition, other health endpoints have been examined and some
research suggests that eating fish may be associated with reduced incidences or severity of a
number of other endpoints. This evidence, along with the superior nutritional value of fish, is
strong enough that public health officials routinely encourage the public to eat more fish.
Consuming uncontaminated fish (or at least fish that are smaller, younger, or in general less
contaminated) may provide health benefits as mentioned above, but without the potential health
risks associated with contamination. The eating of such “cleaner” fish rather than more
contaminated fish, would maximize the net benefit of fish consumption, as we show specifically
for low versus high concentrations of chemicals in fish, for those chemicals that either
bioaccumulate or not, or for fish contaminated with more that one chemical.
This framework is an initial attempt to evaluate risks and benefits (qualitatively and
quantitatively) on a common scale. Constructing this framework has identified numerous areas
that need further research and development. Two needs seem paramount. First, better
estimations of benefits are needed for the general population and its sensitive subgroups.
Although information in this text is highly suggestive of the protective effects of eating fish and
allows some quantification, more definitive work is needed to support or modify our chosen
quantitative values. Second, better risk information is needed on the chemicals that commonly
contaminate fish. Sufficient knowledge on the toxicity of most of these pollutants exists, on
which noncancer risks could be quantified. Both sets of information are essential for this
framework to be most effective.
Toxicology Excellence for Risk Assessment xvi 8/06/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
1 Introduction
Toxic chemicals from point sources such as industrial or municipal discharges, and from non-
point sources such as agricultural runoff have contaminated some surface waters and their
sediments across the United States (U.S. EPA, 1992; Schmitt and Brumbaugh, 1990; Schmitt et
al., 1990). In addition, naturally occurring chemicals such as mercury can also contaminate
waters and sediments. Many of these pollutants concentrate in fish tissues by accumulating in fat
or binding to muscle. These contaminants found in fish may pose health risks to people eating
the fish. Those eating higher than average amounts of fish, such as sport and subsistence
anglers, are at a potential greater risk from eating contaminated fish than the general population.
In an effort to protect public health, state, local, and federal agencies and tribes issue fish
consumption advisories, when necessary, that usually recommend limits on the number of fish
meals which can safely be consumed within a specified time period (U.S. EPA, 1997a; Reinert et
a!., 1996; Dourson and Clark, 1990). These advisories are often issued for certain species of fish
from specific bodies of water, to address local contamination.
Fish consumption advisories are the current method for consumers to gain information on health
risks of contaminated fish. It is States and Tribes that issue fish consumption advisories and they
use varying methods and scientific judgments in reaching their conclusions. In addition, policy
issues may also be considered in setting these advisories, leading to greater difficulties for
individuals trying to determine their personal risks (Kamrin and Fischer, 1999).
While these advisories are generally based solely on considerations of the potential adverse
effects posed by the chemicals in fish, these same fish are an excellent source of low-fat protein
and may provide additional health benefits. Some recent publications have suggested that the
health benefits of eating even contaminated fish may outweigh the potential risks caused by the
presence of contaminants (e.g., Anderson and Weiner, 1995).
Fish consumption advisories, however, are not regulations and compliance with the
governmental advice varies (e.g., May and Burger, 1996; Knuth, 1995). It is individuals who
make the decision whether and how much fish to eat. Anglers, fishery experts, and health care
experts have all identified the importance of having information about how risks change with
different levels of fish consumption (Velicer and Knuth, 1994). Studies have demonstrated that
some anglers do respond to health risk information by changing their fishing-related behavior.
Changes include eating less sport-caught fish, changing fish-cleaning methods, changing fishing
locations, changing species eaten, changing the size of fish eaten, and changing cooking methods
(Connelly et al., 1992). Connelly et al. (1996) provided evidence that fish consumption
suppression (anglers eating less fish than they would in the absence of health advisories) was
prevalent among Lake Ontario anglers.
Studies of licensed anglers have indicated the perceived importance of health advisory
information on potential health benefits and risks associated with fish consumption. These same
studies also note that anglers recognize the importance of how risks change as more or less fish
is eaten, and compare the health risks of eating fish with the risks from other protein sources
(e.g., Connelly c i al., 1992; Connelly and Knuth, 1993).
Toxicology Excellence for Risk Assessment 1-1 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Evaluating the potential risks (and benefits) requires information on contamination levels and
consumption rates. Surveys of anglers and their families have shown that rates of fish
consumption vary widely among subpopulations by race or ethnicity, age, sex, income, fishing
mode, region of the country and other demographic variables (CAL EPA, 1997). For example,
regional surveys of sport fishing populations report overall mean rates for consumption of sport
fish ranging from 12.3 to 63.2 glday (CALEPA, 1997), while U.S. EPA estimates a fish
consumption rate for the general population for all fish of 20.1 g/day (uncooked weight) (U.S.
EPA, 1997b). Studies among tribal and subsistence fishing populations have found much higher
levels of consumption (see for example Toy et al., 1996, CALEPA 1997 and U.S. EPA, 1997b).
This wide variability in consumption rates and patterns reinforces the necessity of evaluating fish
pollutants and consumption on a case-by-case or local basis.
While contaminants in fish pose a public health risk, fish is also an excellent source of protein
and provides additional health benefits not available from other foods. It has been recognized for
over a decade that a need exists to evaluate the benefits of fish as a food source, as well as the
risks from contaminants, when setting fish consumption advisories (CDHS, 1988; Kimbrough
1991; Egeland and Middaugh, 1997). The California Department of Health Services sponsored a
workshop in 1988 called “Balancing the Scales: Weighing the Benefits and Risks of Fish
Consumption.” Speakers addressed the nutritional composition of fish, cardiovascular effects
from n-3FA and benefits of fish oil consumption, along with exposures and health risks. Over
ten years later there is more scientific data on potential health benefits of eating fish.
Putting risks into perspective is even more important when the fish are a part of a traditional
subsistence diet, which is important to a group’s cultural identity (Egeland et al., 1998). In
addition, for some communities, alternate foods are not readily available or affordable.
The need to consider the beneficial aspects of fish consumption has also been recognized by the
Federal-State-Tribal Fish Forum sponsored by EPA (AFS, 1997). This group of federal, state
and tribal scientists and public health officials has identified consideration of benefits from fish
as one of their top issues needing research and guidance. The research discussed in this
document is a direct result of this group’s request.
When advisories are issued and suggestions made to reduce consumption of contaminated fish,
individuals may respond in a number of ways. They may follow the fish advisory and reduce
their consumption of that particular type of fish, they may reduce exposure to contaminants by
selecting a less-contaminated fish or preparation method, they may stop eating fish, or, they may
ignore the advice and eat without regard to the advisory. Ideally, by selecting and eating the
least contaminated species, one can enjoy fish and its benefits without the health risks of
contaminants. However, if individuals do reduce their consumption of contaminated fish and
replace it with other non-fish foods; depending on the food choices made, these dietary changes
may not reduce overall health risks and may actually result in greater overall health risks.
Situations of subsistence populations, who have limited alternatives to a contaminated fish
source, may encounter this dilemma of needing to weigh the benefits and risks. To fully evaluate
the risks and benefits, one needs to examine the target risk - that is the adverse health effect from
eating fish with chemical contaminants - as well as the countervailing risks, such as the
consequences of reducing fish consumption and the potentially reduced nutritional or health
benefits of the substituted foods.
Toxicology Excellence for Risk Assessment 1-2 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Graham and Wiener explore the issues of target and countervailing risks for a number of public
health issues in their book Risk vs. Risk: Tradeoffs in Protecting Health and the Environment
(1995). For example, if a fish consumption advisory recommends reducing consumption of fish
contaminated with a particular chemical, and
• the fish in the diet is replaced with a large amount of fruits and vegetables, the consumer
may trade a decreased cancer risk from contaminants in fish (the target risk, i.e., the risk the
advisory is designed to reduce) for an increased cancer risk from increased ingestion of
anthropogenic and natural pesticides (the countervailing risk, i.e., the risk that may increase
as a result of the advisory).
• the fish in the diet is replaced with red meat, the consumer may be trading a decreased risk
of mortality from cancer (target risk) for an increased risk of mortality from heart disease
(countervailing risk), due to an increased consumption of saturated fat.
• the consumption of local fish high in PCBs, such as salmon, is replaced with an increased
consumption of canned tuna high in methylmercury, the consumer may be trading increased
risk of developmental toxicity and cancer from PCBs (target risk), for an increased risk of
neurological disease from methyl mercury (coantervailing risk).
In one chapter of this book Anderson and Wiener (1995) concluded that the protective effect of
increasing fish consumption on chronic heart disease far outweighed the increased cancer risk
posed by contaminants in fish. Using U.S. EPA’s cancer slope factors and assuming that fish
contained the FDA limits of 6 common fish contaminants’, Anderson and Wiener (1995) found
that the cancer risk associated with eating 1 gram of fish per day for a 70-year lifetime was 5 x
10-4. Based on their analysis, increasing consumption of fish from 0 to 40 grams per day would
increase the average American’s risk of dying by 2 percent from cancer 2 . However, the same
increase in fish consumption would decrease the average American’s risk of dying from heart
disease by 35 percent. Thus, public health officials and consumers might want to evaluate a
broad range of dietary infonnation before making decisions regarding consumption of
contaminated fish.
Countervailing risks can go beyond the health implications of food substitutions and include
social, economic, religious and cultural impacts (Wheatley and Paradis, 1996). Harris and
Harper (1997) have explored how to evaluate impacts other than direct risk to health. They have
developed a Native American exposure scenario that identifies parameters for evaluating
countervailing impacts on cultural and religious activities. These may affect quality of life,
which in turn impacts both individual and community health and well-being.
The direct benefits of fish consumption can be thought of as arising from two sources. The first
relates to the change in the incidence of a particular health outcome as related to fish
consumption rate (e.g., decrease in heart disease with increasing fish consumption). The results
‘Chiorpyrifos, Chiordane, DDT, Dioxin, PCBs and Methylmercury
2 Anderson and Weber (1995) recognized that EPA’s cancer risk method predicts upper bound incidence of cancer,
but assumed that cancer incidence was the same as cancer mortality in order to err on the side of protecting public
health
Toxicology Excellence for Risk Assessment 1-3 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
of these studies cai be used to derive a dose-response relationship between fish consumption rate
and the health outcome being investigated (within limits imposed by the data). The second
relates to how general nutritional status changes as fish is substituted for some other source of
protein or is removed from the diet.
There is some evidence for an association between decreased risk of coronary heart disease
(CHD) or myocardial infarction (MI) and consumption of small amounts of fish, including
mainly lean (non-fatty) fish. In addition, other health endpoints have been examined and some
research suggests that eating fish may be associated with reduced incidences or severity of a
number of other endpoints. The possible benefits in the form of reduced risk of particular
diseases are discussed in Chapter 2.
There are many nutritional benefits associated with eating fish, regardless of the species type.
Perhaps, unlike red meats, eggs and dairy products, fish provides very high quality protein and a
“heart healthy” combination of fatty acids. Further, fish (both lean and fatty) is one of the few
foods that contain omega-3 (n-3) fatty acids, a class of fatty acids that are essential for the
development of the nervous system and that may have other beneficial health effects. Fish
supplies a number of vitamins and minerals that tend to be low in the U.S. diet, including
calcium, iron, zinc, vitamin A, niacin, vitamin B6 and vitamin D, in additIon to others. The
nutritional advantages of fish compared to other protein sources are discussed in Chapter 3.
Fish consumption advisory programs have traditionally focussed on assessing the potential
human health risks from eating contaminated fish and estimating safe consumption limits.
Chapter 4 discusses potential health risks for a number of common contaminants and discusses
the methods for estimating risk used later in this document.
Food, and fish in particular, may also be an important part of a culture, serving economic, social,
aesthetic, and religious functions. Specific foods are often seen as having special nutritional or
medicinal qualities, and methods of food preparation are frequently part of one’s cultural
identity. These cultural factors may need to be considered in evaluating risks and benefits from
consumption of contaminated fish for some subpopulations. Chapter 5 outlines the social and
cultural importance of fish to particular groups of people.
Chapter 6 develops the comparative dietary risk framework which compares the possible health
risks of consuming contaminated fish, while considering the potential health benefits lost by not
eating fish. Example outputs using hypothetical data and two case studies with actual exposure
scenarios are also included. The result of using the framework is the fish consumption index
(FCI), which is a crude quantitative representation of the net risk (or benefit) associated with
eating contaminated fish. It provides a mechanism by which users can weigh the possible health
risks versus the possible health benefits of eating contaminated fish. Cultural benefits of
catching and eating fish (or detriments of not being able to fish or consume fish) may also be
considered, however the current version of the framework does not attempt to quantify these
benefits.
Because of the data intense process and results of the FCI, a solid risk communication program is
necessary to insure successful usage of the information generated. Chapter 7 summarizes key
Toxicology Excellence for Risk Assessment 1 4 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
elements of the risk communication process as applied to the comparative dietary risk
framework, emphasizing that risk communication is a process of information exchange between
the target audience and the risk communicator. Although the framework provides a mechanism
for comparing risks and benefits associated with fish consumption, it is not a justification for
accepting fish consumption risks as long as there is a net benefit. Decisions about acceptable
risks and distribution of risks and benefits throughout society is a social decision, to be made
collectively by the communities affected. Rather, the framework helps make the tradeoffs
between risks and benefits more transparent.
When alternatives to consumption of contaminated fish are not available or desired, it may be
appropriate to weigh the risks of eating less contaminated fish with the benefits gained from
eating more of these same fish. The framework developed here can crudely compare these risks
and benefits. However, this framework has a number of significant data gaps, which are
discussed in Chapter 8. These gaps are sufficiently large so as to prevent any definitive
conclusions from this study or any overall recommendations regarding existing fish consumption
advisory programs of the U.S. or other countries. Further work is needed to confirm and extend
these preliminary findings.
The purpose of the current research is to develop an understanding and framework by which to
evaluate the comparative risks posed by dietary changes as a result of fish consumption
advisories. This research builds upon previous work from a series of documents developed by
the U.S. EPA on “Guidance for Assessing Chemical Contaminant Data for Use in Fish
Advisories.” The four-volume set includes Volume 1-Fish Sampling and Analysis (1995a),
Volume 2-Risk Assessment and Fish Consumption Limits (1997a), Volume 3-Overview of Risk
Management (1996), and Volume 4-Risk Communication (1995b). The results of this research
can lead to a better understanding of the effects that fish consumption advisories have on diet and
public health. We anticipate that public health officials and consumers may use this increased
understanding to evaluate a broad range of dietary information before making decisions about
whether or not to eat contaminated fish.
1.1 References
AFS (American Fisheries Society). 1997. Recommendations for the Second Federal State
Action Plan for Fish Consumption Advisories. A report to the U.S. EPA by the Water Quality
Section, American Fisheries Society. Bethesda, MD.
Anderson, P.A. and J.B. Wiener. 1995. Eating Fish. In: Risk vs. Risk: Tradeoffs in Protecting
Health and the Environment. J.D. Graham and J.B. Wiener. eds. Harvard University Press,
Cambridge, Massachusetts. pp 104-124.
CAL EPA. 1997. Consumption of fish and shellfish in California and the United States. Final
draft report, Chemicals in Fish, Report No. 1. Pesticide and Environmental Toxicology Section,
Office of Environmental Health Hazard Assessment.
CDHS (California Department of Health Services). 1988. Balancing the scales: weighing the
benefits and risks of fish consumption. Proceedings of a workshop held on October 20, 1988 in
Toxicology Excellence for Risk Assessment 1-5 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Concord, California.
Connelly, N.A., B.A. Knuth, and C.A. Bisogni. 1992. Effects of the health advisory and
advisory changes on fishing habits and fish consumption in New York sport fIsheries. HDRU
Series No. 92-9. Department of Natural Resources, Cornell University. Ithaca, NY.
Connelly, N.A., B.A. Knuth, and J.E. Vena. 1993. New York State angler cohort study: health
advisory knowledge and related attitudes and behavior, with a focus on Lake Ontario. HDRU
Series No. 93-9. Department of Natural Resources, New York State College of Agriculture and
Life Science, Cornell University. Ithaca, NY.
Connelly, N.A., B.A. Knuth, and T.L. Brown. 1996. Sportfish Consumption Patterns of Lake
Ontario Anglers and the Relationship to Health Advisories. North American Journal of Fisheries
Management. 16: 90-101.
Dourson, M.L. and J.M. Clark. 1990. Fish consumption advisories: toward a unified,
scientifically credible approach. Regul. Toxicol. Pharmacol. 12: 161-178.
Egeland, G.M. and J.P. Middaugh. 1997. Balancing fish consumption benefits with mercury
exposure. Science. 278: 1904-1905.
Egeland, G.M., L.A. Feyk, and J.P. Middaugh. 1998. The use of traditional foods in a healthy
diet in Alaska: risks in perspective. Section of Epidemiology, Alaska Division of Public Health,
Department of Health & Social Services. State of Alaska.
Graham, J.D. and J.B. Wiener, eds. 1995. Risk vs. Risk: Tradeoffs in Protecting Health and the
Environment. Harvard University Press, Cambridge, Massachusetts.
Hanis, S.G. and B.L. Harper. 1997. A Native American exposure scenario. Risk Anal. 17(6):
789-795.
Kamrin, M.A. and L.J. Fischer. 1999. Current status of sport fish consumption advisories for
PCBs in the Great Lakes. Regul. Toxicol. Pharmacol. 29(2 Pt. 2): 175-181.
Kimbrough, R.D. 1991. Consumption of fish: benefits and perceived risk. J. Toxicol. Environ.
Health. 33: 81-91.
Knuth, B.A. 1995. Fish consumption health advisories: who heeds the advice? Great Lakes
Res. Rev. 1(2): 36-40.
May, H. and J. Burger. 1996. Fishing in a polluted estuary: fishing behavior, fish consumption,
and potential risk. Risk Anal. 16(4): 459-47 1.
Reinert, R.E., B.A. Knuth, M.A. Kamrin, et al. 1996. A review of the basic principles and
assumptions used to issue fish consumption advisories. American Fisheries Society Symposium.
16: 98-106.
Toxicology Excellence for Risk Assessment 1-6 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Schmitt, C.J. and Brumbaugh, W.G. 1990. National contaminant biomonitoring program:
concentrations of arsenic, cadmium, copper, lead, mercury, selenium, and zinc in U.S. freshwater
fish, 1976-1984. Arch. Environ. Contam. Toxicol. 19: 731-747.
Schmitt, C.J., J.L. Zajicek, and P.H. Peterman. 1990. National contaminant biomonitoring
program: residues of organochlorine chemicals in U.S. freshwater fish, 1976-1984. Arch.
Environ. Contam. Toxicol. 19: 748-781.
Toy, K.A., N.L. Polissar, S. Liao, et al. 1996. A fish consumption survey of the Tulalip and
Squaxin Island Tribes of the Puget Sound Region. Tulalip Tribes, Department of Environment,
7615 Totem Beach Road, Marysville, WA 98271.
U. S. EPA. 1992. National study of chemical residues, Vol. 2. Office of Science and
Technology, Standards and Applied Science Division. EPA 823-R-92-008b.
U.S. EPA. 1995a. Guidance for assessing chemical contaminant data for use in fish advisories,
Volume I. Fish sampling and analysis, 2nd ed. Office of Water. EPA 823-R-95-007.
U.S. EPA. 1995b. Guidance for assessing chemical contaminant data for use in fish advisories,
Volume IV. Risk Communication, 2nd ed. Office of Water. EPA 823-R-95-O01.
U.S. EPA. 1996. Guidance for assessing chemical contaminant data for use in fish advisories.
Volume 3. Overview of Risk Management. Office of Water. EPA 823-R-95-001.
U.S. EPA. 1997a. Guidance for assessing chemical contaminant data for use in fish advisories.
Vol. 2: Risk assessment and fish consumption limits, 2nd ed. Office of Science and Technology,
Office of Water. EPA 823-B-97-009.
U.S. EPA. 1997b. Exposure Factors Handbook, Vol. II: Food Ingestion Factors. Office of
Research and Development. EPA 6001P-95/OO2Fb.
Velicer, C.M. and B.A. Knuth 1994. Communicating contaminant risks from sport-caught fish:
the importance of target audience assessment. Risk Anal. 14(5): 833-84 1.
Wheatley, B. and S. Paradis. 1996. Balancing human exposure, risk and reality: questions
raised by the Canadian Aboriginal Methylmercury Program. Neurotoxicology. 17(1): 241-250.
Toxicology Excellence for Risk Assessment 1-7 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
2 Health Benefits From Eating Fish
2.1 Introduction
In addition to providing high quality protein, essential fatty acids, and other nutrients required
daily in the human diet (discussed in Chapter 3: Nutritional Benefits of Eating Fish Compared to
Other Protein Food Sources), fish consumption is also associated with certain health endpoints
over the longer term. This chapter provides a brief overview of health endpoints that have been
shown, or are hypothesized, to be associated with fish consumption. In some cases the weight of
evidence supports the relationship between eating fish and a lowered risk of disease (e.g.,
coronary heart disease or CHD). For other health endpoints, the link is more controversial (e.g.,
arthritis) and more research studies are needed. This report refers to the changes in health
endpoints associated with fish consumption as benefits, because they generally involve a
reduction in the risk of chronic disease. The chapter begins with an overview of the major
studies that have examined the association between fish consumption and CHD, both those
studies that have found associations and those which have not. It then continues with a brief
description of studies that have looked at fish consumption in relation to several other endpoints.
The concept that eating fish may reduce the risk of CHD apparently originated from reports on
the small population of non-acculturated Eskimos in arctic Greenland, where high consumption
of marine animals (e.g., seal, fish) was observed (Bang et al., 1971, 1980). It was claimed that
coronary rates were low, but available data were — and remained — limited and tenuous. The
inhabitants of the Japanese island of Okinawa were also observed to have low CHD mortality
rates, and they too consume high amounts of fish (Kagawa et al., 1982). The “Seven Countries
Study t ’ (Keys, 1980) conducted by Dr. Keys found that rates of CHD and myocardial infarctions
(MI) were lower in southern Italy, Spain and Greece than rates in the United States, the
Netherlands, and other countries. This raised the question of whether the Mediterranean diet,
which includes fish, red wine, olive oil, nuts and legumes, was partly responsible for the
findings.
These observations prompted epidemiologic investigations -- the first from the Netherlands
(Kromhout et al., 1985) -- on the relationship of fish consumption to CHD and MI. When the
results of the long-term, prospective studies became available in the mid-1980s and 1990s, they
provided strong evidence that higher levels of fish consumption among middle-aged men, free of
CHD at baseline examination, were associated with a lower risk of mortality from CHD. Early
hypotheses on the protective biological mechanism of fish consumption focused on the very long
chain, polyunsaturated fatty acids of the omega-3 class (n-3 FA), since fish are one of the few
good dietary sources of these types of fatty acids. It was thought that the n-3 FA of fish
contributed to more healthful ratios of blood lipids (fats, including cholesterol), and this reduced
one of the major risk factors for CHD: high blood levels of cholesterol and saturated fats. Later
studies focused on the factors that cause blood platelets and other materials to clump together
and deposit on the inside of artery walls, and specifically, the role that n-3 FA have in the
synthesis of prostaglandins, which reduce platelet aggregation.
The American Heart Association has issued a statement regarding fish consumption, fish oil,
lipids, and CHD. The primary focus of this statement, however, is on the evidence of health
Toxicology Excellennce for Risk Assessment 2-1 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
benefits from n-3 FAs. It concludes that it is premature to recommend general usage of fish oil
supplements at this time, but that the “inclusion of marine sources of the n-3 PUPA in the diet
seems reasonable because they are good sources of protein without the accompanying high
saturated fat seen in fatty meat products” (American Heart Association 1996). For reasons that
will be discussed further in this section, it is unlikely that n-3 FA are the primary factor
responsible for the observed association between fish consumption and lowered risk of CHD.
In addition to the relationship between fish consumption and sudden death from CHD and MI,
other health endpoints have been examined. Scientists have investigated possible associations
between eating fish and inflammation-related diseases such as rheumatoid arthritis, since the n-3
FA in fish oils are believed to have anti-inflammatory effects (reviewed by Simopoulos, 1991).
Relatively few studies have been done, however, and their findings are still controversial. Other
research has suggested that eating fish may be associated with reduced incidences or severity of
asthma, psoriasis, gastrointestinal diseases, as well as lung damage caused by smoking. (Of
course, even if fish consumption does protect against these diseases, it may exert its protective
effect through mechanisms other than the reduction of inflammation by n-3 FA.) Recently,
possible associations between fish consumption and other health endpoints for pregnant women,
unborn babies and young infants have been reported. For example, there is evidence that
consuming substantial amounts of fish during pregnancy may lengthen gestation (Olsen, Hansen
& Sorensen, 1986), thereby resulting in higher birthweights. The n-3 fatty acids that fish provide
have important roles in the development of the retina, brain and other central nervous system
tissues in the unborn and infant (up until 12 months of age), as well.
The following sections briefly review available results of research on the association between
fish consumption and several endpoints, and the possible health benefits of eating fish for adults,
pregnant women, developing fetuses and infants. The nutritional contributions of fish to the
diets of children are also highlighted.
2.2 Health Benefits Associated with Fish Consumption
2.2.1 Coronary Heart Disease (CHD) and Myocardial Infarction (MI)
Table 2-1 presents a summary of the samples, designs and main findings of 13 published reports
examining the association between fish consumption and CHD. Studies 1 to 9 and study 13 are
prospective studies, a strong epidemiological study design in which individuals who have no
observable symptoms of CHD are categorized according to their level of fish consumption, and
then followed for long periods of time to observe who later develops CHD and who does not.
Studies 10 and 11 are case-control designs, whereby those who already have CHD are matched
on important characteristics with those who do not, and retrospective information about their
diets is collected. Study 12 is a randomized, controlled, clinical trial, the only study included
here that contains an intervention. A brief summary of the findings of each of these 13 reports
follows the table.
Toxicology Excellence for Risk Assessment 2-2 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
Table 2-1. Studies of Fish Consumption and Coronary Heart Disease (CHD){PRIVATE }
Study,
Year of
No. of Participants,
Diet
Stratification of
CHD End
Main Findings,
Country,
Baseline &
Gender, Baseline
Assessment
Fish Intake
Points,
Fish Intake and CHD
Type of Study,
Follow-Up
Ages, Background
Method
(no. or % of
Number
Reference No.
Duration
persons, or
person-years)
of Events
1- Zutphen, The
1960
852 men ages 40-59;
in-depth
0 g/d (159), 1-14
CHD
significant independent
Netherlands,
20 years
urban community
cross-check
(283), 15-29
death;
inverse relationship;
prospective
general population
diet history
(215), 30-44
78 deaths
RR values for strata:
epidemiologic
(116), >45 (79),
1.00, 0.60. 0.57, 0.46,
(Kromhout et al,
mean: 0, 8, 22,
0.42, p for trend <0.05
1985)
36, 67 g/d, about
2/3 lean fish
significant independent
inverse relationship;
2- Rotterdam,
1971
292 men and women
in-depth
non-consumers
CHD
RR = 0.51, 0.41 for
The
17 years
(137 & 135) ages 64-
cross-check
(about 40%) and
death;
men, 0.64 for women;
Netherlands,
85; urban general
diet history
consumers; for
58 deaths
no relation of fish
prospective
practitioners list
latter, mean: 24
intake to all causes
epidemiologic
g/d (21.6 g/d
death (RR = 0.96)
(Kromhout et
lean fish)
al., 1995)
3- Swedish
1967-68
14 years
10,966 men and
women ages 40-70;
self-
administered
none (few
people) + low
CHD and
MI death;
CHD: age-sex adjusted
RR (95% CI) -- 1.00,
0.94 (0.83-1.06), 0.85
(0.69-1.06); MI 1.00,
0.91 (0.76-1.08), 0.70
Twins,
nationwide
questionnaire
(12,315 p-y),
800 and
(0.50-0.98); results
prospective
population-based
moderate
395 deaths
similar with
epidemiologic
twin registry
(70,848 p-y),
multivariate
(Norell eta!.,
high (57,084 p-
adjustment, and for 2
1986)
y)
sexes separately
Toxicology Excellence for Risk Assessment 2-3 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Table 2-1. continued
Bergen, 1967 11,000 men, 65% postal Fish index CHD and CHD, MI (also all
Norway, 14 years ages 55+, 30% ages questionnaire, approximating MI death causes death): no
prospective 65+: urban including 3 no. of times fish before age relationship; for men
epidemiologic community general questions on eaten/month; 0-4 80; 967 ages <45 at entry,
(Norell eta]., population fish intake (642), 5-9 CHD inverse relationship,
1986) (2242), 10-14 deaths, 301 fish and CHD death, p
(4412), 15-19 inhealthy =0.058
(1726), 20-24 subcohort,
(1497), >25 22 in men
(482) ages <45 at
entiy
4- Honolulu 1965-68 7,615 Japanese- questionnaire from usual CHD no significant
Heart Program, 12 years American men ages on usual frequency incidence relationships
Hawaii, U.S.A., 45-68; urban frequency of questionnaire: and CHD
prospective community general eating various almost never death; nos.
epidemiologic population foods; 24- (32), <2 of events
(Curb et al., hour dietary times/wk (4143), not stated
1985) recall 2-4 times/wk
(2884), almost
daily (545),
>once/day (10);
from 24-h recall:
0 g/d (4232), 28-
56 (1374), 84-
112 (1092), 140-
168 (505), >168
( 412 ) =
Toxicology Excellence for Risk Assessment 2-4 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
Table 2-1. continued
5- Adventist 1976 26,473 non- 65-item semi- never (43%), Definite multivariate adjusted
Health Study, 6 years Hispanic white quantitative 1/wk (10%) MI, 134 1.66), 1.04 (0.55-
U.S.A., Seventh-Day frequency events; 1.96); 1.00, 1.01
prospective Adventists ages questionnaire Definite (0.76-1.35), 0.74
epidemiologic 25+, 10,003 men, mailed to Fatal CHD (0.42-1.33); 1 00. 1.10
(Fraser et al., mean age 51, cohort (clinical, (0.89-1.37), 1.09
1992) 16,470 women, autopsy- (0.73-1.61)
mean age 53 based), 260
deaths;
Fatal CHD
(death
certificate),
463 deaths
6- Physician’s 1982 21,185 U.S. male self- <1 meal/wk nonfatal multivariate adjusted
Health Study, 4 years physicians administered (4,501), 1 MI and all RR, nonfatal MI: 1.0,
U.S.A., nationwide, ages semi- (8,156), 2-4 MI; 259 1.4, 1.2, 0.8; all MI
prospective 40-84, free of quantitative (7,455), >5 and 281 1.0, 1.5, 1.3, 0.9 --
epidemiologic history of major food (1,073) events (i.e., nonsignificant; p for
(Morris et al., disease frequency 22 MI trend 0.78, 0.72
1995) questionnaire deaths)
Toxicology Excellence for Risk Assessment 2-5 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Table 2-1. continued
12 years
1986
6 years
7- Physician’s
Health Study,
U.S.A., (Albert
eta]., 1998)
8- Health
Professionals
Follow-Up
Study, U.S.A.,
prospective
epidemiology
(Ascherio et al.,
1995)
20,551 U.S. male
physicians from
above cohort
44,895 male health
professionals
nationwide, ages
40-75, free of
known CVD
same
self-
administered
semi-
quantitative
food
frequency
questionnaire
<1 meallmonth
and 1÷; also, <1,
1-3, 1-<2/wk, 2-
<5/wk, 5+/wk
servings of fish
<1/month, 0 g/d
(2,042); 1-
3/month, 7 g/d
(3,314); 1/wk,
18 g/d (12,296);
2-3/wk, 37 g/d
(16,920); 4-
5/wk, 69 g/d
(6,271); >6/wk,
119g/d (4,052)
sudden
death
(within 1
hour of
symptom
onset); 115
deaths
fatal CHD
(264),
nonfatal
MI (554),
any MI
(811),
CABG
(735), any
CHD
(including
CABG)
(1543)
<1 vs. 1+ meallmonth:
unadjusted RR 0.44
(0.22-0.9 1), p 0.03;
multivariate adjusted
RRO.51 (O.25-l.O5),p
0.07; age-adjusted RR
for 5 strata: 1.00, 0.71
(0.29-1.77), 0.44
(0.20-0.94), 0.43
(0.15-0.98), 0.39
(0.15-0.98)
for fatal CHD,
multivariate adjusted
RR 1.00, 0.74, 0.86,
0.71, 0.54, 0.77, all
95% CIs include 1.00;
for RR 0.54, 95% CI
0.29-1.00; p for trend
0.14; for CABG, 1.0,
1.31, 1.43, 1.40, 1.71,
1.65; all 95% CIs
include 1.00; p for
trend 0.02; for other
end points, no
significant
relationship
Toxicology Excellence for Risk Assessment
2-6
8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Table 2-1. continued
Jan. 1985-
Feb. 1989
9- Northern Italy
hospital-based
case-control
study (Gramenzi
eta]., 1990)
287 women ages
22-69 with MI
identified in CCUs
of 30 hospitals; 649
controls ages 21-
69, in-hospital with
acute disorders
other than CHD
interview
with use of
structured
questionnaire
on frequency
of
consumption
of individual
foods and
beverages
tertiles of
portions
consumed per
week: <1, 1, >1-
- for cases: 148,
81. 58; for
controls: 270,
220, 159
nonfatal
MI
age-adjusted odds
ratio: 1.0, 0.7, 0.6, p
<0.05
Oct. 1988 to
July 1994
10- Seattle and
suburban King
Co., Washington
case-control
study (Siscovick
eta]., 1995)
334 men and
women with
primary cardiac
arrest, ages 25-74,
mean age 59, 80%
men, 493
population-based
controls, age-sex
matched
primary
cardiac
arrest
quantitative
food
frequency
questionnaire,
including 25
fish and 10
shellfish,
consumption
during prior
month;
spouse as
proxy
respondent
multivariate adjusted
odds ratio (OR): 1.0,
0.9 (0.8-1.0), 0.7 (0.6-
0.9), 0.5 (0.4-0.8), 0.4
(0.2-0.7)
no seafood
intake, and
quartiles of
long-chain n-3
polyunsaturated
fat intake,
equivalent to 0
servings/month
of fresh salmon
(0 g/m), 0.6
(15g), 2.0 (45g),
3.7 (84g), 9.1
(207g)
Toxicology Excellence for Risk Assessment
2-7
8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
Table 2-1. continued
11-DART 2 years 2,033 men <70 dietary 1015 men, nonfatal + for CHD incidence end
secondary years old, counseling randomly fatal CHD, point, multivariate
prevention recovered from assigned, asked fatal CHD adjusted RR 0.84, 95%
randomized acute MI, to consume at CI 0.66-1.07; for CHD
controlled trial, diagnosed at 21 least 2 portions death, unadjusted RR =
Wales, U.K., hospitals, randomly per week (200- 0.68, p <0.01; for all
factorial design assigned to 1 of 8 400 g) of fatty causes death, 0.71
(Burr eta]., groups, with no fish, plus advice (0.54-0.93), p <0.05
1989) dietary advice, fat on 2, 1, or no
modification, other dietary
increased fiber factors; intake of
intake from cereals, eicosapentanoic
increased intake of acid estimated
fatty fish for 2 groups to
be 2.3&0.7
g/week
12- Chicago 1957-58 1,822 men ages 40- in-depth cross- 0 g/d (189), 1-17 Fatal CHD, significant independent
inverse relationships;
Western Electric 30 years 55; urban workers check diet (646), 18-34 (430), fatal for fatal CHD, RR
Study, U.S.A.
(Daviglus eta!., history (745), a 35 MI (293),
1997) (245) nonsudden 1.00, 0.88, 0.84, 0.62, p
for trend 0.040; for
death (> 12 fatal MI, RR 1.00,
hours of
0.88, 0.76, 0.56 p for
symptom trend 0.017; for
onset)
nonsudden MI, RR
1.00, 1.04, 0.76, 0.33, p
for trend 0.007
Toxicology Excellence for Risk Assessment 2-8 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
2 2.1.1 Kromhout, Bosschieter and Coulander, 1985. “The Zutphen Study” (1)
This was the first prospective study to create great interest in the topic of fish consumption and
CHD risk reduction. The investigators followed 852 men for 20 years: there were 78 CHD
deaths with 100% follow-up. The results showed an inverse relationship between the amount of
fish reported and the relative risk of death from CHD (p
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
classified as high, moderate or low fish consumers based on ratios of average amounts of fish
eaten in relation to other foods. Individuals who reported that they did not eat fish were included
in the “low consumption” group, a limitation of this study. By combining the “unexposed”
subjects and “low exposure” subjects into one statistical group, this “may have led to an
underestimation of the strength of the inverse relation between fish consumption and death from
MI and CHD” (p.426).
The results showed that 800 people died of coronary heart disease, 395 of whom had fatal
myocardial infarctions. When the relative risk of death from each was calculated, controlling for
gender and age, there was a dose-response relationship: those in the highest category of fish
consumption had a relative risk of .85 for CHD death, and .70 for MI death when compared to
the low/no fish group. The investigators report that when the data were controlled using
multivariate analysis techniques the results were similar.
2.2.1.4 Vollset, Heuch and Bjelke, 1985. “Norway Postal Dietary Survey” (4)
In a letter to the editor of the New England Journal of Medicine, Voilset and his colleagues
reported findings from a subset of 11,000 middle-aged and older men from their Norwegian
prospective study, who were followed for 14 years. These men reported their smoking and
“selected cardiovascular symptoms” in 1964, and their fish consumption in 1967. The
investigators used the latter information to construct “a fish index approximating the number of
times fish was eaten per month” (p.820). Overall, there were 967 deaths, 301 of which were due
to fatal myocardial infarctions. The distribution of deaths observed, according to six categories
of number of fish meals per month, was not significantly different from the predicted
distribution. When the analyses excluded men who had reported CHD symptoms (such as
angina) in 1964 and used only data from the 301 who died from acute MIs, there was still no
significant relation between fish consumption and CHD death. However, the analyses did not
control for a number of possibly important confounding factors (e.g., level of physical activity
level, body fatness, high blood pressure, parental history of early death from CHD, diabetes
mellitus, etc.).
2.2.1.5 Curb and Reed, 1985. “Honolulu Heart Program” (5)
The NEJM letter by Voilset and colleagues was followed by another letter to the editor reporting
the relationship between fish consumption and CHD. Curb and Reed provided the results of a
12-year study following 7,615 Japanese men participating in the Honolulu Heart Program. The
ages of the men upon enrolment into the study were not given, but the men were “without
prevalent atherosclerotic disease”. The results showed few differences in the rates of CHD and
of fatal CHD across the categories of frequency of fish meals and amounts eaten in the previous
24 hours, and there were no statistically significant trends. The authors did not mention if they
controlled for other risk factors for CHD in their analyses. Given that this population had high
mean levels of fish consumption (--44% reported eating fish during the previous 24 hrs, with
portion sizes ranging from 28 to 476 grams--) with few subjects reporting that they never ate fish
(n=32) compared to other similar studies, the authors wondered if “maximal benefit” from fish
consumption was already being obtained.
Toxicology Excellence for Risk Assessment 2-10 8/6/99
-------
Coo:krative Agreement with U S EPA on Comparative Dietary Risk
2.2.1.6 Fraser, Sabate, Beeson, and Strahan, 1992. “The Adventist Health Study” (6)
A cohort of 26,473 non-Hispanic Caucasian Seventh Day Adventists aged 25 years or older in
1974, was followed for 6 years. The sample was well educated and concerned about health;
participants tended to abstain from smoking and the use of alcohol, and follow lacto-ovo-
vegetarian diets (i.e., diets that include dairy products and eggs, but exclude meat, fish and
poultry). The criteria for participation in this study were stricter than most previously conducted:
those with a known history of heart disease, or whose history for CHD had not been assessed, as
well as individuals with diabetes, were excluded. Baseline data were collected using a mailed
self-administered questionnaire in 1976, and unlike other prospective studies, yearly data on
hospital admissions (with access to medical records) and the development of CHD symptoms
were also collected. Death due to CHD was strictly defined and three end-points were used:
definite nonfatal MI (134 events), definite fatal MI (260 cases), and confirmed fatal CHD (463
cases).
The adjusted relative risks from multivariate analyses for three levels of fish consumption (none,
less than once/week, once a week or greater) suggested protective effects for those consumers
eating fish once a week or more, only for death from MI (adjusted RR = 0.74; 95% C.I = 0.42-
1.33). The group eating fish less than once a week had similar adjusted relative risks to the
group eating no fish, for all 3 end-points. Given that only 10% of the largely vegetarian sample
ate fish once a week or more, and that the sample was followed for only 6 years and was still
relatively young (mean age of 51.3 years for men and 53.2 years for women), these effects may
have been underestimated.
2.2.1.7 Morris, Manson, Rosner, Buring, Willett and Hennekens, 1995. “The US
Physicians’ Health Study: 4 years” (7)
This prospective study followed 21,185 U.S. physicians aged 40-84 years, with no history of MI
or other cardiovascular disease, cancer, liver or renal disease, peptic ulcer, gout or use of certain
drugs (including aspirin), from 1983 to 1987. As with the previous study, information on CHD
symptoms was collected annually. What is different from other prospective studies, however, is
that annual information was also collected on fish consumption, foods high in saturated fats, and
parental MI events; as well, use of aspirin and beta-carotene were controlled for in analyses.
Several CHD endpoints were used, along with all stroke events, and these were confirmed by
ECG and enzyme test records, other patient records or autopsy. The relative risk values for all
cardiovascular and stroke events (fatal and nonfatal) for subjects who ate 1 meal, 2-4 meals, and
5 or more fish meals per week were not significantly different from values for subjects who ate
less than 1 meal of fish weekly, and there were no statistically significant trends across the four
levels of fish consumption. The number of subjects at each level of fish consumption for each of
the cardiovascular disease outcomes was relatively small, though (e.g., stroke and cardiovascular
deaths each had only 7 subjects who consumed 5 or more fish meals per week), and some
confidence intervals were wide (e.g., 95% CI = 0.8-5.9). As with the Swedish Twins Study
above, grouping low consumers with consumers who never ate fish may have underestimated the
benefits of eating modest amounts of fish (one or two meals weekly, such as in the Zutphen
Study). As well, four years is a relatively short period of follow-up.
Toxicology Excellence for R sk Assessment 2-11 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
2.2.1.8 Albert, Hennekens, O’Donnell, Ajani, Carey, Willett, Ruskin and Manson, 1998.
“The US Physicians’ Health Study: 12 years” (8)
This was a continuation of the previous study, and examined the association between fish
consumption and the risk of sudden cardiac death in men over an 11-year period. Eighty percent
of the sample consumed fish between 1 and 4 times/week; the high fish consumers tended to be
those who were at risk for cardiovascular disease (e.g. family history of CHD) and, being
physicians, were aware of this risk. After controlling for age, and aspirin and beta carotene use,
risk of sudden cardiac death was inversely related to fish consumption, and showed a significant
decline (p<.O5) across the five levels of fish consumption. Physicians who ate 1-2 fish
meals/week had a significantly lower risk (RR=0.42, p=O.O2) of sudden death compared to those
who ate fish <1/month. The magnitude of difference in risk did not change significantly with
higher consumption, suggesting a threshold effect. After adjusting for coronary risk factors and
prior cardiovascular disease, the decline in risk was no longer significant across the five
categories of fish consumption but remained significant across three of the categories (<1/month,
1-3/month, and 1/week). Although fish consumption was inversely associated with sudden
cardiac death, it was not related to non-sudden cardiac death, risk of coronary heart disease or
total cardiovascular death.
2.2.1.9 Ascherio, Rimm, Stampfer, Giovannucci and Willett, 1995. “The Health
Professionals Follow-Up Survey” (9)
This six year prospective study followed a cohort of 44,895 male health professionals who were
aged 40 to 75 in 1986. Men who reported MI, angina, stroke, transient ischemic attack,
peripheral artery disease, coronary artery surgery, diabetes, high blood pressure, high blood
cholesterol, or who knew their blood cholesterol at baseline (and thus may have altered their
lifestyles or diets to reduce their ‘high risk’), were excluded from analyses, as were men who had
CHD events during the first 4 years of follow-up. Questionnaires were sent every 2 years to ask
for recent information on CHD events. Follow-up was complete for 94% of subjects. Endpoints
were fatal CHD, nonfatal MI, coronary-artery bypass grafting (CABG) and angioplasty, and
were confirmed using international criteria, patient records (ECG & cardiac enzyme results), and
autopsy in addition to death certificates.
This study assessed the association of both fish intake and n-3 FA intake in relation to CHD
endpoints. For analyses, fish intake was divided into 6 categories ranging from less than once
per month (n 2,042) to 6 or more times weekly (n=4,052) with sufficientnumbers of subjects in
each category for multivariate analyses by CHD endpoints. No association was found between
n-3 FA intake or fish oil supplements and risk of CHD disease. Compared with men who ate
little or no fish (less than 1 serving per month), the relative risk of fatal CHD (after adjusting for
many potential confounders) for men who ate fish 1 to 3 times a month was 0.74. This relative
risk remained fairly constant as fish consumption increased (ranging from 0.86 to 1 meallweek to
0.54 for 4-5 meals/week) but confIdence levels were very wide. These results suggest that as
little as one fish meal a week could have a protective effect against death from CHD, and
increasing fish consumption above this would likely not confer additional benefits. For other
CHD endpoints, there were no apparent relationships with fish consumption, and the authors
Toxicology Excellence for Risk Assessment 2-12 8/6/99
-------
Coop* rative Agreement with U S. EPA on Comparative Dietary Risk
suggested that eating fish might reduce the likelihood of death from a myocardial infarction but
not reduce the risk of a MI event.
2.2.1.10 Gramenzi, Gentile, Fasoli, Negri, Parazzine and La Vecchia, 1990. (10)
This retrospective case control study matched 287 Northern Italian women aged 22 to 69 years
who had suffered acute myocardial infarctions, with 649 controls hospitalised for conditions
unrelated to CHD, cancer, smoking, alcohol, or digestive, hormonal, or reproductive disorders.
The investigators relied upon participants’ assessments of their level of fish consumption (low,
intermediate, high consumption). Age adjusted odds ratios using ‘less than 1 fish meal weekly’
as the reference catego:y showed that the risk of MI was 0.7 for those eating 1 fish meal per
week, and dropped to 0.6 for more than 1 fish meal weekly; this trend was statistically significant
at p
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
or eat more dietary fiber. Those in the fish group who disliked fish were given 3 fish oil capsules
daily instead. A subset of 25 study participants completed 7 day weighed food records, and
subjects in the fat and fish advice groups had serum cholesterol and fatty acid profiles done, as
cross-checks for compliance. After two years, the groups receiving fat and fiber advice showed
no significant differences in death rates. Although there is a confounding of fish consumption
with fish oil consumption, the fish advice group showed a 29% lower risk of death from all
causes than the group which received no fish advice, even after analyses were adjusted for 10
potentially confounding factors. This was the first randomised, controlled trial to investigate the
effectiveness of increased fish consumption on the secondary prevention of MI.
2.2.1.13 Daviglus et al., 1997. “The Western Electric Study” (13)
The relationship between baseline fish consumption and the 30-year risk of CHD was assessed in
1822 men aged 40 to 55 years, as part of the Chicago Western Electric Study. Fish consumption
was determined from a detailed diet history and was stratified into 4 categories: 0, 1-17 g/day,
18-34 g/day, and >35 glday. Annual examinations, conducted for the first 10 years, and mailed
questionnaires or telephone interviews done over the next 15 years, were used to obtain
information on the status of the study participants. Dietary information was collected at the first
and second annual examination. During the 31st year vital status was determined from the
National Death Index, the Health Care Financing Administration, and surviving participants.
Deaths due to CHD were classified as death from MI (sudden or nonsudden) or death from other
coronary causes. Cox proportional hazards regression was used to estimate the RR of death for
each of the four levels of fisl consumption, after controlling for 13 possible confounders. Age-
adjusted death rates from MI, coronary heart disease, cardiovascular disease, and all causes were
the lowest in men who had the highest consumption of fish. The relative risks of death from any
MI (sudden or nonsudden), sudden MI, and nonsudden MI were 0.56, 0.68, and 0.33 respectively
for men who consumed >35 g of fish per day, compared to the group who consumed no fish.
There was a significant trend towards a lower relative risk as the level of fish in the diet for
nonsudden MI (p=O.0O7) and all CHD (p=O.O1O) but not for sudden MI. The results of this study
indicated a significant inverse relationship between fish consumption and 30-year risk of death
from coronary heart disease, including nonsudden MI.
2.2.1.14 Conclusions and Weight of Evidence for an Association between Coronary Heart
Disease and Fish Consumption -
The 13 studies chosen for this review included 1 clinical (randomised, controlled) trial and 12
epidemiological studies (2 case-control and 10 prospective cohort studies) with strong designs
and large sample sizes. All the prospective studies had good rates of follow-up (i.e., they had
few study dropouts), although there were large differences in the lengths of follow-up. Earlier
studies relied upon information on death certificates: later studies used international guidelines
for diagnostic criteria, autopsies, physician and hospital records (ECG results, cardiac enzymes,
blood lipid profiles), and interviews with relatives, to improve the accuracy of CHD endpoints
diagnosed. Later studies also expanded their exclusion criteria and did not recruit subjects with
pre-existing or early symptoms of CHD (e.g., angina pectoris, angioplasty procedures, coronary
artery bypass grafting, ischemic stroke, strokes due to injury or tumors, silent infarctions, etc.).
More recent prospective studies have collected information during the follow-up period, not just
Toxicology Excellence for Risk Assessment 2-14 8/6/99
-------
Coop ’rative Agreement with U S EPA on Comparative Dietary Risk
at baseline, to determine if and when health conditions, family history, or diet changes. They
also controlled for substances and lifestyle factors believed to lower CHD risk, such as aspirin,
anti-oxidant vitamins, alcohol, and regular exercise, in addition to other potential confounding
health and socio-demographic variables. Dietary assessment methods have improved as well,
with the use of instruments with known validity and reliability, and crosschecks for assessing the
accuracy of recalls or compliance with dietary advice.
Findings from five (of eight) prospective population studies B Zutphen (Kromhout eta]., 1985),
Rotterdam (Kromhout, 1995), Sweden (Norell et al., 1985), U.S. physicians 12-year study
(Albert eta]., 1998), and Chicago Western Electric (Daviglus eta!., 1997) are broadly
concordant in showing a significant inverse relation between fish intake and risk of CHD
mortality, as are also results from the two case-control studies (Gramenzi et al., 1990, Siscovick
eta!., 1995) and the one intervention trial (Burr eta]., 1989). For the specific endpoints of non-
sudden vs. sudden CHD death, findings in the Physicians’ Health Study (Albert et a!., 1998) and
the Seattle Study (Siscovick eta!., 1995), appear different from those in the Chicago Western
Electric Study (Daviglus eta!., 1997). Both of the former found an association between reported
fish intake and rate of sudden death, whereas the Chicago Western Electric data found a lower
rate of non-sudden MI death (not sudden MI death).
Three prospective studies, the Bergen Norway (Vollset eta]., 1985), Hawaii (Curb & Reed,
1985), and U.S. Health Professionals (Ascherio eta]., 1995) studies, did not find a relationship
between fish intake and CHD-MI. There were many differences among the prospective
epidemilogic investigations, one or more of which may account for this apparent inconsistency in
results:
1. different methods to assess diet and to array men by fish intake;
2. different distributions of reported fish intake, such that in some cohorts (e.g., Bergen and
Hawaii) there were few or no people in the group consuming little or no fish; hence, there
was no fully suitable reference group;
3. different study sites and times, with populations that have quite different diets (resulting
in different dietary intakes of cholesterol, saturated fats, antioxidants, fiber, etc.). For
example, the cohorts of Norwegian, Hawaiian Japanese-American and U.S. health
professionals were studied during the 1980s and 1990s, at a time of widespread
awareness of general dietary advice about ‘heart healthy eating’, as well as some
awareness of the idea that fish may “protect” against CHD;
4. the possibility of bias due to this recent awareness, i.e., people who know they have some
risk factors for CHD may have differentially become greater fish eaters in the 1980s. This
could potentially cause an inversion of the fish-CHD relationship, i.e., the people who eat
more fish have higher relative risks of death from CHD (e.g., the Health Professionals
Follow-Up Study finding re: fish and CABG?);
Toxicology Excellence for Risk Assessment 2-15 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
5. different durations of follow-up, ranging from decades to 4-6 years (e.g., findings in
Physicians’ Health Study with 4- and with 12-year follow-up, compared to findings in the
Chicago Western Electric Study with 30 year follow-up -- Table 2-1);
6. different CHD endpoints, with only a few studies reporting on fish and fatal MI, and only
two prospective studies (Chicago Western Electric Study and Physicians’ Health Study)
reporting on fish and the suddenness of CHD death;
7. differences in the interpretation of findings (e.g., in the Health Professionals Follow-up
Study, are data on CHD death more soundly interpreted as indicating no relation or an
inverse relation of fish intake to this end point?);
8. chance, that is, random variation across studies in results.
Although the data available at this time do not allow us to say definitively that these factors
account for the apparent discrepancies in findings, the ‘weight of evidence’ supports an
association between fish consumption and lower risk of CHD in men with no previous history or
symptoms of CHD. It appears that this relation is evident with as little as one to two meals of
fish (lean or fatty) per week.
This inverse relation seems unlikely to be due to n-3 FA content of fish, for several reasons.
First, lean fish have lower levels of n-3 FA than fatty fish, yet both appear to protect against
sudden cardiac death. Second, fish oil supplementation trials have observed effects at high
doses, the equivalent of enormous amounts of fish in the diet; however, there is evidence of an
association between a reduced risk of sudden death from CHD and MI when as little as one meal
of fish a week is consumed. It is possible that a component or combination of substances in fish,
through some mechanism not yet discovered, confers the cardio-protective benefits.
With the exception of the DART study (Burr et al., 1989), a randomised controlled clinical trial
which examined the effects of three different interventions on the secondary prevention of MI,
studies have not been designed to suggest a ‘cause and effect’ relationship between eating fish
and rates of CHD. These epidemiological studies can indicate associations between variables,
but randomised controlled clinical trials are needed in order to answer the question, ‘Does eating
fish lead to a lower risk of CHD?’
2.2.2 Studies of Other Possible Health Effects of Fish Consumption
The following sections briefly describe investigations of benefits of fish consumption with
respect to other health endpoints (Table 2-2). These endpoints have not been studied as well as
coronary heart disease. Table 2-2 summarizes this research.
Toxicology Excellence for Risk Assessment 2-16 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Table 2-2. Studies of Fish Consumption and Other Endpoints
Condition Key Studies Evidence to Date
Smoking-Related Chronic ARIC Study (Shahar et a)., fish consumption associated with better lung function for
Obstructive Pulmonary Disease 1994) Whites who currently or formerly smoked, but not Blacks
Lung Damage from Smoking ARIC Study (Shahar et al frequent fish consumption may protect the lungs of 35+ yr.
1994) and Honolulu Heart smokers from damage; smokers of 30+ cigarettes daily not
Program (Sharp et a)., 1994) protected though
Rheumatoid Arthritis Shapiro et al., 1996 2 or more servings of broiledlbaked fish/week reduced risk
of rheumatoid arthritis (OR=0.57); no association for other
types of fish (fried, shellfish, canned tuna)
Hodge etaL, 1996 (fish)
Childhood Asthma Hodge eta), 1998 (n-3 oil) 1996 study of 71 children with asthma found 1+ meals of
Thien 1996 oily fish/month reduced risk of asthma (OR=0.26); not
supported by Nurses’ Health Study (Thien, 1996); 1998
study using n-3 oil found no effect on seventy of asthma
Collier eta)., 1993
Plaque Psoriasis small clinical trial (cross-over design) with diets including 6
oz oily or 6 oz lean fish/day; 11-15% improvement of
psonasis symptoms with oily fish
Nurses Health Study (Willett
Colon Cancer eta)., 1990) no association with fish alone, but ratio of>5.2 red
meat:fish + chicken was 2.5 x more likely to get colon
cancer than ratio of <1.2 red meat: fish ÷ chicken
[ reviewed in O’Keefe, 1996]
Gastrointestinal Disease ulcerative colitis and adenomatous polyps successfully
treated with fish oil supplements; no data re: intake of fish
and incidence/severity
Toxicology Excellence for Risk Assessment 2-17 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Condition Key Studies Evidence to Date
Dyslipidemia in Non-Insulin- Dunston etal., 1997 NIDDM patients who ate 1 fish meal/day as part of a low
Dependent Diabetes Mellitus fat diet had reduced triglyercide and increased HDL levels,
but poorer glycemic control; when exercise was added,
glycemic control was maintained
Antioxidant Levels Anttolainen et a!., 1996 comparison of 82 Finns with very high fish consumption to
group eating <1 meal/month showed similar levels of
aritioxidants (beta-carotene & vitamin E)
Toxicology Excellence for Risk Assessment 2-18 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
2 2.2.1 Smoking- Related Chronic Obstructive Pulmonary Disease (COPD)
Shahar et al. (1994) looked at COPD in current and former smokers participating in the
Atherosclerosis Risk in Communities (ARIC) study. They defined COPD as chronic bronchitis,
physician-diagnosed emphysema, or spirometrically detected COPD. Current or former smokers
in the third and fourth quartiles of fish consumption ( 1.5 servings/week) were significantly less
likely to have COPD than those who ate little or no fish. Fish consumption was also associated
with improved lung function (greater forced expiratory volume in one second [ FEy 1 ] and greater
FEV 1 fForced vital capacity LFVC]) in white current and former smokers. Oddly, this
relationship was not seen in black participants. Fish consumption was not related to lung
function in ARIC participants who had never smoked, suggesting that fish consumption does not
actually improve lung function directly, but rather protects the lungs from damage caused by
smoking.
2.2.2.2 Lung Damage from Smoking
Like the results of the A RIC study summarized above, the results of the Honolulu Heart Program
(HHP) study (Sharp et al., 1994) suggest that frequent fish consumption may protect the lungs of
long-term cigarette smokers from damage. Male cigarette smokers who reported eating fish at
least 2 times/week showed less of a relationship between duration of tobacco exposure and
reduced FEV 1 than did smokers who ate fish less frequently. However, fish consumption only
seemed to be protective among smokers who had smoked for more than 35 years. There was
also some evidence that heavy smokers (>30 cigarettes/day) were not protected by frequent fish
consumption.
The authors of both the ARIC and the HHP studies speculate that fish consumption may exert its
protective effect by inhibiting the production of various mediators of lung inflammation, some of
which have been associated with cigarette smoking. Supplementation of the diet with fish oil has
been shown to inhibit the production of a number of known and putative mediators of lung
inflammation, and fish itself might produce similar effects. However, the average intake of n-3
fatty acids by study participants was much lower than the dosage contained in fish oil
supplements, so some factor other than n-3 fatty acids may be at work.
2.2.2.3 Rheumatoid Arthritis
In a case-control study by Shapiro eta]. (1996), consumption of broiled or baked fish was
associated with a decreased risk of rheumatoid arthritis in women. The adjusted OR for >2
servings of broiled or baked fish/week, compared with <1 serving, was 0.57 (95%CI=0.35-0.93).
However, consumption of other types of fish (fried fish, canned tuna, and shellfish) was not
associated with rheumatoid arthritis, and neither was a combined measure including all types of
fish consumed. In addition, there was no significant association between the estimated amounts
of n-3 fatty acids in the participants’ diets and rheumatoid arthritis when all cases were included
in the analysis. When the analysis was restricted to cases who were rheumatoid factor positive
(RF÷), the association with consumption of broiled or baked fish remained significant, and there
was also an association between n-3 fatty acid consumption and reduced risk of arthritis.
Toxicology Excellence for Risk Assessment 2-19 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
However, the latter association appeared to be weaker than the former: the association between
n-3 fatty acid consumption and rheumatoid arthritis was only statistically significant when
participants in the top 10% of n-3 consumption were compared to those in the first quartile.
2.2.2.4 Childhood Asthma
In a study of 71 children with asthma and 263 controls, Hodge eta!. (1996) found that
consumption of fresh, oily fish was associated with a reduced risk of current asthma. The
adjusted OR for those who ate oily fish at least monthly versus those who never ate oily fish was
0.26, with a 95% C.I. of (0.09-0.72). There was no statistically significant association between
asthma and consumption of non-oily fish or of canned or processed fish. .Thien et al. (1996)
were somewhat dubious about these results, since other studies had suggested that fish in the diet
would have no effect on risk of asthma or on reducing the severity of asthma. For instance, the
Nurses’ Health Study had shown no relationship between adult-onset asthma and fish in the diet,
and most clinical trials of fish oil supplements as a treatment for asthma had yielded
disappointing results. In fact, a small clinical trial by Hodge eta!. (1998) showed that n-3 oil
supplements had no effect on the severity of symptoms in asthmatic children. However, it is
possible that the association observed in the 1996 study was genuine, and that some compound in
fish other than n-3 fatty acids was responsible for the protective effect.
2.2.2.5 Plaque Psoriasis
Collier et a]. (1993) carried out a small clinical trial to examine whether a diet containing 6 oz. of
oily fish/day could improve psoriasis symptoms: the control diet contained 6 oz. of white
fish/day. Patients on the oily fish diet showed a small (between 11% and 15%) but statistically
significant improvement after 6 weeks on the diet, and symptoms worsened again when patients
switched from the oily fish diet to the white fish diet. The oily fish diet contained high levels of
both vitamin D and n-3 fatty acids, and the authors speculated that one or both of these factors
might have been responsible for the beneficial effect of the diet.
2.2.2.6 Colon Cancer
There was no significant relationship between fish consumption and colon cancer among the
Nurses’ Health Study cohort (Willett eta!., 1990). However, the total amount of chicken and
fish in the diet was associated with a reduced risk of colon cancer, and the’ratio of red meat to
chicken and fish in the diet was associated with an increased risk of colon cancer. Women who
ate 5.2 times more red meat than chicken and fish (the highest quintile) were 2.5 times more
likely to have colon cancer than women in the lowest quintile (< 1.2 times more red meat than
chicken and fish). In this case, it seems likely that fish and chicken exerted their protective
effect by substituting for red meat, thus decreasing the amount of animal fat and other potentially
hazardous components of red meat in the diet.
2.2.2.7 Gastrointestinal Disease
O’Keefe (1996) cites a number of studies in which fish oil supplements were successfully used to
treat patients with ulcerative colitis or adenomatous polyps. However, there seem to be no
Toxicology Excellence for Risk Assessment 2-20 8/6/99
-------
Cooj erative Agreement with U S EPA on Comparative Dietary Risk
published studies of the effect of fish in the diet on the incidence or severity of colitis or other
gastrointestinal disordei s.
2.2.2.8 Dyslipidemia in Non-Insulin-Dependent Diabetes Mellitus
Dunston etal. (1997) conducted a small clinical trial to test whether fish consumption (1
meallday) could improve serum lipid levels in dyslipidemic NIDDM patients without increasing
levels of plasma glucose and glycated hemoglobin. The addition of one fish meal per day to a
low-fat diet k 30% of energy intake) did cause triglyceride levels to fall and HDL 2 levels to rise,
but it also caused glycated hemoglobin and serum glucose to rise. However, in patients who ate
one fish meal per day and participated in an exercise program (riding a stationary bicycle for 30’
at 55-65% Of V 02 m ), plasma lipid levels improved with no deterioration in glycemic control.
2.3 Antioxidant Levels
Several researchers have speculated that a high-fish diet might reduce antioxidant levels in the
diet or in the blood, since n-3 fatty acids have a tendency to oxidize in vitro, and since fish oil
contains less of the antioxidant vitamin E than vegetable oils. Anttolainen et al. (1996) studied
82 Finns to determine whether men and women with very high levels of fish consumption do in
fact have lower levels of antioxidants in their diet and in their plasma than people who eat little
or no fish. When compared to participants who ate fish less than once per month, participants
who ate fish every day or almost every day had equivalent or higher levels of vitamin E, vitamin
C, 13-carotene, and selenium in their diets. Similarly, levels of plasma tocopherol (vitamin E) and
B-carotene were very similar in the high-fish and low-fish groups. Although this study could not
measure the vitamin E siored in tissue, the results do suggest that a diet high in fish does not
dangerously reduce antioxidant levels.
2.4 Health Benefits During Pregnancy, Lactation and Infancy
As mentioned earlier, fish is a good dietary source of the omega-3 fatty acids docosahexaenoic
acid (DHA) and eicosapentaenoic acid (EPA). DHA is incorporated into the cell membranes of
the retina, brain and other parts of the central nervous system. Several animal studies have
demonstrated that if DHA levels are low when these tissues are developing, vision and learning
problems may result (see Neuringer, Reisbick and Janowsky, 1994 for a review). This is most
important during the thu d trimester of pregnancy when the brain of the unborn child is rapidly
developing, but is also important during the first year after birth when the brain continues to
grow. Thus, there has been an interest in the possible health benefits of consuming fish during
pregnancy. The unborn child may benefit from the omega-3 fatty acids from fish the mother
consumes during pregnancy, and may be born with greater body stores of these fatty acids to
draw upon during the first year of life.
Human milk contains many long-chain polyunsaturated fatty acids, including DHA and
arachidonic acid (AA). DHA accounts for 0.1% to 1.5% of total fatty acids in human milk
depending upon the amount of pre-formed n-3 fatty acids, the main source of which is fish, in the
mother’s diet. Currently, commercial infant formulas in the U.S. contain the n-6 fatty acid
linoleic acid (precursor to arachidonic acid), and the n-3 fatty acid linolenic acid (precursor to
Toxicology Excellence for Risk Assessment 2-21 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
DHA & EPA), but do not contain AA, DHA or EPA (Auestad etal., 1997). Thus, mothers who
consume fish while provide pie-formed n-3 fatty acids to their infants through breast milk,
during the first year of life when the brain is still developing.
In the Faroe Islands study, women who ate high amounts of marine fish and marine animals
during their pregnancies were observed to have longer gestations and correspondingly heavier
babies (Olsen, Hansen and Sorensen, 1986). A subsequent study of Danish women hypothesized
that it was the n-3 fatty acids in marine fish and animals that conferred these benefits; indeed, the
results showed that women who received fish oil supplements in their third trimesters had
pregnancies that were an average of four days longer than women who received olive oil
supplements or no supplements (Olsen, eta!., 1992). Comparisons among the three groups of
mothers showed that those given fish oil supplements had significantly higher levels of n-3 fatty
acids, particularly docosahexaenoic acid (DHA), in their umbilical cord blood than the two other
groups (Van Houwelingen, Sorensen, Hornstra eta!., 1995). Thus, the extra n-3 fatty acids
consumed by the supplemented pregnant mothers led to higher n-3 fatty acid levels in their
babies at birth. The authors concluded that “it is, indeed, possible to interfere with the DHA
status at birth: children born to mothers supplemented with fish oil in the last trimester of
pregnancy start with a better DHA status at birth, which may be beneficial to neonatal
neurodevelopment” (Van Houwelingen et aL, 1995, p.723).
A study of 300 Canadian Inuit women provides some preliminary evidence that consuming
marine fish and mammals during pregnancy may reduce the likelihood of pregnancy-induced
hypertension for women at risk (Popeski eta!., 1991). Women from communities harvesting
large amounts of marine foods had significantly lower diastolic blood pressure levels in the last
six hours of their pregnancies, compared with women from communities where less marine fish
and sea animals, and more caribou, were eaten. Among the 53 women who developed
hypertension, 12 were from communities with high fish and sea mammal consumption and 41
from communities eating lower amounts. Although these differences in blood pressure levels
could reflect other dietary differences not assessed in this study (e.g., sodium intake), levels of
the n-3 fatty acids eicosapentaenoic acid (EPA) and DHA in cord blood were higher in the
communities with more marine foods in their diets, lending support to the hypothesis that the
differences were related to diet.
2.5 Health Benefits for Children Consuming Fish
All fish is a good source of high-quality protein, essential fatty acids, and minerals such as iron
and zinc (see Chapter 3 Nutritional Benefits of Eating Fish Compared to Other Protein Sources),
and thus is an important food in the diets of growing children. Consuming other protein foods,
particularly fast foods that are popular among children and teens, can result in more saturated
fats in a meal. Given the increasing rates of obesity among American children, and the evidence
that obese children tend to become obese adults, high dietary fat intakes among children are a
public health concern. Learning to enjoy a healthful diet and to maintain an appropriate body
weight during childhood may help children to reduce their risks of cardiovascular disease, Type
II diabetes mellitus, and some forms of diet-related cancers, as adults.
Toxicology Excellence for Risk Assessment 2-22 816/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
Some children and adolescents follow strict vegetarian (vegan) diets, or variations that include
dairy products and eggs (lacto-ovo-vegetarians), or dairy products, eggs and fish (lacto-ovo-
pesco-vegetarians or semi-vegetarians). A recent Slovakian study examined the fatty acid
profiles of children following such diets and children who were omnivores (Krajcovicova-
Kudlackova eta!., 1997). The children following semi-vegetarian, lacto-ovo-vegetarian, and
vegan diets had significantly lower blood levels of saturated fatty acids compared to the
omnivore children. The semi-vegetarians, who consumed fish but not poultry or meats, had
significantly higher levels of EPA and DHA than lacto-ovo-vegetarians; vegans had the lowest
levels of both n-3 fatty acids. There were no significant differences among the groups in the
long-chain n-6 AA or in monounsaturated fatty acids. The authors warned that the “significantly
reduced n-3 fatty acid content and significantly higher ratio n-61n-3 may represent a health risk
in vegans,” with respect to cardiovascular diseases. However, the high EPA and DHA levels and
low n-6/n--3 ratio found in semi-vegetarians, who reported eating an average of 1.9 fish meals
weekly, was considered important for the prevention of cardiovascular diseases.
While children in our culture seem to love hamburgers, hot dogs, and macaroni and cheese by an
early age, they may not appear to have the same preference for a lower-fat meal of baked, broiled
or steamed fish. Leann Birch (1996), an authority in children’s food acceptance patterns, notes
that some taste preferences are “built-in” and “unlearned,” infants have a preference for sweet
and salty tastes, and an aversion to sour and bitter tastes. Other food likes, such as for fatty,
crunchy, or creamy foods, get established through repeated experiences with the food.
Therefore, “early experiences have a profound effect on food preferences” and “repeated
opportunities to taste the food enhance food acceptance” (Birch, 1996; p.235). Children can
learn to accept and like fish (or any other new food) if they are introduced to it at an early age
and are given time to become familiar with its taste and texture. If children have ample
opportunities to become familiar with fish while they are young, chances are greater they will
learn to like it and continue eating it as an adult. Thus, if we are interested in promoting the
cardiovascular and other nutritionalThealth benefits of eating fish to adults, we should encourage
parents to serve fish to young children so they can develop a liking for it and a willingness to
include it in their diets at an early age.
2.6 Conclusions and Research Needs
The above data provide some evidence for an association between decreased risk of CHD or MI
and consumption of small amounts of fish, including mainly lean (non-fatty) fish. However, it
seems unlikely that decades-long intake of small amounts of fish protect, if fish is indeed
etiologically protective, via the very small amounts of omega-3 long-chain polyunsaturated fatty
acids so ingested. The resolution of this issue has important implications for public health and
nutritional recommendations. Thus, further studies -- observational and interventional,
particularly trials - - are needed to resolve whether there is an etiologically significant protection
against CHD or MI afforded by regular ingestion of modest amounts of fish. Similarly, more
research is needed on the relationship of fish intake and health endpoints other than CHD or MI.
Toxicology Excellence for Risk Assessment 2-23 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
2.7 References
Albert, C.M., J.E. Manson, C.J. O’Donnell, eta]. 1998. Fish consumption and risk of sudden
death iii the Physician’s Health Study. Circulation. 94: 3382 (abst).
American Heart Association. 1996. Fish consumption, fish oil, lipids, and coronary heart
disease. Circulation. 94: 2337-2340.
Anttolainen, M., L.M. Valsta, G. Alfthan, et al. 1996. Effect of extreme fish consumption on
dietary and plasma antioxidant levels and fatty acid composition. Eur. J. Clin. Nutr. 50: 741-
746.
Ascherio, A., E.B. Rimm, M.J. Stampfer, et al. 1995. Dietary intake of marine n-3 fatty acids,
fish intake, and the risk of coronary disease among men. N. Eng. J. Med. 332: 977-82.
Auestad, N., M.B. Montalto, R.T. Hall, eta]. 1997. Visual acuity, erythrocyte fatty acid
composition, and growth in term infants fed formulas with long chain polyunsaturated fatty acids
for one year. Pediatr. Res. 41: 1-10. -
Bang, H.O., J. Dyerberg, and A.B. Nielsen. 1971. Plasma lipid and lipoprotein pattern in
Greenlandic west-coast Eskimos. Lancet. 1: 1143-6.
Bang, H.O., J. Dyerberg, and H.M. Sinclair. 1980. The composition of the Eskimo food in
North Western Greenland. Am. J. Clin. Nutr. 33: 2657-266 1.
Birch, L.L. 1996. Children’s food acceptance patterns. Nutrition Today. 31: 234-240.
Burr, M.L., A.M. Fehily, J.F. Gilbert, et al. 1989. Effects of changes in fat, fish and fiber
intakes on death and myocardial reinfarction. Diet and Reinfarction Trial (DART). Lancet.
2(8666): 757-61.
Collier, P.M., A. Ursell, K. Zaremba, eta]. 1993. Effect of regular consumption of oily fish
compared with white fish on chronic plaque psoriasis. Eur. J. Clin. Nutr. 47: 25 1-254.
Curb, J.D. and D. Reed. 1985. Fish consumption and mortality from coronary heart disease
(letter). N. Eng.J. Med. 313: 821-2.
Daviglus M.L., J. Stamler, A.J. Orencia, et a]. 1997. Fish consumption and the 30-year risk of
fatal myocardial infarction. N. Eng. J. Med. 336: 1046-53.
Dunston, D.W., T.A. Mori, I.B. Puddey, et al. 1997. The independent and combined effects of
aerobic exercise and dietary fish intake on serum lipids and glycemic control in NIDDM.
Diabetes Care. 20: 913-921.
Toxicology Excellence for Risk Assessment 2-24 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Fraser, G.E., J. Sabate, W.L. Beeson, et al. 1992. A possible protective effect of nut
consumption on risk of coronary heart disease: The Adventist Health Study. Arch. Intern. Med.
152: 1416-24.
Gramenzi, A., A. Gentile, M. Fasoli, eta!. 1990. Association between certain foods and risk of
acute myocardial infarction in women. Br. Med. J. 300: 77 1-3.
Hodge, L., C.M. Salome, J.K. Peat, eta!. 1996. Consumption of oily fish and childhood asthma
risk. M.J.A. 164: 137-140.
Hodge, L., C.M. Salome, J.M. Hughes, eta). 1998. Effect of dietary intake of omega-3 and
omega-6 fatly acids on severity of asthma in children. Eur. Respir. J. 11: 361-365.
Kagawa, Y., M. Nishizawa, M. Suzuki, eta). 1982. Eicosapolyenoic acid of serum lipids of
Japanese islanders with low incidence of cardiovascular disease. J. Nutr. Sci. Vitaminol.
(Tokyo). 28: 441-53.
Keys, A. 1980. Seven Countries: a Multivariate Analysis of Death and Coronary Heart Disease.
Harvard University Press. Cambridge, MA.
Krajcovicova-Kudlackova, M., R. Simoncic, A. Bederova, et al. 1997. Plasma fatty acid profile
and alternative nutrition. Ann. Nutr. Metab. 41: 365-370.
Kromhout D, E.B. Bosschieter, and C. de Lezenne Coulander. 1985. The inverse relation
between fish consumption and 20-year mortality from coronary heart disease. N. EngI. J. Med.
312: 1205-9.
Kromhout D, E.J. Fesken, and C.H. Bowles. 1995. The protective effect of small amount of fish
on coronary heart disease mortality in elderly population. mt. J. Epidemiol. 24: 340-5.
Morris, M.C., J.E. Manson, B. Rosner, eta!. 1995. Fish consumption and cardiovascular
disease in the Physicians’ Health Study: A prospective study. Am. J. Epidemiol. 142: 166-75.
Neuringer, M., S. Reisbick, and J. Janowsky. 1994. The role of n-3 fatty acids in visual and
cognitive development: current evidence and methods of assessment. J. Pediatr. 125: S39-47.
Norell, S.E., A. Ahlbom, M. Feytching, eta). 1986. Fish consumption and mortality from
coronary heart disease (letter). Br. Med. J. 293: 426.
O’Keefe, S.J.D. 1996. I Jutrition and gastrointestinal disease. Scand. J. Gastroenterol. 220(31
Suppl): 52-59.
Olsen, S.F., H.S. Hansen, T.I.A. Sorensen, eta!. 1986. Intake of marine fat, rich in (n-3)-PUFA,
may increase birth weight by prolonging gestation. Lancet. 2(8503): 367-69.
Toxicology Excellence for Risk Assessment 2-25 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Olsen, S.F., J.D. Sorensen, N.J. Secher, eta]. 1992. Randomized controlled trial of effect of
fish-oil supplementation on pregnancy duration. Lancet. 339: 1003-07.
Popeski, D., L.R. Ebbeling, P.B. Brown, et al. 1991. Blood pressure during pregnancy in
Canadian Inuit: community differences related to diet. Can. Med. Assoc. J. 145: 445-454.
Shahar E., A.R. Folsom, S.L. Melnick, et a]. 1994. Dietary n-3 polyunsaturated fatty acids and
smoking-related chronic obstructive pulmonary disease. N. Eng. J. Med. 331(4): 228-233.
Shapiro, J.A., T.D. Koepsell, L.F. Voigt, eta]. 1996. Diet and rheumatoid arthritis in women: a
possible protective effect of fish consumption. Epidemiology. 7: 256-263.
Sharp, D.S., B.L. Rodriguez, E. Shahar, etal. 1994. Fish consumption may limit the damage of
smoking on the lung. Am. J. Respir. Critic. Care Med. 150: 983-987.
Simopoulos, A.P. 1991. Omega-3 fatty acids in health and disease and in growth and
development 1-4. Am. J. Clin. Nutr. 54: 438-463.
Siscovick, D.S., T.E. Raghunathan, I. King, eta]. 1995. Dietary intake and cell membrane
levels of long-chain n-3 polyunsaturated fatty acids and the risk of primary cardiac arrest.
J.A.M.A. 274: 1363-7.
Thien, F.C.K., R.K. Woods, and E.H. Walters. 1996. Oily fish and asthma--a fishy story?
M.J.A. 164: 135-136.
Van Houwelingen, A.C., J.D. Sorensen, G. Homstra, eta]. 1995. Essential fatty acid status in
neonates after fish-oil supplementation during late pregnancy. Brit. J. Nutr. 74: 723-73 1.
Vollset, S.E., I. Heuch, and E. Bjelke. 1985. Fish consumption and mortality from coronary
heart disease (letter). N. Eng. J. Med. 313: 820-1.
Willett, W.C., M.J. Stampfer, G.A. Colditz, eta!. 1990. Relation of meat, fat, and fiber intake to
the risk of colon cancer in a prospective study among women. N. Eng. J. Med. 323: 1664-1672.
Toxicology Excellence for Risk Assessment 2-26 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
3 Nutritional Aspects of Fish Compared with Other Protein Sources
3.1 Introduction
This chapter presents an overview of the consumption rates and nutritional benefits of eating
fish. While its specific contributions to the nutritional quality of the diet depend upon the amount
of fish (versus other foods) and species (fatty versus lean) consumed, it is most valued as a
“protein food”. The Biological Value and Protein Efficiency Ratio, indices of the amino acid
profile and ability to support growth, are higher for fish than for beef, pork, chicken and milk
proteins. In addition, the types and proportions of dietary fats are generally more “heart healthy”
than the fats found in other protein foods. Approximately 50% of the fatty acids in lean fish
(e.g., walleye and yellow perch) and 25% in fattier fish (e.g., channel catfish and rainbow trout)
are polyunsaturated fatty acids. The amount of saturated fatty acids, associated with increased
risk of heart disease, tends to be relatively constant across fish species, at about 25% (Sabry,
1990). In contrast, the polyunsaturated and saturated fatty acids in beef are 4-10% and 40-45%,
respectively, of the total fatty acids present. Fish is also valued as a source of omega-3 (n-3) fatty
acids, very long chain polyunsaturated fatty acids which are critical for the development of the
brain and retina, and which may be protective of some chronic diseases. Eicosapentanoic acid
(EPA) (20:5 n-3) and docosahexanoic acid (DHA) (22:6 n-3), which account for approximately
90% of the polyunsaturated fatty acids in fish species from the North Atlantic and North Pacific
(Sabry, 1990), are absent or present in much lower amounts in other foods. The amount of
cholesterol found in fish is comparable to levels in beef, pork, and chicken. Fish is an excellent
source of the B vitamins niacin and B12, and in general is a better source of Vitamins D and A
than beef, pork or chicken. Fish can also contribute appreciable amounts of dietary calcium,
heme iron and zinc, nutrients that tend to be low in people’s diets. Fish is among the best sources
of dietary selenium.
These nutritional benefits are examined from a population health perspective (i.e., what is
relevant to the healthy ‘general population’), rather than a high-risk approach (which is primarily
interested in individuals at highest risk of disease). The nutrient profiles of commonly consumed
sport-caught fish are also compared with those of other protein sources, and discussed in terms of
current population intakes and recommendations. A table that summarizes the nutrient content
and contaminant concentrations of various species of fish and other foods is found at the end of
this chapter. Finally, a brief discussion of the effects on nutritional quality and contaminant
levels from how the fish is prepared and cooked is provided.
3.2 Per Capita Consumption of Fish (FinfIsh and Shellfish)
In the Continuing Survey of Food Intakes by Individuals (CSFII), the USDA regularly collects
information from large numbers of respondents across the United States, about foods eaten.
These consumption data (recorded as weights of the food item consumed for three consecutive
days) provide estimates of the average amounts of finfish and shellfish eaten daily by various
population sub-groups.
In 1996, the U.S. EPA published Daily Average Per Capita Fish Consumption Estimates Based
on the Combined USDA 1989, 1990, and 1991 CSFIIin their Exposure Factors Handbook (U.S.
Toxicology Excellence for Risk Assessment 3-1 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative DietaryRisk
EPA, 1997a). It contains summaries and weighted population estimates based on data collected
from the 11,912 participants in these three national surveys. The U.S. EPA considers this “the
key study for estimating mean fish intake” (U.S. EPA, 1997a), stating that.the data “are probably
adequate for assessing fish ingestion exposure for current populations.”
These food consumption data are summarized for the survey respondents and weighted for
extrapolation to the U.S. general population. Estimates of both uncooked fish weights and
cooked fish weights are provided. The average cooked weight of fish (fInfish and shellfish) eaten
from freshwater and estuaries was estimated to be 4.7 grams (90% C.I. = 4.2-5.3) per person per
day for the U.S. population. Among the 18.5% of survey participants who reported eating
freshwater and estuarine fish, the average cooked weight consumed was estimated to be 68.0
g/day (90% C.!. = 61.9-74.1; Tables 10-11 and 10-21, p. 10 - 38 and 10-44). The average per
capita intake of marine fish is 10.9 glday (cooked weight; 90% C.I. = 10.1-11.7) for the U.S.
population, and 87.8 g/day (90% C.I. = 83.7-91.8) among those 30.1% of survey participants who
reported eating marine fish (Tables 10-11 and 10-22, p. 10-38 and 10-44). Overall, 37% of
individuals reported eating fish (from all sources); on average, they ate an estimated 100.6 grams
of (cooked) fish a day (Table 10-23, p.10-45). Perhaps not surprisingly, males ate higher
amounts of freshwater/estuarine, marine and all fish (77.5 g/day, 98.6 g/day, and 114.2 g/day
respectively) than females (58.8 glday freshwater/estuarine; 78.5 g/day marine; 88.5 g/day all
fish).
The U.S. per capita consumption of fish since 1977/78 has been approximately 11 glday (U.S.
EPA, 1997a, p.10-5), ranging from 4 glday for children aged 0-5 years, to 12 g/day and 15 g/day
among females and males aged 20 and older, respectively (Table 10-46, p.10-56).
Geographically, slightly higher rates of fish consumption are found in New England and the
Atlantic states, than in the rest of the U.S.
Preliminary analyses conducted on data from the most recent CSFII survey (1994-96) suggest
that the per capita consumption of fish may have fallen slightly, from 11 to 9 g/day (Borrud et al.,
1996). This is, on average, about the same as pork with a per capita consumption rate of 11
g/day (down considerably from 20 g/day in 1977/78). Beef consumption has declined
dramatically over the past 20 years, from 52 g/per person/per day to 24; this is approximately the
same as chicken, with an average per capita consumption rate of 23 g/day.
These shifts in dietary patterns have led to a drop in the percentage of energy (calories) from fat
(from 40% in 1977-78 to 34%) and saturated fat (from 12 to 11%), and an increase in the
percentage of energy from carbohydrates (from 43 to 5 1%). While these changes still do not
meet current recommendations for “no more than 30% of energy from fat,-10% of energy from
saturated fat, and at least 55% of energy from carbohydrates”, they are important and healthful
changes. However, the same data suggest that average intakes of vitamin B6, calcium, iron, zinc,
and some other nutrients are below the Recommended Daily intakes (RDAs). As the study
authors explain, “as the percentage of the population with intakes below 100% of a given RDA
increases, so does the likeithood that some people are at nutritional risk. Only 21% of the
women in the 1994 CSFII had diets that met the RDA for calcium.., and 17% for zinc” (Borrud et
al, 1996, p.17).
Toxicology Excellence for Risk Assessment 3-2 8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
3.3 Nutritional Content and Contaminant Levels for Fish and Other Protein Sources
3.3.1 Selection of Nutrients, Foods and Contaminants for Tables 3-1 and 3-2
Tables 3-1 and 3-2 (found at the end of this chapter on pages 3-14 and 3-18) were designed to
facilitate the comparison of the nutritional components and contaminant levels in several fish
species, to levels present in other commonly consumed foods of high protein quality. Ocean fish
and seafood species were chosen to represent commonly consumed store-bought fish (e.g. halibut
and tuna) as well as species commercially or sport caught in specific geographical regions (e.g.,
crab). Freshwater fish species selected represent commonly consumed sport-caught fish
occurring in various geographical locations across the United States, and include species of
varying fat content. Other foods present in the table, such as beef tenderloin, pork loin, and
chicken breast, were chosen as lean protein sources that could be compared with lean fish species
(e.g. walleye). Fast foods and processed foods such as fried chicken, hamburgers, hotdogs, and
fish sticks are important sources of protein for a large segment of the population. Other foods,
such as tofu and refried beans, were included because they may be important protein sources for
specific ethnic groups.
The nutrients listed in Tables 3-1 and 3-2 represent the most important nutrients commonly
obtained from fish, which could be regarded as public health concerns. In other words, fish
would be considered good sources of these nutrients, which, according to U.S. population
surveys, are often not consumed in recommended amounts. Fish do supply other important
micronutrients and compounds, such as fluorine and copper, which are not currently public
health nutrition concerns in the U.S. A detailed discussion of specific nutrients can be found in
subsequent sections of this chapter. By comparing the protein (and associated indices such as
protein efficiency ratio), vitamin, mineral and lipid (e.g. omega-3 fatty acids) levels in fish to
those of other protein food sources, the nutritional benefits of consuming various protein sources
can be assessed.
The nutrient compositions of the foods listed in the table were obtained from the USDA’s food
composition database (USDA, 1998). To create a nutrition profile for a given food, data are
compiled from several sources. For example, the data on fish are obtained from the scientific
literature, government agencies outside the USDA, USDA contracts, and industry and trade
associations (J. Exler, personal communication, 26 April, 1999). Several sources of data are
compiled to help make he samples more representative of fish obtained throughout the U.S.
Values for cooked fish were calculated from data on raw fish. Fish species that are listed as
‘wild” were obtained from commercial sources or, if the wild form of the species was not
available, the fish was caught by the group contracted to provide the sample. All values for the
fish listed in the tables aie for the “wild” form. For processed fish and other foods, data are
available on the most popular brand names.
For the development of the framework, six chemical contaminants were chosen to represent toxic
substances that may be present in various protein sources. By comparing the levels of
contaminants in commonly consumed fish species to those in other commonly consumed protein
food sources, the relative risk of exposure to contaminants through the consumption of various
foods can be assessed. Contaminant levels in foods, including wild fish species, are monitored to
Toxicology Excellence for Risk Assessment 3-3 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
identify foods that may contain unsafe levels. While published reports of contaminant levels in
fish and other foods have been released as recently as 1998, the data used for many of these
reports were collected in the late 1970s and early 1980s. Since the level of many of these
contaminants in foods declined during those years (and the decline most likely continued
throughout the 1990s) references that contained data collected in the late 1970s and early 1980s
were not used in Tables 3-1 and 3-2. Data for a number of these chemical contaminants was not
available.
3.3.2 Substituting Other Foods for Fish: Effects on Macronutrient Profiles
While it can be useful to compare the nutrients found in equal weights of various foods, it is the
impact of substituting one for food another on the total diet that is important. For example, 24-
hour food record for a 45 year old female of Asian descent, who had eaten 150 grams of rainbow
trout for dinner was analyzed by a nutrient software program (Candat, 1994), and the results are
shown in Table 3-3. Her total protein intake (from the fish and other foods eaten) on that
particular day accounted for 24% of her energy intake (calories), while 37% of her calories came
from fat and 39% came from carbohydrates. (Although this dietary profile does not meet current
recommendations for diets composed of 30% or less of energy from fat and 55% or more of
energy from carbohydrates, it is a typical North American dietary profile). Table 3-3 shows the
effect of substituting 150 grams of perch (a lean herbivorous fish), skinless chicken breast, or hot
dogs for the rainbow trout. The rainbow trout, a fattier fish species, and the skinless chicken
breast produce daily dietary profiles that are similar. The best macronutrient profile is obtained
with perch (i.e., the lowest % of energy from fat, highest % of energy from carbohydrates, and
lowest total number of kcal), while substituting 150 grams of hot dogs produces the worst dietary
profile. Thus, substituting a fatty or a lean species of fish for hot dogs at one meal can make a
noticeable improvement in the profile of the whole day’s diet, even if no other dietary changes are
made.
Table 3-3. Percent of energy (calories) from macronutrients based upon one day’s diet which
included a 150 gram serving of fish, chicken or hotdogs.’ (calculations based on Candat, 1994)
Rainbow
Trout
Perch Mixed
Species
Chicken
Breast (no
skin)
Hotdog
%energyfrom
Protein
24%
26%
27%
15%
%energy from
Fat
37%
33%
34%
52%
%energyfrom
Carbohydrates
39%
41%
39%
33%
Total energy
intake for 24-hr
1148 kcal
1099 kcal
1171 kcal
1396 kcal
based upon an actual 24-hr food record of a 45-year-old female
Toxicology Excellence for Risk Assessment
3-4
8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
3.4 Fish as a Protein Source
3.4.1 Protein Quality
As Groff et al. (1995) states, “The importance of protein in nutrition and health cannot be
overemphasized.” Protein is composed of amino acids; nine of these (leucine, isoleucine, valine,
lysine, tryptophan, threonine, methionine, phenylalanine, and histidine) are considered
nutritionally essential or “indispensable” in the human diet, because they cannot be synthesized
by the body. Protein also supplies nitrogen, for the internal synthesis of other amino acids
required by the body. Different categories of proteins include enzymes, required as catalysts in
most of the body’s chemical reactions, peptide hormones (such as insulin, thyroid hormones, and
the growth hormone somatotropin), structural proteins (in muscle and connective tissue),
transport proteins (e.g., albumin, transferrin, hemoglobin), and immunoproteins or antibodies.
The quality of a protein is mainly determined by the specific amounts and relative proportions of
its essential amino acids, their availability to the body, and to a lesser extent, the protein’s
digestibility. While the amount of protein required by individuals depends on their body weight
and height, energy (calorie) intake, and physiologic condition (e.g., infancy, pregnancy), it is the
quality of the protein which is most important in determining daily requirements.
3.4.2 Fish Protein vs. Other Dietary Protein Sources
There are many ways of evaluating the quality of the protein provided by different foods, for
comparison purposes. In general, protein from animal foods (e.g., dairy products, eggs, meats,
fish, and poultry) is of higher quality than protein from plant foods (e.g., pasta, rice, fruits, and
vegetables).
The Chemical Score or Amino Acid Score compares a food’s amino acid pattern to that of whole
egg protein (with a score of 100), considered to be have the ‘ideal’ reference composition. The
Chemical Score of finfish is 70, an indication of its high quality; beef is 69 and cow’s milk is 60
(Sabry, 1990).
The Biological Value (By) of a protein is calculated by measuring the body’s nitrogen balance--
nitrogen ingested (from the protein of interest), absorption and use for synthesizing new amino
acids, and losses (through urine and feces). The percentage of the absorbed nitrogen which is
retained by the body for tissue growth and maintenance, is the BV:
BV = (nitrogen retained/nitrogen absorbed) x 100
As shown in Tables 3-1 and 3-2, the BV of fish (76%) is slightly higher than that of beef (74.3),
pork (74.0) and chicken (74.3), but all are somewhat lower than egg (93.7).
The Protein Efficiency Ratio (PER) is another measure of protein quality, usually calculated by
putting young animals on diets with various test proteins, and monitoring their growth. The PER
Toxicology Excellence for Risk Assessment 3-5 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
is a ratio of the gain in weight divided by the weight of the protein consumed:
PER = gain in body weight (in grams)/grams of protein consumed
The PER of fish (3.55) is higher than beef (2.30) and milk proteins (casein = 2.50), and close to
that of egg (3.92).
3.5 Fish as a Source of Essential Fatty Acids
Although high fat intakes have been associated with an increase in the risk of several chronic
diseases, a certain amount of fat is necessary for the body to function normally. Triglycerides
provide the body with a continuous fuel source, supply heat to the body, protect the body from
mechanical shock, and certain fatty acid components are important building blocks for several
hormone regulators (Whitney and Rolfes, 1996). Phospholipids and sterols are a major
component of cell membranes, and sterol cholesterol provides the building blocks for some
hormones, Vitamin D, and bile.
Food is composed of saturated, monounsaturated and polyunsaturated fatty acids. In lean finfIsh
(e.g., walleye and yellow perch) polyunsaturated fatty acids account for approximately 50% of
the total fatty acid content in the flesh. Saturated and monounsaturated fatty acids each comprise
approximately 25% of the total fatty acid content. For fattier finfish (e.g., channel catfish and
rainbow trout) polyunsaturated and saturated fatty acids each constitute approximately 25% of
the total fatty acid content. The amount of saturated fatty acids tends to be relatively constant
across fish species with proportions of polyunsaturated fatty acids being lower in fish with higher
levels of monounsaturated fatty acids (Sabry, 1990).
Finfish tend to have higher levels of polyunsaturated fatty acids and lower levels of saturated
fatty acids than other meat sources of protein. The proportion of saturated, monounsaturated and
polyunsaturated fatty acids found in beef and pork are, respectively, approximately 40-45%,
50%, and 4-10% (Sabry, 1990). The fatty acid profile of chicken (30-35% saturates, 35-40%
monounsaturates, and 25-30% polyunsaturates) falls between that of fish and beef and pork.
Dairy products (e.g., cheese and eggs) have a much higher saturated fat component (40-65%),
similar monounsaturated levels (30-45%), and much lower polyunsaturated levels (5%) than fish
Fish also contain long chain polyunsaturated fatty acids of the n-3 (or omega-3) series which are
not commonly found in other food sources (Sabry, 1990). Eicosapentanoic acid (EPA) (20:5n-3)
and docosahexanoic acid (DHA) (22:6n-3) are the most common n-3 fattyacids, and account for
approximately 90% of the total polyunsaturated fatty acids in fish species from the North Atlantic
and North Pacific (Sabry, 1990). Linolenic acid (18:3n-3), linoleic acid (18:2n-6) and
arachidonic acid (20:4n-6) are also present in fish, although in much smaller proportions (i.e., 1-
2%). In lean and lower fat fish (e.g., walleye and yellow perch), the n-3 fatty acid content (EPA
+ DHA) is often less than 0.5g/lOOg fish. For higher fat fish (e.g., coho salmon and rainbow
trout) the fatty acid content is often between 0.8-1.OgIlOOg fish. These fatty acid components are
absent or present in much smaller amounts in other protein food sources.
Toxicology Excellence for Risk Assessment 3-6 8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
Linoleic and linolenic acid are essential fatty acids that are not produced in the body and must be
obtained from dietary sources (Whitney and Rolfes, 1996). These fatty acids are an important
part of the structural component of cell membranes, and are necessary for the formation of
eicosanoids which assist in blood pressure regulation, blood clot formation, maintenance of
blood lipid levels, and assist in the body’s immune response.
Because studies have linked the type and amount of dietary fat to various diseases (e.g.,
cardiovascular disease), the fatty acid composition of fish has been of great interest to researchers
for the past several years. The benefits of consuming the various types of fatty acids (and their
component parts) found in fish are discussed in Chapter 2.
3.6 Cholesterol
Cholesterol forms the building blocks of several compounds (e.g., bile, sex hormones, adrenal
hormones, and Vitamin D) with important physiological functions, and is a major structural
component of cell mem ranes (Whitney and Rolfes, 1996). Although food provides an
important source of cholesterol, endogenous sources contribute much higher amounts.
Cholesterol is synthesized in the liver, through the production of bile, and in the intestine by de
novo synthesis and desquamation of mucosal cells. The amount of endogenous cholesterol
produced in the liver is dependent on the amount of raw materials available (i.e., carbohydrate,
protein, and fat), the extent of bile production and availability of regulating hormones (e.g.,
insulin).
The amount of cholesterol found in finfish varies from approximately 50-100 mg/100 gram
portion. This is comparable to the amounts found in beef (84 mg), pork (79 mg), and chicken
(85 mg), and is somewhat lower than the levels found in cheddar cheese (105mg) and eggs (424
mg). The method of food preparation will also affect cholesterol levels. Deep frying, compared
with dry heat cooking, increased the cholesterol level of channel catfish by approximately 10%
(USDA, 1998).
3.7 Vitamins
3.7.1 Vitamins B3, B6, and B12
Vitamins B3 (niacin, nicotinic acid, nicotinamide) and B6 (pyridoxine, pyridoxal, and
pyridoxamine) are water soluble organic compounds that are absorbed into the portal blood and
stored only briefly in the body; thus, they must be supplied in the human diet every day. Vitamin
B12 (cyanocobalamin), also water-soluble, is available from animal foods and can also be re-
absorbed into the bloodstream from bile and secretions in the small intestine (i.e., via
enterohepatic circulation). It can be stored in the liver and other tissues for years.
Niacin is involved in hydrogen transfer reactions (as part of coenzyrnes NAD and NADP) and a
deficiency results in diarrhea, dermatitis and dementia, a condition known as Pellagra. Adults
require approx. 15-19 mg per day, according to the 1989 US Recommended Dietary Allowances.
Sixty milligrams of the amino acid tryptophan are considered to be equivalent to 1 mg niacin
and expressed as a niacin equivalent (NE).
Toxicology Excellence for Risk Assessment 3-7 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Fish is an excellent source of niacin. A 100 gram portion of canned tuna supplies 13.280 mg of
niacin, which compares favorably with another good source, chicken (13.712 mg). The same
portion size of channel catfish or coho salmon supplies 7.95 mg of niacin, in contrast to beef
tenderloin (3.92 mg), pork loin (5.243 mg), fortified pasta (1.672 mg), or egg (0.064 mg).
Vitamin B6 exists as several different chemical structures and their phosphorylated forms.
Different forms of Vitamin B6 serve as important co-enzymes in transamination,
decarboxlyation, and transulfhydration and desuithydration reactions. The RDAs for adult males
and females are 2.0 mg and 1.6 mg, respectively, and deficiencies are rare in North America.
The pyridoxine forms are found only in plant foods (especially bananas, navy beans, and
walnuts), while the phosphorylated pyridoxal and pyridoxamine forms arefound in animal foods,
particularly coho salmon (0.568 mg/100 g) and roast chicken breast (0.600 mg/100 g). As shown
in Table 3-1, 100 grams of fresh-water drum and rainbow trout (both 0.346 mg), halibut (0.397
mg) or canned tuna (0.350 mg) contain about the same amount of Vitamin B6 as pork (0.277 mg
for a shoulder cut and 0.492 mg for a loin cut), beef (0.440 mg for tenderloin), ham (0.340 mg)
or fast food chicken (0.350 mg). Even fish with lower amounts, such as channel catfish (0.106
mg) and northern pike (0.135 mg), compare favorably with the Vitamin B6 content of egg (0.121
mg) and are better sources than hot dogs (0.050 mg), cheddar cheese (0.074 mg), pasta (0.035
mg), and rice (0.093 mg).
Vitamin B12 or cyanocobalamin is produced by microorganisms in animals, and does not occur
naturally in plant foods. It is important in three enzymatic reactions: the conversion of (1)
homocysteine into methionine; (2) L-methylmalonyl CoA to succinyl CoA; and (3) the formation
of leucine aminomutase. Little Vitamin B12 is lost through urine or feces; most is excreted into
bile and then re-absorbed in the ileum. A deficiency occurs if absorption is impaired (e.g., with
doses of 500 mg or more of Vitamin C), or after many years on a strict vegan diet, and results in
megaloblastic anemia and neuropathy.
Sport-caught fish are among the best dietary sources of Vitamin B12; e.g.. rainbow trout, coho
salmon and channel catfish (6.3 ug, 5 ug, and 2.9 ug per 100 gram portions, respectively) provide
more than beef (2.57 ug), pork (1.06 ug), chicken (0.34 ug), or egg (1.10 ug).
3.7.2 Vitamin A
Vitamin A (retinol, retinal, retinoic acid) is a fat-soluble vitamin, meaning that its absorption,
transport and storage are linked to lipids (i.e., it requires bile salts for absorption, is transported
as chylomicrons, and stored in fatty tissues). It is critical for good vision, growth, bone
development and maintenance, T-lymphocyte function and antibody response, among other
things.
Provitamin A refers to carotenoids found in plants and converted into retinol. Carotenoids such
as lycopene and beta-carotene, which function as anti-oxidants, are thought to play a role in the
prevention of some cancers. One Retinol Equivalent (RE) is the same as 1 microgram of all-
trans retinol, 6 micrograms of all-trans beta-carotene, 12 micrograms of other provitamin A
carotenoids and 3.33 International Units (Hi) of pre-formed Vitamin A. According to the 1989
Toxicology Excellence for Risk Assessment 3-8 816/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
RDAs, adult males and females require 1,000 and 800 micrograms RE, respectively, each day.
The vitamin A content of sport-caught fish depends on the fattiness of the species and the type of
preparation for cooking (i.e., skin on vs. off, internal organs consumed). Relatively lower-fat
species, such as yellow perch, contain smaller amounts of Vitamin A (10 RE per 100 g portion)
in comparison with higher-fat species, such as coho salmon (39 RE), bass (35 RE) and fresh-
water drum (59 RE). In general, fish is a better source of this vitamin than beef, pork, or chicken.
3.7.3 VitaminD
Vitamin D is a fat-soluble sterol that occurs naturally in many forms. It is ergosterol when found
in plants, while in animals and humans, it is synthesized in the skin as 7-dehydrocholesterol and
upon exposure to sunlight is subsequently converted to precalciferol, then cholecalciferol. This
is later converted by the liver to the active form, calcitriol. Vitamin D behaves like a steroid
hormone to control blood levels of calcium, and thus has effects on bone, kidney, and intestinal
tissues. A deficiency of Vitamin D interferes with the body’s absorption of calcium and perhaps
also phosphorus, which results in bone demineralization.
Fish is among the best food sources of Vitamin D, but few data are available in food composition
tables. A 100 g portion of herring and tuna provide 22 and 6 micrograms, respectively, of
Vitamin D. In contrast, the same amount of Vitamin D-fortified milk and liver provide only 1
and 0.1-0.2 micrograms, respectively.
3.8 Minerals
While fish is known to contain many important trace elements, only those minerals that are
frequently lacking in the diets of healthy populations are presented here.
3.8.1 Calcium
An estimated 99% of the body’s calcium resides in the teeth and bones, where it is extracted and
re-deposited as needed to keep blood levels of calcium constant. The one percent found in the
blood, lymph and other body fluids is critical to the intracellular and extracellular environments
of all living cells. The parathyroid hormone, calcitonin, calcitriol (Vitamin D) and other
hormones help to regulate levels by releasing calcium from bone and controlling its absorption
from the intestine and excretion in urine, feces and sweat. Bone loss occurs naturally with age, in
both males and females: therefore, it is important that children and young adults achieve their
optimum bone density so that age-related losses will not result in osteoporosis.
Many dietary factors affect the bioavailability of calcium; dietary fiber, phytic acid, uronic acid
and oxalic acid are believed to reduce intestinal absorption. Calcium is poorly absorbed (5%)
from spinach, for example, which is high in oxalic acid, but more readily available from kale,
which is low in oxalic acid. Dairy products provide about one-half of dietary calcium in the U.S.
Fish with soft bones can also be important dietary sources; for example, walleye, bass and yellow
perch provide 141, 103 and 102 mg of calcium, respectively, in a 100 gram portion. Small fish
eaten whole, such as sardines and smelts, as well as canned fish with bones, such as salmon, also
contribute appreciable amounts of dietary calcium.
Toxicology Excellence for Risk Assessment 3-9 8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
3.8.2 hon
In North America and worldwide, iron deficiency is the most prevalent nutritional deficiency
(Yip and Daliman, 1996), particularly among young children and premenopausal women. Men
typically have approximately 3.8 grams of iron in their bodies, one-third of which is stored as
ferritin and hemosiderin in the liver, bone marrow and spleen, and two-thirds of which is
functional iron, mostly in the form of hemoglobin and myoglobin. Women’s bodies have about
2.3 grams of iron, and only about one-eighth (0.3 grams) is in storage. The body’s increased
demand for iron in pregnancy (1.0 gram) is much greater than the average woman’s iron stores.
In situations such as this, when the body’s functional needs for iron outstrip bodily stores, iron
deficiency may result. Measures of serum ferritin indicate when body stores of iron are low or
depleted (iron depletion). When transferrin saturation falls, erythrocyte protoporphyrin levels
rise, but hemoglobin levels are normal, resulting is iron deficiency without anemia. Iron-
deficiency anemia occurs when blood hemoglobin levels have also dropped below normal values.
On average, men absorb only 6% of the dietary iron they consume; the rest is not soluble and
thus not bioavailable. Premenopausal women absorb about 13% of their dietary iron, which
helps to offset their smaller body stores. Fish and other animal foods contain heme iron, which
only accounts for about 15% of dietary iron (85% is nonheme) but which is absorbed at over
twice the rate of nonheme iron. Further, the presence of fish, meat, or poultry in a meal greatly
increases the bioavailability of the nonheme iron provided by plant foods.
As seen in Table 3-1, fish species differ in the amount of dietary iron they provide--channel
catfish has 0.35 mg/100 grams while the same size portion of bass contains 1.91 mg. In general,
fish contains less iron than beef, but comparable levels to pork and chicken. Although dietary
iron levels are similar in eggs, the iron has poor bioavailability.
3.8.3 Zinc
Zinc is part of many enzymes, biomembranes, and is involved in RNA transcription, among other
activities too numerous to mention here. It has held tremendous public health significance in
developing countries since the 1960s, when zinc deficiency was linked to stunted growth and
delayed sexual maturation. Inadequate intakes of zinc are common in North America,
particularly among vegetarians and adult women (e.g., Borrud et. al, 1996). While quite widely
distributed among plant and animal foods, zinc often has low bioavailability because of its
interactions with copper, iron, and other food components such as phytates.
Animal foods provide approximately 70% of the dietary zinc in the U.S., and in general, foods
rich in protein are also good sources of zinc. As shown in Tables 3-1 and .3-2, sport-caught fish
contain from 0.51-1.43 mg of zinc per 100 gram edible portions, comparable to levels in eggs,
tofu, refried beans, pasta and rice.
Toxicology Excellence for Risk Assessment 3-10 8/6/99
-------
Cooj erative Agreement with U S. EPA on Comparative Dietary Risk
3.8.4 Selenium
Selenium is found in difi’erent forms in plants (mainly as selenomethionine) and animals (mainly
as selenocysteine). Animals and humans rely exclusively on dietary sources of the mineral,
which is an important component of the enzyme glutathione peroxidase and some transfer RNAs,
and functions as an anti oxidant. There is also some evidence that it may form complexes with
mercury, cadmium, and other toxic heavy metals (Levander & Burk, 1991).
Seafood and organ meats (e.g., liver) are the best dietary sources of selenium (40-150
micrograms per 100 gram portions). Beef and pork have lower levels (10-40 micrograms), and
are followed by cereals and grains, <10 to >80 depending upon the soil content; dairy products,
<10 to 30; and fruits and vegetables <10 (Levander & Burk, 1991).
3.9 Effects of Food Preparation Methods on Nutritional Benefits
The type of cooking method may affect some nutritional components, while other components
remain unaffected. During the cooking process fish flesh loses moisture, with the amount lost
dependent on the fish species, the size of the piece and cooking method (Sabry, 1990). For
finfish, cooking with dry heat results in a moisture loss of 22% of the original weight compared
to a 21% loss with moisr heat, or an eight percent loss with bread-fried cooking. Microwave
cooking results in even higher moisture losses (i.e. 30-35%) than dry or moist heat cooking.
Because the proportion of solids is increased with moisture loss, the concentrations of certain
nutrients tend to be higher in cooked relative to raw fish. After adjustments for water loss have
been made, however, cooking does not appear to affect the protein or total lipid levels. With the
exception of poaching, which may cause a loss of some dissolved minerals when cooking water
is discarded, the mineral content of cooked fish is usually not affected by the cooking process.
Some vitamins may be destroyed during the cooking process but this is dependent on the method,
duration, and temperature of cooking. For the B vitamins (i.e. thiamin, riboflavin, niacin, B6,
and B12) a loss in the range of 0- 30% has been reported (Sabry, 1990). For vitamin A, a loss of
5-15% has been observed due to the cooking process. In general, cooking with high temperatures
for long periods of time tends to cause the greatest loss of vitamins.
The fat content of fish nfay also be altered depending upon how it is prepared and cooked. A
100-gram portion of channel catfish that was breaded and fried had twice the amount of energy
(calories), and four times the amount of fat than a similar portion that was cooked using dry heat
(USDA, 1998). Breading and frying also significantly altered the proportions of lipids. The
amount of saturated fat, monounsaturated fats, and polyunsaturated fats increased by 400 - 500%
in the breaded and fried fish. Levels of omega-3 fatty acids also increased in the breaded and
fried fish, although not as dramatically (i.e. 10-100%). When comparing the nutritional benefits
of consuming various types of fish and other protein sources, it is important to consider the
method of food preparation and any additional ingredients (e.g., oil for frying) that may be added
to the final cooked product.
Toxicology Excellence for Risk Assessment 3-11 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
3.10 Effects of Food Preparation Methods on Contaminant Levels
The way fish is prepared and cooked can modify the amount of chemical contaminants
consumed. Appendix E of the U.S. EPA Guidance for Assessing Chemical Contaminant Data for
Use in Fish Advisories, Volume 2 (U.S. EPA, 1997b) discusses the available data; some of this
information is presented here.
The degree to which contaminants bioaccumulate in different fish species is dependent on their
methods of feeding, the ability of the fish to metabolize the contaminants and the fat content of
the fish (U.S. EPA, 1997b). Trimming the fat, and removing the skin and the internal organs will
help decrease the amount of lipophilic contaminants but will not reduce exposure to those
contaminants that concentrate in muscle and other protein-rich tissues (e.g. mercury).
The method of preparing the fish for cooking and eating appears to be a major factor in reducing
the amount of certain contaminants in standard fish fillets. For example, trimming and cooking
brown trout reduced the Mirex and PCB content by 74% and 78%, respectively (U.S. EPA,
1997b). However, broiling reduced the Mirex content by 26% but resulted in no reduction in the
PCB levels. Removing the skin before cooking also resulted in a further 17% reduction in alpha-
chlordane levels in chinook salmon fillets, compared to fillets baked with the skin on.
Toxaphene levels were reduced by 40% when lake trout fillets had the skin removed and were
charbroiled. For smallmouth bass, DDE levels were reduced by 54% when the fillets were
trimmed. For certain ocean fish such as bluefish, the PCB content of the fillets was decreased by
27% after cooking and removal of skin and oil drippings (Trotter et al., 1988). For five fish
species from the Great Lakes region, Zabik and Zabik (1995) found a 30-100% reduction in
dioxin levels depending on the type of cooking method used. Smoking removed the greatest
amount of dioxin (i.e. 100%) while salt boiling resulted in the smallest reduction (i.e. 30%).
Preparation methods have a significant influence on the amount of various contaminants present
in a cooked fish fillet.
3.11 Conclusions and Research Needs
There are many nutritional benefits associated with eating fish, regardless of the species type.
Unlike red meats, eggs and dairy products, fish provides very high quality protein anda heart
healthy’ combination of fatty acids. Further, fish (both lean and fatty) is one of the few foods that
contain n-3 (omega-3) fatty acids, a class of fatty acids that are essential for the development of
the nervous system and that may have other beneficial health effects. Calcium, iron, zinc,
vitamin A, niacin, vitamin B6, and vitamin D tend to be low in U.S. diets; fish supplies all of
these vitamins and minerals, in addition to others.
Fish is known to be a good dietary source of selenium, but few reference data are published;
more research into the role of selenium in human health is also needed. Nutrient databases
contain a wide range of fish species, but samples used to obtain nutrient values are composites of
cooked fish from various unknown locations. Nutrient values are generally expressed on the
basis of a 100 gram cooked fish portion. This limits the extent to which comparisons can be
made with contaminant data, which are usually based on raw tissue samples of wild fish gathered
from specific geographic areas, and expressed as concentrations rather than on a weight basis.
Toxicology Excellence for Risk Assessment 3-12 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
We know that different methods of preparing and cooking fish will alter some of the
organochiorine contaminant levels. Ideally, the same samples of prepared and cooked fish would
be sent for both contaminant and nutrient analysis. and weighed records of amounts of the fish
consumed would be kept to enable researchers to better assess the physiological risks and
benefits to humans.
Toxicology Excellence for Risk Assessment 3-13 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Table 3-1 Nutrition Values and Contaminant Levels in Fish - values for 100 g edible portion
Coho
Salmon
Rainbow
Trout
Northern
Pike
Walleye
Rainbow
Smelt
Fresh-water
Drum
Bass Mixed
Species
Channel
Catfish
SunfIsh
Perch
Mixed
Species
Protein
23.4g
229g
247g
245g
226g
22.5g
242g
185g
248g
249g
Protein Efficiency
Ratio’
3 55
3.55
3.55
3 55
3 55
3.55
3.55
3.55
3.55
3.55
Biological Value 2
76 0
76.0
76 0
76 0
76 0
76.0
76 0
76.0
76 0
76 0
Calories
139 kcal
l5Okcal
113 kcal
Il9kcal
124 kcal
153 kcal
l46kcal
IO5kcal
ll4kcal
ll7kcal
Totalfat
43g
5.8g
0.9g
16g
3lg
6.3g
47g
2.9g
0.9g
12g
Saturatedfat
hg
1.6g
02g
03g
0.6g
14g
1.Og
O7g
02g
0.2g
MUFA=s
1.6 g
1 8 g
0.2 g
0.4 g
0.8 g
2.8 g
1.8 g
1.1 g
0.2 g
0 2 g
PUFA=s
13g
18g
0.3g
06g
hg
l.5g
1.4g
0.6g
03g
05g
18:ZLinoleic
0056g
0.288g
0041g
0033g
0058g
0.199g
0.112g
0142g
O.019g
0014g
18:3 Linolenic
0055 g
0.187 g
0 027 g
0018 g
0063 g
0.146 g
0 142 g
0096 g
0.013 g
0.015 g
20:4AA
0022g
0.120g
0036g
0074g
0071g
0287g
0.l85g
0087g
OlOig
0067g
20:5 EPA
0401g
0468g
0042g
OllOg
0353g
0295g
0305g
OlOOg
0047g
OlOig
22:6DHA
0658g
0520g
OO95g
0288g
0536g
0368g
0458g
0137g
0092g
0223g
Cholesterol
55 mg
69 mg
50 mg
110 mg
90 mg
82 mg
87 mg
72 mg
86 mg
115 mg
Zinc
0 560 mg
0 510 mg
0.860mg
0 79 mg
2 12 mg
0 850 mg
0 83 mg
0 61 mg
1.99 mg
1 43 mg
Toxicology Excellence for Risk Assessment 3-14 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Iron
0 610 mg
0 380 mg
0 710 mg
1 67 mg
115 mg
115 mg
1.91 mg
0 35 mg
154 mg
116 mg
Calcium
45mg
86mg
73mg
141 mg
77mg
77 mg
103mg
11 mg
103 mg
102mg
Vitamin A
39 RE
15 RE
24 RE
24 RE
17 RE
59 RE
35 RE
15 RE
17 RE
10 RE
Vitamin B 3
7 950 mg
5 770 mg
2.800 mg
2.810 mg
1 766 mg
2 862 mg
1 522 mg
7 950 mg
1.460 mg
1 900 mg
Vitamin B 6
0 568 mg
0 346 mg
0 135 mg
0 138 mg
0.170 mg
0 346 mg
0 138 mg
0.106 mg
0.138 mg
0.140 mg
Vitamin B 12
5 00 ug
6 30 ug
2.30 ug
2 31 ug
3.97 ug
2.31 ug
2.31 ug
2.90 ug
2 30 ug
2 20 ug
Chordane
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
DDT
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Chiorpyrifos
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Dioxins
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Lindane
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
PCBs
NA
NA
NA
92-1933
ng/ wet
WI.
NA
NA
NA
NA
NA
NA
Mercury
NA
NA
NA
187-793
ng/ wet
WI
NA
NA
NA
NA
NA
NA
Protein Efficiency Ratio - gain in weight divided by weight of protein consumed
2 Biological Value - the percentage of absorbed nitrogen retained
3 Dellinger, J A et al. 1996 The Ojibwa Health Study fish residue comparisons for Lakes Superior, Michigan, and Huron Tox. Ind Health 12 393-402
NA indicates Not Available
Toxicology Excellence for Risk Assessment
3-15 8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
Table 3-1 (Cont) Nutrition Values and Contaminant Levels in Fish - values for 100 edible portion, cont.
Tuna
Canned in
water
Cod
(Atlantic)
Halibut
FlatfIsh
(Flounder
& Sole)
Haddock
Mackerel
Mixed
Species
Orange
Roughy
Ocean
Perch
Sardines
Blue
Crab
Northern
Lobster
Protein
25 5 g
22 8 g
26.7 g
24 2 g
24.2 g
25 7 g
18.9 g
23.9 g
24.6 g
20 2 g
20 5 g
Protein Efficiency
Ratio’
3 55
3.55
3 55
3 55
3 55
3 55
3.55
3.55
3 55
NA
NA
Biological Value 2
76.0
76.0
76 0
76 0
76.0
76 0
76.0
76.0
76 0
NA
NA
Calories
116 kcal
105 kcal
140 kcal
117 kcal
112 kcal
201 kcal
89 kcal
121 kcal
208 kcal
102 kcal
98 kcal
Totalfat
08g
0.9g
29g
15g
09g
10. lg
0.9g
21g
11.5g
l.8g
06g
Saturatedfat
0.2g
02g
0.4 .g
04g
02g
29g
0.02g
0.3g
15g
0.2g
0.107g
MUFA=s
02g
Olg
1.Og
0.2g
02g
3.4g
06g
08g
3.9g
03g
0.160g
PUFA=s
0.3g
03g
0.9g
06g
O.3g
25g
0.02g
05g
51g
0.7g
0.091g
18:2 Linoleic
0.009 g
0 006 g
0 038 g
0 014 g
0.012 g
0 149 g
0.009 g
0.036 g
3.543 g
0.028 g
0 005 g
18 3 Linolenic
0002 g
0 001 g
0 083 g
0.016 g
0003 g
0 064 g
0.002 g
0.073 g
0.498 g
0 021 g
0
20 4 AA
0 034 g
0 028 g
0 178 g
0.048 g
0 029 g
0 104 g
0 002 g
0 005 g
0
0.084 g
0
20 5 EPA
0.047 g
0 004 g
0 09] g
0 243 g
0 076 g
0.653 g
0 002 g
0.103 g
0.473 g
0.243 g
0.053 g
22 6 DHA
0.223 g
0 154 g
0 374 g
0 258 g
0.162 g
1.195 g
NA
0 271 g
0.509 g
0.231 g
0.031 g
Cholesterol
30 mg
55 mg
41 mg
68 mg
74 mg
60 mg
26 mg
54 mg
142 mg
100 mg
72 mg
Zmc
0 77 mg
0 58 mg
0 53 mg
0.63 mg
0 48 mg
0 86 mg
0.96 mg
0.61 mg
1.31 mg
4 22 mg
2.92 mg
Iron
1 53 mg
0 49 mg
1.07 mg
0.34 mg
1 35 mg
1 49 mg
0,23 mg
1.18 mg
2 92 mg
0.91 mg
0.39 mg
Toxicology Excellence for Risk Assessment
3-16 8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
Table 3-1 (Cont) Nutrition Values and Contaminant Levels in Fish - values for 100 g edible portion, cont.
Calcium
11 mg
14 mg
60 mg
18 mg
42 mg
29 mg
38 mg
137 mg
382 mg
104 mg
61 mg
Vitamin A
17 RE
14 RE
54 RE
11 RE
19 RE
14 RE
24 RE
14 RE
67 RE
2 RE
26 RE
Vitamin B 3
13 280 mg
2 513mg
7 123
mg
2.179 mg
4.632 mg
10.667mg
3 654 mg
2436 mg
5 245 mg
3.300mg
1 070 mg
Vitamin B
0 3c0 mg
0 283mg
0.397
mg
0 240mg
46 mg
0381 mg
0346mg
0270 nig
0 167 nig
0 180mg
0077 ‘f g
VitaminB , 2
2990ug
1.048ug
1366ug
2.509ug
1.387ug
4230ug
2310ug
1154ug
8940ug
7300ug
3llOug
Chordane
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
DDT
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Chiorpyrifos
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Dioxins
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Lindane
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
PCBs (Fotal)
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Mercuiy
<0.10-075
ppm
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
‘Protein Efficiency Ratio - gain in weight divided by weight of protein consumed
2 Biological Value - the percentage of absorbed nitrogen retained
3 Dellinger, J A. et al 1996 The Ojibwa Health Study: fish residue comparisons for Lakes Superior, Michigan, and Huron Tox. md. Health 12 393-402
NA indicates Not Available
Toxicology Excellence for Risk Assessment
3-17 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Table 3-2 Nutrition Values and Contaminant Levels in Other Protein Sources - values for 100 g edible portion
Chicken
(brst, roast)
Lamb
(shank &
sirloin)
Beef
(tenderloin)
Pork
(loin)
Pork
(shoulder)
Ham
(sliced)
Sausage
(turkey)
Sausage
(pork link)
Bacon
Fish Sticks
Protein
310g
28.3g
283g
286g
256g
17.6g
144g
163g
30. Sg
157g
Protein Efficiency
Ratio’
NA
NA
230
NA
NA
NA
NA
NA
NA
NA
Biological Value 2
74 3
NA
74 3
74.0
74 0
74 0
NA
74.0
74 0
NA
Calories
165 kcal
191 kcal
222 kcal
210 kcal
259 kcal
182 kcal
160 kcal
343 kcal
576 kcal
272 kcal
Totalfat
36g
7.7g
ll.2g
9.8g
166g
106g
96g
30.5g
49.2g
122g
Saturatedfat
1.Olg
2.8g
42g
3.6g
6.Og
34g
2.7g
107g
174g
3.lg
MUFA=s
12g
34g
4.2g
45g
7.4g
5.Og
36g
14,8g
237g
51g
PUFA=s
08g
0.5g
0.4g
08g
15g
12g
2.7g
37g
58g
32g
18:2 Ltholeic
0.590 g
0.4 10 g
0 340 g
0 680 g
1 260 g
1.040 g
2 420 g
3 130 g
4 890 g
2 738 g
18:3 Linolemc
0.030 g
0.050 g
0 040 g
0020 g
0050 g
0.170 g
0.260 g
0 550 g
0 790 g
0 172 g
20:4 AA
0 060 g
0 050 g
0.050 g
0040 g
0.070 g
0
NA -
NA
0 130 g
0.018 g
20:5 EPA
0.010 g
0
NA
0
0
0
NA
NA
0
0 086 g
22:6 DHA
0.020 g
0
NA
0
0
0
NA
NA
0
0 128 g
Cholesterol
85 mg
89 m
84 mg
79 mg
95 mg
57 mg
64 mg
77 mg
85 mg
112 mg
Zinc
1 00 mg
4.94 mg
5 59 mg
2.48 mg
4 51 mg
2 14 mg
2 15 mg
2 60 mg
3 26 mg
0.66 mg
Toxicology Excellence for Risk Assessment 3-18 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
Iron
1 04 mg
2 12 mg
3 58 mg
0 91 mg
1 40 mg
0 99 mg
1.38 mg
1 72 mg
1 61 mg
0.74 mg
Calcium
15 mg
8 mg
7 mg
17 mg
36 mg
7 mg
26 mg
16 mg
12 mg
20 mg
VitaminA
6RE
0
0
2RE
3RE
0
0
0
0
31RE
Vitamin B 3
13.712 mg
6 340 mg
3 920 mg
5 243 mg
4.070 mg
5 251 mg
NA
NA
7 322 mg
2 129 mg
Vitamin B 6
0 600 mg
0 170 mg
0.440 mg
0 492 mg
0 277 mg
0 340 mg
NA
NA
0 270 mg
0 060 mg
Vitamin B, 2
0 340 ug
2.640 ug
2 570 ug
0 720 ug
1.060 ug
0 830 mg
NA
NA
1 750 ug
1 797 ug
Chordane
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
DDT
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Chiorpyrifos
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Dioxins
NA
NA
0.6 ppt, wet
wt 3
59.3 ppt
wetwt 3
NA
593 ppt
wetwt 3
NA
NA
NA
NA
Lindane
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
PCBs (Total)
416 pg
wet wt
528 pg/ag
wet wt
672 pg/ g
wet wt
NA
NA
NA
NA
NA
N
Meicury
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
‘Protein Efficiency Ratio - gain in weight divided by weight of protein consumed
2 Biological Value - the percentage of absorbed nitrogen retained
3 Research Triangle Institute. 1997 Toxicological profile for Chlorinated Dibenzo-p-dioxins (Draft) p 384
4 Newsome, W H et al 1998 Residues of polychlorinated biphenyls (PCB) in fatty foods of the Canadian diet Food Addit & Contam 15(1) 19-29
NA indicates Not Available
Toxicology Excellence for Risk Assessment 3-19 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Table 3-2 Nutrition Values and Contaminant Levels in Other Protein Sources - values for 100 g edible portion
Hot Dog
(plain)
Burger
(double)
Chicken
(fast food)
Fish
Sandwich
(fast food)
Kidney
Beans
Refried
beans
Pasta
Rice
Tofu
(firm)
Yogurt
(plain)
Cheddar
Cheese
Egg
(boiled)
Protein
10.6g
17.Og
21.9g
10.7g
5.2g
5.5g
48g
27g
158g
35g
249g
126g
Protein Efficiency
Ratio’
NA
230
NA
NA
NA
NA
NA
NA
NA
NA
NA
392
Biological Value 2
NA
743
74 3
NA
NA
NA
NA
NA
NA
NA
NA
93.7
Calories
247 Iccal
309 kcal
303 kcal
273 kcal
81 kcal
94 kcal
141 kcal
130 kcal
145 kcal
61 kcal
403 kcal
155 kcal
Totalfat
148g
l5.9g
18.lg
144g
0.3g
1.3g
07g
03g
8.7g
3.3g
331g
106g
Saturatedfat
52g
59g
48g
33g
005g
05g
Olg
Olg
13g
21g
21.lg
3.3g
MUFA=s
7.Og
69g
7.5g
4.9g
0.02g
06g
Olg
0.lg
19g
09g
94g
4lg
PUFA=s
17g
13g
42g
5.2g
02g
02g
03g
0.lg
49g
Olg
09g
14g
18 2 Linoleic
1 310 g
NA
3.840 g
4.821 g
0 067 g
0 132 g
0 249 g
0 062 g
4 339 g
0.065 g
0.577 g
1188 g
l83Linolenic
0432g
NA
0214g
0399g
0.105g
0021g
0024g
00l3g
0582g
0.027g
0365g
0035g
204AA
NA
NA
0.083g
NA
NA
0
0
0
NA
0
0
0149g
20 5 EPA
NA
NA
0.00 1 g
NA
NA
0
0
0
NA
0
0
0.005 g
22 6 DHA
NA
NA
0023 g
NA
NA
0
0
0
NA
0
0
0.038 g
Cholesterol
45 mg
56 mg
91 mg
35 mg
0
8 mg
0
0
0
13 mg
105 mg
424 mg
Zinc
202 mg
3 25 mg
095 mg
063mg
0 55 mg
117 mg
053 mg
0.49 mg
1.57 mg
059 mg
3 11 mg
1 05 mg
Toxicology Excellence for Risk Assessment 3-20 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
‘Protein Efficiency Ratio - gain in weight divided by weight of protein consumed
2 Biological Value - the percentage of absorbed nitrogen retained
3 Newsome, W H et al 1998 Residues of polychiorinated biphenyls (PCB) in fatty foods of the Canadian diet
NA indicates Not Available
Iron
2 36 mg
2 59 mg
0 91 mg
1 65 mg
1 23 mg
1 66 mg
1 40 mg
1 2 mg
10 47 mg
0 05 mg
0 68 mg
119 mg
Calcium
24 mg
49 mg
37 mg
53 mg
27 mg
35 mg
7 mg
10 mg
205 mg
121 mg
721 mg
50 mg
Vitamin A
0
0
36 RE
19 RE
0
0
0
0
17 RE
30 RE
278 RE
168 RE
Vitamin B 3
3 720 mg
4 690 mg
7.35 mg
2 15 mg
0 502 mg
0.315 mg
1 672 mg
1.476 mg
0 381 mg
0 075 mg
0 080 mg
0 064 mg
Vitamin B 6
0 050 mg
0 180 mg
0.350 mg
0070 mg
0 069 mg
0.143 mg
0 035 mg
0 093 mg
0.092 mg
0 032 mg
0 074 mg
0 121 mg
Vitamin B 12
0 520 ug
1.660 ug
0 410 ug
0 680 ug
0
0
0
0
0
0.372 ug
0 827 ug
110 ug
Chlordane
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
DDT
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Chiorpyrifos
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Dioxins
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Lindane
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
PCBs (Total)
678 pg/p
wet wt.
428 pg/p
wet wt
454 pg/ 3 g
wet wt.
NA
NA
NA
NA
NA
NA
NA
NA
867 pg/ag
wet wt
Mercury
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Food Addit. & Contam 15(1).19-29.
Toxicology Excellence for Risk Assessment
3-21 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
3.12 References
Borrud, L., EC. Wilkinson, and S. Mickle. 1996. What we eat in America: USDA surveys food
consumption changes. Food Review: 14-19.
Candat. 1994. Candat Nutrient Calculation System User’s Manual. Godin, London, Inc.
Dellinger, J.A. et al. 1996. The Ojibwa Health Study: fish residue comparisons for Lakes
Superior, Michigan, and Huron. lox. md. Health 12:393-402.
Exier, J. 1999. Personal communication with Elaine Murkin. 26 April.
Groff, J.L., S.S. Gropper, and S.M. Hunt. 1995. Advanced Nutrition and Human Metabolism.
West Publishing Co. Minneapolis/St. Paul, MN.
Levander, O.A. and R.F. Burk. 1994. Chapter 12: Selenium. In: M.E. Shils, J.A. Olson, & M.
Shike, (eds). Modern Nutrition in Health and Disease, 8th ed. Williams & Wilkins. Baltimore,
MD.
Newsome, W.H. et a!. 1998. Residues of polychiorinated biphenyls (PCB) in fatty foods of the
Canadian diet. Food Addit. & Contam. 15(1):19-29.
Research Triangle Institute. 1997. Toxicological profile for Chlorinated Dibenzo-p-dioxins.
(Draft). p.384.
Sabry, J.H. 1990. Nutritional aspects of fish consumption. A report prepared for the National
Institute of Nutrition. Ottawa, Canada.
Trotter, W.J., P.E. Corneliussen, R.R. Laski, et a!. 1988. Levels of polychlorinated biphenyls
and pesticides in bluefish before and after cooking. J. Assoc. Anal. Chem. 72: 50 1-503.
USDA. 1998. Nutrient Data Laboratory, Agricultural Research Service, Beltsville Human
Nutrition Research Center. Online at: http://www.nal.usda.gov/fnic/foodcomp/
U.S. EPA. 1997a. Food ingestion factors. Exposure factors handbook, Vol. II. Office of
Research and Development. EPA/600fP-95/OO2Fb.
U.S. EPA. 1997b. Appendix E: Dose modifications due to food preparation and cooking. In:
Guidance for assessing chemical contaminant data for use in fish advisories, Volume II. Risk
assessment and fish consumption limits, 2nd ed. Office of Water. EPA 823-B-97-009.
Whimey, E.N. and S.R. Rolfes. 1996. Understanding Nutrition. West Publishing Co. St. Paul,
MN.
Yip, R. and P.R. Dailman. 1996. Chapter 28: Iron. In E.E. Ziegler and L.J. Filer (eds). Present
Knowledge in Nutrition, 7th ed. International Life Sciences Institute. Washington, DC.
Toxicology Excellence for Risk Assessment 3-22 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Zabik, M.E. and M.J. Zabik. 1995. Tetrachlorodibenzo-p-dioxin residue reduction by
cooking/processing of fish fillets harvested from the Great Lakes. Bull. Environ. Contam.
Toxicol. 55: 264-269.
Toxicology Excellence for Risk Assessment 3-23 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
4 Health Risks from Eating Contaminated Fish
4.1 Introduction
Assessing and quantifying the potential risks to human health from eating contaminated fish is
essential to evaluating both the target risks from consuming contaminated fish and the
countervailing risks that may result from consumers following fish advisory advice. Adverse
health effects from contaminants in fish range widely and may include cancer, developmental
and reproductive toxicity, and other systemic effects. The occurrence and severity of the effects
will depend upon the amount to which a person is exposed, and characteristics of the individual,
including genetic makeup and life stage.
Traditionally, risk assessors have calculated estimates of an individual or population’s risk for
getting cancer from exposure to chemicals, while for non-cancer endpoints, a reference dose or
concentration (for contaminants in air) is identified at which one would not expect to see adverse
effects in a population (including sensitive subgroups). Cancer slope factors are estimates of risk
that are derived from dose-response data from laboratory animal or human epidemiology studies.
Traditionally, a linearized multi-stage model has been used to extrapolate from what is observed
at high experimental concentrations to lower environmental exposure levels. This cancer
potency is estimated as the 95% upper confidence limit of the slope of the dose-response curve in
the low dose region. Th s is an upper estimate of risk and the actual risk may be much lower or
even approach zero. EPA proposed revised cancer guidelines in 1996 (U.S. EPA 1996) and
additional proposed guidance in 1998 (U.S. EPA, 1998), which recommend that the mode of
action be considered. The guidance recommends that a linear extrapolation should be used if the
chemical is believed to act via a genotoxic mode of action, if the mode of action is expected to be
linear at low doses, or (as a default) if no mode of action data are available. The guidance also
recommends that a non-linear approach to extrapolation to low doses should be used when
sufficient information on mode of action warrants. For non-cancer effects, a single estimate of a
“safe” dose is identified from animal or human data, using the No Observed Adverse Effect
Level (NOAEL) divided by uncertainty factors to account for extrapolation from animals to
humans, variability in the human population, and deficiencies in the database of studies on the
substance. The resulting Rf [ ) is defined as “an estimate (with uncertainty perhaps spanning an
order of magnitude) of a daily exposure to the human population (including sensitive subgroups)
that is likely to be without an appreciable risk of deleterious effects during a lifetime” (U.S. EPA
1999).
These cancer risk estimates and reference doses have been used to estimate consumption rates of
contaminated fish which would be without appreciable risk to humans following exposure over a
lifetime (U.S. EPA, 1997b). However, in order to compare risks to risks and risks to benefits for
both cancer and non caIlcer endpoints, we thought it necessary to express both types as a risk
estimate. A method to determine how likely it is that adverse effects will occur above the RID
(or with a hazard index (HI) greater than 1 for a mixture of chemicals) is needed. In this
document we use a method for estimating non-cancer risk above the reference dose (RID)
developed by the U.S. EPA and the ChemRisk Division of McLarenlHart (Price et al., 1997).
This is an evolving area of research, however, which needs further development.
Toxicology Excellence for Risk Assessment 4-1 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
To assist in illustrating the framework, six target substances were chosen to develop estimates of
risk above the RID (see Table 4-1). The six were selected based upon the frequency of detection
in a national study of chemical residues in fish (U.S. EPA 1992), and the number of states that
have issued advisories for that substance. Of these six substances appropriate data for
developing these non-cancer estimates were available for methylmercury and chlordane.
Table 4-1. Frequency of residue presence in fish, and the number of states that have issued
advisories for the chosen chemicals.’
Chemical Name
Number of states with
advisories
Percent of sites sampled were
comFound was detected in
fish
DDT and Metabolites
9
99%
Methylmercury
27
92%
Dioxin (TCDD)
22
70%
PCBs
31
91%
Chlordane
24
61%
Chlorpyrifos
--
26%
SourceS U S EPA (1992),
were collected from 388 sites throughout the United States. Sites included target areas near point and nonpoint
sources, sites representative of background levels, and USGS NASQAN sites. A subset of 103 sites were sampled
during the National Dioxin Study (U.S. EPA, 1987) that had been analyzed only for 2,3,7,8 TCDD, but were
reanalyzed for all dioxin/furan congeners and other xenobiotic compounds
This chapter briefly explains the available methods for estimating risk above the RfD and
resulting risk information for the six selected chemicals. More information on traditional risk
assessment methods and calculating risks from contaminants in fish is found in EPA’s Guidance
for Assessing Chemical Contaminant Data for use in Fish Advisories: Volume 2 (U. S. EPA
1 997b).
4.2 Calculating Risk above the Reference Dose for Noncancer Endpoints
The RID approach and measures of hazard such as the Margin of Exposure (MOE) and Hazard
Index (HI) are useful to determine unsafe doses or aid in the evaluation of mixtures and
comparison of chemicals; however, these approaches provide no guidance on risk. The
framework developed in this document looks at both the noncancer and carcinogenic risks from
chemicals together to balance these with the health benefits of consuming fish. This necessitates
using a method to estimate health risk for noncancer endpoints when the RfD or “safe” level is
exceeded.
Unlike for carcinogenicity, there is no noncancer dose-response model commonly accepted for
estimating risks above the RID. For the purposes of illustrating this framework, we estimated
risk using the approach developed by the U.S. EPA and ChemRisk (Price et al., 1997). This
approach was selected because of its ease of use and fidelity to existing EPA sources of
information. Two other possible approaches are briefly presented below. Estimating noncancer
risk is an area that needs further research and development.
Toxicology Excellence for Risk Assessment
4-2
8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
4.2.1 EPA/Ch ’mRisk Model
This approach is the result of a four-year collaboration between the ChemRisk division of
McLarenlHart and the U.S. EPA (Price etal., 1998; Swartout etal, 1998b). It builds on the
existing RfD framework requires minimal new information, and quantitatively deals with
uncertainty and variation in response. The basic premise is that there is generally more
information available than is used to derive the RID and that the same method that estimates the
RfD can be applied to other points on the dose-response curve for humans. This method
postulates a conservative linear threshold model for the dose-response relationship of
noncarcinogenic agents. The approach uses Monte Carlo simulation of uncertainty factor
distributions to develop a family of potential dose-response curves reflecting the uncertainty in
interspecies extrapolation, interindividual variability, extrapolation from subchronic to chronic
responses, extrapolation from LOAEL to NOAEL, and database uncertainty.
The approach is based on three concepts:
• Use the RID as a model of a zero (or minimal) risk (ED 0 - effective dose zero) in humans.
• Use the current syst.-’m of uncertainty factors to predict the dose that causes an effect in a
typical human (ED 50 ),
• Use a linear model as an upper bound to the actual dose response for doses between the ED 50
and ED 0 .
To model the ED 0 , one can use the definition of the RfD where ED 0 is equal to the NOAEL, or
Benchmark Dose, divided by the uncertainty factors for various extrapolations (e.g., intra- and
interspecies extrapolation). These uncertainty factors are defined as distributions
The concept of estimating the dose that causes an effect in a “typical individual” is not usually
considered in noncancer risk assessment; in this approach the dose causing a response in a typical
individual is conceptually similar to a chronic ED 50 . To calculate the ED 50 for humans one could
take the ED 50 in animals (ED 50 a) and divide by the uncertainty factor for interspecies
extrapolation (UFA). In this case the intraspecies uncertainty factor (UFH) is not applied. This
UFA is traditionally vie ed as representing interspecies differences in the NOAEL, not in the
ED 505 The ED 50 a shouL be based on all adverse effects in the test animals, not just the critical
effect.
The third concept -- use of a linear model -- assumes that the fraction of the population that
responds at doses between EDOh and ED5Oh is a linear function of dose in excess of the EDOh.
This assumption will be conservative for compounds with sublinear dose-response curves.
In this model the uncertainties in factors such as UFA and UFH are expressed as distributions.
The model predicts the dose response by randomly selecting values for UFA and UFH and
calculating a response to the dose. The process is repeated several thousand times and a range of
response values is produced for each dose. There is a limitation to this approach in that the
Toxicology Excellence for Risk Assessment 4-3 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
assumption that a linear response is conservative only holds for doses below the ED 50 . The
proposed model cannot be used, therefore, to predict the dose that causes responses greater than
50 percent.
The predicted response should be viewed as a conservative (health protective) estimate of the
probability of one or more adverse effects occurring in an exposed individual. Due to limited
evidence of concordance between effects in animals and humans, however, the effect in humans
should not be assumed to be the same. The linear response will result in an overestimate of risk.
This approach has several strengths: it can produce quantitative estimates of risk; the analysis is
independent of the actual dose-response curve, and the analysis is conservative (linear model
assumption).
Because of the uncertainty in the level of risk at the RID, this model should be used to estimate
doses associated with low risk above the RID with some caution. An estimate of a dose causing
a 5% response (ED 05 ), however, would not be expected to change much. The assumptions upon
which this approach is based are not without controversy. Many toxicologists and risk assessors
are not comfortable with the assumption that the RID is a measure of the ED 0 , while others
believe that the RID is a subthreshold dose. For certain uses, this model is insensitive to whether
the RID compares to zero risk or merely very low risk (ED 0 ooo ). This proposed method
produces a constant measure of risk, is consistent with the RID, and differentiates between
uncertainty and variability.
To illustrate this framework, calculation of risks above the RID were attempted for the six
selected chemicals (DDT, methylmercury, dioxin, PCBs, chiordane, and chlorpyrifos). The risks
are based on the critical effect given in EPA’s Integrated Risk Information System (IRIS). Dose
response modeling was used to estimate the dose causing a 50% response (ED 50 ). Probability
distributions were employed for each uncertainty factor (UF) used by EPA in setting the RID.
Where an uncertainty factor of ten was used in the RID derivation, the reference distribution
developed by Swartout et al. (1998b) to represent the uncertainty factor was used. Where a value
of three was used, the square root of the values from the Swartout et al. (1998b) distribution was
used. In cases where another factor was used, an alternative distribution was used (e.g.,
methylmercury). The specific distributions used are detailed in the sections on individual
chemicals below.
The results of this analysis is an estimate of the dose response for a substance between the
threshold (or minimal risk level, such as the ED 01 ) and the ED 50 . The model assumes that the
dose response is linear over this range, and where a compound has a sub-linear response this
approach will over-estimate the risk. The approach also characterizes the uncertainty in the dose
response that occurs because of the uncertainty in the estimates of the threshold and the ED 50 . In
this analysis, this uncertainty is presented in terms of the median estimate of a response for a
given dose (i.e., the response that has a good chance of being above or below the true risk) and
the 90 percent confidence limits for the response.
Analyses of risks above the RID for chlordane and methylmercury were completed. The critical
studies and UFs for chlorpyrifos are also presented below; however, the analysis could not be
Toxicology Excellence for Risk Assessment 4-4 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
completed due to unavailability of necessary data from the critical studies. Risks above the RfD
could not be estimated for dioxin, DDT, and PCBs due to limitations in the available data.
Details regarding the limitations of the critical study/effect data are provided below.
4.2.2 Other Approaches to Calculate Risk above the RID
Another method to potentially estimate risk above the RID is to adapt the use of a benchmark
dose (BMD). EPA (1995) has defined the BMD as a statistical lower confidence limit for a dose
that produces a predetermined change in response rate of an adverse effect compared to
background. The BMD method attempts to use more of the available dose-response information
by fitting a mathematical model to the data and then determining the dose associated with a
specified response rate of an adverse effect. The resulting BMD value is used as a substitute for
the no effect level and divided by appropriate uncertainty factors to estimate a RID
A number of decisions n ed to be made in applying the BMD method to estimate RIDs and
likewise apply to estimating risks above the RID. For example, for risk above the RID decisions
must be made on which mathematical model to use, what confidence limit to use, and what effect
to model. Furthermore, the choice of uncertainty factor must also be incorporated into the
estimation of risk and model should have some way to approximate the RID as zero risk.
A third method that has 5een proposed for quantitative dose-response analysis for noncancer
toxicity data is that of categorical regression. This involves statistical regression on severity
categories of overall toxicity (Hertzberg and Miller 1985; Hertzberg and Wymer 1991; Hertzberg
1991). By assigning severity categories, all adverse effects may be taken into account rather than
focusing on the critical effect only, as in the previous two approaches. In addition, toxicity data
from multiple studies caii be used in this approach. [ Categorical regression also has the added
advantage of incorporating a severity ranking into its determination, thereby avoiding
unnecessary criticism of the scale severity that we highlight in Chapter 6.]
The results of the regression can then be used to estimate risk above the RID, by providing
information about increasing toxicity with increasing dose rate. Categorical regression may be a
preferred approach for calculating risk above the RfD because it uses more data than the other
approaches. However, the approach is data and resource intensive and has not been done for the
most significant fish contaminants. We recommend that the categorical regression approach be
used for common pollutants found in fish in order to better estimate risks and allow consideration
of benefits.
Confidence in this approach was enhanced by the close proximity of the data to the RID.
Confidence in using categorical regression or BMD modeling to estimate the risk above the RID is
increased when the RID is based on human data (and thus a small uncertainty factor was used).
Greater caution would be needed in the estimation of risks further from doses at which data exist.
BMD and categorical regression are newer dose response modeling techniques for use with
noncancer toxicity data. One can with caution extend the modeling below the data to regions above
the RID — and the closer the extrapolation is to the data, the more confidence one has in the results.
Toxicology Excellence for Risk Assessment 4-5 8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
4.3 Dose Response Information for the Six Selected Target Substances
The framework relies upon estimates of risk for both cancer and non-cancer endpoints. For
illustration in this document, EPA risk estimates from the IRIS have been used (along with the
estimates of risk above the RfD calculated for this project, which were based on IRIS RfDs and
their corresponding principal studies).
Volume 2 of U.S. EPA’s series on Guidance forAssessing Contaminant Data for Use in Fish
Advisories provides guidance on chemical contaminant data for use in fish advisories and on the
development of risk-based meal consumption limits for 25 high-priority chemical contaminants,
referred to as target analytes (U.S. EPA, 1997b). These 25 target analytes were identified by
EPA’s Office of Water as significant based on documented occurrence in fish and shellfish,
persistence in the environment, potential for bioaccumulation, and oral toxicity to humans.
Volume 2 contains a toxicological profile summary for each of the target analytes is provided and
consumption limit tables for adults and children are presented. Instructions for modifying the
consumption limit tables to reflect local site-specific conditions for populations of concern are
given. Separate tables are provided for women of reproductive age for methylmercury and
polychlorinated biphenyls (PCBs). Additional information on risk assessment methods,
population exposure, fish consumption patterns consumption surveys, risk reduction through use
of various preparation and cooking procedures, and risk characterization is presented. Unless
otherwise noted the toxicity data in this chapter are summarized from Volume 2.
In the development of the following framework, a hypothetical example was used which included
estimates for risks above the RfD for methylmercury and chlordane. The details of these
estimations are found below, and were derived using methods summarized below. In addition,
cancer risk was estimated for chiordane since an EPA slope factor is available. Several of the
other chemicals have cancer slope factors, but for none of the others have risk above the RfD
calculations been developed.
4.3.1 DDT and Metabolites (DDE and DDD)
DDT is an organochiorine pesticide that in experimental animals causes cancer, liver damage,
and to a lesser extent leukocytosis and decreased hemoglobin levels. Adverse developmental and
immunological effects have been shown as well as estrogen-like effects on the developing
reproductive system with chronic exposure. Prenatal exposure in experimental animals also
evokes latent effects such as altered learning ability and permanent structural changes in the
brain. Immunological effects have been also observed after short exposures. Some groups of
people may be at greater risk, including children, those with cardiac disease, diseases of the
nervous system or liver, and nursing infants (due to increased exposures).
EPA’s IRIS classifies DDT (and its metabolites DDE and DDD) as B2, probable human
carcinogens by the U.S. EPA. This classification is based upon studies in various mouse strains
and studies in rats. IRIS reports a medium confidence RID of 0.0005 mg/kg-day for liver lesions
based upon a 1950 dietary study with rats. For cancer, the slope factor for oral exposure to DDT
is 0.34 per (mg/kg)/day. This corresponds to a risk specific dose (RSD) of 0.00003 mg/kg-day at
Toxicology Excellence for Risk Assessment 4-6 8/6/99
-------
Cooi rative Agreement with U S. EPA on Comparative Dietary Risk
the 1 in 100,000 risk le ’ l. These values should be used for the sum of the 4,4’, and 2, 4’
isomers of DDT, DDE, and DDD.
To estimate risk from exposures above the RfD, the IRIS RfD was examined. This RID was
derived from a dietary study in weanhing rats in which animals (25/sex/group) were fed
commercial DDT at levels of 0, 1, 5, 10, or 50 ppm for 15-27 weeks (Laug eta!., 1950). The
critical effect was defined by EPA as liver lesions described as hepatocellular hypertrophy,
especially centrilobularly, increased cytoplasmic oxyphilia, and peripheral basophilic
cytoplasmic granules (based on H and E paraffin sections). The NOAEL and LOAEL were
defined as 0.05 mgfkg-day (1 ppm) and 0.25 mgfkg-day (5 ppm), respectively. A total UF of 100
was used representing values of 10 each for interspecies extrapolation and interindividual
variation.
Unfortunately, dose response modeling for noncancer endpoints could not be performed and risks
above the RID for DDT “ould not be quantified because sufficient data were not available from
the critical study (Laug U a!., 1950) regarding the incidence of animals with liver lesions in each
dose group. It is also important to note that the liver changes observed in this study were
subsequently suggested to be adaptive in nature and not representative of actual liver toxicity
(Ortega, 1966). The noncancer risks for DDT may be overestimated by using a RID based on
this study.
4.3.2 Methylmercury
Chronic exposure to merhylmercury (MeHg) produces impairment of nervous system
development in human fetuses, with exposure at sufficient levels evoking cerebral palsy-like
symptoms. Prenatal exposure to lower doses shows more subtle retardation of infant
development. In postnatal chronic exposure from fish consumption, neurological effects are also
exhibited. Symptoms include visual and aural impairment, numbness in the extremities and
around the mouth, impairment of fine motor functions such as writing, speaking, and walking,
and mental disturbances. Chronic oral risk values have been developed by EPA (1999), ICF
Kaiser (ITER, 1999) and the Agency for Toxic Substances and Disease Registry (ATSDR, 1999).
Mice exposed to methylmercury developed kidney tumors in males but not females. However,
carcinogenic effects in mice were observed in the presence of extensive tissue damage.
For methylmercury, the U.S. EPA’s IRIS reports a medium confidence oral RID for chronic
exposure of 0.000 1 mg/kg-day based upon neurological effects on 81 fraqi children who had
been exposed in utero. The mothers had consumed methylmercury-contaminated grain. The
RID is based on a level of exposure estimated by determining the mercury level in hair associated
with the 95% lower confidence limit on a benchmark dose (BMDL) of 10%. The BMDL of 11
ppm maternal hair is adjusted by a dose conversion equation used to relate exposure level to the
concentration of methylmercury in blood and hair, and an uncertainty factor of 10. The U.S.
EPA has not evaluated the risk of carcinogenic effects because of insufficient data.
A site-specific RID distribution (1-99 percentiles) of 0.0003 to 0.00 1 mg/kg-day, based upon a
cohort study of mother- iiifant pairs in the Seychelles Islands has been developed by ICF Kaiser
Toxicology Excellence for Fisk Assessment 4-7 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
and peer reviewed by an independent group of scientists (ITER, 1999). The RfD is based on a
distribution of intakes associated with a BMDL of 10% of 21 ppm maternal hair, a
physiologically-based pharmacokinetic (PBPK) model, and an uncertainty factor of 3.
The Agency for Toxic Substances and Disease Registry (ATSDR, 1999) updated its earlier
Toxicological Profile on Mercury and developed a minimal risk level (MRL) of 0.0003 mg/kg-
day based upon a cohort study of mother-infant pairs in the Seychelles Islands, a NOAEL of
0.0013 mg/kg-day, an uncertainty factor of 3, and a modifying factor of 1.5. ATSDR noted high
confidence that the MRL of 0.0 003 mg/kg-day “is protective of the health of all potentially
exposed human populations” (ATSDR, 1999; p. 258). In its announcement of this new MRL,
ATSDR advises fish consumers, states and other agencies not to revise their existing fish
advisories based on the ATSDR updated profile.
Developing fetuses and individuals with impaired central nervous system (CNS), kidney, or liver
function are particularly susceptible to adverse effects from exposure to methylmercury.
Individuals with inadequate levels of zinc, glutathione, antioxidants and/or selenium are also at a
higher risk.
EPA issued interim guidance in April 1999 to EPA managers directing them to continue to use
the RfD on IRIS until the National Academy of Sciences completes its report in 2000. Therefore,
to calculate the risk above the RID, the data behind the Rif) on IRIS were examined. The
methylmercury RID is based on an evaluation of mother-child pairs from Iraq who were exposed
to methylmercury in grains and bread (Marsh etal., 1987; Seafood Safety; .1991). The critical
effect used by EPA to derive the RfD were developmental effects in infants, including delayed
onset of walking, delayed onset of talking, mental symptoms, seizures, and neurological scores
based on clinical evaluations. Continuous data from this study were placed in 5 different dose
groups and incidence rates were determined. The RID (0.000 1 mg/kg-day) was based on a
benchmark dose of 11 ppm methylmercury in maternal hair. A pharmacokinetic algorithm was
used to transform the concentration in maternal hair to the daily intake of 1.1 ugfkg-day. The
following quanta! data were used to calculate the benchmark dose and the ED 50 :
Dose (ppm in hair) Incidence of Developmental Effects
1.37 5/27
10.0 3/14
52.5 6/13
163 8/12
437 14/15
The MLE (maximum likelihood estimate) of the ED 50 for this data is 117 ppm MeHg in maternal
hair or 0.0011 mg/kg-day. This value was derived using a Weibull model with the threshold set
as zero (Swartout, 1998a).
A total UF of 10 was used for the IRIS RID, which represents values of 3 each for human
variability and database uncertainty, respectively. The UP of 3 for human variability represents
the uncertainty in the ratio of daily intake to hair levels due to human variability in
Toxicology Excellence for Risk Assessment 4-8 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
05
Figure 4-1. Dose-Response Curves for Methylmercury
0.45
04
035
03
0
0
02
0 15
01
0 05
0
00001
methylmercury pharmacokinetics. This uncertainty was characterized by EPA using a one-
compartment pharmacokinetic model and data on interindividual variation in the mode! inputs
(EPA, 1997a). Therefort , the distribution used to characterize this UF was based on specific data
on methylmercury pharmacokinetics (Swartout, 1998a). The square root of the reference
distribution was used to establish the distribution of the database UF.
Figure 4-1 shows the 5 th, 50 ih and 95 th percentile dose-response curves for methylmercury.
Responses at these percentiles for given multiples of the methylmercury RfD are a!so presented
in Table 4-2. Figure 4- and Table 4-2 suggest a much steeper dose-response curve for
methylmercury than for chlordane (see below). As the table shows, at a dose on!y five times
higher than the RfD, the 90% confidence limit suggests that as many as 4% of the population
might respond. At 50 times the RID, there is 50% probability that 44% of the population would
respond, with 90% certainty that the response would range from 28% to greater than 50%.
001
0001
Dose (mg/kg-day)
Toxicology Excellence for Risk Assessment
4-9
8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
Table 4-2: Methylmercury Responses at Multiples of the Reference Dose’
Response
Multiple of
Reference Dose
Dose (mg/kg-
day)
5th Percentile
50th Percentile
95th
Percentile
-
1
0.00011
0.0%
0.0%
0.0%
5
0.00055
0.0%
0.0%
3.9%
10
0.0011
0.0%
4.3%
12.1%
50
0.0055
28.2%
43.7%
>50%
100
0.011
>50%
>50%
>50%
‘Based on the RfD and underlying data as on EPA ’s IRIS (U.S. EPA, 1999). The use of other, more recent, data is
also possible which may lead to a different estimation of the risk above the RfD
4.3.3 Dioxin
Dioxin is a generic term that is used for 2,3,7 ,8-tetrachlorodibenzo-p-dioxin (TCDD). However,
seventeen 2,3,7,8-substituted dibenzo-p-dioxin compounds are grouped together in the interest of
simplicity (U.S. EPA 1987). Dioxin is extremely toxic and targets multiple organ systems in
experimental animals. Some of these effects have also been seen in humans. Effects observed in
animal studies include teratogenicity, fetotoxicity, reproductive dysfunction, carcmogenicity, and
immunotoxicity. Wide differences in the toxic responses to dioxin are seen among species.
There is a great deal of concern over the health effects of TCDD because of its persistence in the
environment, its potency as a carcinogen, and its potential for bioaccumulation (U.S. EPA 1987).
Dioxin is currently under reassessment by the U.S. EPA; for the purposes of this report
information from U.S. EPA (1987) is summarized below.
Dioxins have the highest cancer potency in animals of any chemicals evaluated by the U.S. EPA.
U.S. EPA (1987) reports a cancer slope factor of 1.56 x i0 5 per (mg/k )/day, based on
experimental animal results. This corresponds to an RSD of 6.4 x o ‘mg/kg-day or 2 x i0
mg/ L drinking water at the 1 in 100,000 risk level. EPA at one time calculated a RID of
0.00000 1 mg/kg-day. EPA is in the process of revising its assessment of the risk from dioxin
exposure. This reassessment may likely change both the cancer and noncancer risk values for
this chemical.
To calculate the risk above the reference dose, the acceptable daily intake (ADD of 0.00000000 1
mg/kg/day from EPA’s Ambient Water Quality Criteria Document for 2,3,7,8-
Tetrachlorodibenzo-p-dioxin (EPA, 1984) was examined. This ADI was based on a reproductive
effects reported in a 3-generation rat study (Murray et al., 1979). A LDAEL of 0.00 1 ug/kg-day
was determined for a reduction in the gestation index, decreased fetal weight, increased liver to
body weight ratio, and increased incidence of dilated renal pelvis. This was based on a re-
evaluation of the data from the original study (Nisbet and Paxton, 1982). A total uncertainty
factor of 1000 was used, consisting of interspecies, interindividual, and LOAEL to NOAEL UFs
of 10 each.
Toxicology Excellence for Risk Assessment
4-10
8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
The gestation survival index was the only critical effect data available from the Murray etal.,
(1979) for which dose response modeling could be performed to generate an ED 50 . Gestation
survival data for the fia and fib generations (offspring of two separate matings of exposed
parents) were combined prior to dose response modeling. Dose response modeling was
performed for this data using the THRESH multistage model (ICF Kaiser, 1997). The gestation
survival data failed the goodness of fit criteria for the model and the ED 50 could not be estimated.
Risks above this ADI could therefore not be determined for TCDD using this data.
4.3.4 Polychlorinated Biphenyls (PCBs)
Polychiorinated biphenyls (PCBs) are mixtures of chlorinated biphenyl compounds manufactured
by under the trade name Aroclor with a numeric designation indicating the chlorine content of
the mixture. Manufacture and use of PCBs was banned 1979, however, PCBs are extremely
persistent in the environment and biomagnify via the food chain (U.S. EPA, 1997b).
The majority of mutagenicity assays for PCBs have been negative, but positive cancer responses
are seen with the highei chlorinated congeners (ATSDR, 1997). EPA classifies PCBs as Group
B2, probable human carcinogens, based liver tumors seen in studies in rats exposed to various
Aroclor mixtures.
Because environmental processes such as degradation and transportation alter the composition of
environmental PCB mixtures compared with commercial mixtures, EPA has used a tiered
approach for assessing cancer potency of PCB mixtures and developed several ranges of risk
values. For the “high risk and persistence” tier, EPA calculated a central-estimate slope factor of
1.0 per (mgfkg)/day and an upper-bound slope factor of 2.0 per (mg/kg)/day, based on several
studies of Aroclor 1254 This corresponds to Risk Specific Doses of 0.00001 and 0.000005
mg/kg-day at the 1 in 100,000 level. This range of risk values should be used when exposure is
likely to be through the food chain, soil ingestion, dust or aerosol inhalation, or whenever there is
potential for exposure ii early life (U.S. EPA, 1999). Evaluating exposure and risk from
contaminated fish would use this range of slope factors.
The U.S. EPA has established a chronic oral RID for both Aroclor 1254 and Aroclor 1016. The
medium confidence RID for Aroclor 1254 is 0.00002 mg/kg-day based upon a LOAEL of 0.005
mg/kg-day for ocular ani immunological effects in monkeys. The medium confidence RID for
Aroclor 1016 is 0 00007 mg/kg-day based on reduced birth weights in a monkey reproductive
bioassay. The ATSDR also has established a minimal risk level (MRL) for Aroclor 1254 of
0.00002 mg/kg-day based upon a LOAEL of 0.005 mg/kg-day for an immunological endpoint
(ATSDR, 1999).
To estimate risk above the RID, the Aroclor 1254 RiD was examined (it is more toxic and the
1016 data did not appear adequate for modeling). This RID is based on dermal/ocular and
immunological changes in rhesus monkeys exposed to dietary concentrations of 5, 20, 40, or 80
ug/kg-day Aroclor 1254 for over 5 years (Tryphonas etai, 1989, 1991a,b; Arnold eta!.,
1993a,b). The lowest do e was designated as a LOAEL for these effects. A total uncertainty
Toxicology Excellence for Risk Assessment 4-11 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietaiy Risk
factor of 300 was used based on an interindividual UF of 10, an interspecies UF of 3, a
subchronic to chronic UF of 3, and a NOAEL to LOAEL UF of 3.
The incidence data for the dermallocular effects in each dose group is not available from the
critical studies. Group mean data are available for some of the immune system parameters
measured (i.e., decrease in antibody response to sheep erythrocytes). However, the data from
each individual monkey was needed for conversion of the data to a quantal form that could be
used in standard dose response modeling to generate the ED 50 . Individual clinical records for
each monkey could be reviewed to generate incidence data for the clinical health findings;
however, this was beyond the scope of the current project. Thus, risk above the RID was not
estimated for aroclor 1254.
To date, U.S. EPA has not evaluated data on Aroclors 1242 or 1260; nor has EPA developed
RfCs for any of the Aroclors. The Great Lakes Task Force has developed an interim position on
the noncancer toxicity of PCBs in fish using an average of the RfDs for Aroclors 1016 and 1254.
This value is 0.00005 mg/kg-day. As mentioned above, the U.S. EPA has also published a
revised position on PCB cancer risk assessment on IRIS (EPA, 1999).
A problem with the available risk values is that they are based on the toxicity of commercial
mixtures, and the environmentally-relevant mixtures have not been tested. While use of a
sufficiently similar mixture may be appropriate, the question of what constitutes “sufficiently
similar” for environmental mixtures of PCBs when compared to commercial mixtures needs to
be answered at each individual site or situation. Toxicity equivalency factors have been
discussed for different PCB congeners and some limited conclusions reached (EPA, 1997b). For
the purposes of this report, however, we assume that the toxicity data on PCBs found on EPA’s
IRIS is relevant for the estimation of potential risks from PCB mixtures found in fish.
The results of human epidemiology studies differ somewhat from the non-human primate
toxicity studies in critical effect. PCB mixtures passed to neonates from in utero exposure and
breast milk have been shown to cause developmental defects including cognitive deficits that
persisted at least until 4 years of age in human infants (ATSDR and U.S. EPA, 1997).
A number of epidemiological studies relating fish consumption with deleterious effects have
been conducted over the past two decades. Several cohort studies in the Great Lakes Basin have
been conducted (the New York State Angler Cohort, the Michigan Sports Fisherman Cohort, the
Michigan Maternal/Infant Cohort, and the Wisconsin Maternal/Infant Cohort). These studies
have focused upon possible adverse effects due to fetal PCB exposure from mothers who
consume fish from the Great Lakes. There have been indications of alterations in birth size,
gestational age, and neurological development in these studies. ATSDR and EPA recently
reviewed studies of exposure to PCBs through fish consumption, particularly for fish from the
Great Lakes. The two agencies prepared a paper entitled “Public Health Implications of
Exposure to Polychiorinated Biphenyls (PCBs)” (ATSDR and U.S. EPA, 1999) which concluded
that the “weight of evidence clearly indicates that populations continue to eat fish containing
PCBs and that significant health consequences are associated with consumption of large amounts
of some fish” (ATSDR and U.S. EPA, 1999, p. 2). Health effects include possible reproductive
Toxicology Excellence for Risk Assessment 4-12 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
function may be disrupt?d, neurobehavioral and development deficits in newborns and school-
aged children exposed in utero, other systemic effects that are associated with elevated serum
levels of PCBs, and increased cancer risks.
Swain (1991) compared all four of the cohorts and concluded that PCB exposure from fish could
be correlated with alterations in neonatal health and health in early infancy with reasonable
certainty (Swain, 1991). Swain (1991) found that effects in infant birth weight, maternal health
condition, gestational age. composite activity ranking and McCarthy memory scale deficits had to
be classified as indeterminate, but could not be negated. Swain (1991) also stated that the
relationship between adverse effects on the health status of neonates and infants and PCB
exposure in the Michigan cohort could be causally affirmed and that data from other geographic
locals only tends to support this hypothesis. The available human data could not be used to
estimate a risk above the RfD.
Contrary to the findings of most other studies, Dar et al. (1992) found increasing birth weight
with increased fish consumption for women who gained less than 34 pounds during pregnancy.
Dar et al. (1992) showed a positive correlation between maternal serum PCB levels, and fish
consumption. However, PCB exposures were lower than those in the other studies (Dar et al.,
1992).
In a recent study (Lonky et al. 1996), 395 infant-mother pairs that consumed Lake Ontario fish
were compared to 164 pairs who did not. The exposed pairs were divided into high and low
exposure groups and examined using the NBAS (Neonatal Behavioral Assessment Scale). The
neonates from the high exposure group scored more poorly on the reflex, autonomic, and
habituation clusters of the NBAS. These results confirm the findings of the Michigan
maternallinfant Cohort.
It may be useful to look at these human data to estimate an RfD for PCB mixtures in fish. If
possible, risk above this new RfD might be very useful in future evaluations of the framework.
4.3.5 Chiordajie
Chiordane exposure affects the liver, nervous system, and immune system. Liver effects include
hepatocellular hypertrophy (swelling), hepatic fatty degeneration, hepatocellular adenomas, and
hepatic necrosis. Neurological effects include grand mal seizures and altered EEG results.
Prenatal and postnatal cnlordane exposure may have permanent effects on the immune system
such as a reduction in the number of stem cells. Early childhood exposure to chiordane has been
associated with prenatal and early childhood neuroblastoma and acute leukemia.
Chiordane is classified as a probable human carcinogen (B2) by the U.S. EPA based on oral
studies in animals. IRIS reports an oral slope factor of 0.35 per (mgfkg)/day, which corresponds
to a RSD of 0.00003 mg/kg-day at the 1 in 100,000 risk level. U.S. EPA’s IRIS (1999) reports a
medium confidence oral RfD of 0.0005 mg/kg-day for chlordane based on a 2-year mouse
NOAEL of 0.15 mg/kg-day for hepatic necrosis.
Toxicology Excellence for Risk Assessment 4-13 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
To calculate risk above the RID the IRIS RID was examined. A chronic feeding study in mice
was used to define the RID for chlordane (Khasawinah and Grutsch, 1989). ICR mice
(80/sex/group) were fed 0, 1, 5, or 12.5 ppm technical grade chiordane in the diet for 104 weeks.
The critical effect is defined in IRIS as liver necrosis in male mice. Other cited effects include
increased liver weight, liver cell hypertrophy and fatty degeneration of the liver. Hepatocellular
adenomas were also observed at the high dose group. Hepatic necrosis was observed in male
mice only at the following incidence rates:
Dose (mg/kg-day) Incidence of Hepatic Necrosis in Male Mice
0 7/80
0.15 8/80
0.75 25/80
1.88 27/80
EPA has defined the NOAEL and LOAEL doses as 0.15 mg/kg-day and 0.75 mg/kg-day,
respectively. A total uncertainty factor of 300 was used, which includes factors of 10 for
interspecies extrapolation, 10 for interindividual variation, and 3 for lack of reproductive studies
(database uncertainty).
The ED 50 was calculated using the THRESH multistage model (ICF Kaiser, 1997). To estimate
risks above the RfD, reference distributions were used for the interspecies and interindividual
UFs. The square root of the reference distribution was used to establish a distribution for the
database UF.
Figure 4-2 shows the 5 , 50 th and 95 th percentile dose-response curves for chlordane. The 50 th
percentile curve can be viewed as the estimate of dose response that is equally likely to over
estimate or under estimate response if the substance followed a linear response between the RfD
and the ED 50 . The remaining two curves can be viewed as the 90 percent confidence limits for
the dose response. That is the true response has a 90 percent certainty of falling between the
values. Responses at these percentiles for given multiples of the chlordane RID are also
presented in Table 4-3. As the figure and table show, at a dose of 0.005 mg/kg-day, or ten times
the chlordane RID, the median estimate is 0% with 90% certainty that less than one percent of
the population would respond. Likewise, at 50 times the RID, the median estimate is less than
2% with 90% certainty that less than 10% of the population would respond.
Toxicology Excellence for Risk Assessment 4-14 8/6/99
-------
Cocperative Agreement with U S EPA on Comparative Dietary Risk
Figure 4-2. Dose-Response Curves for Chiordane
05
0.45
0.4
0.35
0.3
0
0.2
0.15
01
0.05
0
00001
0001 0.01 01
Dose (mg/kg-day)
Toxicology Excellence for Risk Assessment
4-15
8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
Table 4-3: Chlordane Responses at Multiples of the Reference Dose
Response
Multiple of
Reference Dose
Dose (mg/kg-
day)
5th Percentile
50th
Percentile
95th Percentile
1
0.0005
0.0%
0.0%
0.0%
5
0.0025
0.0%
0.0%
0.2%
10
0.005
0.0%
0.0%
0.9%
50
0.025
0.0%
1.5%
7.2%
100
0.05
1.0%
3.8%
15.0%
500
0.25
8.7%
22.2%
>50%
4.3.6 Chlorpyrifos
Chiorpyrifos is known to cause cholinesterase inhibition in human blood serum. Other CNS
effects do not occur at similar or lower doses than that causing cholinesterase inhibition. For this
reason, some controversy exists as to whether cholinesterase inhibition in and of itself should be
considered an adverse effect. In experimental animals, cholinesterase inhibition is seen at low
doses and more severe neurotoxicity is seen at higher doses. It is also thought to be fetotoxic.
For chlorpyrifos, IRIS provides a medium confidence oral RID of 0.003 mg/kg-day based on a
NOEL of 0.03 from a 20-day human study reported in 1972 for plasma cholinesterase inhibition
in adult males after 9 days of exposure. The U.S. EPA has not developed a risk value for the
carcinogenicity of chlorpyrifos.
To estimate risk above the RID, the IRIS RfD was examined. The critical study for the RID was
conducted in human volunteers that were treated with chlorpyrifos (4/dose group) at doses of 0,
0.0 14, 0.03, or 0.1 mg/kg-day (Dow Chemical Company, 1972). Doses were administered in a
capsule for 20 days at the low and mid dose and 9 days at the high dose. Treatment at the high
dose was discontinued after 9 days due to runny nose and blurred vision in one individual. Mean
plasma cholinesterase (ChE) in the high dose group was 65% of control. No effect on RBC ChE
was observed at any dose. Decreased plasma ChE after 9 days was considered the critical effect
and EPA defined the NOAEL and LOAEL as 0.03 mg/kg-day and 0.1 mg/kg-day, respectively.
A UF of 10 was used as the standard factor for the range of human sensitivity to ChE inhibition.
The detailed dose response data for the critical study is not available from IRIS. The Dow
Chemical Company report (1972) presumably contains the pretreatment and 9-day plasma ChE
level for each individual subject, but it was not available at the time our analysis was performed.
Incidence data could be generated from this study by assuming that individuals with a greater
than 20% decrease in the plasma ChE level at 9 days are adversely impacted (Dourson et al.,
1997). The quantal incidence data derived from this study could be used to calculate the ED 50 .
Toxicology Excellence for Risk Assessment 4-16 8/6/99
-------
Coo k’rative Agreement with U S EPA on Comparative Dietary Risk
The reference distribution for interindividual uncertainty would be used to estimate risks above
the RfD.
It should be noted that the time scale of the effects of chiorpyrifos is much shorter than chiordane
and methylmercury. Both of those compounds must accumulate in the body and reach a critical
level in order to produce adverse effects in individuals or their offspring. In contrast,
chlorpyrifos is rapidly metabolized and does not accumulate in humans to any great extent. The
adverse effects on which the RID is based are from short term (9-20 days) toxicity studies in
humans. As a result, it may be appropriate to evaluate the risks based on acute exposures (the
day the fish was consumed) rather than annual average doses.
4.4 Multigenerational Study of Great Lakes Salmon Fed to Rats
The vast majority of studies on health effects of contaminants involve single compounds due to
the complexities of testing mixtures of chemicals. However, many human exposures, including
exposure to contaminated fish, involve combinations of chemicals. To help address this issue,
Health Canada initiated a multigenerational study of the health effects in rodents consuming diets
containing Lake Ontariu or Lake Huron chinook salmon (Oncorhynchus.tsawytscha) (Arnold et
a]., 1998a; Arnold eta!., 1998b; Feely and Jordan, 1998; Tryphonas eta!., 1998a; Tryphonas et
a]., 1998b; Pappas eta]., 1998; Seegal eta!., 1998; Iverson etal., 1998). Contaminant levels
exceeded existing standards for commercial fish and seafood for PCBs, dioxins, mirex,
chiordane, and mercury (Feely eta!., 1998). Results of this study are summarized here. It is the
only experimental dose-response study with environmentally relevant exposures to mixtures of
chemicals of interest in fish.
The chinook salmon filk ts used in the dietary formulations for this study were lyophilized and
analyzed for concentrations of the following contaminants: polychiorinated biphenyls (PCBs),
dibenzodioxins (PCDDs), dibenzofurans (PCDFs), polycyclic aromatic hydrocarbons (PAHs),
organochiorine pesticides, metals, volatile organics, chlorinated phenols, and benzenes (Feely
and Jordan, 1998). Levels of contaminants were obtained before and after lyophilizing.
Lyophilized salmon fillets were then incorporated into the normal rat diet in varying proportions
(control, 5% salmon, 10% salmon, 15% salmon, 20% salmon for five different dose groups
including one control. After 70 days on the diet males and females (F o) were mated on a one-to-
one basis within each group (Arnold eta!., 1998b). The (F 1 ) pups were weaned from the dam
after 21 days and then fed the diets for 13 weeks. Seventy days after weaning, one (F 1 ) male and
one (F 1 ) female within each dose group were mated. The (F 2 ) pups were then treated similarly to
the (F 1 ) pups. Randomly selected (F 0 ), (F 1 ), and (F 2 ) adults and neonates were necropsied
(Arnold eta!. 1998b). The study included a reversibility group (F 1 -R) in which the rats were
switched to the control diet after 13 weeks of exposure for 13 weeks (Tryphonas et a]., 1998a).
Increased relative liver i’id kidney weights were observed in both generations and both sexes fed
diets containing 20% Lake Huron or Lake Ontario salmon (Arnold eta]., 1998b). Tryphonas et
a]. (1998a) reported on additional effects. Reduced thymus weights were observed in the Lake
Ontario (20%) female (F 1 -R) reversibility group. Increased growth rates in the (F 1 ) male rats
were observed in those consuming the Lake Huron diets compared with those consuming the
Lake Ontario diets. Reduced, reversible, decreases in counts of red blood cells, white blood
Toxicology Excellence for R!sk Assessment 4-17 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
cells, neutrophil, lymphocytes, and monocytes in the fish-fed (F 1 ) females were seen. This
reduction in counts was greater in the Lake Ontario salmon fed females than the Lake Huron fed
females. Red blood cell, white blood cells, and lymphocyte counts were decreased in the (F 2 )
male rats fed the Lake Ontario (20%) diets compared to the Lake Huron (20%) diets.
Although quantitative aspects of the immune system were affected by the treatments, no
significant effect on its function was observed (Tryphonas et al, 1998b). No significant
behavioral effects were observed in any of the treatments except for one effect observed in the
20% (Fi) Lake Ontario and the (F 2 ) Lake Huron males (Pappas etal., 1998). These males
showed reduced performance in the reference/working memory version of the radial arm maze.
Frontal cortex dopamine concentrations were significantly reduced in all of the fish fed rats
(Seegal et al., 1998). Caudate nucleus dopamine levels were also reduced in all fish fed groups
(Seegal et al., 1998). However, decreases in dopamine levels in the substantia nigra were only
observed in the Lake Ontario (20%) fed rats (Seegal et a]., 1998). Significant effects on all of the
fish fed groups except the Lake Ontario 5% group in levels of norepinephrine concentrations
were observed in the substantia nigra (Seegal, 1998). The same was true for 3,4-
dihydroxyphenylacetic acid (Seegal, 1998).
Overall, the authors concluded that the consumption of the fish diets by rats of two consecutive
generations resulted in a variety of effects that can be described as adaptive responses or of
limited biological significance (Feeley eta]., 1998). Exceptions to their general statement
include potential modification of working and reference memory, an effect on thymus weights
noted in the first generation and an effect of lymphocyte numbers in the second generation. All
of these exceptions occurred at the highest dose. The authors concluded that the risk presented
by the complex mixture of contaminants in salmon collected from two locations in the Great
Lakes could be considered minimal, especially if sport fish consumption advisories are followed.
4.5 Breast Milk as a Source of Contaminants
Breast milk is the ideal source of nutrients for newborns. However, breast milk is also a route of
excretion for some toxic substances and an extremely important route of exposure for the nursing
child. Substances are secreted into the milk by simple diffusion (Klaassen; 1991). Several
factors influence excretion of substances in breast milk.
(1) Three to four percent of human milk consists of lipids. Lipid soluble compounds diffuse
into the mammary gland along with fats from the plasma.
(2) Milk is more acidic than plasma. Therefore, basic compounds are concentrated in the
milk, while acidic compounds have lower concentrations in the milk than in the plasma
(Findlay, 1983; Wilson, 1983).
(3) Compounds that are chemically similar to calcium such as lead and substances that form
complexes with calcium are excreted into the milk (Klaassen, 1991).
(4) Differences in excretion between mammalian species depends upon the amount of lipid
secreted into the mammary gland from the plasma vs. the amount of lipid synthesized de
novo in the mammary gland.
Toxicology Excellence for Risk Assessment 4-18 8/6(99
-------
Coc ierative Agreement with U S EPA on Comparative Dietary Risk
Mercury is an example of au environmental contaminant that transfers to mothers milk. In mice,
the transfer of inorganic mercury from plasma to milk is greater than the transfer of
methylmercury to the milk. However, neonate uptake of methylmercury is greater than uptake of
inorganic mercury. In humans exposed to mercury via dental amalgam and contaminated fish in
Sweden, milk levels of mercury were approximately 30% of plasma levels. Exposure to
methylmercury from rer nt consumption of fish was reflected in the plasma but not in the milk
(Oskarsson et al., 1996)
4.6 Conclusions and Research Needs
For the framework to be most useful, noncancer risks above the RfD must be estimated for all
significant critical effects of chemicals that contaminate fish, in particular, for the contaminant
PCBs. For example, the case study of the Vietnamese immigrant women consuming Lake
Ontario sportfish (discussed later) was severely hampered by our inability to estimate the risks
above the RD for PCBs; this was critical because some exceedances of the RID were as much as
40-fold. Other chemicals need similar investigation.
RfDs are designed to be protective of the critical effect. This means that as long as doses remain
below the RfD, neither the critical effect, nor any other adverse effect associated with the
chemical is expected to manifest itself in the population. When doses exceed the RfD, as the
framework assumes they could, then the critical effect may begin to manifest itself in the exposed
population. The framework uses dose-response information on the critical effect to predict the
increased incidence of the critical effect. But in addition to the critical effect, other effects may
also be seen at higher doses. Some of these may be more severe than the critical effect. At
present, EPA has not de”eloped dose-response relationships for non-critical effects in humans.
For the framework to fully characterize potential risks, and the net possible health benefit of
eating contaminated fish, dose-response relationships for non-critical effects should also be
developed.
Moreover, the method that we chose for determining these risks above the RD (Price et al.,
1997) should be more closely examined. This method has the advantages that it is more
generally applicable than categorical regression and is less resource intensive. It can be used
directly from the existing data as on EPA’s IRIS. However, it is not the only approach to the
problem of risk above the RfD, and as demonstrated, the method does not work for all chemicals.
4.7 Referenr’s
Arnold, D.L., F. Bryce, K. Karpinski, etal. 1993a. Toxicological consequences of Aroclor 1254
ingestion by female rhesus (Macaca mulatta) monkeys. Part 1A. Prebreeding phase: clinical and
health findings. Food Chem. Toxicol. 31(11): 799-810.
Arnold, D.L., F. Bryce, K. Karpinski, et al. 1 993b. Toxicological consequences of Aroclor 1254
ingestion by female rhesus (Macaca mulatta) monkeys. Part lB. Prebreeding phase: clinical and
analytical laboratory findings. Food Chem. Toxicol. 31(11): 811-824.
Toxicology Excellence for Risk Assessment 4-19 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary i isk
Arnold, D.L., F. Bryce, P.F. McGuire, eta]. 1995. Toxicological consequences of Aroclor 1254
ingestion by female rhesus (Macaca mulatta) monkeys. Part 2: Reproduction and infant
findings. Food Chem. Toxic. 33(6): 457-474.
Arnold, D.L., R. Stapley, F. Bryce, eta]. 1998a. A multigeneration study to ascertain the
toxicological effects of Great Lakes salmon fed to rats: Study overview and design. Regul.
Toxicol. Pharmacol. 27: S1-S7.
Arnold, D.L., F. Bryce, D. Miller, et al. 1998b. The toxicological effects following the ingestion
of chinook salmon from the Great Lakes by Sprague-Dawley rats during a two-generation
feeding-reproduction study. Regul. Toxicol. Pharmacol. 27: S18-S27.
ATSDR. MRL for Aroclor. ATSDR website. www.atsdr.cdc.gov/mrls.html
ATSDR. 1999. Toxicological profile for mercury. Update. Atlanta, GA.
ATSDR. 1997. Toxicological profile for polychlorinated biphenyls. Draft for public comment.
Atlanta, GA.
ATSDR and U.S. EPA. 1999. Public Health Implications of Exposure to Polychionnated
Biphenyls (PCBs). Online at: http://www.epa.gov/ostwater/fishlpcb99.html
Dar, E., M.S. Kanarek, H.A. Anderson, et aL 1992. Fish consumption and reproductive
outcomes in Green Bay, Wisconsin. Environ. Res. 59: 189-201.
Dourson, M.L., L.K. Teuschler, P.R. Durkth, et aL 1997. Categorical regression of toxicity data:
a case study using aldicarb. Regul. Toxicol. Pharmacol. 25: 121-129.
Dow Chemical Company. 1972. Accession No. 112118. Available from EPA. Write to FOl,
EPA, Washington DC 20460. (As cited on U.S. EPA’s IRIS database)
Feely, M.M., and S.A. Jordan. 1998. Dietary and tissue residue analysis and contaminant intake
estimations in rats consuming diets composed of Great Lakes salmon: a multigeneration study.
Regul. Toxicol. Pharmacol. 27: S8-S17.
Feely, M.M., S.A. Jordan, and A.P. Gilman. 1998. The Health Canada Great Lakes
multigeneration study - summary and regulatory considerations. Regul. Toxicol. Pharmacol. 27:
S91-S98.
Findlay, J.W.A. 1983. The distribution of some commonly used drugs in human breast milk.
Drug Metabol. Rev. 14: 653-686. (As cited in Klaasen, 1991).
Hertzberg, R.C., and M. Miller. 1985. A statistical model for species extrapolating using
categorical response data. Toxicol. md. Health. 1(4): 43-63.
Toxicology Excellence for Risk Assessment 4-20 8/6/99
-------
Coo ii ’rative Agreement with U S EPA on Comparative Dietary Risk
Hertzberg, R.C. 1991. Quantitative extrapolation of toxicological findings. In: Statistics in
Toxicology. D. Krewski, and C. Franklin, eds. Gordon and Breach Science Publishers. New
York, NY.
Hertzberg, R.C, and L. Wymer. 1991. Modeling the severity of toxic effects. Proceedings of the
84 th Annual Meeting and Exhibition of the Air and Waste Management Association. Vancouver,
B.C , Canada.
ICF Kaiser. 1997a. THRESH: A computer program to compute a reference dose from quanta!
animal toxicity data using the benchmark dose method. KS Crump Division. Ruston, LA.
Iverson, F., R. Mehta, L Hierlihy, et al. 1998. Microsomal enzyme activity, glutathione-s-
transferase-placental form expression, cell proliferation, and vitamin A stores in livers of rats
consuming Great Lakes salmon. Regu!. Toxicol. Pharmacol. 27: S76-S89.
Khasawinah, A. M. and J.F. Grutsch. 1989. Chlordane: 24-month tumongenicity and chronic
toxicity test in mice. Regul. Toxicol. Pharmacol. 10: 244-2 54.
Klaassen, C.D. 1991. Casarett and Doull’s Toxicology: The Basic Science of Poisons. 5 th ed.
McGraw-Hill. New York, NY.
Laug, E.P., A.A. Nelson, O.G. Fitzhugh, et al. 1950. Liver cell alteration and DDT storage in
the fat of the rat induced by dietary levels of 1-50 ppm DDT. J. Pharmaco!. Exp. Ther. 98: 268-
273. (As cited in U.S. EPA, 1999)
Marsh, D.O., T.W. Clarkson, C. Cox, et al. 1987. Fetal methylmercury poisoning: relationship
between concentration a single strand of maternal hair and child effects. Arch. Neurol. 44:
1017- 1022.
Murray, F.J., F.A. Smith, K.D. Nitschke, et a]. 1979. Three-generation reproduction study of
rats given 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) in the diet. Toxicol. Appi. Pharmacol.
50: 241-252.
Nisbet, I.C.T, and M.B. Paxton. 1982. Statistical aspects of three-generation studies of the
reproductive toxicity of TCDD and 2,4,5-T. American Statistician. 36(3-2): 290-298.
Ortega, P. 1966. Light and electron microscopy of dichiorodiphenyltrichioroethane (DDT)
poisoning in the rat liveF. Lab. Invest. 15(4): 657-679.
Oskarsson, A., A. Schuiz, S. Skerfving, eta]. 1996. Total and inorganic mercury in breast milk
and blood in relation to fish consumption and amalgam fillings in lactating women. Arch.
Environ. Health. 51(3): 243-41.
Pappas, B.A., S.J. Murtha, G.A. Park, et a]. 1998. Neurobehavioral effects of chronic ingestion
of Great Lakes chinook salmon. Regu!. Toxico!. Pharmacol. 27: S55-S68.
Toxicology Excellence for Risk Assessment 4-21 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative DietaryRisk
Price, P., R. Keenan, J. Swartout, eta!. 1997. An approach for modeling noncancer dose
responses with an emphasis on uncertainty. Risk Anal. 17(4): 427-437.
Seafood Safety. 1991. Chapter on Methylmercury. Committee on Evaluation of the Safety of
Fishery Products, FDA Risk Assessment and Current Regulations. National Academy Press.
Washington, D.C. p. 196-22 1.
Seegal, R.F., B.A. Pappas, and G.A. Park. 1998. Neurochemical effects of consumption of
Great Lakes salmon by rats. Regul. Toxicol. Pharmacol. 27: S68-S75.
Swain, W.R. 1991. Effects of organochiorine chemicals on the reproductive outcome of humans
WHO consumed contaminated Great Lakes fish: an epidemiological consideration. J. Toxicol.
Environ. Health. 33: 587-639.
Swartout, J.C. 1998a. Personal communication with Mike Dourson. December.
Swartout, J.C., P.S. Price, M.L. Dourson, et al. 1998b. A probabilistic framework for the
reference dose. Risk Anal. 18(3): 271-282.
Tryphonas, H., M.I. Luster, G. Schiffman, et al. 1991a. Effect of chronic exposure of PCB
(Aroclor 1254) on specific and nonspecific immune parameters in the rhesus (Macaca mulatta)
monkey. Fundam. Appl. Toxicol. 16: 773-786.
Tryphonas, H., M.I. Luster, K.L. White, eta!. 1991b. Effects of PCB (Aroclor 7 1254) on
non-specific immune parameters in rhesus (Macaca mulatta) monkeys. mt. J.
Immunopharmacol. 13(6): 639-648.
Tryphonas, H. 1995. The use of non-human primates in the study of PCB immunomodulation.
Hum. Exp. Toxicol. 14: 107-110.
Tryphonas, H., M. Fournier, F. Lacroix, et al. 1998a. Effects of Great Lakes fish consumption
on the immune system of Sprague-Dawley rats investigated during a two-generation reproductive
study: body and organ weights, food consumption, and hematological parameters. Regul.
Toxicol. Pharmacol. 27: S28-S39.
Tryphonas, H., M. Fournier, F. Lacroix, et a]. 1998b. Effects of Great Lakes fish consumption
on the immune system of Sprague-Dawley rats investigated during a two-generation reproductive
study: quantitative and functional aspects. Regul. Toxicol. Pharmacol. 27: S28-S39.
U.S. EPA. 1984. Ambient Water Quality Criteria for 2,3,7,8-Tetrachlorodibenzo-p-dioxin.
Office of Water Regulations and Standards. EPA 440/5-84-007.
U.S. EPA. 1987. National dioxin study. Tiers 3,5,6 and 7. Office of Water Regulations and
Standards. EPA-440/4-87-003.
Toxicology Excellence for Risk Assessment 4-22 8/6/99
-------
Coop3ratlve Agreement with U S EPA on Comparative Dietary Risk
U.S. EPA. 1992. National Study of Chemical Residues in Fish. Office of Science and
Technology. EPA-823-R-29-008a.
U.S. EPA. 1995. Use of the benchmark dose approach in health risk assessment.
EPA/630/R-94/007.
U.S. EPA. 1996. Proposed Guidelines for Carcinogen Risk Assessment. Office of Research and
Development. EPA/6001P-92/003C.
U.S. EPA. 1997a. Mercury Study Report to Congress. Volume I: Executive Summary. Office
of Research and Development. EPA-4521R-97-003.
U.S. EPA. 1997b. Risk assessment and fish consumption limits. Guidance for assessing
chemical contaminant data for use in fish advisories: Volume H, 2h1( ed. Office of Water. EPA-
823-B-97-009.
U.S. EPA. 1998. Carcinogen risk assessment guidelines. Draft for SAB discussion. Online:
http : //www.epa . gov/nceawww 1/raft SABamtg.pdf
U.S. EPA. 1999. Integrated Risk Information System (IRIS). National Center for
Environmental Assessment. Online: http\\www.epa.gov/iris.
Wilson, J.T. 1983. Determinants and consequences of drug excretion in breast milk. Drug
Metab. Rev. 14: 619-65?.
Toxicology Excellence for Risk Assessment 4-23 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
5 Socio-Cultural Considerations of Fish Consumption
5.1 Introduction
This chapter discusses several different ethnic and other groups of people who either consume
more fish than others, consume different parts of fish, or who may fish more contaminated
waters. Included below are discussions on Asian-Americans, Native Americans, subsistence
anglers, and low-income, urban anglers (including African-American and Latino anglers)’
These groups have special behaviors in regard to fish consumption that should be considered in
evaluating risks and benefits of fish consumption. Fish advisories can impact social, cultural,
religious, andlor economic aspects of life that may affect an individual or group’s health and well
being. A framework for evaluating risks and benefits of fish consumption needs to consider
these impacts.
Food, as an important part of a culture, serves economic, social, aesthetic, ceremonial, and
religious functions. Food is used to solidify social ties. Specific foods are often seen as having
special nutritional or medicinal qualities, such as the belief that the consumption of oysters
improves libido. Foods often serve as social class or status markers. Foods are important gift
items. Specific foods, and methods of food preparation, are frequently part of one’s cultural
identity.
Patterns of food consumption are often very resistant to change. When new immigrants arrive in
the U.S. (or elsewhere), many aspects of cultural identity change rather quickly. It is common
for country-of-origin lai guage fluency to be lost by the second or third generation in immigrant
families, for example. However, along with religious practices, food habits are among the most
resistant to change. They often act in a powerful way to build and/or maintain cultural identity.
The use of food to maintain cultural identity is of particular importance for ethnic groups for
whom the consumption of fish is a long-standing tradition.
Fish, as an important cultural resource, may contribute to community well being and
cohesiveness. Fish may hold a prominent place in religious and social ceremonies and rituals.
Fishing activity often involves the intergenerational transfer of knowledge, and may contribute to
sharing and social bonding within the family and community. For some, the consumption of
self-caught fish is an important means of augmenting family food supplies; it has important
economic impacts. In icolated, rural communities, alternate food sources may not be readily
available. In poorer communities, families may lack sufficient income to purchase alternate
foods. For some ethnic groups, especially certain Native American communities, fish hold
‘These group designations are not mutually exclusive, but rather refer to dominant cultural identity (in the case of
the first two) or socioecononuc and residence status (in the case of the last two). “Asian-American/Pacific Islander,”
as used here, also includes individuals who trace their ancestry to the indigenous peoples of Australia and New
Zealand “Native American,’ as used here, includes all indigenous peoples of the Americas, including the Inuit
Subsistence anglers are considered to be those who fish primarily to meet or supplement household food needs
Low-income, urban anglers are included because these form a special at-risk population The small amount of
research data discovered regarding Latino and African-American fishing and consumption behavior is included in
the section on low income, uiban anglers. “Latino,” as used here, refers to individuals whose primary language is
Spanish or Portuguese andlo to those individuals who trace their ancestry to predominantly Spanish or Portuguese-
speaking populations
Toxicology Excellence for Risk Assessment 5-1 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
important cultural meaning and are inextricably linked to traditional religious and cosmological
concepts regarding the place of humans within local ecosystems and the health of the
environment. If fish foods are culturally revered, it may be difficult to conceive of these foods as
“hazardous,” particularly if immediate negative health effects are not perceived to occur. Certain
ethnic groups may also have special concern with the preservation of fishery resources for future
generations and with questions of social inequality (environmental justice) in regard to these.
Fish advisories, especially if they result in rapid culture change, or loss of cultural identity, in a
community, can have numerous catastrophic effects. These may include loss of self-esteem, loss
of social solidarity (sense of community), loss of culturally meaningful social activities, a switch
to less healthy foods, worsened economic conditions, and increases in violence and substance
abuse. Consequently, it is not uncommon for individuals in such communities to resist changing
their consumption habits. Some, especially those living in isolated locations with limited
incomes, may have few if any viable options to consuming locally procured fish.
Certain ethnic groups are also subject to special health concerns. Many individuals of Asian,
African, and Native American ancestry are lactose intolerant, so dairy products are not viable
alternate protein sources. African-Americans have a higher rate of heart disease than the general
population. Switching from fish to meats with a high-fat content would be deleterious for many
of these people. Many Native American communities suffer from exceptionally high rates of
diabetes and obesity. Reduced fish consumption could exacerbate these problems. Chapter 2
discusses possible health benefits of consuming fish, although most studies have not been
conducted to determine differences between sub-populations of the U.S., or other groups.
Because foods, and fish in particular, are incorporated into a complex of socio-cultural
phenomena, changes in eating habits often have multiple, complex (and sometimes unforeseen)
consequences within a particular family, community, or ethnic group. For this reason, special
consideration should be given to subcultural groups known to have special behaviors in regard to
fish consumption. In the U.S., these include Asian-Americans, Native Americans, subsistence
anglers, and low-income, urban anglers (including African-American and Latino anglers).
5.2 Asian-Americans and Fish
While the designation, “Asian-American,” encompasses a diverse range of particular ethnic
populations, several studies have delineated general fish consumption characteristics that
distinguish this group from the majority U.S. population. Allen eta!. (1996), as part of a general
study of fishing behavior and fish consumption in the Santa Monica Bay area of California,
found that Asian-Americans (including Pacific Islanders) exhibited higher rates of fish
consumption, and were more likely than other ethnic groups to eat whole (gutted) fish.
Rockfishes and chub mackerel were the most preferred finfish. Asian-Americans were also
found to consume a greater variety of species and to consume more fish body parts than other
groups.
In a recent study of San Francisco Bay fishing for food activities (Wong, 1997), the majority of
anglers (70%) were persons of color. Asian-American males were most numerous. The average
rate of fishing activity increased with age. An estimated 90% of respondents consumed more
Toxicology Excellence for Risk Assessment 5-2 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
than the recommended amounts of fish from the Bay. Commonly consumed species included
perch, striped bass, white croaker, salmon, and smelt, often cooked whole with the skin on. Crab
was the most common shellfish consumed. Many respondents were unaware of current fish
advisories. On average, Asian-Americans consumed approximately three times the
recommended quantities of fish recommended by the current advisories.
A similar study focused upon Asian-Americans and Pacific Islanders in the Puget Sound area of
Washington State (Nakano, 1996). Recent immigrants (first or second generation) were found to
be most at risk because they are most likely to practice seafood collection, preparation, and
consumption habits closely resembling those in their native countries. The majority of these
individuals do not speak English well, experience high levels of trauma or stress (many came to
the U.S. as refugees), and cannot compete well in the U.S. job market. They are limited in
obtaining adequate access to environmental health-related educational materials and
environmental protection. Many new immigrants identify fishing and self-collection of seafood
as ways to maintain activities familiar to them from their native lands. These activities provide a
sense of cultural continuity. Seafood harvesting may be regarded as a coping mechanism to ease
the oftentimes painful and difficult transition into U.S. society and culture.
Nakano (1996) found that the Asian-Americans and Pacific Islanders they studied had a
preference for a wide range of seafood, including species likely to experience higher levels of
contamination, especially shellfish, bottom feeders like catfish and sole, and sea cucumber. The
consumption of fish heads, internal organs, skin, and cooking water is common among some
Asian-Americans. There is also a clear cultural preference for some body parts that are known to
have higher contaminant concentrations, such as crab hepatopancreas.
Hutchison and Kraft (1994), studying Hmong consumers in the Midwest, found that fishing was
predominantly a male activity. Hmong tend to consume the more easily caught species, and
those that do not require expensive tackle to catch, especially white bass, perch, and, to a lesser
extent, trout. They did not, however, fish for carp or catfish, species otherwise commonly caught
by other lower income anglers. While the researchers state that Hmong fishing (in the Midwest)
is primarily “recreational,” 90% of fish caught were consumed, and consumption rates were
above the norm for this part of the country.
In Laos, fishing is a common, learned activity used to supplement the family diet. This
conception of fishing is, according to Hutchison and Kraft (1994), carried over to the U.S.
Hmong and consumption probably exceeds state guidelines. White bass may be preferred
because they are most analogous to species common in Laos, unlike walleye and other spiney-
rayed fish.
Story and Harris (1989), studying Cambodian and Hmong residents in the U.S., observed that
fish foods serve as a link with the past, ease the shock of entering a new culture, and provide a
means to maintain ethni ’ identity. Consumption of fish, however, was much less than in
respondents’ Asian home lands, with red meats (especially beefsteak) being the preferred
substitute foods.
Toxicology Excellence for Risk Assessment 5-3 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
5.3 Native Americans and Fish
In Native American communities, tribal (ethnic) identity includes culture, religion, and place.
Traditional tribal cultural practices have evolved over long periods of time in tandem with
sustainable associations between humans and other species and their environment. Breaking the
links between an indigenous people and their environment negatively impacts the culture and
religion of these people. Tribal identity is often inseparable from place. Full and safe access to
places and their resources is often necessary to preserve cultural values (Harper, 1997).
In many Native American communities, the perceived sacredness or purity of a place and its
resources is extremely important. Environmental contaminants and the existence of health
advisories may harm this sacredness or purity, making a place and its resources impure. This
negatively impacts cultural, religious, and aesthetic beliefs, sentiments, and values. In many
Native American cultures, personal identity or sense of self is derived more from group
identification than may be true for European-Americans. Native American group identity is
generally very strongly associated with place. Negative impacts upon the resource base of an
indigenous community are likely to have considerable negative impact upon individual identity
and self-worth.
The use of traditional foods can help “boost a lagging cultural morale” (Kuhnlein, 1989, p. 102).
Traditional foods can provide nutrients otherwise lacking in the diet of impoverished peoples. In
general, Native Americans are the poorest ethnic group in U.S. society. They may not have the
means to acquire healthy alternative foods. Many Native Americans have only seasonal
employment; the off-season procurement of traditional foods (like fish) can help families and
communities economically (Kuhnlein, 1989).
Taste, availability, and harvest time help to determine the particular fish species preferred in
Native American communities. Peoples of the Northwest Coast, for example, make considerable
use of ooligan or candlefish (Thaleichthyspacificus). Coastal inhabitants relish this oily food. It
is especially rich in retinol and tocopherol. The fish seasonally migrates up particular coastal
streams in large numbers, making it a very efficient food source. Large quantities can be caught
with relatively little effort. The same is true of the many salmon species native to this region. In
fact, the native peoples of the Northwest are among the most fish-dependent populations in
North America.
According to Berkes (1990), in those Native American communities highly dependent on fish as
a food source, there often exists extensive local traditional knowledge regarding the distributions
and life cycles of particular fish species. A reduction in fishing activities may endanger such
valuable bodies of knowledge. For many indigenous communities, fishing is “a critical
economic activity, not an incidental cultural remnant from the past” (Berkes, 1990, p. 41).
Even when fishing for subsistence purposes is no longer the norm, fish may still serve important,
beneficial social functions. Among the Chippewa of Wisconsin, for example, spring
spearfishing for walleye is an important communal activity. The fishing is done primarily by
men. The season is short (2-3 weeks), but highly productive. Traditional feasts are held, and the
Toxicology Excellence for Risk Assessment 5-4 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
widespread sharing of tI;e catch is an important social activity during this time of the year,
helping to increase social cohesion and cooperation (Peterson et al., 1994).
In several Canadian Native communities, especially among the Cree people, advisories to
suspend the consumption of fish from certain bodies of water (due to high methylmercury levels)
resulted in serious negative sociocultural and health-related impacts. Increased levels of
diabetes, obesity, community and family violence, alcoholism, drug abuse, and suicide have been
reported. The social and cultural disruption that followed the advisories seems to have had more
deleterious impacts than would the continued consumption of locally-procured fish (Wheatley
and Paradis, 1996).
5.4 Subsistence Fishing
In some rural areas of the country, and especially in a large portion of Alaska, fishing for
subsistence purposes is quite common. What has been termed the “mixed subsistence-market
economy” is important in many rural communities. Subsistence fishing and other subsistence
activities may be important domestic (family-based) economic activities. This style of life, in
which part of a household’s needs is met via subsistence activities, is highly valued in rural
Alaska (Wolfe and Walker, 1987). A subsistence-based lifestyle has positive impacts upon an
individual’s self-definition, and sense of self-determination (Egeland et al., 1998)
In Alaska, fish is a primary food staple throughout the state. Salmon species constitute the
majority of fish caught and consumed, but others, including arctic grayling, herring, flounder,
pike, smelt, whitefish, and cod, make substantial contributions to local diets as well (Egeland et
a!., 1998). In a survey of Alaskan Native American communities, it was found that 30-45% of
calories consumed cam from local, self-procured food sources. In Alaska, fish consumption is
six times the national average; the majority of protein consumed is derived from local fish
sources. Social aspects of sharing the fish harvest are very important. Traditional harvesting
activities provide meaningful work, promote self-reliance, help maintain social bonds, provide
economic benefits, enhance cultural identity, and help to sustain the intergenerational transfer of
local knowledge (Egeland eta!, 1998).
5.5 Low-Income, Urban Anglers
A number of recent studies have focused attention upon the fishing and consumption behavior of
low-income, urban anglers because this group may be at higher risk of exposure to fish-borne
contaminants. West (1992), studying fishing along the Detroit River, found that “non-white”
anglers (in this case, mostly African-American anglers) were more likely to view fishing as a
food source than were “white” anglers. They were also more likely to eat species with higher
contaminant levels, such as white bass and sheepshead.
Belton eta!., (1986) observed finfishing and crabbing activities in the New York City area.
Most anglers were older (over 50), employed in blue-collar jobs or service occupations, and
“white.” One-third were retired. The most common species sought by these anglers were
snapper Øuvenile bluefish) and blue crab. These were both among those species contaminated
above FDA tolerance levels. Other species caught included fluke, bluefish, striped bass, and
Toxicology Excellence for Risk Assessment 5-5 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
flounder. Most anglers, nearly 60%, reported eating some of their catch, and many shared with
families, friends, and neighbors. Most often consumed species included blue crab, flounder,
fluke, and snapper, followed by striped bass and bluefish. Most respondents had consumed fish
from these (polluted) waters for 10 years or more. Crabs were most often boiled, fInfish mostly
fried. No respondents reported eating the crab hepatopancreas and crab cooking water was
always discarded.
In terms of local risk perception, two-thirds of the anglers interviewed thought that their catch
was totally safe. About one-fifth saw their catch as slightly polluted but not harmful. Those who
considered their catch fairly polluted said that the fish would, nonetheless, not hurt them, or that
they rarely ate the fish. Most acknowledged that the local waters are polluted. However, some
thought that crabs could rid themselves of pollutants. Others said that the fin fish had moved in
from cleaner waters. Thus, most anglers were able to explain away the risk (Belton et a., 1986).
Belton et a]. (1986) also found that there seemed to be much misunderstanding andlor ignorance
of local fish advisories. One-quarter of the anglers felt that they could effectively assess the
safety of consuming particular animals based upon visual inspection, observation of the animal’s
behavior, smell, and taste. Many felt that washing, cleaning, andlor cooking could make the fish
safe to consume. Approximately one-half said that if you eat fish and one or two days later are
not sick, then the fish was safe to consume.
Burger et al. (1993) reported results of surveys in the New York City area. They found that most
anglers equated unsafe fish with lesions, discoloration, or odor, not with undetected chemical
contaminants. A majority of subjects in the study were African-American or Latino. Most
believed that designated contaminated sites were actually safe, and that the fish caught in these
places were safe to eat. Possible reasons cited for the widespread ignoring (or ignorance) of
health advisories include a low literacy rate among the subject population, language problems
(for Latino anglers), and the inadequate dissemination of health advisory information. Burger et
al. (1993) posited that people underestimate risks associated with voluntary, necessary, and/or
familiar hazards, and overestimate risks from involuntary, unusual, andlor unexpected hazards.
In a similar study, also in the greater New York City area, May and Burger (1996) found that
anglers underestimated the risks of consuming self-caught fish because this was an enjoyable,
voluntary, familiar activity. They observed a common optimistic bias. While an individual
angler might acknowledge that a hazard existed, she or he would feel that she or he had a less
than average chance of experiencing the hazard. There was a common mistrust of government
sources of information, and, faced with uncertain risks, many chose to ignore the risks.
The May and Burger (1996) study involved more African-American respondents. Most
respondents consumed fish fried, and frequently whole. Crabs were generally boiled. There was
a common belief that fish are less contaminated than the waters in which they live. Crabs were
also believed to be able to filter out pollutants. Most anglers admitted that they fished even
though they were aware of the local health warnings. Many preferred self-caught fish because
they were fresher than those purchased in stores.
Toxicology Excellence for Risk Assessment 5-6 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Burger and Gochfeld (1991) examined fishing at a lagoon with high methylmercury levels near
San Juan, Puerto Rico. They discovered that only one person had reported sickness from eating
locally caught fish. Almost all of the anglers questioned were aware of the mercury problem.
Most felt, however, that there was not any serious pollution in the particular places on the lagoon
where they fished. Crab, tarpon (Megalops atlanticus) , and Tilapia mossambica were most
commonly caught. In general, Puerto Ricans fish more, for more of the year, and consume more
fish than U.S. mainland residents. Fish heads and crab hepatopancreas are more often consumed.
Fish and crab are frequently prepared in stews and soups, so essentially all body parts are
consumed.
In Michigan, Smith and Thompson (1989) witnessed persistent angling occurring in the face of
strong warnings to avoid a contaminated portion of the Tittibawassee River. Most anglers were
low-income, blue-collar, and unemployed or underemployed. Just over two-thirds had
completed high school. Most were aware of health concerns, but fished and consumed fish from
the river regardless. Sortie respondents cited no adverse health impacts from past consumption
of locally caught fish. Others expressed fatalistic sentiments. According to Smith and
Thompson, reasons for noncompliance with the posted advisories include possible denial of
recognized truths, fatalism, and alienation. Information overload, in the form of so much
negative news in the mass media, may also have resulted in anglers ignoring one more piece of
“bad” news about their river. For these anglers, fishing in the local river constituted one of the
few recreational outlets hat they had and that they could afford. A general distrust of
government authorities may also contribute to noncompliance with fish consumption advisories.
5.6 Conclusions and Research Needs
In using the framework outlined in this report, it is important to consider how socio-cultural
factors impact the relative risks and benefits of fish consumption. This should include not only
consideration of health-related risks and benefits, but also those related to the economic, social,
and cultural well being of particular communities. Among isolated and/or lower-income groups,
fish may represent an important economic resource, and a source of needed high-quality protein,
that is not easily replaced. For others, especially certain Native American and Asian-American
communities, fish may have special cultural and/or religious significance. In such communities,
advisories designed to limit consumption of fish may have unforeseen detrimental socio-cultural
impacts. These potential consequences need to be considered when assessing the risk and
benefits of fish consumption.
A participatory approach to incorporating socio-cultural factors into frameworks for assessing
the risks and benefits of fish consumption in local communities or among specific target
populations needs to be adopted by risk managers. Socio-culttiral risks and benefits are
ultimately based upon shared community values, and these can be best understood by actively
including members of affected groups in the planning process.
For example, Harper and Harris (Harper, 1999) are developing a process to estimate cultural
consequences of contamination to specific locations or resources. In addition, they are also
working on a universal harm scale, which could be used to help normalize the severity of
Toxicology Excellence for Risk Assessment 5-7 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
disparate risks. This work has not yet been published, but ultimately might be very useful in
conjunction with the framework we propose in the next chapter.
While considerable scientific attention has focused on the biological health risks and benefits of
fish consumption, a relative paucity of concomitant research has been conducted examining
concurrent sociocultural risks and benefits. As delineated in this chapter, for a number of
specific human populations, fish serve important social, cultural, religious, economic, and
aesthetic functions. Fish are integrally positioned within a matrix of shared beliefs, norms, and
behaviors. The more central the position of fish within the social fabric of a community, the
greater the number of these social interconnections.
As the relative importance of fish, fishing, and fish consumption behaviors varies markedly
among the many ethnic and socioeconomic groups comprising the larger U.S. population, there
is a clear need for more comprehensive, comparative analysis of the sociocultural risks and
benefits of fish consumption. More quantitative information needs to be amassed on specific
consumption behaviors, with the aim of more productively combining sociocultural data with
biological data in developing risk assessments and consequent risk management strategies.
There is a need for more detailed empirical data differentiating specific ethnic populations within
larger culture groups, for example, data on the consumption behaviors of Japanese-Americans as
compared to Chinese-Americans, in contrast to data only on Asian Americans in general. The
development of measurement tools (typologies, scales, indices) that will allow for better
comparisons of various sociocultural groups should be a high priority in future research. These
might lead to the eventual development of a theoretical model for better predicting the outcomes
of advisories on specific human populations. There is also a definite need for more research on
environmental justice issues in regard to fish consumption, and on the relationship between fish
consumption and group sovereignty issues, especially in regard to Native American
communities. A related concern that has received only limited attention to date is the influence
of past government relations on the current acceptance of advisories, and other risk management
communication, by specific ethnic and socioeconomic groups.
5.7 References
Allen, M.J., P.V. Velez, D.W. Diehl, etaL 1996. Demographic variability in seafood
consumption rates among recreational anglers of Santa Monica Bay, in 1991-1992. Fisheries
Bulletin 94(4): 597-610.
Belton, T., R. Roundy, and N. Weinstein. 1986. Urban fishermen: managing the risks of toxic
exposure. Environment. 28(9): 19-20, 30-37.
Berkes, F. 1990. Native subsistence fisheries: a synthesis of harvest studies in Canada. Arctic.
43(1): 35-42.
Burger, J. and M. Gochfeld. 1991. Fishing a Superfund site: dissonance and risk perception of
environmental hazards by fishermen in Puerto Rico. Risk Anal. 11(2): 269-277.
Toxicology Excellence for Risk Assessment 5-8 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Burger, J., K. Staine, an:1 M. Gochfeld. 1993. Fishing in contaminated waters: knowledge and
risk perception of hazards by fishermen in New York City. J. Toxicol. Environ. Health. 39: 95-
105.
Egeland, G.M., L.A. Feyk, and J.P. Middaugh. 1998. The use of traditional foods in a healthy
diet in Alaska: risks in perspective. Section of Epidemiology, Alaska Division of Public Health
& Social Services. Anchorage, Alaska.
Harper, B.L. 1997. Incorporating tribal cultural interests and treaty-reserved rights in risk
management. : Fundamentals of Risk Analysis and Risk Management. V. Molak, ed. CRC
Lewis Publishers. Boca Raton, FL:
Harper, B.L. 1999. Personal communication with Michael Dourson. TERA. June.
Hutchison, R. and C.E. ICraft. 1994. Hmong fishing activity and fish consumption. Journal of
Great Lakes Research. 0 (2): 47 1-478.
Kuhnlein, H.V. 1989. Factors influencing use of traditional foods among the Nuxalk people. J.
Can. Diet. Assoc. 50(2): 102-106.
May, H. and J. Burger. 1996. Fishing in a polluted estuary: fishing behavior, fish consumption,
and potential risk. Risk Anal. 16 (4): 459-47 1.
Nakano, C. 1996. Asian and Pacific Islander seafood consumption study: exposure information
obtained through a comr 1 iunity-centered approach. Seattle, WA. U.S. EPA. EPA 910/R-96-007.
Peterson, D.E., M.S. Kanarek, M.A. Kuykendall, et al. 1994. Fish consumption patterns and
blood mercury levels in Wisconsin Chippewa Indians. Arch. Environ. Health. 49 (1): 53-58.
Smith, B.F. and W.N. Thompson. 1989. Environmental sociology: fishermen of the
Tittabawassee. Environment. 26(5): 5, 43.
Story, M. and L.J. Hams. 1989. Food habits and dietary change of Southeast Asian refugee
families living in the United States. J. Am. Diet. Assoc. 89 (6): 800-803.
West, P.C. 1992. Invitanon to poison? Detroit minorities and toxic fish consumption from the
Detroit River. In: Race and the Incidence of Environmental Hazards. B. Bryant, P. Mohai, eds.
Westview Press. p 96-99.
Wheatley, B. and S. Paradis. 1996. Balancing human exposure, risk and reality: questions
raised by the Canadian Aboriginal Methylmercury Program. Neurotoxicology. 17(1): 241-250.
Wolfe, R.J. and R.J. Walker. 1987. Subsistence economies in Alaska: productivity, geography,
and development impacts. Arctic Anthropol. 24(2): 56-81.
Toxicology Excellence for Risk Assessment 5-9 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Wong, K. 1997. Fishing for food in San Francisco: part II, with an analysis of the Bay
Protection & Toxic Cleanup Program. San Francisco Bay Association. Oakland, CA. 40 p.
Toxicology Excellence for Risk Assessment 5-10 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
6 Framework and Case Studies
6.1 Introduction
This chapter presents an initial comparative dietary risk framework (referred to as the
framework) that combines and compares the potential benefits and potential risks associated with
eating contaminated fish. The results of this framework are imprecise, due to the multi-factorial
analysis involved. Thus, while the framework is a quantitative representation of the net risk (or
benefit) associated with eating contaminated fish, it should be used to investigate and compare
various alternative fish protein sources, including perhaps other non-fish proteins. The
framework should not be used in its present form for decisions regarding the merit of specific
fish consumption advisories.
The output of the framework is referred to as the fish consumption index (FCI). The FCI is an
estimate of relative risk. It is not an estimate of absolute risk. In other words, it does-not provide
users of the framework with an estimate of their increased or decreased incidence of a particular
health outcome. It simply provides a mechanism by which users can weigh the health risks
versus the health benefits of eating contaminated fish. Alternate net health risks or benefits of
various food alternatives can then be compared. Cultural benefits of catching and eating fish (or
detriments of not being able to fish or consume fish) may also be considered, however this
framework does not attempt to quantify these benefits.
The framework provides information for a range of fish consumption rates. This allows a user to
determine the range of consumption rates at which he or she may have the largest benefit, the
largest risk, a “net” benefit, or a “net” risk. The user can also determine the fish consumption
rate at which benefits are first affected by the health risks, or the consumption rate at which there
is no net change in the health index. The user can also compare an FCI from one type of
contaminated fish to another.
The framework was designed to be flexible. It can account for multiple health benefits for which
dose response information is available and for as many different health endpoints as information
exists’. When estimating the potential risk associated with chemicals in fish, the framework
considers both cancer and non-cancer effects and is able to consider the presence of multiple
chemicals in fish. Because some health endpoints are considered less severe than others (e.g.,
developing arthritis versus dying of coronary heart disease), a method of incorporating a
modifier to account for the biological differences in the severity of different health endpoints is
needed. The framework also can accommodate a factor to account for personal perceived
differences in severity, and for culture-related benefits of fish consumption, if desired. However,
we did not develop a method for estimating cultural benefits or personal perception of severity in
this project.
The remainder of this chapter describes the goals of the framework and its inputs, and
demonstrates how it could be used with both hypothetical examples and two case studies.
‘However, there are limited quantitative data available on health benefits of consuming fish See Chapter 2 for a
discussion of available data.
Toxicology Excellence for Risk Assessment 6-1 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
6.2 Goals of the Comparative Dietary Risk Framework
This section presents the goals of the framework. As described in Chapter 2, substantial data
exist suggesting that consumption of fish leads to a reduction in the relative risk of several
adverse health endpoints. At the same time, analyses of fish in water bodies throughout the
United States have confirmed the presence of environmental chemicals in fish (AFS, 1997; U.S.
EPA, 1995). In many cases the concentrations of chemicals have been high enough to warrant
the posting of fish consumption advisories by state governments. Although some effort is made
in some of these advisories to describe the benefits of eating fish, the actual advisory is usually
based solely on the potential adverse effects posed by the chemicals in fish, and not on a
consideration of any potential nutritional or health benefit.
Weighing the benefits of fish consumption in setting of advisories is not straightforward. Prior
to having knowledge about the benefits, all one needed to do was estimate the potential risk at
various consumption rates and then select a maximum allowable consumption rate that
corresponded to an allowable risk level. Since consumption of fish also confers health benefits
to people, incorporating information about potential health benefits might be helpful for fish
consumers. One of the goals of this framework is to provide an approach to quantitatively
compare the potential risks and benefits of eating contaminated fish.
The publications to date that have quantified the risks and benefits of eating contaminated fish
have focussed primarily on the increased incidence of cancer and not on other adverse health
effects (e.g., Anderson and Wiener, 1995). Including adverse effects otherthan cancer will
likely increase the estimates of health risk from eating contaminated fish. Thus, another goal of
the framework is to include adverse health outcomes in addition to cancer to more accurately
represent the overall risk. This is especially important because some chemicals for which
advisories exist are judged not to be carcinogenic (e.g., methylmercury). A framework that is
not able to weigh non-cancer risks versus benefits would be of little help to someone evaluating
risks and benefits of fish consumption for such a chemical.
Anderson and Wiener (1995) compared the risks and benefits to adults of eating contaminated
fish. Because they focussed on cancer (for which average daily dose over a lifetime is assumed
to be relevant) as the adverse health effect, estimating risk for adults was appropriate. However,
exposure periods considerably shorter than lifetime and exposures of children, infants and
fetuses (via the mother) are also relevant. Doses received by children (or breast-fed infants
whose mothers are eating contaminated fish) over a short period of time are important to
consider when setting fish consumption advisories for non-cancer health endpoints. Some
existing advisories differentiate between adults, women of childbearing age, and children to
reflect this differences in risk relative to consumption (e.g., Minnesota, 1998). Similarly, the net
benefit of eating fish may differ among these groups (e.g., differences in genetic susceptibility to
cardiovascular disease) and this should be taken into consideration.
Once more than one health endpoint is included in the comparison (whether two or more risk
endpoints, two or more benefit endpoints or different risk and benefit endpoints), a mechanism
must be developed to account for differences in the biological severity and perceived severity of
the different health endpoints. For example, it may be appropriate to treat mortality from cancer
Toxicology Excellence for Risk Assessment 6-2 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
and coronary heart disease as being equally severe biologically, but the perceived severity by
individuals or subpopuhtions of these two outcomes might differ. Thus, another goal of the
framework is to incorpo: ate measures of biological and perceived severity of different health
outcomes in the weighir.g of risks and benefits. We discuss several way to incorporate biological
severity, and show where a scale for personal perceived severity may fIt.
The issuance of fish advisories may pose a risk to the livelihood of certain cultures and
subpopulations whose existence or cultural wellbeing depend upon catching and eating certain
species of fish (see Chapter 5). Ideally the framework would be able to quantitatively account
for the cultural benefits associated with catching and eating fish, in order to weigh these against
the risks.
The final goal of the framework is that it be flexible so that it can be used in a variety of
situations. It should be able to compare the risks and benefits of fish consumption over a wide
range of fish consumption rates, different fish species, different bodies of water, and different
mixtures of chemicals. People using the framework should be able to apply it to a variety of
contaminants and contaminant concentrations within a species. The framework should be able to
easily incorporate new data on either health benefits of eating fish or adverse health outcomes
associated with chemicals in fish.
6.3 Inputs for the Comparative Dietary Risk Framework
6.3.1 Potential Health Benefits of Fish Consumption
Researchers have identified numerous potential health benefits associated with eating fish that
are discussed in Chapter 2. Evidence of benefits can be thought of as arising from two sources.
The first source consists of studies that look at how the change in the incidence of a particular
health outcome is related to fish consumption rate. The results of these studies can be used to
derive a dose-response relationship between fish consumption rate and the health outcome being
investigated, within the limits imposed by the research results. The second source results from
investigation of how ge iieral nutritional status changes as fish is substituted for some other
source of protein or removed from the diet as discussed in Chapter 3. Often, the change in the
incidence of a particular health outcome cannot be quantified from these latter investigations.
This is because the studies conclude that a particular nutritional component (i.e. high density
cholesterol) either increases or decreases with the change in dietary pattern, but they do not tie
the change in the nutritional parameter to a change in a specific health outcome (i.e. incidence of
coronary heart disease)., The absence of a quantitative relationship among fish consumption,
changes in nutritional pcfl ameters, and changes in specific health outcomes makes it more
difficult to incorporate information from these latter types of studies into the framework.
When incorporating fish consumption benefits information, the framework relies primarily upon
results from the first type of study (i.e., Chapter 2). Because change in fish consumption rate
affects many measures of general nutritional health, this report also presents a summary of
nutritional content of numerous protein sources (i.e., Chapter 3). These data are presented to
provide additional perspective about how to interpret the results of the framework. For example,
the framework may indicate that a net health benefit exists when eating contaminated fish at a
Toxicology Excellence for Risk Assessment 6-3 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
particular rate; however, additional nutritional information may suggest that skinless chicken
confers many of the same nutritional benefits as eating fish, but with perhaps a lower level of
contaminants. Such information may be especially useful to segments of the population that are
monitoring one or more nutritional parameters (e.g., cholesterol intake).
For several health endpoints quantitative dose-response data are available. These allow
development of dose-response curves for benefits that relate the change in relative risk of the
health outcome to change in fish consumption rate. These are the data the framework relies upon
to develop an estimate of the net benefit (or risk) of eating contaminated fish. These data are
more fully discussed in Chapter 2.
Because the framework is concerned with the decrease or increase in risk that can be attributed to
consumption of fish, it is the attributable risk, not the relative risk that is desired. The
attributable risk (AR) estimates the excess rate of disease among the exposed and non-exposed
individuals that is attributable to the exposure, while the relative risk (RR) estimates the
magnitude of an association between exposure and disease. The RR also indicates the likelihood
of developing the disease in the exposed group relative to those who are not exposed. Another
way to look at these differences is that RR is the ratio between two incidence rates (exposed and
non-exposed) while AR is the difference between these two incidence rates.
Unfortunately, most of the published data report results as relative risk ratios. Therefore, relative
risk ratios were used in the analysis of the framework in this report. For the purposes of
developing the framework, we chose the relative risk ratios as shown in Table 6-1.
Please note that other values could have been selected. Further study to determine the relative
risks of eating specific types of fish is needed.
Table 6-1. Relative Risks for Various Endpoints listed in Table 2-1.
Health Endpoint
Background
Incidence (B)
Consumption Rates (grams/day)
Relative Risk (RR)
6.5 grams/day 60 grams/day
Coronary Heart Disease
0.32
0.6 0.45
Stroke
0.07
0.85 0.55
Arthritis
0.13
0.92 0.57
6.3.2 Measuring Severity of Health Outcomes and Magnitude of ‘Health Benefits
6.3.2.1 Introduction
The biological severity of a toxic response, based on pathological staging of a disease or
collection of symptoms, must be considered in any framework that attempts to compare the
responses of often disparate effects. However, no one approach can be expected to account for
the totality of the observed effect and the results are only crude approximations of the underlying
biology, subject to change with additional data and judgment.
Toxicology Excellence for Risk Assessment 6-4 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
In addition, the concept of severity also has a societal or personal perception component. Quite
simply, some individuals might rather suffer one type of health effect than another-- -heart
disease versus cancer, for example -- despite the fact that when judged from the biological
perspective of overall impact on the organism, these effects might be considered similar. This
personal perception of severity is important in any comparison of health effects, but is not
considered further in this text other than to show where it can be used as a possible modifier of
the framework results.
In the development of this framework, biological severity is considered directly in the
development of a health index. Several approaches to address the biological severity of toxic
effects have been published and are actively used in several environmental assessment programs,
although not without controversy. For example,
• Within the Superfund office of the U.S. EPA, a 10-value scheme for severity of toxic effect
is used to determine Reportable Quantities (RQs) for noncancer health effects (e.g., DeRosa
eta]., 1985). This scheme has been used since 1983 to determine RQs that are used to
determine the responses to environmental spills in the U. S. Hartung and Durkin (1986) have
also published on the merits of this approach, and suggest ways to make it more general and
usable. Some scientists believe, however, that this scheme incorporates both pathological
staging of severity (the biological component) and personal perception of severity.
• In the development of RiDs and RfCs by EPA and MRLs by ATSDR, a simpler severity
scheme is employed whereby no observed adverse effect levels (NOAELs), lowest observed
adverse effect levels (LOAELs), less serious and serious LOAELs, or Frank Effect Levels
(FELs) are identified (Dourson eta]., 1985; Jarabek, 1994; Pohl and Abadin, 1995). The
identification of these levels is not often recognized as a severity approach per Se, but it does
reflect a crude tool to gauge pathological staging of different environmental effects. One
advantage of this approach is that NOAELs, LOAELs and FELs have been identified for
hundreds of chemicals in the supporting documentation of risk assessment values for these
U.S. agencies. In addition, similar schemes are used by other world health organizations
(e.g., Health Canada; Meek eta!., 1994), and similar lists of NOAELs, LOAELs and FELs
have also been compiled.
• An approach has also been proposed for the effects caused by drugs (Tallarida et a!., 1979).
These investigators 3ssign relative weights to adverse effects of increasing severity based on
physicians’ judgmeii s. This judgment in turn is based on the acceptability that the adverse
effect is likely to be associated with a dose that has a specified probability of curing a disease
of a different severity. This scheme has been considered for use with environmental agents
byDurkin (1999).
• An approach for the development of fish consumption advisories has also been proposed
which incorporates the severity of the effect and the years of life affected, while also
considering the beneficial effects of eating fish (Ponce eta!., 1998). Here investigators use
the benchmark dose to defme a risk curve and a logit model for defining the benefits curve.
A judgment is made as to the “severity” of both risk and benefit on a scale of 0 to 1 (where 0
indicates no significance and 1 indicates loss of life). This severity score is then multiplied
by the number of years of life through which the individual must suffer the risk or enjoy the
Toxicology Excellence for Risk Assessment 6-5 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
benefit. This latter multiplication, often referred to as Quality-Adjusted Life Years (QALYs),
is also being considered by the U.S. Environmental Protection Agency in its deliberations of
comparative risk for disinfectant byproducts (U.S. EPA, 1998). The result of this approach is
similar to the framework proposed in this text.
6.3.2.2 Incorporation of Severity into the Framework
For this framework, we use the severity approach of EPA and ATSDR for estimating RfDs/RfCs
and MRLs. This approach has the advantages of simplicity, familiarity and consistency with the
use of information from EPA’s IRIS, and of ATSDR information found itt its toxicology profiles.
The adaptation of this approach into a multiplier factor for use in the framework is shown in
Table 6-2.
A shortcoming to this approach is the implied equal spacing between levels. There is no
scientific or mathematical justification proposed for a FEL being considered thrice as “severe” as
a less serious LOAEL. This is a disadvantage of the Ponce et al (1998) and DeRosa etal. (1985)
severity schemes as well. Tallarida et aL (1979) addresses this concern somewhat through the
use of physicians’ judgments. Other caveats associated with this choice of severity scale are
shown in Table 6-3.
In like fashion, some modifier to the magnitude of benefits accrued from eating fish needs to be
used in order to roughly compare to the risk of different health endpoints. Such an approach has
been developed for risk/benefit tradeoffs in clinical medicine (Tallarida et al., 1979), and the
scheme by Ponce et al. (1998) uses such a modifier to the magnitude of benefits. For this
framework, we chose to use a simple, scheme that matches the choice of severity ranking for
health risks. Thus, we also rank severity of health outcome avoided (e.g., coronary heart disease)
as none, minimal, moderate or severe, as shown in Table 6-2. As with health risks, we are using
these qualitative labels that are being used in a quantitative fashion in the framework. This is not
an ideal situation.
However, none of the proposed comparative risk schemes solve this problem directly. This is
because the effects of concern in overt clinical disease are not easily comparable with the effects
of concern from widespread environmental exposures. For example, Durkin (1999), has studied
the similarities and differences of effects between clinical disease and environmental exposures
and states that all of the clinical effects covered in the Tallarida et aL (1979) scheme are by
defmition, effects associated with signs or symptoms of toxicity. Thus, these effects would be
classified as FELs or serious LOAELs in environmental parlance. In environmental exposures,
however, anticipated effects are generally not overt (e.g., minimal fatty infiltration of the liver),
or are less severe, adaptive or compensatory. These effects would be classified as less serious
LOAELs or NOAELs.
Toxicology Excellence for Risk Assessment 6-6 8/6/99
-------
Coop.rative Agreement with U.S. EPA on Comparative Dietary Risk
Table 6-2. Severity Ranking of Effects and Benefits and Resulting Multipliers for the
Frameworka
EPA Severity Ranking of Effects Multiplier to the Incidence of EffectlBenefIt
NOEL or NOAEL 0
Less serious LOAEL 1
More serious LOAEL 2
FEL 3
“ Severity” Ranking of Benefits I
None 0
Minimal 1
Moderate 2
Maximum 3
a Please note the intended asst ciation of the term “severity” with “benefits.” In order to balance risks with benefits
within a framework that was z asy to implement, a comparable scaling and terminology was chosen.
The scores in Table 6-2 are multiplied by the available quantitative information on risks and
benefits to yield a modified risk or benefit curve. These modified curves are only expected to be
crude approximations of reality. A number of caveats must be considered before such modifiers
could be used in making final judgments (see Table 6-3). However, this approach was a starting
point that allowed us to develop the framework.
The resulting health scores from use of these multipliers in Table 6-2 have not been further
modified with QALYs. The decision to withhold the use of QALYs was based on practicality.
Quite simply, we chose to see if a framework could be developed using the simplest information
available. If appropriate, the use of QALYs can be added later. The anticipated effect of adding
QALYs on the modified risk or benefit curve is expected to be minimal, however, because the
typical effect or benefit used in the framework is expected to generally occur over a large portion
of an individual’s lifespan. If the comparable risk and benefits occur over significantly different
portions of lifespan, thei3 the lack of use of QALYs becomes more important, and the results of
the framework would need additional study.
Other severity schemes could be used -- and in fact are proposed for comparing the health risks
and benefits of fish consumption. For example, Ponce eta!. (1998) uses further distinctions
among effects and bene ts of different severity than shown in Table 6-2, which necessitates
additional judgment regarding the appropriate severity level of both the critical effect and
benefit. Ponce eta!. (1998) also incorporates the concept of duration of the effect or benefit
through the use of QALYs. However, we do not perceive a great difference between the results
of Ponce et al. (1998) and what is proposed here. If benefits and risks were matched in these
other schemes similarly to what we propose here, the resulting health scores would also be
similar. Moreover, the framework can encompass other severity schemes as appropriate.
Toxicology Excellence for Risk Assessment 6-7 8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
Table 6-3. Caveats with the Use of Severity Schemes Shown in Table 6-2 for Adjusting
Quantitative Information on Risks and Benefits.
Caveat Description
Scheme is too simple The suggested severity scheme is so simple that distinctions are not
possible among, for example, survival of an individual versus survival
of the species through reproduction; such a scheme should enfold
additional complexit?.
Multipliers cannot The use of multipliers implies that effects of a given severity are simple
address severity multiples (or divisors) of other seventies; with a limited severity scale,
this lead to comparisons that do not always make biological sense.
Scheme cannot use Error bars around the “net” health score are not possible because of the
error bars arbitrary value of the multiplier; this makes interpretation of the
appropriate “net” score difficult.
Health scales do not Health benefits and risks are equally matched through the use of the
match same “severity” ranking; this may not be appropriate for effects or
benefits that occur over different durations
Benefits data lack Benefits of fish consumption have been observed in populations
contamination history consuming fish with an unknown contamination level; thus, the net
benefit score may be inappropriately low if all other items are equal.
a See for example the severity scheme for reportable quantities (DeRosa et al., 1985) which gives specific values for
developmental and reproductive toxicity
b Note the method of Ponce etal (1998) specifically addresses the duration issue through the use of number of years
affected by the health endpoint.
6.3.3 Estimates of Human Health Risk
Chapter 4 provides details on estimates of cancer risk, reference doses for non-cancer endpoints,
and calculation of risk above the RfD. These are the inputs needed for the framework. Dose
response information for six common contaminants found in fish (DDT and metabolites,
methylmercury, dioxin, PCBs, chlordane and chlorpyrifos) is provided. EPA’s Integrated Risk
Information System (IRIS) (U.S. EPA, 1999) is the source of RfDs and cancer estimates.
Estimates of risk above the RID were calculated specifically for this project using data from
IRIS.
6.3.4 Dietary Considerations
In order to assess changes in risk to an individual or population with varying consumption of
chemically contaminated fish, a common measure of health is needed. For this framework a
“disability,” or health, scale is the measure against which relative comparisons are made with
regard to chemical contamination and health benefits of fish consumption.
Toxicology Excellence for Risk Assessment 6-8 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Figure 6-1 shows a hypothetical plot of health status with varying protein intake as a percent of
diet. The expected U-shaped dose response curve is presented for protein intake as a percent of
diet, spanning disease on both the high and low ends of protein intake, and normal health status
in between (see curve B).
For a number of reasons, it is difficult to quantify the specific values of the health status scale.
The primary reason is the lack of a good single indicator of health status. However, the lack of a
good measurement does not preclude the use ofjudgment to distinguish the likely effects of how
this function would change if different types and quality of protein were consumed. For
example, if total protein were to come from a source high in saturated fats and salt, and low in
other nutrients, it would be easy to envision a curve similar to C as shown in Figure 6-1.
Alternatively, if the total protein were to come from a source low in unsaturated fats and salt, and
high in other nutrients, it would be easy to envision a curve similar to A as shown in Figure 6-1.
In fact, such curves might be very representative of sole protein sources such as hot dogs (curve
C) or fish (curve A) when compared to an average mixed diet (curve B).
The contamination of these same protein sources with chemicals adds another layer of
complexity to this analysis, but one that can be investigated at least theoretically. For example,
if chemical contamination of hot dogs was low, but of fish was high, then the expected curves of
health status would move towards one another, that is, curves A and C would move closer
together. Although the direction of movement is known, the degree of movement and the
determination of whether the resulting health curves overlap, would necessitate a uniform scale
for health effects.
Such a uniform scale for health effects has been proposed, where organism disability is shown as
a function of target organ impairment (DeRosa eta!., 1989). An adaptation of this curve is
shown in Figure 6-2, where organism disability as a function of target organ impairment is
shown for both insufficient and excess protein intake. Curves for different protein sources (as in
Figure 6-1) could also be drawn here. This single curve given in Figure 6-2 might represent a
balanced (as to source) protein intake. This uniform scale ties in nicely with the proposed
severity modifiers that we discussed in sections 6.3.2.2 (Table 6-2).
Toxicology Excellence for Risk Assessment 6-9 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Figure 6-1. Hypothetical curve: No data are presented nor is the scale likely to be correct. Disability scale as a
function of amount and quality of protein intake as a percent of diet. Curve A is protein intake that is low in fat &
salt, and high in nutrients. Curve B is mixed protein intake (perhaps a normal average diet). Curve C reflects
protein intake that is high in fat & salt, and low in nutrients.
Disease
Disturbed
Health
Normal
Health
Enhanced
Health
0
>
—
.
V)
0
Increasing Protein Intake (% of diet)
100
Toxicology Excellence for Risk Assessment
6-10
8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
Adaptive
Adaptive
Protein Intake (Percent of Diet)
Figure 6-2. Hypothetical curve: No data are presented nor is the scale likely to be correct. Organism disability as a function of
target organ impairment. A uniform scale of NOELs, NOAELs, (LO)AELs and FELs is proposed. Figure adapted from
DeRosa et al (1989).
a)
U
(I )
>..
—
C,)
(LO)AELs
FELs
Limit of compensatory
processes
C
0
U
C
•0
a)
.0
Limit of compensatory
processes
(LO)AELs
Normal function maintained
without significant cost
4,
0%
Adverse Compensatory
NOEL
NOAELs
NOEL
Compensatory
Adverse
+ 100%
Toxicology Excellence for Risk Assessment
6-11
8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
Figure 6-3 Relative risk as a function of intake rate and source of protein Hypothetical curves no
data are presented nor is the scale likely to be correct
Figure 6-3 presents yet another idea for a uniform scale with relative risk r n the y-axis and
intake on the x-axis. The solid line on top indicates “enhanced” health from consuming fish. A
“normal” health status is the solid line in the middle from normal protein intake, and the risk
curve from the chemical in fish is the lowest solid line indicating “decreased” health. The
broken lines in between indicate a hypothetical benefit and risk for pork as an alternative protein
source to fish. Net changes in benefits and risks (shown as dotted lines) might then be compared
amongst protein sources.
Such comparison of net benefits from different protein sources as shown hypothetically in
Figures 6-1 and 6-3 might be considered ideal, because trade-offs among protein sources are
quantifiable. Unfortunately, chemical contamination of different protein sOurces is generally not
known for many chemicals (Chapter 3). Nor are quantifiable benefits datareadily available for
protein sources other than fish (Chapter 2). Because of this, further use of hither of these adapted
scales to compare chemical contamination was not further investigated. This remains a viable
area for future study.
6.3.5 Cultural Considerations
In developing the framework, it is important to consider that social and cultural factors may also
impact the relative risks and benefits of fish consumption. One must consider not only health-
related risks and benefits, but also aspects related to the economic, social, ‘religious, and cultural
well being of particular communities. For example, among isolated andlor lower-income groups,
fish may represent an important economic resource, and a source of needed high-quality protein,
I Benefit from Fish —
Benefit from Pork
Health,
Relative INormal protein intake I
Risk Net benefit/risk for porki
JRisk from chemicals in Pork
LI__________________
skfro he al$i h
Intake
Meals or glday
Toxicology Excellence for Risk Assessment
6-12
8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
that is not easily replaced. For others, particularly Native American tribes or Asian American
communities, fish may have special cultural significance. In such communities, advisories
designed to limit consumption of fish may have unforeseen detrimental socio-cultural impacts.
These potential consequences or countervailing risks need to be considered when assessing the
risks and benefits of fish consumption. Socio-cultural considerations were discussed further in
Chapter 5.
A modifying factor for considerations could be incorporated into the framework below.
However, the magnitude of this factor and how much impact these considerations have on a
community or individual must be assessed on a case-by-case basis and ideally by the community
members themselves. A process and scales for assessing socio-cultural impacts and weighing
them against other health risks and benefits is not available, and developing one was beyond the
scope of this project. In particular, if a cultural modifying factor is employed and a more refined
derivation methodology developed, explicitly including the perspectives and concerns of the
culture in question is strongly recommended. The cultural modifying factor should not be
imposed upon a culture without their consent or involvement. It is expected that the population
will agree with the use of such a factor. Harper and Harris (1999) are developing a cultural
impact scale that normalizes disparate kinds of risk, but this has not yet been published.
6.4 The Benefit/Risk Framework
The simplest representation of health risks and benefits associated with eating contaminated fish
is shown in Figure 6-4. This figure presents the change in several health benefits in the top part
of the figure and the change in health risk as a function of fish consumption rate for several
endpoints in the lower part of the figure. Several measures of benefit and risk are plotted on the
y-axis and fish consumption rate (shown as grams/day) is plotted on the x-axis.
The top part of the fIgure presents the change in benefit, specifically the decrease in risk for
coronary heart disease, arthritis, and stroke with increasing fish consumption. Thus, the curve
(labeled “CHD (Upper Bound)”) indicates that people eating about 20, about 35 and about 60
grams of fish per day had a 12% lower, 16% lower and 38% lower (relative risks of 0.88, 0.84
and 0.62, respectively) incidence of CHD than people consuming 0 grams of fish per day. The
dose response curves shown for these endpoints are based upon results from human
epidemiological studies. These endpoints have been selected because quantitative
epidemiological data ar& available that relate changes in these endpoints to changes in fish
consumption rate. Because the incidence of many of these effects is assumed to decrease with
increasing fish consumption, reductions in incidence can be viewed as examples of the benefits
of eating fish. For each endpoint, an upper bound and lower bound curve are presented to
provide a sense of the range of a particular health benefit These are not statistical upper and
lower bounds (i.e., they are not the upper and lower 95% confidence interval of a relative risk
ratio). Rather they represent the range of best-estimate responses reported by different studies or
of different populations of people within a single study. In the case of coronary heart disease
(CHD), the upper bound represents the best-estimate change in the adjusted relative risk of death
from all causes of CHD reported by Daviglus eta!. (1997). The lower bound represents the best-
estimate change in the crude risk ratio
Toxicology Excellence for Risk Assessment 6-13 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
1E+O1
1E+0O
1 E-O1
1E-02
1E-03
Cancer
Incidence
(log scale)
1E-04
175
Figure 6-4. Relative risk of benefits and toxicity as a function of different amounts of
fish consumed assuming contamination with 2.1 ppm methylmercury and 12 ppm
chiordane. Note different scales for non-cancer and cancer toxicity.
0 25 50 75 100 125 150
Fish Consumption Rate (giday)
Toxicology Excellence for Risk Assessment
6-14
8/6/99
-------
Coo rative Agreement with U.S. EPA on Comparative Dietary Risk
of death due to CHD during a 20 year-long follow-up period reported by Kromhout et. a!.
(1985)2. The upper and lower bound changes in incidence of stroke represent the best-estimate
adjusted relative risk of acute stroke in men and women, respectively, between the ages of 45
and 74 (Gillum eta)., 1996). For rheumatoid arthritis, the upper bound is the change in the best-
estimate adjusted odds ratio for all types of fish consumed by the subject population while the
lower bound is the best ‘estimate change in the adjusted odds ratio when considering only broiled
or baked fish (Shapiro eta)., 1996). Chapter 2 provides further details about these studies.
It is important to recognize that the benefit curves shown in Figure 6-4 are based upon a
selection of the available quantitative data. They do not represent the conclusions of an in-depth
review of all available quantitative health benefit data. Use of data from other studies would
have produced alternative benefits curves (and some studies may not show a benefit at all [ e.g.,
Siscovick eta)., 1995]). Because the studies used to develop the benefits curves shown in Figure
6-4 are for illustrative purposes only, the results shown in this report should also be considered
illustrative and not definitive.
The framework examples in this report and the case studies use the best estimates (i.e. 50 th
percentile of population response) of potential non-cancer risk and health benefit to predict the
net change in health associated with eating contaminated fish. To estimate excess lifetime
cancer risk, the framewerk uses the EPA cancer slope factor (CSF) that represents a 95% upper
bound of the distribution of CSFs calculated by the linearized multistage model. Use of the
upper bound CSF will cause an underestimate of the net benefit (or overestimate of risk) because
an upper bound estimate of risk (derived using the standard conservative toxicity assumptions
employed by EPA) is being compared to a best estimate of benefit. This is recognized as a
conservative bias. However, the framework uses the upper bound because it is what EPA has
available for the majorit of chemicals. In the future, use of the best estimate of the CSF to
calculate cancer risk is preferred in order to derive a more reasonable comparison of health risk
and health benefit data.
Other comparisons are possible to address this bias. For example, the upper bound of potential
risk could be compared lo the greatest estimate of potential benefit to derive an alternative
estimate of potential bejiefIt.
The five curves in the lower portion of Figure 6-4 (originating from the x-axis) present the
change in noncancer risk associated with methylmercury (assumed to be present in fish at 2.1
ppm) and chlordane (assumed to be present in fish at 12 ppm). Noncancer risk is expressed as
the change in incidence of a particular effect in the exposed population. Thus, the upper 95%
confidence bound of the mercury dose-response curve of Figure 6-4 (labeled “Mercury 95%”),
indicates that at a consumption rate of about 60 grams of fish per day, twenty-five percent (0.25)
of the exposed population would be expected to experience the critical effect associated with
methylmercury.
2 Note that if the adjusted risk ratios (instead of crude ratios) from the Kromhout et al study had been used, the
reduction in death due to Cl-it’ would have been slightly less than shown in Figure 6-4 at low fish consumption rates
and greater than that shown at high fish consumption rates (Kromhout eta], 1985)
Toxicology Excellence for Risk Assessment 6-15 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary ‘Risk
Unlike most non-cancer risk assessments, which simply assume that exceedence of the RID is
unacceptable and do not estimate the incidence of non-cancer effects above the RID, the five
lower curves on the this graph estimate the incidence of adverse non-cancer effects caused by
both methylmercury and chiordane. The method for calculating risks above the RID is described
in Chapter 4, which also presents the actual calculations for several chemicals of interest. For
methylmercury, three curves are shown: the upper 95% bound of population response (labeled
“Mercury 95%”); the best estimate of population response (labeled “Mercury 50%”); and, the
lower 5% bound(labeled “Mercury 5%”). For chlordane, the upper 95% bound (labeled
“Chiordane 95%”) and best estimates (labeled “Chiordane 50%”) are shown. The bounds refer
to the lower 5%, best estimate (i.e., 50%) and upper 95% bound of the dos response curve for
the critical noncancer effect associated with methylmercury and chlordane. It is important to
recognize that, with the exception of the noncancer risks curves discussed above, uncertainty in
health benefits and risks is not dealt with explicitly by this initial version of the framework. An
important future refinement of the framework would be explicit consideration and quantification
of uncertainty surrounding estimates of potential health risk and benefit. These uncertainties
could be addressed by considering different benefit curves than the ones we chose or varying the
chemical concentrations or mixtures of chemicals in fish. Some of this variations are shown later
in this chapter.
The potential cancer risk associated with chlordane is also presented in Figure 6-4. It is shown
as the straight line in the middle of the graph labeled “Chiordane Upper 95% Cancer Risk”.
Note that the scale for increased cancer risk is shown on the right-hand side of Figure 6-4. Thus,
at a consumption rate of about 20 grams per day of fish, the increased cancer risk is about 1x10 3
and the increased risk approaches 1x10 2 as the consumption rate approaches 200 grams per day.
The change in cancer risk is shown on a separate scale because it would not have been visible on
the scale used for the other non-cancer endpoints. Changes in non-cancer effects represent an
percent level increase in a person’s risk of manifesting the critical effect associated with a
chemical, while an excess cancer risk of even as high as one in one thousand (1x10 3 ) represents
an increase in risk of only a tenth of a percent.
Figure 6-4 illustrates the complexity of capturing the relative changes in risk or benefit as a
function of fish consumption. Note that the six benefit curves on the top portion of the figure are
independent of the concentrations and types of chemicals in fish, to the extent that the chemical
contamination of the fish in these studies was generally not known. Thus, hey are assumed to
represent fixed health benefits associated with eating fish. 3 The five noncapcer risk curves and
one cancer risk curve on the lower portion of the graph will change as the types and
concentrations of chemicals change. The illustration presented in Figure 6-4 estimates potential
risk from just two chemicals (methylmercury and chlordane) at fixed concentrations in the fish of
2.1 mg/kg and 12 mg/kg, respectively. The chemicals and concentrations,do not represent any
particular site. They were chosen simply to provide an example of how the framework can be
Actually, they include any potential adverse effect associated with chemicals in the fish, though information about
chemical concentration in fish is not available for most benefit studies. To the extent such chemicals are present and
that they directly impact the change in benefit incidence, these benefits curves might reprrsent net benefit already;
fish with less chemical contamination might be associated with even greater benefits
Toxicology Excellence for Risk Assessment 6-16 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
used. The shape and slope of the cancer and noncancer risk curves is a direct result of the types
and concentrations of chemicals in fish.
Figure 6-4 is already quite complicated and yet it only presents the benefits and risks associated
with consuming fish containing a specific set of chemicals at specific concentrations over a range
of fish consumption rates. Figure 6-4 also does not capture all of the possible health benefits
information available (see Chapter 2). Nor does it capture situations where the identity of
chemicals and their concentrations vary. Indeed, it is very difficult to combine all this
information to determine whether a net benefit exists. This problem becomes more complex
when fish of different chemical concentrations are considered, because multiple versions of
Figure 6-4 could then be drawn. In other words, a public health official modifying an existing
risk-based advisory might have difficulty deciding whether to modify the advisory and if so, by
how much, based on Figure 6-4, or its many versions. Nonetheless, for the framework to be of
greatest use, the multiple benefits and risks need to be combined and a net health outcome needs
to be derived. We approach this problem by developing separate algorithms of benefit, risk and
their combination.
6.4.1 Algorithm for Health Benefits
For each health endpoint where fish consumption has been shown to improve health, we develop
a quantitative algorithm for estimating the benefit. The benefit is a function of the background
incidence of that health endpoint in the U.S. population, the relative reduction in risk of that
endpoint caused by eating fish, the biological severity of that health endpoint, and the amount of
fish eaten. The equation used to calculate the benefit for any particular endpoint at a given fish
consumption rate is:
[ B, x (1-RR )] x S 1 = Benefit,
Where:
B 1 is the background incidence of health endpoint i (see Chapter 2);
RR 1 is the relative risk of health endpoint i at the given consumption rate (see
Table 6-1);
S i is the biological “severity” of health benefit endpoint i (see Table 6-2);
and,
Benefit 1 Is the possible benefit for health endpoint i associated with eating a given
amount of fish.
Background incidences of various health endpoints are available from a variety of sources.
Relative risks associated with fish consumption are summarized above (Table 6-1) and all
readily available quantitative data are presented in Chapter 2. As described above, the benefits
from fish consumption for different health endpoints will vary in their biological and perceived
“severity” (health risks that were the basis for the calculation of risk above the RfD vary in
severity in a like manner). For the purposes of illustrating this framework, we assigned a score
to the biological “severity”, or magnitude of the disease avoided using the values presented in
Table 6-2 and discussed earlier. Severity of benefits (and risks) must be included in the
Toxicology Excellence for Risk Assessment 6-17 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
calculation of the FCI in order to add benefits and risks of disparate effects and diseases. For the
presentation of the framework here, we did not attempt to incorporate personal or societal
perception of severity. The biological “severity” scores used for the benefits in the framework
range from 0 to 3, with a higher score being assigned to reduction of more severe disease.
The health benefit associated with eating fish is expressed as a unitless positive number and is
plotted on a health scale. The number is positive because a reduction in an adverse effect is
assumed to represent an improvement, as opposed to a decrement, in health. As described
below, risk from consuming chemicals in fish is expressed as a negative number to connote an
anticipated decrement in health.
When a benefit associated with fish consumption exists for more than one health endpoint the
framework calculates a total benefit by summing the benefits associated with each individual
health endpoint using the equation shown below:
n
{ [ B 1 x (1-RRJ] x S 1 } = E Benefit
i=1
The framework can also be modified to account for the cultural benefits of eating fish as
described below.
n
{ [ B, x (1-RR 1 )] x S 1 } x C = E Benefit
i=l
All the parameters are the same except for the addition of a cultural factor “C”. The cultural
factor represents the cultural value associated with fish consumption (this ,could also represent
religious or social benefits). For use in the framework, the cultural value is expressed relative to
the health benefits because it modifies the predicted total health benefit. Thus, if a particular
subpopulation decides that the cultural benefits of eating fish are equal to the health benefits,
then the total benefit of eating fish would be twice the health benefit alone and “C” in the above
equation would be assigned a value of 2. Other ways to incorporate the cultural benefits of
eating fish are also possible. For example, instead of multiplying the total benefit by “C”, the
constant “C” could be added to the health benefits. Addition of “C” suggests that the cultural
value of fish consumption is constant across all fish consumption rates while multiplication (as
shown in the above equation) connotes that cultural benefit follows health benefit and increases
with increasing fish consumption rate. A third alternative is to have cultural value be very high
at low consumption rates but decrease with increasing fish consumption rate. Such a relationship
may represent a situation where fish are essential in ceremonies that mark a subpopulation’s
continued existence but do not have to be a large fraction of that particular culture’s daily diet.
An “objective” scale that can be applied to measure cultural benefits has not been developed for
this framework. This must be determined on a case-by-case basis, ideally by the individuals and
populations themselves. Obstacles to developing such a quantitative factor include measurement
of physical, emotional and mental well being with the disruption or enhancement of a “cultural”
Toxicology Excellence for Risk Assessment 6-18 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
practice such as catching or consuming fish. Quantitative data are not available, but the
population itself may have a qualitative judgment about the negative or positive consequences of
a cultural practice. For example, a tribe that relies heavily on locally caught fish, could examine
the consequences to the population’s health (e.g., effects of use of replacement foods), or to the
continuation of its traditional lifestyle. From the perspective of the cultural value of fish, the key
aspect of the framework is that it contains the flexibility to incorporate the cultural importance of
fish and to weigh that importance against potential health risks. There are many possible ways
this important parameter could be included. A specific approach for estimating “C” has not been
developed for this project, although others are investigating ways to estimate cultural
consequences (Harper, 1999).
6.4.2 Algorithm for Health Risk
The process used to derive a single estimate of risk from chemicals in fish parallels that used to
derive a single estimate of the benefit associated with eating fish. For each chemical and single
adverse effect it causes, the increased risk associated with contaminated fish is calculated using
the following equation:
(R 1 x S 1 ) x (-1) Risk 1
where:
R 1 is the increased risk of health endpoint i associated with a particular fish
consumption rate,
S 1 is the biological severity of health endpoint i; and,
Risk is the decrease in health (because of the increase in risk of health endpoint
i) associated with eating a given amount of fish.
Risk (R 1 ) is the increased risk of health endpoint “i”, above the background incidence, which is
assumed to be caused by exposure to chemicals in fish (see Chapter 4). The severity score (S,) is
the same as described above (see Table 6-2). Risk 1 is the change in health associated with eating
fish containing a chemical that causes an increase in endpoint i and is expressed as a unitless
negative number. The number is negative because an increase in an adverse effect leads to a
decrement in health.
When a risk associated with fish consumption exists for more than one chemical, or a chemical
causes more than one adverse effect, the framework calculates a total risk by summing the risks
associated with each individual chemical (or for each endpoint caused by a single chemical)
using the equation shown below:
n
[ (R 1 x S 1 ) x (-1)] = Risk
1=1
Note that both cancer and noncancer risks are added after adjustment by the biological severity
index S 1 . For example, the increased incidence of the critical (noncancer) effect associated with
Toxicology Excellence for Risk Assessment 6-19 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
methylmercury is added to the increased risk of cancer and noncancer effects from chiordane.
Once severity is considered, the resulting risk curve cannot be viewed as the possible increased
incidence of a specific effect in the exposed population or an individual’s increased risk of
manifesting a specific effect.
6.4.3 Algorithm for the Fish Consumption Index (FCI)
To estimate the net health effect of eating contaminated fish, the framework sums the total
benefit and total risk to derive the Fish Consumption Index (FCI) using the following equation:
Benefit + Risk = FCI
The FCI is plotted over a range of fish consumption rates to establish the relationship between
change in health and fish consumption (Figure 6-5). As described above, the FCI is an estimate
of relative risk. It is not an estimate of absolute risk. Nor does it provide users of the framework
with an estimate of their increased or decreased incidence of a particular health outcome.
However, the FCI does provide a simple mechanism by which users can weigh the health risks
versus the health benefits of eating contaminated fish. It also accounts for differences in severity
of the different endpoints. Because the framework provides this information for a range of fish
consumption rates, users will be able to determine the range of consumption rates at which they
may have the largest benefit, and the largest risk. Consumers will also know the possible net risk
or net benefit across consumption rates, or the consumption rate at which the benefits of fish
consumption are first affected by the health risk. 4
Note that if cultural benefits or personal perception of severity are included in the framework, the
FCI is not strictly a health index, but rather represents a combination of health risks and benefits,
personal perception, and cultural benefits and risks.
6.5 Demonstrating the Framework
This section presents a quantitative hypothetical example of how the framework can be applied.
The example is hypothetical and is selected to illustrate particular aspects of the framework that
may be useful. Other hypothetical examples are presented to illustrate various aspects of the
framework, including impacts of changing levels in contaminant concentrations, evaluation of
different subgroups, consideration of mixtures of chemicals and multiple endpoints, and
inclusion of cultural benefits. Detailed examples of applying the framework to real world
situations are presented in section 6.6.
In addition, it provides the user with the tool for comparing risks from different diet options; however, lack of
contaminant data in other foods currently limits one’s ability to do this
Toxicology Excellence for Risk Assessment 6-20 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Figure 6-5. Health Scale as a Function of Fish Consumption Rate. Data are Derived from Figure 6-4
as Explained in the rext. Dashed lines are Extrapolated Values.
Risk from eating fish
-05
‘ .4
0 25 50 75 100 125 150
Fish Ccnsunition Rate (g/d)
1
075
05
025
-025
-075
-1
Toxicology Excellence for Risk Assessment
6-21
8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary’ Risk
6.5.1 Quantitative Example of the Framework
As described above, the hypothetical example used in the this document assumes fish contain 2.1
ppm (mg/kg) of methylmercury and 12 ppm (mg/kg) of chlordane. 5 Consumption of fish is
assumed to decrease the incidence of three health endpoints: coronary heart disease, arthritis, and
stroke. The magnitude of the reduction (i.e., of the relative risk) depends upon the rate of fish
consumption and is based upon data discussed in Chapter 2. As described above, the health
benefits assumed by the framework to be associated with increased fish consumption are based
upon a somewhat arbitrary choice of studies from the literature. They do not represent the
conclusion of an in-depth evaluation of all the available data. Arthritis is judged to be the least
severe of the three endpoints and is assigned a severity score of 1. Coronary heart disease and
stroke are judged to be the most severe and are assigned a severity score of 3. Cultural benefits
are not included in this example. The calculations used to develop this example are shown
below in Table 6-4.
Table 6-4. Input Parameters To Estimate Benefits
Health Endpoint
Background
Incidence (B)
“Severity”
Score (S)
Consumption Rates (grams
/day)fRelative Risk (RR)
Coronary Heart Disease
0.32
3
6.5/0.6 and 60/0.45
Stroke
0.07
3
6.5/0.85 and 60/0.55
Arthritis
0.13
1
6.5/0.92 and 60/0.57
Risk also depends upon the rate of fish consumption. Increased lifetime cancer risk is estimated
using standard EPA exposure and toxicity assumptions (e.g. a CSF of 0.35 per mg/kg-day for
chiordane, a body weight of 70 kg, and a 70-year exposure duration). Increased risk of the non-
cancer effects of chlordane and methylmercury are estimated using exposure assumptions
identical to those used to estimate increased cancer risk combined with the”risk above reference
dose” technique described in Chapter 4. Table 6-5 below summarizes the inputs used to estimate
risks for this hypothetical example.
Note that the benefit curve shown in Figure 6-5 becomes flat at a fish consumption rate of about
60 grams per day and is drawn as a dashed line for higher consumption rates. This is because
feW studies have quantified the benefits of fish consumption at specific consumption rates of
greater than about 60 grams per day. Most studies report the maximum consumption rate as
greater than some specific rate (i.e., more than two meals per week). The benefit curves assumed
the highest consumption rate to be equal to the “greater than” consumption rate reported by a
particular study. Based upon the absence of specific data on high consumption rates and some
minimal evidence that some benefits appear to be leveling off (Figure 6-4) at the higher
5 We chose these values for no particular reason Other values could be used in the development of this hypothetical
example
Toxicology Excellence for Risk Assessment 6-22 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Table 6-5. Inputs Parameters To Estimate Risks
ChemicallHealth Endpoint
Chlordane/cancer
Severity
Score (S)
3
Consumption Rate
(grams/day)/Increased Risk (R)
6.5/1x10 3 and 60/0.0 1 and 120/0.02
Methylmercury/nori-cancer:
neurological_abnormalities
1
6.5/0.02 and 60/0.12 and 120/0.30
Chlordane/non-cancer: hepatic
necrosis
3
6.5/0 and 60/0.001 and 120/0.01
consumption rates, it is assumed that the possible benefits remain constant at higher consumption
rates. This is recognized as a conservative assumption, and it is important to recognize that
possible benefits at high consumption rates may be underestimated by the framework.
Conversely, by ascribing the benefits reported for consumption rates of “greater than 60 grams
per day” to a consumption rate of 60 grams per day (see Figure 6-5), the framework may be
overestimating benefits at an actual consumption rate of 60 grams per day. The increased
benefits reported by a study for the population of people eating “greater than 60 grams per day”
may be occurring in people who are actually eating 90 or 100 grams per day.
Note that unlike the benefits curve, the slope of the risk curve becomes steeper at higher
consumption rates. This is based on the slope of the non-cancer dose response curves that
become steeper with increasing dose. The FCI, therefore, generally decreases after about 60
grams of fish per day in this hypothetical example.
The benefits and risks from each chemical can then be summed to derive the FCI and the result
plotted against fish consumption rate (Figure 6-5). In this hypothetical example, the FCI
increases from 0 at a consumption rate of 0 grams per day, reaches its maximum at a
consumption rate of about 60 grams per day and then begins decreasing at higher consumption
rates. The FCI becomes 0 at about 140 grams per day and is negative at higher fish consumption
rates.
Different users of the framework may be interested in different portions of the FCI curve. For
example, someone may decide to select a consumption rate where benefits equal risks (i.e., the
point at which fish does not pose an increased risk above background). Alternatively, someone
else may decide to focus on the consumption rate at which the FCI (overall health) is maximized.
In the case of the hypothetical example used here, no net change in health outcome occurs at
about 140 grams per day. while maximum benefit is realized at about 60 grams per day.
Yet another use of the FCI curve is to compare it to the benefit curve. The benefit curve can be
viewed as the best representation of an ideal health benefit associated with eating fish. The FCI
represents the possible health benefit when potential risks from chemicals are included. The
difference between the two curves is the reduction in benefit caused by the chemical
contamination. Note too, that the FCI can be used in a similar way to estimate the effect of very
restrictive fish consumption advisories, in terms of unrealized health benefits. For example,
when the PCI associated with setting an advisory at 5 grams per day is compared to the benefits
Toxicology Excellence for Risk Assessment 6-23 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
associated with setting an advisory at 60 grams per day. The FCI curve could also be used to
compare different fish of the same species to find out the quantitative benefit of eating smaller,
less contaminated fish.
A number of assumptions and estimates have been folded together to create this FCI and the
resulting net health risk curves. However, the sometimes disparate information used in the
development of the FCI does not lend itself to an easy estimation of error. One approach to
seeing how such potential error might affect the use of the framework is to suppose that a set of
FCI values, for example that of 0.25 to —0.25, defines a range of reasonable error. Thus,
someone for this example might consider 140 grams per day as the consumption limit for adults
because that is the consumption rate at which risks and benefits are equalS However, someone
else might consider a value of 125 grams per day as the consumption limit for adults because that
is the approximate rate at which the risks and benefits are at a value of 0.25.
It is notable that in this example, typical risk assessment techniques indicate that the upper bound
cancer risk from chiordane alone equals one in one thousand (1x10 3 ) at aconsumption rate of
about 25 grams per day and contamination of 12 ppm (mg/kg) (Figure 6-4). In the absence of
the benefit information and based upon the results of a typical risk assessment, it might be that an
advisory for the fish used in this example would restrict consumption to rates much lower than
the either of the choices given above.
6.5.1.1 Calculations for Estimating Benefits
In the hypothetical example given above, benefits are predicted using the following equation
(described in Section 6.4.1 of the framework):
[ B x (1-RR )1 x S 1 = Benefit
Where:
B 1 is the background incidence of health endpoint i (see Table 6-4 and
Chapter 2);
RR is the relative risk of health endpoint i at the given consumption rate (see
Table 6-4 and Chapter 2);
S 1 is the biological “severity” of health benefit endpoint i (see Table 6-2);
and,
Benefit 1 is the benefit for health endpoint i associated with eating a given amount
of fish.
The hypothetical example calculates benefits at two unique consumption rates (6.5 grams per day
and 60 grams per day). Because data about benefits do not exist beyond a consumption rate of
60 grams per day, benefits are assumed to remain constant at higher consumption rates. Of
course, this assumption breaks down as the percent of protein in diet approaches 100 (see Figure
6-1). However, for purposes of this framework example, the assumption is very reasonable
because of the amount of fish consumed is a smaller part of the total daily food consumption (for
example, 10 to 200 grams of fish is only approximately ito 20% of a daily food intake of 1 kg).
Toxicology Excellence for Risk Assessment 6-24 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Table 6-4 gives values for background incidence, severity ratings and relative risks, and are
shown below. Please note that the severity ratings reflect our judgments. Other judgments may
be appropriate.
Using the above equations, at 6.5 grams per day:
CHD benefit: (0.32 x (1-0.6)) x 3 = 0.38;
Stroke benefit: (0.07 x (1-0.85)) x 3 = 0.03; and,
Arthritis benefit: (0.13 x (1-0.92)) x 1 = 0.01.
The total benefit is derived by summing the benefit for each health endpoint using the following
equation (described in Section 6.4.1 of the framework):
n
{ [ B 1 x (1-RR )] x S } = Benefit
i=1
Thus, the total benefit at 6.5 grams per day of fish consumption is 0.42.
At 60 grams per day:
CHD benefit: (0.32 x (1-0.45)) x 3 = 0.53;
Stroke benefit: (0.07 x (1-0.55)) x 3 = 0.09;
Arthritis benefit: (0.13 x (1-0.57)) x 1 = 0.06; and,
the total benefit is: 0.68.
6.5.1.2 Calculations for Estimating Risks
In the hypothetical example given above, risks from fish consumption are estimated using
standard risk assessment.equations (EPA 1989). As indicated before, fish in this hypothetical
example are assumed to contain 2.1 mg/kg of methylmercury and 12 mg/kg of chlordane.
People are assumed to weigh 70 kilograms and eat fish at a specified rate for their entire lifetime.
6.5.1.2.1 Excess Lifetime Cancer Risk
The equation used to estimate increase in excess lifetime cancer risk is:
R=axbxcxd÷e:
Where:
R = excess lifetime cancer risk;
a = concentration of chemical in fish (mg/kg);
b = consumption rate of fish (g/person-day);
c = cancer slope factor (per mg/kg-day);
Toxicology Excellence for Risk Assessment 6-25 8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Wsk
d = conversion factor (kg/bOO g); and,
e = body weight (kg/person).
Using the above assumptions and equation and assuming chiordane has a CSF of 3.5 x10’
results in an upper bound excess lifetime cancer risk of:
4 x iO’ at 6.5 grams per day;
4 x i0 3 at 60 grams per day; and,
7 x i0 3 at 120 grams per day.
6.5.1.2.2 Excess Lifetime Non-Cancer Risk
Non-cancer risk is estimated by first calculating the daily exposure and then comparing that
exposure to the dose response data for non-cancer effects presented in Chapter 4. The
comparison requires determining how many times greater than the Rfl) the estimated dose is,
and then estimating the response for that exceedence of the Rfl) (from the risk above RfD dose-
response data).
Daily dose is estimated using the following equation:
D=axbxc—d:
Where:
D = daily dose;
a = concentration of chemical in fish (mg/kg);
b = consumption rate of fish (g/person-day);
c = conversion factor (kg! 1000 g); and,
d = body weight (kg/person).
Using the above assumptions and equation, the daily doses of chlordane at three different
consumption rates are:
1.1 x i0 3 mg/kg-day at 6.5 grams per day;
1.0 x 102 mg/kg-day at 60 grams per day; and,
2.1 x 102 mg/kg-day at 120 grams per day.
Similarly, the daily doses for methylmercury are:
2.0 x i0 mg/kg-day at 6.5 grams per day;
1.8 x i0 3 mg/kg-day at 60 grams per day; and,
3.6 x i0 3 mg/kg-day at 120 grams per day.
Using the dose-response information for chiordane for the percentage of the population predicted
to manifest an effect, the best estimate ( 50 th percentile) of the increased incidence of the critical
effect is:
Toxicology Excellence for Risk Assessment 6-26 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
o at 1.1 x i0 3 mg/kg-day;
0.005 at 1.0 x 10 2 mg/kg-day; and,
0.01 at 2.1 x 102 mg/kg-day.
For mercury the best estimate ( 50 th percentile) of the increased incidence of the critical effect is:
0 at 2.0 x i0 4 m Ikg-day;
0.12 at 1.8 x 10 mg/kg-day; and,
0.30 at 3.6 x 10 mg/kg-day.
Risk (R), as used in the framework, is then calculated using the following equation (described in
Section 6.4.2 of the framework):
(R x S 1 ) x (-1) = Risk 1
At 6.5 grams per day:
Chiordane cancer risk:
Chiordane non-cancer risk:
Methylmercury non-cancer risk:
-(4x 10 4 x3) =-lx i0 3 ;
-(0x3) =0; and,
-(Ox 1)=O.
The risk for each health endpoint and chemical is summed using the equation shown below
(described in Section 6.4.2) to arrive a total risk of -0.001 at 6.5 grams per day:
n
-1 x (R x S 1 ) = Risk
1=1
At 60 grams per day:
Chiordane cancer risk:
Chiordane non-cancer risk:
Methylmercury non-cancer risk:
the total risk is:
At 120 grams per day:
Chiordane cancer risk:
Chiordane non-cancer risk:
Methylmercury non-cancer risk:
the total risk is:
-(4x 10 3 x3) = i x iO ;
-(0.005 x 3) = -0.0 15;
-(0.12 x 1) = -0.12; and
-0.15.
-(7x 10 3 x3) =-Zx 102;
-(0.01 x 3) = -0.03;
-(0.3 x 1) = -0.30; and,
-0.35
Toxicology Excellence for Risk Assessment
6-27
8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
6.5.1.3 Estimating the FCI
The FCI is derived by combining the total benefit (B) and the total risk (R) for each consumption
rate using the equation shown below (described in Section 6.4.3 of the framework):
Benefit + Risk = FCI
Thus:
at 6.5 grams per day the FCI is equal to 0.42
(the total benefit of 0.42 plus the total risk of -0.001);
at 60 grams per day the FCI is equal to 0.53
(the total benefit of 0.68 plus the total risk of -0.15); and,
at 120 grams per day the FCI is equal to 0.33
(the total benefit of 0.68 plus the total risk of -0.35).
The benefit (B), risk (R) and FCI are plotted in Figure 6-5.
6.5.2 Impacts from Changes in Contaminant Concentrations -
Changes in chemical concentration in fish and the type of health endpoint (i.e., cancer or non-
cancer) a chemical causes will have a substantial effect on the FCI. Three observations about the
interaction of chemical concentration and type of effect are important to recognize.
First, as evident from the quantitative example presented above (Section 6.5.1) even relatively
large increases in excess lifetime cancer risk (large when evaluated using typical allowable risk
levels of 1x10 6 to 1x10 4 ) have a relatively small effect on the FCI. This is consistent with the
results of Anderson and Wiener (1995) and is shown in Figure 6-6. As concentration of a
chemical increases, the excess lifetime cancer risk also increases, but because increase in cancer
risk is assumed to be linear for environmental exposures, the change in FCI remains relatively
small. A comparison of Figure 6-6 (low concentration of a carcinogenic chemical) to Figure 6-7
(a four-fold increase in chemical concentration) reveals that the general shape of the risk (R) and
FCI curves is not dramatically different. (Note that the benefit curve (B) remains the same
because benefits depend only upon the amount of fish eaten and not the concentration of
chemicals in fish.) In general, it appears that only in those instances where either people eat
extraordinarily high amounts of fish or where the fish have very high levels of many
carcinogenic chemicals, will the potential cancer risk associated with contaminated fish be
greater than the potential benefits as identified for this example.
A second observation is that accounting for non-cancer effects can have a substantial effect on
the shape of the risk curve (R) and the FCI. This difference occurs because the estimated risk
from noncancer effects for these chemicals is on the order of a few percent compared to i0
to10 4 risk from cancer at the doses of interest. At low concentrations of a chemical in fish, non-
cancer effects may not manifest themselves until large amounts of fish are eaten (see Figure 6-8)
Toxicology Excellence for Risk Assessment 6-28 .. 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
1
95
0
-o 5
-1
0
Figure 6-6. Low Concentration Carcino n
0 25 50 75 100 125 150 175
Fish Consumption Rate (g/d)
0
(0
(. 1
I n
5
(0
0
I
1
Figure 6-7. High Concentration Carcinogen
05
0 )
U
U,
5
C
0
I
1) 5
-1
0
25
50 75 100 125
Fish consumption R,ate (gid)
150
175
Toxicology Excellence for Risk Assessment
6-29
8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative DietaryRisk
Figure 6-8. Low Concentration Non-Cancer
05
C)
U
U,
0 25 50 75 100 125 150 175
Fish Consumption Rate (g/d)
Figure 6-9. High Concentration Non-Cancer
B
C )
‘3
U,
C)
I
1
05
0
-05
-1
RN
.FCIN
0 25 50 75 100 125
Fish Consumption Rate (9/cl)
150 175
Toxicology Excellence for Risk Assessment
6-30
8/6/ 99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Figure 6-10. Low Concentration Cancer &
Non-Cancer
ii
05
0
-0 5
1
05
0
-0 5
-1
Figure 6-11. High Concentration Cancer &
Non-Cancer
0 25 50 75 100 125
Fish Consumption Rate (g/d)
0 25 50 75 100 125
Fish Consumption Rate (gid)
150 175
C,
U,
C,
I
C)
U
U,
4S
C,
I
Rc&N
FCIC&N
150 175
Toxicology Excellence for Risk Assessment
6-31
8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
and the FCI may remain positive even at high consumption rates. When the concentration of a
chemical in fish increases (in the case of this hypothetical example, by four-fold), the risk curve
(R) shifts to the left and causes the FCI curve to do the same (compare Figures 6-8 and 6-9).
Contrary to the observations made regarding concentrations of chemicals assumed to cause
cancer, changes in the concentration of chemicals assumed to cause non-cancer effects could
lead to substantial changes in the FCI.
In this hypothetical example, the combined cancer and non-cancer risk (R) and the FCI are
dominated and largely determined by non-cancer effects (see Figures 6-10 and 6-1 1). This
appears to be the case at both low (Figure 6-10) and high concentrations (Figure 6-11) of a
chemical in fish. This hypothetical example may or may not apply to real situations, but it
emphasizes the importance of evaluating noncancer effects for chemicals that cause cancer.
Exceptions to this finding would be a chemical that is a highly potent carcinogen and causes no,
or very minimal, non-cancer effects. None of the six chemicals currently included in the
framework has this set of characteristics.
Part of this behavior can be explained by the use of a severity scheme that only allows
differences of 1, 2, or 3 to effects of different biological severity. If a different quantitative scale
is used, for example 1, 3, and 10, a different outcome might be expected. The suggested
framework can use different severity scales if needed.
6.5.3 Evaluation of Different Subgroups
The framework has been designed to allow evaluation of the benefits and risks to multiple
subgroups exposed to chemicals in fish. For example, children, teenagers, and adults may be
exposed to chemicals in fish via direct consumption while a breast-fed infant may be exposed to
chemicals in its mother’s milk. If the chemicals in fish bioaccumulate in mother’s milk, a breast-
fed infant’s exposure may be greater than an adult’s for any given concentration of a chemical in
fish (for a brief discussion of this issue, please see Chapter 4). In addition, differences in body
weight among people who eat fish will result in differences in exposure. Dividing a population
into subgroups allows one to estimate the exposure for each subgroup and the framework can
calculate a unique FCI for each subgroup.
Figures 6-12 through 6-15 show how the potential risk and resulting FCI change for adults and
infants with different concentrations of non-bioaccumulative or bioaccumulative chemicals in
fish. Fish consumption rate is shown on the horizontal axis and the health scale is show on the
vertical axis. Several curves are shown on each figure. Curve “B” represents the benefit
associated with eating fish and remains constant for all subgroups and in all figures. Curve “R”
represents the potential risk associated with eating fish. Separate risk curves are shown for
adults (R A3 and infants (R 1 ). Figures 6-12- and 6-13 show two infant curves, one for low
concentration in breast milk (R 1 low) and one for high concentration (R 1 , high). The curves labeled
FCIA and FCI 1 shows the FCI for adults and infants, respectively.
These four hypothetical examples present FCIs for an adult and breast-fed infant. One example
each is presented for fish with a low (Figure 6-12) and high (Figure 6-13) concentration of a non-
bioaccumulative chemical and for fish with a low (Figure 6-14) and high (Figure 6-15)
Toxicology Excellence for Risk Assessment 6-32 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
concentration of a bioaccumulative chemical. The framework assumes that the same toxicity
benchmarks (CSFs and risk above the RfDs) can be used to estimate risk and calculate FCIs for
different individuals. Given this assumption, the differences in risk and FCI are solely a function
of differences in estimated dose.
Non-bioaccumulative Chemicals. For a chemical that does not bloaccumulate in breast milk
(such as methylmercury) the differences in the infant and adult FCI may not be large (see Figures
6-12 and 6-13) regardless of the chemical concentration in fish. Any differences in FCI between
these two subgroups arise from differences in dose. If a chemical is not readily transferred to
breast milk or in situations where breast milk comprises a small fraction of an infant’s diet, the
infant FCI may be higher than the adult FCI (curve FCI 1 , i 0 on Figures 6-12 and 6-13). This
would mainly be due to the fact that the infant is exposed to less chemical on a per kilogram
body weight basis. In such a scenario the adult is the more exposed individual.
Alternatively, if a chemical is readily transferred (but not bioaccumulated) to breast milk and if
the majority of an infant’s diet is comprised of breast milk, then the infant’s FCI may be lower
(i.e., more negative) than the adults (Figures 6-12 and 6-13). This would be mainly due to the
fact that the infant is exposed to more chemical on a per kilogram body weight basis. When this
occurs, a fish consumption advisory could be set to protect the infant and adult separately. This
could be accomplished by selecting two sets of consumption rate limits, one for breast-feeding
(or soon to be breast-feeding) mothers and another for other fish consumers. As an example, for
the scenario shown in Figure 6-13, the FCIA and FCI 1 curves could be used to guide a decision-
maker in setting appropi late levels. 6
Bioaccumulative Chemicals. For chemicals that bloaccumulate in breast milk (chlorinated
pesticides for example), the infant FCI may be much lower (i.e., more negative) than the adult
regardless of the concentration of the chemical in fish (Figures 6-14 and 6-15). At low
concentrations the adult FCI may remain positive (i.e., fish consumption leads to a net health
benefit) and perhaps even at very high consumption rates, while the breast-fed infant FCI may
become negative when the mother eats even moderate amounts of fish (Figure 6-14). At high
concentrations, the infant’s FCI may become negative when the breast-feeding mother eats low
amounts of fish (Figure 6-15).
Thus, for bioaccumulative chemicals the FCI may differ substantially between adults and breast-
fed infants. It is important to note that the consumption limits derived using the framework
apply to the people eating the fish (i.e., older children, teenagers, and adults eating a particular
amount of fish per day): Calculating a breast-feeding infant FCI depends upon estimating the
infant’s exposure through breast milk, which in turn requires conversion of the breast-feeding
mother’s fish consumption exposure into a breast milk concentration. This can be done using
empirical data that relates an infant’s exposure (consumption and breast milk concentration) to a
mother’s exposure or by using pharmacokinetic models that predict breast milk concentrations
6 Note that the framework does not consider how many months prior to beginning breast-feeding a mother should
restrict her consumption of fish This issue arises whenever setting advisories to protect breast-fed infants and
depends upon the pharmacokinetics of the chemicals being evaluated. The same methods used to derive traditional
fish consumption advisories can be used in the framework
Toxicology Excellence for Risk Assessment 6-33 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
based upon maternal exposure. The Everglades case study presented later in this chapter
illustrates the latter approach.
6.5.4 Mixtures of Chemicals and Multiple Endpoints
Fish can, and often do, contain more than one chemical. The framework has been designed to
consider this. In addition, more than one non-cancer effect could be possible after exposure to
chemicals. Many uncertainties and complexities arise when assessing exposures to mixtures or
evaluating multiple endpoints.
Data on the toxicity of a specific mixture of chemicals in fish will generally not be available In
the absence of such toxicity data, the framework, like most other mixture risk assessments,
defaults to an additivity approach, as per EPA guidelines (U.S. EPA, 1986; 1988). Cancer risk is
estimated for each chemical individually and the risk from each chemical is then added together
to derive a total risk associated with the mixture of chemicals. For non-cancer endpoints (with
similar mechanism of action or at least target organ) the daily dose is divided by the RfD and the
resulting fractions are summed for all chemicals to calculate a Hazard Index (HI). As long as the
HI is at or below one, no hazard is assumed; a HI above one may be cause for concern, but
cannot be interpreted in a quantitative fashion.
In this framework, as the potential risk from each successive chemical is combined, the total risk
increases and the FCI decreases (Figure 6-16). However, the benefit curve remains the same
whether there is one chemical or multiple chemicals present. Here fish consumption rate is
shown on the horizontal axis and the Health Scale is shown on the vertical axis. Curve “B”
represents the benefit associated with eating fish and remains constant regardless of how many
chemicals are included in the analysis. Curves “R” and “FCI” represent the risk and FCI,
respectively, associated with eating fish. Separate risk and FCI curves are shown for chemical A
(RA FCIA), chemicals A and B combined (RAB FCIAB), and chemicals A, B and C combined
(R c FCI c). As discussed above, a parallel but opposite change in the FCI might occur if
new or greater benefits associated with fish consumption (e.g., cultural benefits) are included in
the framework.
Figures 6-17, 6-18, and 6-19 show how risk and FCI might change as risks from additional
endpoints, which were not the basis for the RfD, are added to the framework. The top part of
each figure shows the hypothetically dose response data for the critical effects and effects A &
B. The low part of each figure shows fish consumption rate on the horizontal axis and the Health
Scale on the vertical axis. Curve “B” represents the benefit associated witt eating fish and
remains constant regardless the number of adverse effects that are included in the analysis.
Curves “R” and “FCI” represent the risk and FCI, respectively, associated with eating fish.
Separate risk and FCI curves are shown for the critical effect only (RCE, FCICE” and all endpoints
(Rati. FCIaii). Figure 6-17 shows an example where the non-critical effects begin to manifest
themselves at doses much greater than the critical effect. Figure 6-18 shows an example where
the non-critical effects manifest themselves at doses similar to the critical effect but their dose
response curve has a much smaller slope than that of the critical effect. Figure 6-19 shows an
Toxicology Excellence for Risk Assessment 6-34 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
a) 05
U
U,
a.
a)
I -05
—1
Figure 6-12. Low Concentration
Non-Bioaccumulative
R Low
B
Ed 1 Low
FCIA
Ed 1 Hugh
0 25 50 75 100 125 150 175
Fish Consumption Rate (gid)
a)
U
C d ,
a.
a)
I
B
-. FCI 1 Low
FCIA
.FCI 1 High
0 25 50 75 100 125 150 175
Fish Consumption Rate (g/d)
Figure 6-13. High Concentration
Non-Bioaccumulative
.0 5
-1
I. Low
R
Figure 6-14. Low Concentration
Bioaccumulative
0 25 50 75 100 125 150 175
Fish Consumption Rate (g/d)
Figure 6-15. High Concentration
Bioaccumulative
1
. 05
U
U,
-1
B
0 25 50 75 100 125
Fish Consumption Rate (gld)
150 175
Toxicology Excellence for Risk Assessment
6-35 8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
Figure 6-16. Change in FCI as more chemicals are evaluated for health risk in fish Figure 6-16.
- - -
1 B
FCIA
0.5 FCIAB
.
U
C /, FCIABC
• 0
RA
I
-0.5 RAB
-1
0
50
Fish Consumption Rate (gid)
100
Toxicology Excellence for Risk Assessment
6-36
8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
Figure 6-17. Non-critical effects begin to manifest themselves at doses much greater than the
critical effect.
qi)
0
Critical El
A
Response
U
U)
I
B
1
05
0
.0 5
—1
FCICE
FCIAII
RCE’
0 50 100 150 200
Fish Consumption Rate (gld)
250
Toxicology Excellence for Risk Assessment
6-37
8/6/ 99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Figure 6-18. Non-critical effects manifested at doses similar to critical effect but dose response
curves are shallower.
a)
0
Critical
Effect A
Effect B
u
0 50 100 150 200 250
Fish Consumption Rate (gid)
Toxicology Excellence for Risk Assessment
6-38
8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
Figure 6-19. Non-critical effects begin at doses similar to the critical effect and their dose
response curves are — similar.
Effect B
0 . ,
C ,,
0
Effect A
Response
z IIII
0 50 100 150 200 250
Fish Consumption Rate (gid)
Toxicology Excellence for Risk Assessment
6-39
8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
example where the non-critical effects begin to manifest themselves at doses similar to the
critical effect and their dose response curve has a slope that is similar to or larger than the slope
of the critical effect.
The framework discussed here differs from most risk assessments in how it estimates risk from a
mixture of chemicals in two important ways. First, most mixture risk assessments estimate non-
cancer risk by combining the hazard quotients only for those chemicals that adversely effect the
same health endpoint. This approach can lead to the estimation of several hazard indices (one
for each health endpoint) for a mixture of chemicals, which are then combined into the overall
Hazard Index (HI). The current framework described in this text combines the potential risks for
all noncancer endpoints (regardless of endpoint) and thus, may predict a greater noncancer risk
from a mixture of chemicals than a traditional risk assessment following U.S. EPA (1986)
mixture guidelines.
Second, traditional mixture risk assessments separate the evaluation of cancer and non-cancer
endpoints. However, because this framework uses a biological severity score, the cancer and
non-cancer risks can be added to estimate the total risk and the FCI. Thus, as with mixtures of
chemicals causing non-cancer effects only, the framework estimates different risks than
traditional risk assessment might, for chemicals and chemical mixtures that cause both cancer
and non-cancer effects.
The current version of the framework highlights another phenomenon that is similar to, but not
related to, the effects of mixtures of chemicals. Namely, chemicals can cause more than one
non-cancer effect. Because of the approach used by the framework in plotting risk above the
RfD relates to only one effect (i.e., the critical effect caused by each chemical), consideration of
non-critical effects has the potential to change the outcome of the framework. This can lead to
an underestimation of adverse effects associated with chemicals in fish.
RfDs are derived to be protective of the critical endpoint (i.e., the first adverse effect or its
known precursor as dose increases). It is assumed that if exposure remains at or below the RID,
then the critical effect will not be manifested, and neither will any other adverse endpoints. Once
exposures exceed the RID, however, the critical endpoint may be manifested, and if the
exceedance is large enough, other endpoints would be expected.
For the most part, risk management decisions based upon the results of typical risk assessments
consider exposures above the RID to be unacceptable. Such a paradigm makes the other adverse
effects associated with exposure above the RID moot. It is essential to appreciate that the
framework described in this text explicitly uses estimates of risks above the RID for the critical
effect of several chemicals. Estimates of the risks from other endpoints that may occur at doses
above the RID are not used here, as the data were not available. Other approaches to estimating
risk above the RID could take multiple possible endpoints into consideration (e.g., categorical
regression). As a result, non-cancer risks associated with doses above the RID may be
underestimated.
Toxicology Excellence for Risk Assessment 6-40 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
The magnitude of this underestimate is unknown. It depends in part on the number of adverse
effects caused by the chemical other than the critical effect and the dose response curves for
these other effects. For example, if dose response curves for these other effects either begin at
doses much greater than the RfD (Figure 6-17), or have a small slope compared to the critical
effect (Figure 6-18), then the current omission of non-critical effects is likely to have little effect
on the results of the framework (i.e., the FCI) — and little effect on conclusions resulting from its
use. Alternatively, if these other non-critical effects begin to manifest themselves at doses
similar to those at which the critical effect is observed, and have dose response curves with
slopes similar to or greater than that for the critical effect, then the risk could be substantially
underestimated. 7 Of course, RfDs are established to be protective, and methods to estimate risks
at exposures above the RID assume that adverse effects occur immediately. This immediacy
may not be correct, but it is in the direction of countering the concern expressed with the lack of
modeling effects other than the critical effect.
Although these concerns would tend to cancel each other out, the resulting uncertainty in the
value of the FCI is increased. This is one reason why risk assessors and managers may wish to
use FCI values in a range, such as 0.25 to —0.25, rather than a single FCI value when making
decisions.
6.5.5 Cultural Benefits
All of the examples presented in the above sections derived FCIs by comparing health risks to
health benefits. For some subgroups, fish are of great cultural importance and their value cannot
be measured as simply a source of protein or a source of important health benefits (see Chapter
5). As described above, the framework has built into it the flexibility to adjust the FCI (the net
benefit of consuming fish) based upon cultural impacts or some other factor not explicitly
accounted for by the risk and benefit equations.
The framework allows for a factor or modifier to adjust the FCI for culture-based impacts. The
value of this factor can be based upon the cultural value of fish and/or fishing-related activities to
the population. As the cultural importance increases, the factor can increase. This leads to an
increase in the benefits associated with fish consumption, which in turn leads to an increase in
FCI (Figure 6-20). However, as described above, the cultural factor may not be a multiple of
health benefits. It could be a constant added to the FCI or some other consumption rate-related
adjustment of the FCI. The current framework does not contain a methodology to derive the
At the present time, existing data have not been used to estimate the dose response curve for each of the non-
cancer effects that may be caused by a particular chemical As resources permit, the framework allows the
incorporation of such information
Toxicology Excellence for Risk Assessment 6-41 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
Figure 6-20. FCI changes when cultural benefits of fish consumption are added.
2
1.5
1
C.,
U)
BH+C
BH
R
FCIH’ FCIH+C
0 50 100
Fish Consumption Rate (gid)
150
Toxicology Excellence for Risk Assessment
6-42
8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
cultural factor. It is assumed that this factor would be developed by public health regulators and
the population for whom fish is of great cultural importance. Such a methodology would need to
evaluate and ideally quantify the physical, emotional and mental well-being aspects along with
the disruption or enhancement of a “cultural” practice such as catching or consuming fish.
Quantitative data are not available, but the population itself may have a qualitative judgement
about the negative or positive consequences of a cultural practice, which they would want to
incorporate.
An important attribute of the framework is that by including cultural importance in the derivation
of the FCI, it provides a basis for responding to the needs of the subpopulation. For example,
both the general population and a subpopulation may be eating the same species of fish from the
same water body. If the subpopulation places great cultural value on fish consumption, the
framework can reflect this. Figure 6-20 illustrates hypothetically how the FCI could change
when cultural benefits of fish consumption are added. Fish consumption rate is shown on the
horizontal axis and the Health Scale is shown on the vertical axis. Curves “BH” and “BH+c”
represent, respectively, the health benefits only and the health and cultural benefits combined.
Curves “R” and “FCI” represent the risk and FCI, respectively, associated with eating fish.
Separate FCI curves are shown for health benefits only (FCIH) and health and cultural benefits
combined (FCIH+C).
The outcome of making this adjustment is that the framework can identify one consumption rate
for the general population and in this hypothetical example, a higher consumption rate for the
subpopulation. In this example, the C factor is equal to the other health benefits combined and
therefore nearly doubles the FCI.
6.5.6 Personal Perception of Severity
As for cultural benefits, a scaler for the personal perception of the severity of an effect or benefit
could be added to the framework. Like for the cultural scaler, the resulting FCI could not be
consider a strictly health .based score. We do not attempt to provide a quantitative handle on the
value of this potential personal perception of severity. If it was used, however, it would appear
to be best placed at the development of the risk scale as shown below:
(R 1 x S ) x (-1) x PPS= Risk 1 pps
Where:
R 1 is the increased risk of health endpoint i associated with a particular fish
consumption rate,
S is the severity of health endpoint i,
PPS personal perception of severity, and,
Risk 1 is the decrease in health (because of the increase in risk of health endpoint
i) associated with eating a given amount of fish.
6-43
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
6.6 Case Studies
6.6.1 Case Study: The Florida Everglades
6.6.1.1 Background
In 1989, a Florida panther was found dead in the Everglades with extremely high mercury levels
in the liver (>100 ppm) (Fleming eta]., 1995). As the apex carnivore in the Everglades
ecosystem, the panther is a good indicator of the potential for biomagnification of
methylmercury. This incident, along with elevated levels of mercury in other wildlife (raccoons,
otters, and alligators) has sparked concern over the potential health effects.on humans who eat
fish from the Everglades. Factors such as wetland morphology, hydroper’iod, water chemistry,
dissolved organic carbon, and bacterial processes in the Everglades have resulted in increased the
methylation and subsequent biomagnification of mercury, although the mechanisms behind these
associations are not fully understood (Choi and Bartha, 1994).
Analyses of freshwater fish from the Everglades revealed the presence of. methylmercury at
concentrations up to 7 ppm (Science Subgroup, 1994). The average concentration ranged from 2
to 3 ppm in freshwater fish, and other wildlife (Fleming eta]., 1995). Florida’s advisory level
for methylmercury is 1.5 ppm (Krabbenhoft, 1996). As a result, the state of Florida issued a
Health Advisory in March 1989 recommending limits on consumption of several fish species
that are caught in the Everglades. No cases of human poisoning due to Everglades fish
consumption have been reported; however, clinical diagnosis of mercury poisoning is difficult.
6.6.1.2 Summary of Existing Data
Fleming et al. (1995) recruited and questioned 1794 people who had consumed Everglades fish
(sport anglers, subsistence fishers, Native Americans and other Everglades residents). Of the
1794 individuals, 405 had eaten fish and/or wildlife from the Everglades. Of these 405
individuals, 55 refused to participate, leaving 350 subjects. No data were collected from those
that refused, so it is not known if their consumption differs from the study population. Fleming
et al. (1995) reported a weekly fish consumption of 1.79 meals per week for all subjects who
consumed fish over the 6-month sampling period.
The subjects completed a questionnaire and provided a hair sample. The hair samples were
analyzed by atomic absorption for total mercury. The detection limit (DL) for total hair Hg was
1.26 ppm. Out of 330 subjects sampled, 119 (36%) subjects had total hair mercury
concentrations above the detection limit. For samples with concentrations above the DL, the
mean level of total Hg in the hair was 3.48 ± 3.01 ppm (Fleming et al. 1995). The highest total
hair mercury concentration measured was 15.57 ppm. Because the mercury concentrations in the
211 hair samples with values below the detection limit were not known, a default value of one-
half detection limit (0.63 ppm) will be used for the purposes of this case seudy. The resulting
mean for all 330 samples is 1.66 ppm, using 0.63 ppm as the default value for all samples below
the detection limit.
6-44
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
This study found that the most exposed groups were men and African-Americans. Within these
groups, those with highest hair Hg levels were mostly subsistence anglers with a small income
and a low level of education (Fleming eta]., 1995).
6.6.1.3 Exposure Assessment
In estimating the risk to the fetus from methylmercury exposure, maternal mercury exposure is
used as a dose surrogate for fetal exposure. U.S. EPA (1999) provides one method for
extrapolating an estimated daily dose of mercury from hair mercury levels. Fleming et al. (1995)
provides a distribution of total Hg in the hair, as well as an estimated mean fish consumption
rate. The estimated dose based upon self-reported consumption can be verified by extrapolating
daily dose from total hair mercury. Please see Section 6.6.1.6 for a detailed description of this
procedure.
Fleming et at (1995) did not report the size of meal that corresponded to the reported mean
consumption of 1.79 meals/week. For the purposes of this case study, we assumed that a meal
consists of 4 oz. of fish; however, the true average portion size may differ from this assumption.
It is important to have an accurate estimate of meal size in order to estimate the average number
of grams of fish consumed per day. An accurate estimate of fish tissue methylmercury
concentrations is also crucial in the resulting estimate of daily methylmercury dose at a given
level of consumption.
Since maternal hair total mercury is used as a dose surrogate for fetal methylmercury, the
average weight for a pregnant woman is used (60 kg) (U.S. EPA, 1998). The dose extrapolated
from maximum reported hair concentration is 1E-3 mg/kg-day (see Table 6-6). We can also use
the range of daily methylmercury dose estimated from consumption of fish at 28 g/day
containing 2-3 ppm meivury. For a 60-kg pregnant woman consuming 28 glday of fish
containing 3 ppm the estimated dose is 1E-3 mg/kg-day. This dose is consistent with the dose
(1E-3 mg/kg-day) extrapolated using the maximum reported hair concentration (15 ppm). This
consistency tends to validate the approaches and assumptions used here.
Table 6-6. Calculation of estimated daily doses using total hair Hg data from Fleming et al.
(1995).
Components in the
Equation
Ch (ug/g)
—250=
Cb (mgfL)
x
b
x
V
(L)
-
A
x
f
=
I
(mg/day)
—6Okg=
D (mgi
kg-day)
Adjusted mean
(0.63 ppm substituted for
values_below_the_DL)
1.66
0.0066
0.0 14
4.9
0.95
0.05
1 E-2
2 E-4
Maximum
15.6
0.062
0.014
4.9
0.95
0.05
9 E-2
1 E-3
Mean of data above DL
3.48
0.014
0.014
4.9
0.95
0.05
2 E-2
3 E-4
6-45
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
6.6.1.4 Calculation of PCI
The critical effects of methylmercury poisoning, on which the RfD is based, occur in the fetus
exposed in utero (See Chapter 4). Risk above this RID is estimated and compared to benefits for
the general population (See Figure 4.1 in Chapter 4). Benefits data are available for the
fetus/children of women who consumed methylmercury contaminated fish in the Seychelles
Islands (Davidson et al., 1998). The incorporation of these data into a benefits curve for
fetuses/infants of mothers who consumed fish has not been attempted; however, because of the
preliminary nature of these data.
The estimate of risk to the adult population (See Figure 4.1 in Chapter 4) may be conservative,
because the critical effect is in the fetus; however, it is also likely that not all of the contaminants
present in the fish have been included here. Adding additional chemicals would reduce the FCI
because additional risk would be added, but the benefits remain constant.
U.S. EPA (1999) reports a RID of 1E-4 for methylmercury (see Chapter 4 for details). The
estimated methylmercury dose extrapolated from the adjusted mean mercury hair concentration
is 2E-4 mg/kg-day. This exceeds the EPA’s RfD for methylmercury by 2-fold. The dose
extrapolated from the maximum hair mercury concentration, and the dose estimated based upon
consumption of fish containing 3 ppm methylmercury (1E-3 mg/kg-day) both exceed the EPA’s
RID by 10-fold. Therefore, a risk of adverse health effects may exist for this population.
In order to apply the framework, risk above the RfD must be calculated for various levels of fish
consumption. Table 6-7 is a summary of calculations of risk above the RfD (as more fully
described in Chapter 4).
Table 6-7. Dose-response estimates for methylmercury (Price et a]. 1997) .
Response
Multiple of RfD
(1E-4 mg/kg-day)
Dose
(mg/kg-
day)
5th
percentile
50th
percentile
95th
tpercentile
.
1
0.0001
0.0%
0.0%
0.0%
5
0.0005
0.0%
0.0%
3.9%
10
0.001
0.0%
4.3%
12%
50
0.005
28%
44%
>50%
100
0.01
>50%
>50%
. >50%
For example, according to this dose response model, a mean risk of 4.3% and an upper 95% limit
risk of 12% exist at a dose 1E-3 mg/kg-day. This value which is 10-fold beater than the RfD,
corresponds to the consumption of 28 g/day of fish containing 3 ppm men!ury.
For the purposes of this case study the FCI will be calculated for 6.5, 60 and 120 g of fish per
day. Benefits from other consumption rates can be determined, if needed, from Figure 6-5.
Daily dose is estimated using the following equation.
6-46
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
D=axbxc—d
Where:
D = daily dose;
a = concentration of chemical in fish (mg/kg),
b = consumption rate of fish (g/person-day),
c = conversion factor (kg/bOO g); and,
d = body weight (kg/person).
For a consumption rate of 6.5 glday at 3 mg Hg/kg fish tissue the dose is approximately 3E-4
mg/kg-day.
D = 3 mg Hg/kg fish x 6.5 glday consumed + 60 kg body weight x 1 kg/bOO mg = 3E-4
mg/kg-day.
These calculations are repeated for several consumption rates.
This dose (3E-4 mg/kg-day) corresponds to a best estimate of relative risk ( 5 Øth percentile) of
0.0. The severity factoi used in the framework for subtle neurodevelopmental defects is judged
to be 1 (other judgments and severity scales are possible). The risk is adjusted by multiplying by
—1 and the severity factur (1). The resulting adjusted risk is 0.0. For the upper bound of risk
( 95 th percentile) the relative risk is about 0.02, and resulting adjusted risk is —0.02.
For a consumption rate of 60 glday at the same mercury concentration the dose is approximately
3E-3 mg/kg-day. This corresponds to a best estimate of relative risk of approximately —0.23.
Since the severity factor for the target endpoint is 1, the adjusted risk is —0.23. The upper bound
of risk is —0.45, and resulting adjusted risk is —0.45.
For consumption of 120 g/day, the dose is approximately 6E-3 mg/kg-day. This corresponds to a
best estimate of relative risk of —0.50 mg/kg-day. The upper bound of risk is greater than —0.50.
The best estimate ( 50 th percentile) of the benefits in adults for fish consumption at 6.5, 60 and
120 g were calculated above to be 0.42, 0.68 and 0.68 respectively. There are no data for
benefits at consumption levels greater than 60 glday. The Price et al. (1997) model only predicts
to the estimated ED 50 in humans. Again the working assumption is that these benefits remain
constant until the percelli protein in the diet approaches 100% at which time the expected
benefits will decrease as shown hypothetically in Figure 6-1. These resulting FCIs are 0.42,
0.45, and 0.18 for 6.5, 60 and 120 glday, respectively, when mean values (best estimates) are
compared. See Figure 6-2 la for comparison of benefits and risk for the general population, and
Figure 6-2 lb for the fetus.
6-47
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary fisk
6.6.1.5 Discussion
Figure 6-2 la illustrates the relationship between relative risk, benefit, and the resulting FCI for
the general population. The results show that the loss of benefits first occurs near 10 g of
fishlday, but that for the entire range of 6.5 to 120 g, the general FCI based on the average values
is positive. However, it should be remembered that fetal endpoints were the critical effect for
development of the RfD and risk above the RfD. For a man, or a woman who is not of child
bearing age, the FCI values may actually be higher.
Deleterious effects of fish consumption predicted by the Price et al. (1997) model used in this
framework are based upon higher exposures to methylmercury from contaminated bread in the
Marsh et al. (1987) Iraqi cohort (See Chapter 4). Recent results from the Seychelles Islands
cohort consists of mothers and infants exposed to methylmercury from fish (Davidson et a!.
1998). These results show increased cognitive performance for four of six measures in children
from mothers with the highest hair mercury levels at 66 months of age after pre- and postnatal
methylmercury exposure (Davidson eta!. 1998). It is unlikely that the methylmercury is the
cause of this increased cognitive performance. However, it might be that the higher levels of
maternal methylmercury are an indicator of more fish consumption, and that it is the increased
consumption of fish is the cause of enhanced performance in the most exposed children.
A quantitative dose-response treatment of this benefit is not attempted here, however, because of
the preliminary nature of the findings. The Faroe Islands cohort studied by Grandjean and
colleagues shows contrasting results in cognitive performance; however pilot whales were the
primary source of methylmercury in the Faroe Islands, from which the mothers were also
exposed to high levels of PCBs (30 ppm in blubber) (Grandjean et al. 1997; Weihe eta!. 1996).
Exposure to PCBs is a potential serious confounder in the Faroe Islands cohort that may also
explain the decreased cognitive performance. In the Seychelles PCBs were not detected (DL 0.2
ng/ml) in the blood of 49 of the children tested at 66 months of age (Davidson et al., 1998).
Alternatively, the Faroe Islands data may serve as a very good case study for combined
exposures.
Since benefits data for the fetus are either preliminary (Davidson eta!., 1998) or not quantifiable
(Chapter 2), only risk can be input into the framework. For the fetus, the FCI is negative for the
entire consumption range (Figure 6-21b). A dose-response relationship for fetal benefits of
maternal fish consumption can be established when the data from Davidson eta!. (1998) are
verified or if quantitative benefits can be derived from the information provided in Chapter 2.
This illustrates an important aspect of this case study. In order to derive a FCI, benefits and
risks should be compared for the same populations. In the case of fetotoxicants more data on
the pre- and postnatal benefits of maternal fish consumption are needed in order to apply the
framework correctly.
Consumption of these fish by women of childbearing age should also be carefully considered.
This is because the benefit that these women may accrue from consumption of fish may also
result in a risk to their offspring (although the preliminary data from the Seychelles Islands
suggest otherwise). Perceived risk may be greater when the risks accrue to the next generation
as opposed to the current generation.
6-48
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
08
06
0.4
0.2
0
-0.2
-04
-06
-0.8
.1
Figure 6-Zla Estimated Risk, Benefit, and PCI for Mercury Contaminated Fish from the Everglades for
the General Population.
0 20 40
140
________T ___&
U
—
4 ?
-, : <
—.-— Risk 50th percentile
—.— Risk 95th percentile
—g-— Benefit 50th percentile
— u-— FCI 50th percentile
60 80
Fish Consumption (g/day)
100
120
6-49
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Figure 6.2 lb Estimated Risk to the Fetus As a Function of Everglades Fish Consumption.
Risk 50th percentile
—.— Risk 95th percentile
FCI
—:— Benefit
20 40 60 80 100 120
Fish consumption (glday)
-0 1
-o z
-o 3
-0 4
(V
U
U,
(V
I
-0 5
-0 6
-0 7
-0 8
-0 9
0
140
6-50
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
It is important to communicate to the fishers in the Everglades these potential risks and benefits.
The highest mercury levels in hair correlated with impoverished men, with little or no access to
health education. For this group, the necessity, and benefits of consuming the fish may outweigh
the risks. It is also important that health effects information reach these individuals, especially if
fish are being taken to pregnant family members, since the risk is greatest for the fetus. An
additional difficulty is the lack of knowledge of the nutritional background of the study
population. Specifically, it is not known whether the RR of coronary heart disease in this
population is representative of the national values used in the calculation of benefits.
It should also be noted that only data concerning methylmercury contamination were available
for this case study. A more complete analysis of the risks to the populations consuming fish in
this area would necessitate a more complete picture of the contamination profiles of the fish
being consumed. With the addition of these contaminants, the FCI at a given consumption rate
may be reduced.
6.6.1.6 A Method for Verifying Fish Consumption Estimates
When using fish concentration data coupled with consumption estimates, it is useful to verify the
daily intake by extrapolating daily mercury dose from total hair mercury and compare to a dose
based upon consumption rate. The first step in this extrapolation is to relate the mercury
concentration in the hair to serum mercury levels. The hair to serum concentration ratio for Hg
varies seasonally, peaking after fishing season in late fall to early winter. This ratio also depends
upon from what part of the body the hair is sampled (Phelps, et a]., 1980). U.S. EPA’s IRIS uses
the ratio 250:1, based upon the results of several studies (Phelps eta!., 1980; Suzuki etai, 1993;
Tsubaki and Irukayama, 1977).
U.S. EPA (1999) uses the following equation to estimate daily dose of mercury from serum
mercury concentration based upon assumptions that steady state conditions exists, and that first
order kinetics for Hg are being followed.
I = (C x b x V) (A x f)
Where:
I = daily intake of mercury,
Cb = serum mercury concentration,
b = elimination constant,
V = blood volume,
A = absorption factor, and
f = fraction of tissue uptake from the serum.
U.S. EPA (1999) reports the elimination constant (b) for Hg to be approximately 0.014 day’
based upon two studies (Cox eta!., 1989; Sherlock eta]., 1982). The volume of blood (V) is
approximately 7% of the body mass. Assuming an average mass of 60 kg (U.S. EPA, 1999), the
average blood volume is approximately 4.9L. The absorption factor (A), is 0.95 assuming
dietary intake of MeHg from fish (U.S. EPA, 1999; Miettinen eta!., 1971; Aberg etai, 1969).
Toxicology Excellence for Risk Assessment 6-51 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
The fraction of daily uptake of mercury from the blood was derived experimentally (WHO,
1990; Sherlock etal., 1982) to be 0.05. The results of this calculation are shown in Table 6-7.
In order to verify this estimate, reported fish consumption levels from Fleming et al. (1995) can
be used in conjunction with fish tissue Hg concentrations to estimate a daily dose for
comparison. Fleming et al. (1995) reported mean fish consumption of 1.79 meals/week.
Assuming that an average meal consists of about 4 ounces ( 0.11 kg) of fish, this is about 0.028
kg of fish per day. Fish fillet concentrations in the area fished were about 2 or 3 ppm (or mg of
chemical per kg of fish). Therefore, the estimated daily Hg intake based upon the given range
concentrations in fish, and the estimated amount of fish consumption, would range from 6E-2
mg/day or 8E-2 mg/day (i.e., 0.028 x 2 or 3 = 6E-2 or 8E-2). Assuming a body weight of 60 kg
this corresponds to a dose of 1E-3 mg/kg-day.
This range of daily Hg doses based on fish consumption falls within the range of the daily doses
determined from hair concentration found in Table 6-7. Both methods resulted in a similar
estimation of daily dose. Therefore, Hg levels found in hair are not inconsistent with the
hypothesis that this Hg is due to the consumption of contaminated fish, at the rate reported by
Fleming et al (1995). This validates the risk, and FCI estimates at a given consumption rate,
since these estimates are related to fish consumption rate, but rely upon known contaminant
intake at a given fish consumption rate. For example, if average meal size was 8 ounces as
opposed to 4 ounces in this case, then by using 4 ounces, and 3 ppm mercury in fish, daily dose
would be underestimated by half. But by verifying meal size as shown above, this error can be
identified before calculating risk
6.6.2 Vietnamese Immigrant Women Consuming Lake Ontario Sportfish
6.6.2.1 Background
Along the accessible shorelines of western Lake Ontario and the Niagara River, Vietnamese
families can be observed fishing together, filling buckets with fish to take home, and
occasionally cooking a meal of fresh fish near the water’s edge. This study of Vietnamese
women arose from concerns about the potential health risks associated with eating fishing from
Great Lakes Areas of Concern (AOCs), designated by the International Joint Commission as
such because of unacceptable levels of persistent toxic substances. Immigrants from Southeast
Asia appear to eat more sportfish than the average consumer in North America (Hutchison and
Kraft, 1994). This would make them potentially at risk of adverse health effects associated with
chemical contaminants, because of their greater exposures. The focus is on women of
reproductive age because of the possible risks associated with eating contaminated fish during
pregnancy; pregnant and nursing women, and unborn babies, are at risk groups because their
physiological and developmental stages may confer greater sensitivity to chemical contaminants
frequently found in fish.
Studies of wildlife show that organochlorines (such as polychiorinated biphenyls) in the Great
Lakes basin interfere with normal reproduction and development, but few studies have
investigated the effects of mixtures specific to Great Lakes fish on humans. Women participating
in the New York State Angler Study were found to have shorter menstrual cycles if they had
Toxicology Excellence for Risk Assessment 6-52 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
been consuming contaminated fish for seven or more years, or had more than one fish meal
monthly (Mendola et al, 1997). However, fish consumption did not appear to interfere with
time-to-pregnancy (Buck et aL, 1997). Jacobson and Jacobson (1996) have shown that women
who ate Great Lakes fish contaminated with PCBs during their pregnancies, gave birth to
children who had poorer growth and memory in infancy and at 4 years of age, and below average
IQ scores at 11 years. Several studies have investigated the levels of PCBs and DDE in human
breastmilk, attempting to determine the toxicity to infants (Dewailly etal., 1991; Dewailly et al.,
1996; Mes and Weber, 1989).
This case study uses data collected from 1996 to 1998 and not yet published. It is hoped that the
following scenario may be typical of others hi states and provinces where multi-cultural
populations eat fish from contaminated ‘hot spots’. It may also have relevance for other groups
who consider themselves ‘subsistence’ fishers, such as Native Americans or Canadian
Aboriginal populations.
6.6.2.2 Summarj of Existing Data
6.6.2.2.1 Descriptive Data
The sample for this case study consists of 27 Vietnamese women of reproductive age (17-4 7
years; mean ± SD = 35.0 ± 7.3) who consume sportfish caught from Lake Ontario AOCs. These
women have spent 2-16 years in Canada (mean ± SD = 6.9 years ± 3.4), and have been eating
Great Lakes fish from 2 to 8 years (mean ± SD = 3.7 years ± 1.7). There was a wide range in
years of schooling: from 4 to 16 years (mean & median = 11 years).
Households ranged in si e from 2 persons to 7 (mean = 4.1), and 82% of the reported household
incomes fell below the Statistics Canada Low Income Poverty cut-off (based on income,
household size, size of city/town in urban or rural area). Poverty is linked to “food insecurity”, a
condition roughly defined as having insufficient nutritious and culturally-appropriate food or the
need to rely on emergency sources of food. Not surprisingly, only 31% (n=8) of these women
reported that their households were food secure. The remaining 69% (n=19) indicated they
experienced some degree of food insecurity: 42% of the sample (n= 11) reported that their
children sometimes were hungry because of a lack of food.
However, fishing was not viewed as an inexpensive means of gathering food, but instead was
considered an activity that promoted good health; one could ease stress, enjoy fresh air, and
spend time with families and friends. Catching fish and giving it to others was an important act
of sharing, and to catch fish but not eat it (particularly if the fish would not survive when thrown
back) was considered a waste of the resource and unethical.
Some of these women fished themselves; others prepared and ate fish that their partners or
friends caught. Most tended to rely heavily on their partners’ judgments about the safety of the
fish, and generally were uninformed about the fish advisories. Thirty-five percent said they
could tell a fish was safe to eat by looking at its skin surface and color: 46% agreed that “I can
tell if a fish is contaminated (not safe to eat) by the way it smells.” Fifty percent agreed with the
statement, “I feel confident that the Great Lakes fish I eat are safe because I catch them myself.”
Toxicology Excellence for Risk Assessment 6-53 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
Burger et al. (1998) have reported similar confidence in self-caught fish among individuals
fishing and crabbing in New Jersey.
Women were asked their perceptions of the risks to their health from eating Great Lakes fish.
Fifty percent felt that any risks were minor compared to other risks to which they were exposed.
Eighty-one percent said they would eat more Great Lakes fish if health risks from chemical
contaminants did not exist. And 73% agreed with the statement, “For me personally, there are
more benefits to my health from eating Great Lakes fish than risks to my health”.
Body weights averaged 53.9 kg (range was 42.2 - 72.2 kg), and body fatness, assessed using the
Body Mass Index, was in the desired 20-25 range for 67% (n=18). Seven women (26%) had less
body fat than generally considered healthy, and two (7%) were considered “overfat”.
The average of individuals’ macronutrient dietary profiles was excellent: protein averaged 19%
of energy (calories), fat was 22%, carbohydrates 59%, and saturated fats were only 6% of
energy. Only one person had usual dietary intakes of saturated fat and total fat above the current
dietary recommendations. On average, this group consumed 45 g of dietary fat and 85 g of
protein daily, and met current recommendations for calories for their gender and age group
(mean ± SD = 1846 ± 775 kcal). However, many had low intakes of nutrients considered
important for women of reproductive age: calcium (n= 18, 67%), vitamin A (n=1 1, 41%), iron
(n=10, 37%), folate (n=7, 26%) and zinc (n=7, 26%).
6.6.2.2.2 Biochemical Data
Although these women had diets low in saturated fat and total fat, and healthy body weights,
several had already been diagnosed with high cholesterol. Blood analyses revealed there were 2
women at high risk, 8 at moderate risk, and 17 had normal blood cholesterol levels. Two
subjects had low HDL (high density lipoprotein) -cholesterol values, 7 had high LDL (low
density protein) -cholesterol values, and 4 had high triglyercides. These biochemical data suggest
that up to 10 women had abnormal blood lipids, which put them at higher risk for heart disease.
Two women had low hemoglobin values, indicating iron-deficiency anemia, likely a result of the
low iron intakes noted above.
Blood plasma values for the omega-3 fatty acid DHA (C22:6N3) ranged from 2.71 to 9.94
(expressed as percent of total plasma lipids) (mean ± SD = 5.80 ± 1.63), and values for EPA
(C20:5N3) ranged from .29 to 3.70 (mean ± SD = 1.13 ± .77). The ratios of omega-3 fatty acids
to the omega-6 fatty acid, arachidonic acid were:
DHA/AA ratio was from .27 to .98 (mean = .56 + .55)
EPA/AA ratio was from .04 to .35 (mean = .11 ± .07)
EPA+DHA/AA ratio was from .32 to 1.15 (mean = .67 ± .21)
Organochlorine residues with higher-than-usually-observed levels were PCBs (n=1 1, 41% of
sample) and beta-BHC (beta-1,2,3,4,5,6-hexachlorocyclohexane)- (n=2 inthe highest 10%). It is
possible that the latter is due to residues in foods eaten or imported from Vietnam (see Kannan et
a!., 1992) and Hong Kong (see Ip, 1990). One individual was in the highest 10% for Mirex, and
Toxicology Excellence for Risk Assessment 6-54 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
this may be related to eating fish from the Niagara River. Eight subjects (30%) had high
mercury values.
6.6.2.3 Exposure Assessment
The total number of all meals of Great Lakes fish during the previous 12 months ranged from 31
to 277 meals (mean ± SD = 99 ± 52.7), averaging 2 meals per week. These women were high
consumers of other types of fish as well; they ate from 5 to 312 meals (mean ± SD = 118 ± 89.9)
of fish from inland locations, and 3 to 306 meals (mean ± SD = 111 ± 80.4) of purchased or
processed fish. Their total fish consumption over 12 months was 83 to 751 meals (mean ± SD =
322 ± 169; median = 306), an average of 6.2 meals of any kind of fish per week.
The percentage of total fish meals that were Great Lakes fish ranged from 9% to 90%; on
average, 39% of all fish meals for this group were sportfish from the Great Lakes (mostly Lake
Ontario). Women were asked to list the top 3 species they consumed most often. They were
rock bass (n=12, 44% of sample), crappie and smalimouth bass (both mentioned by 10 women or
30% of sample), largemouth bass (n=7, 26%), white bass (n=6, 22%), and channel caffish and
freshwater drum (n=5 each, 19% each).
For the 3 species listed, ihe subject was asked which parts of the fish were consumed and how
the fish were cooked. Only 3 women (11%) stated that they discard the belly fat, but 14 (55%)
discard fat from around organs and 19 women (70%) will puncture or cut the skin. Four women
said they eat the fish eggs. The most common ways to prepare the 3 most frequently consumed
species were stir-frying/frying (93%), and using in soups and stews (82%) where the liquid/sauce
would also be consumed. None reported baking, boiling or smoking fish.
Portion sizes depended upon the species and the way the fish were prepared--i.e., as fillets,
pieces, or used whole. None of the women reported eating their top 3 species as fish steaks, only
3 ate any as fillets and 3 ate some as pieces. Every subject reported eating the whole fish, many
for all 3 of their top species. The average portion size when the whole fish was used in a dish,
was 268 grams. At two meals per week, this is approximately 38.3 grams per day.
For estimates of contaminant levels in freshwater fish, values from salmon caught during the
spawning run of fall 1991 in the Credit River near Toronto published by Feely and Jordan (1998)
were used. These data were chosen to estimate contaminant concentration because of the
location of the Credit River on the northwestern shore of Lake Ontario, and the large number of
contaminants analyzed. Contaminant data for small mouth bass and rock bass in the Niagara
River below Niagara Falls were obtained from the New York State Department of
Environmental Conservation, Bureau of Habitat (N.Y. DEC, 1994). There are data for both
rockbass and smallmouth bass for 1993-1994. The fish were not analyzed for a comprehensive
number of contaminants, but they allow for the incorporation of species and geographic variation
in FCI. Table 6-8 lists concentrations from the Feely and Jordan (1998) values for salmon.
Table 6-9 compares hazard indices for each contaminant and mixtures H.I.s for the total mixture
and by target organ. Table 6-10 shows cancer risk, benefit and FCI based solely on cancer risk.
Risk estimates for PCBs were not available, however dose exceeded the RfD at every
consumption level. Tables 6-11 and 6-12 present dose, RfD, HI, cancer risk, methylmercury
Toxicology Excellence for Risk Assessment 6-55 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
risk, total risk, and FCI for smalimouth bass (6-11) and rockbass (6-12). fhe results in these
tables are further explained in the next section.
6.6.2.4 Calculation of FCI
6.6.2.4.1 Salmon from Credit River
Table 6-8 shows the dose of several contaminants as a function of fish consumption. Note that
the doses for PCBs exceed the RID at every consumption level shown, if the value of the Aroclor
1254 RID of EPA is used. At the average consumption for the study population (38g/day),
exposure to PCBs is 25-fold the RID. Currently risk estimates for exposure to PCBs above the
RID are not available (See Chapter 4). Table 6-9 shows hazard indices for individual
compounds, all compounds, and by critical organ/effect. A hazard index is calculated by
dividing the exposure level by the RID. A hazard index greater than one indicates the possibility
of adverse effects. The total hazard index at 38 g/day fish consumption approaches 30. This
indicates that there is a strong possibility that adverse health effects due to the contaminants
present might be observed in the study population. The hazard index for the liver begins to
exceed a value of one around 60 glday of consumption. The hazard index for PCBs, the only
chemicals with an immunological effect is 25.
Table 6-10 shows the estimated ( 95 th percentile) risk of increased cancer incidence. These
estimations use the cancer slope factors (CSFs) published in EPA’s IRIS (U.S. EPA, 1999). Risk
in terms of the framework incorporates a severity factor of three. The severity factor used to
describe coronary heart disease is also three. Figure 6-22 illustrates the relationship between
benefit, risk, and PCI. In general, increases in cancer risk only marginally affect the increase in
benefits due to fish consumption. This general behavior would change if different severity
scores were used for cancer and CHD, but the change would not be dramatic. Figure 6-23 shows
the relative contribution of each contaminant to total cancer risk.
Toxicology Excellence for Risk Assessment 6-56 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Dose (mg/kg-day) as a Function of Fish Consumption Rate and RfDs for Contaminants from EPA (1999).
rncentrations of contaminants are taken from Feely and Jordan (1998) based upon salmon from the Credit River in Ontario.
trations in ppb except TCDD (ppt)
‘r PCBs shown here is for Aroclor 1254 for which the critical effect is immunosuppression. The critical effect for Aroclor 1016 is a developmental effect.
s found in fish. The RID for Aroclor 1254 is used as a surrogate for the mixture found in fish
lazard Indices Assuming Additive Toxicity for Salmon taken from the Credit River. Calculations for Individual, all Compounds, and by Target Organ or Criti’
ates Possibilit ‘of Toxic Effect.
Total
PCB
Biphenyl
DDT
Aldrin
Dieldrin
Heptachior
epoxide
Mirex
Cadmium
Mercury
Arsenic
Total HI
Liver HI
CNS HI
Immune
HI
6.5
5E+00
1E-4
5E-3
7E-3
6E-3
6E-2
6E-2
9E-6
4E-4
2E-5
5
0.1
4E-4
5
38
3E-i-01
8E-4
3E-2
3E-2
3E-2
3E-1
3E-1
5E-5
2E-3
1E-4
30
0.8
2E-3
30
60
4E+0l
1E-3
4E-2
7E-2
4E-2
5E-1
5E-1
9E-5
3E-3
2E-4
40
1
3E-3
40
120
5E÷0l
2E-3
8E-2
1E-1
1E-1
1
1
2E-4
7E-3
4E-4
50
2
6E-3
50
Immune
Kidney
Liver
Liver
Liver
Liver
Liver
Kidney
CNS
Skin
Total
Liver
CNS
Immune
LTotal
TCDD ]
Total
PCB
Biphenyl
Phen-
anthrene
DDE
DDT
Aldrin
Dieldrin
Heptachlor
epoxide
Mirex
Trans-
nonachlor
Cadmium
Lead
Mercur
Ion*
45.08
834.58
64
13
200
23
1.8
.
2.4
6.7
110
27
0.043
1
o.:
-
SE-9
9E-5
7E-6
— 1E-6
2E-5
2E-6
2E-7
3E-7
7E-7
1E-5
3E-6
5E-9
1E-07
4
3E-8
5E-4
4E-5
8E-6
1E-4
1E-5
1E-6
2E-6
4E-6
7E-5
2E-5
3E-8
6E-07
2
5E-8
8E-4
6E-5
1E-5
2E-4
2E-5
2E-6
2E-6
7E-6
1E-4
3E-5
4E-7
1E-06
3
9E-8
2E-3
1E-4
2E-5
4E-4
5E-5
4E-6
5E-6
1E-5
2E-4
5E-5
9E-7
2E-6
7
NA
2E-5t
5E-2
NA
NA
5E-4
3E-5
5E-5
1E-5
2E-4
NA
5E-4
NA
1
NA
Immune
Kidney
NA
Liver
Liver
Liver
Liver
Liver
Liver
NA
Kidney
CNS
CNS
e at
ay
NA
25 fold
No
NA
NA
No
No
No
No
No
NA
No
NA
No
EPA does not currently have an RI
xcellence for Risk Assessment
6-57 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
Table 6-10. Cancer Incidence, Cancer Risk (Including Severity Factor), Benefit (Including Magnitude) and FCI for
Salmon Taken from the Credit River.
Cancer
Incidence
Total
PCB
DDE
DDT
Aidrin
Dieldrin
Heptachior
Epoxide
Arsenic
Total Risk
Total
Benefit
FCI
Cancer
Slope
Factor
2.OOE÷0
0
3.40E-01
3.40E-01
1.70E+01
1.60E+01
9.1OE÷0O
1.50E+00
6.5 g/day
1.8E-04
7.3E-06
8.4E-07
3.3E-06
4.1E-06
6.6E-06
9.8E-09
2.OE-04
38 g/day
1.1E-03
4.3E-05
5.OE-06
1.9E-05
2.4E-05
3.9E-05
5.7E-08
1.2E-03
60 g/day
1.7E-03
6.8E-05
7.8E-06
3.1E-05
3.8E-05
6.1E-05
9.OE-08
1.9E-03
lZOgIday
3.3E-03
1.4E-04
1.6E-05
6.1E-05
7.7E-05
1.2E-04
1.8E-07
3.8E-03
Risk x
Severity
R b
RDDE
RDDT
R JieIdnn
Rhepta
R -
ZR 1
ZB 1
PCI
6.5 g/day
-5.4E-04
-2.2E-05
-2.5E-06
-1.OE-05
-1.3E-05
-2.OE-05
-2.9E-08
-6.1E-04
0.42
0.42
38 glday
-3.2E-03
-1.3E-04
-1.5E-05
-5.8E-05
-7.3E-05
-1.2E-04
-1.7E-07
-3.6E-03
0.68
0.68
60 g/day
-5.OE-03
-2.OE-04
-2.4E-05
-9.2E-05
-1.2E-04
-1.8E-04
-2.7E-07
-5.6E-03
0.68
0.67
120 g/day
-1.OE-02
-4.1E-04
-4.7E-05
-1.8E-04
-2.3E-04
-3.7E-04
-5.4E-07
-0.01
0.68
0.67
Toxicology Excellence for Risk Assessment 6-58 8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
Consumption
PCB dose
DDT
dose
DDE
dose
Mirex
dose
Hexachioro
benzene
dose
mercury
dose
Cancer
incidence
Cancer
risks
Mercury
nsk 55
Total nsk
Total
benefit
FCI
6.5 glday
1.E-04
1.E-06
2.E-05
7.E-06
2.E-07
4.E-05
3.E-04
-8.E-04
0
-8.E-04
0.42
0.42
38 g/day
8.E-04
7.E-06
1.E-04
4.E-05
1.E-06
2.E-04
2.E-03
-5.E-03
0
-5.E-03
0.68
0.68
60 g/day
120 g/day
1.E-03
1.E-05
2.E-04
6.E-05
2.E-06
3.E-04
3.E-03
-8.E-03
-0.0!
-0.02
0.68
0.66
2.E-03
2.E-05
3.E-04
1.E-04
4.E-06
7.E-04
5.E-03
-2.E-02
-0.08
-0.09
0.68
0.59
RfD
2.E-05
5.E-04
NA
2.E-04
8.E-04
1.E-04
Exceedence?
YES
NO
NA
NO
NO
YES
Hazard Index
Range
40 to
100
2 to 7
Cancer incidence calcuiated using EPA slope factors (U S EPA, 1999) Mercury risk estimated using Price etal. (1997)
* severity factor of I incorporated.
Severity factor of 3 incorporated.
Table 6-12 Dose (mgfkg-day) of chemicals detected in rockbass taken from the Niagara River as a function of fish
consumption (g/da).
Consumption
PCB
dose
DDT
dose
DDE
dose
Mirex
dose
Hexa
Chioro
Benzene
dose
Mercury
dose
Cancer
incidence
Cancer
risks
Mercury
nsk 55
Total risk
Total
benefit
FCI
6.5 g/day
3.E-05
3.E-06
6.E-06
2.E-06
2.E-07
4.E-05
7.E-05
-2.E-04
0
-2.E-04
0.42
0.42
38 g/day
2.E-04
2.E-05
4.E-05
1.E-05
1.E-06
2.E-04
4.E-04
-1.E-03
0
-1.E-03
0.68
0.68
60 glday
3.E-04
3.E-05
6.E-05
2.E-05
2.E 06
4.E-04
7.E-04
-2.E-03
-0.01
-0.01
0.68
0.67
120 g/day
6.E-04
5.E-05
1.E-04
4.E-05
4.E-06
7.E-04
1.E-03
-4.E-03
-0.03
-0.03
0.68
0.65
RID
2 E-05
5.E-04
NA
2.E-04
8.E-04
1.E-04
Exceedence?
YES
NO
NA
NO
NO
YES
Hazard Index
range
1.5 to
30
2 to 7
Cancer incidence calcutated using EPA slope factors (U.S. EPA, 1999)
* severity factor of 1 incorporated.
*5 Severity factor of 3 incorporated.
Mercury risk estimated using Price eta! (1997)
Table 6-11. Dose (mg/kg-day) of chemicals detected in smalimouth bass taken from the Niagara River as a function of fish
consumption (g/da ).
Toxicology Excellence for Risk Assessment
6-59
8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
6.6.2.4.2 Rockbass and Smailmouth Bass from the Niagara River
Tables 6-11 and 6-12 show doses, risk, benefit, and FCI for smalirnouth bass (6-11) and rockbass
(6-12). The FCI for rockbass and smailmouth bass at 38 g/day (the study population average
consumption) is 7E-1. PCB intake exceeds the RfD by 10-fold for rockbass; however, the RfD
was exceeded by 40-fold at the same consumption rate for smailmouth bass. If risk above the
RfD estimates were available for PCBs, there would be a difference in FCI between species. As
for the Credit River Salmon FCI, risk estimates as a function of PCB consumption are needed.
Unlike the estimated intake of salmon from the Credit River, methylmercury intake exceeded the
RID at the two highest consumption levels. Methylmercury risk was estimated using the Price et
al. (1997) model as discussed in Chapter 4. At 38 g/day the best estimate ( 50 th percentile)
methylmercury risk was approximately zero for rockbass and smailmouth bass. Methylmercury
risk for Niagara River fish consumption begins to appear at 60 g/day. The best estimate ( 50 th
percentile) of risk at 60 glday is 1% for both species (See Figures 6-25 and 6-26).
6.6.2.5 Discussion
Cancer risk is far outweighed by health benefits from eating Lake Ontario and Niagara River
fish; however, non-cancer risks from PCB mixtures, and to a lesser extent methylmercury are the
primary hazard in this instance. Unfortunately, since the noncancer risks from PCBs could not
be determined (at least during this present effort), the calculation of an FCI as shown in Figure 6-
22 is misleading, in fact hypothetical Figure 6-24 gives an idea of just how misleading Figure 6-
22 can be. Without calculations of risk above the RfD for PCBs it is difficult to calculate an FCI
for the study population. Please note that neither Figures 6-25 nor 6-26 include risks above the
RID for PCBs. Until such information is developed, the risks from these case studies cannot be
fully appreciated; however, the exposure levels here fall within the range of exposure at which
lower scores in reflex, autonomic and habituation were observed in infants from the Lonkey et
al (1996) study (see Chapter 4).
However, this case study illustrates the versatility of the framework. The framework can
incorporate as many chemicals and effects as necessary. Although at 38 glday there was no risk
of subtle neurological effects due to methylmercury intake from Niagara River fish, at 60 g/day
risks begin to appear. This case is also instructive, because it shows how cancer and noncancer
risks are combined. Especially in the case of smailmouth bass, it is apparent that when intake
levels exceed oral RiDs, the noncancer endpoints will rapidly overtake any benefit from eating
fish. For cancer, there is a steady, but small, increase in the risks incurred and decrease in
benefits.
This case study is far more comprehensive than the simple example presented in the Everglades.
It incorporates both cancer and noncancer risks, and compares FCIs for different species and
different bodies of water. The estimated FCI was approximately equal for all three analyses (See
Table 6-13). However, a large difference in PCB exposure exists for which the noncancer risk
could not be quantified.
Toxicology Excellence for Risk Assessment 6-60 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Table 6-13. FCI at 38 g/day for Salmon Rockbass and Smalimouth Bass.
Salmon
Rockbass
Smalimouth
bass
Niagara River
6.8E-1
6.8E-1
Credit River 6.8E-1
Table 6-14. Hazard Index for PCBs at 38 glday.
Niagara River
Salmon
Rockbass
10
Smailmouth
40
bass
Credit River
25
Overall, given the available information, equal FCIs (See Table 6-13) for each species and
location, and the disparity in total PCB hazard index (See Table 6-14), rockbass from the
Niagara River are probably a better source of fish of the three species analyzed here in terms of
minimizing risk. PCB tissue concentrations are the most important factor in the determination of
the FCI, yet as explained in Chapter 4, the data were insufficient to model risk above the RfD for
this case study. This is, and will continue to be, a critical data gap in any application of the
framework in PCB contaminated waters and should be a priority research need. The framework
illustrates the importance of dose response modeling of noncancer health endpoints in
comparative dietary risk assessment
6.7 Overall Cànclusions and Research Needs
This chapter has outlined an approach to evaluate the potential health benefits of consuming fish
against the potential health risks of eating contaminated fish. Consuming uncontaminated fish
(or at least fish that are smaller, younger, or in general less contaminated) may provide health
benefits, but without the, potential health risks associated with contamination. The eating of such
“cleaner” fish rather than more contaminated fish would maximize the net benefit of fish
consumption. This is shown specifically in Figures 6-6 to 6-16 for low versus high
concentrations of chemicals in fish, those chemicals that either bioaccumulate or not, or for fish
contaminated with more that one chemical.
Toxicology Excellence for Risk Assessment 6-61 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
0.30
0.20
0.10
Total Risk
_____________________________ +
20 40 60 80
Consumption (glday)
100 120
• Total PCB
D DDE
DDT
) ( Aldrin
Dieldrin
• Heptachior epoxide
I Arsenic
- Total risk
Benefit
0 FCI
140
Figure 6-22. Risk, Benefit, and FCI as a Function of Consum ption of Salmon from the Credit River.
0.80
0 70
0.60
0 50
0.40
Benefit
..
FCt
I-
0
C.,
En
-I
a,
=
0.00
.010
0
Toxicology Excellence for Risk Assessment
6-62 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
3.OOE-03
2 50E-03
Figure 6-23. Total Cancer Risk and Individual Components for Salmon Taken from the Credit River.
3.50E-03---
200E-03---
1.50E-03 -
1.OOE-03 -
5 OOE-04 -
100 120 140
4 OOE-03
.
C,
c
0
C,
E
U,
w
—4-—Total PCB
—B-- DDE
— --DDT
—)f—Aldrin
—*-— Dieldrin
—-- Heptachior epoxide
—4— Arsenic
—Total risk
000E+00
0
20
40
60
Consumption (giday)
80
Toxicology Excellence for Risk Assessment
6-63 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Figure 6-24 Hypothetical Risk, Benefit, and FCI Assuming that the Shape of the Noncancer Dose-
Response Curve for PCBs is the Same as that for Methylmercury for Salmon from the Credit
River.
0.8
06
04
C,
U
4-
C D
C 0
=
-0 2
-0 4
-0 6
140
0 20 40 60 80 100 120
• Total Cancer
• Mercury surrogate for PCBs
(risk for x Fold RfD)
A Benefit
) ( FCI
Consumption (g(day)
Toxicology Excellence for Risk Assessment
6-64 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
0 20 40 60 80 100 120
Figure 6-25. Risk, Benefit, and FCI as a Function of Niagara River Rockbass Consumption.
08
0,7
06
05 -
04
03
02
01
0-
-0 1
Benefit
• -- -- - --------- - - - ------ - - - - - -- -- --
- el 1::i: i i:i
FCI
Risk
0 Cancer risk
M Mercury risk”
—0——Total Risk
—0-—Total Benefit
—-A— FCI
* 95 Lh Percentile
50 th Percentile
Consumption (giday)
140
Toxicology Excellence for Risk Assessment
6-65 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Figure 6-26. Risk, Benefit, and FCI as a Function of Niagara River Smalimouth Bass
Consumption.
O 80
Ca ncer risk
0 Mercury risk”
0
0 Total risk
• 030
0 Total benelit
020 * FCI
010
- -0 0
Risk
-0 20
Fush consumption(glday)
* 95 th Percentile
** 5O Percentile
Toxicology Excellence for Risk Assessment 6-66 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
When alternatives to th eating of contaminated fish are not available, it may be appropriate to
weigh the risks of eating less of these contaminated fish with the benefits gained from eating
more of these same fish. The framework developed here can crudely compare these risks and
benefits. However, this framework has a number of significant data gaps. These gaps are
sufficiently large so as to prevent any definitive conclusions from this study or any overall
recommendations regarding existing fish consumption advisory programs of the U.S. or other
countries. Further study is needed to confirm and extend these preliminary findings.
This framework is an initial attempt to evaluate risks and benefits (both qualitatively and
quantitatively) on a common scale. Constructing this framework has identified numerous areas
that need further research and development. Two needs seem paramount. First, better
estimations of benefits are needed for the general population and its sensitive subgroups.
Although information in this text is highly suggestive of the protective effects of eating fish and
allows some quantification, more definitive work is needed to support the quantitative values
shown in Table 6-1. Second, better risk information is needed on the chemicals that commonly
contaminate fish. Indeed, we have sufficient knowledge on the toxicity of most of these
pollutants that quantifying risks above the RfD should be done. This information is essential for
this framework, or any other construct, to be effective.
Specific conclusions and research needs are summarized below.
o Incorporate full range of benefits data: The examples of benefits that are presented in the
framework are representative based on the available data. However, they do not
incorporate the entire quantitative benefits data (see Table 2-1). At a minimum, all the
data sets supporting, or contradicting, the existence of a particular health benefit should
be further summarized and discussed, and data should be presented for any endpoint
having quantitative benefits information. A meta-analysis might be considered for each
endpoint supported by more than one data set. This might allow the development of a
single dose-response curve for each health endpoint. Such single dose-response benefit
curves would make the framework easier to use.
o Reconsider severity schemes: For this framework, the severity approach of EPA and
ATSDR for estimating RfDs/RfCs and MRLs (Table 6-2) was used to modify the health
risks associated with chemical exposure. This approach has the advantages of simplicity,
familiarity and consistency with the use of information from EPA’s IRIS, and of ATSDR
information found in its toxicological profiles. One shortcoming of this approach is the
implied equal spacing between levels. There is no scientific or mathematical justification
proposed for a FEL being considered three times as “severe” as a less serious LOAEL.
Other caveats were discussed in Chapter 6 (see Table 6-3).
In like fashion, a modifier to the magnitude of health benefits accrued from eating fish
was used to roughly compare with the risk of different health endpoints. This modifier of
health benefits (e.g., coronary heart disease avoided) was ranked as none, minimal,
moderate or maximum. This modifier has the advantages of simplicity and consistency
with the use of information for health risks. As for health risks, however, the scheme for
health benefits is being used in a quantitative fashion in the framework, and this results in
several shortcomings which were discussed in Chapter 6 (see Table 6-3).
Toxicology Excellence for Risk Assessment 6-67 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary’Risk
Other severity schemes should be explored for comparing the health risks and benefits of
fish consumption. The results are likely to be more complex, however. Several of these
schemes will necessitate additional judgment regarding the appropriate severity level of
both the critical effect and benefit. At least one of these schemes (i.e., Ponce et aL, 1998)
also incorporates the concept of duration of the effect or benefit through the use of
QALYs. Every attempt should be made to see if these more complex severity schemes
add value when compared to the simpler one, which was used here.
Explicitly incorporate uncertainty: It is important to recognize that with the exception of
noncancer risks (see Figures 4-1 and 4-2), uncertainty in health benefits and risks is not
dealt with explicitly by the framework in its current version. Moreover, the uncertainty
surrounding the estimates of the different benefits and risks associated with eating fish
are unlikely to be the same. For example, the uncertainty surrounding estimated cancer
risks based on animal toxicity data is likely much greater than the uncertainty
surrounding estimated coronary heart disease benefits based on human data.
An important future refinement of the framework would be explicit consideration and
quantification of uncertainty surrounding estimates of potential health risk and benefit,
because both have the potential to alter the interpretation of the framework and the
resulting PCI. Future efforts should be devoted to this area.
• Conduct a sensitivity analysis: The current version of the framework uses fixed inputs
for most of the variables that determine potential risk, potential benefits and the FCI.
Such fixed information helped develop the framework and also allowed for exploration of
a number of issues associated with its use. However, many of these fixed parameters can
and do vary, and additional work is needed to investigate how the FCI changes when
these parameters are changed. Such a sensitivity analysis would greatly improve
interpretation of the framework results and perhaps help focus future work on the input
variables that have the greatest potential to affect the FCI.
• Evaluate additional mixtures of chemicals: The framework and ca e studies used only a
few chemicals and concentrations to examine the relationship between potential risks and
benefits of eating contaminated fish. While the choice of these chemicals reflected the
frequency of residues and number of fish consumption advisories (Table 4-1), other
chemicals are also found in fish. While the analysis of a limited number of chemicals is
useful for the development of the framework and its application, the choice of
concentrations could perhaps better reflect those typically observed in waters of the U.S
(the example concentrations presented here were much higher thanaverage). Based upon
comments from the Advisory Committee, methylmercury, PCBs and dioxin are the
chemicals for which advisories are most commonly needed and typical high
concentrations might vary between 0.2 and 1 mg/kg for methylmercury and PCBs, and be
around 1 ngfkg for dioxin toxic equivalents.
• Develop risk curves for non-sensitive groups: For health risks, specific risk curves for
non-sensitive members of the population could also be developed. This would avoid
Toxicology Excellence for Risk Assessment 6-68 8/6/99
-------
Coofieralive Agreement with U S EPA on Comparative Dietary Risk
matching the he2dth risk for the sensitive individual with the health benefit to the average
individual. For example, with methylmercury the risk curve is based on risk to the infant
and fetus, whereas the benefit curve was for the adult. Use of an adult risk curve would
have changed the conclusions of the Florida Everglades case study.
Develop risk curves for doses above the RID for selected pollutants, in particular for
PCBs: It comes as no surprise that PCBs are a common pollutant in fish and one that
needs to be better studied. As amply demonstrated by the Vietnamese case study,
however, the need for determining the risk above the PCB RfD is paramount. Quite
simply, this case study is woefully deficient without this determination, as demonstrated
by reference to the differences between Figures 6-22 and 6-24.
6.8 References
Aberg, B., L. Ekman, R. Falk, eta]. 1969. Metabolism of methyl mercury (203Hg) compounds
in man. Arch. Environ. Health. 19: 478-484. (As cited in U.S. EPA, 1999)
AFS (American Fisheries Society). 1997. Proceedings of the 1997 American Fisheries Society
forum on contaminants in fish. December 8-10. EVS consultants. Seattle, Washin ton.
Anderson, P.A. and J.B. Wiener. 1995. Eating Fish. Ep: Risk vs. Risk: Tradeoffs In: Protecting
Health and the Environment. J.D. Graham and J.B. Wiener, eds. Harvard University Press,
Cambridge, Massachusetts. pp. 104-124.
Buck, G.M., L.E. Sever, and P. Mendola. 1997. Consumption of contaminated sport fish from
Lake Ontario and time-to-pregnancy. Am. J. Epid. 146: 949-954.
Burger, J., J. Sanchez and M. Gochfeld. 1998. Fishing, consumption, and risk perception in
fisherfolk along an east coast estuary. Environ. Res. 77A: 25-35.
Choi, S.C., R. Bartha. 1994. Environmental factors affecting mercury methylation in estuarine
sediments. Bull. Environ. Contam. Toxicol. 53(6): 805-12
Cox, C., T.W. Clarkson, D.O. Marsh, et al. 1989. Dose-response analysis of infants prenatally
exposed to methylmercury: An application of a single compartment model to single-strand hair
analysis. Environ. Res. 49: 318-332. (As cited in U.S. EPA, 1999)
Davidson, P.W., G.J. Myers, C. Cox, et al. 1998. Effects of renatal and postnatal
methylmercury exposure from fish consumption on neurodevelopment: Outcomes at 66 months
of age in the Seychelles hild Development Study. J.A.M.A. 280(8): 701-707.
Daviglus, M.L., J. Stamler, A.J. Orencia, eta]. 1997. Fish consumption and the 30-year risk of
fatal myocardial infarction. N. Engl. J. Med. 336(15): 1046-1053.
DeRosa, C.T., J.F. Stara, and P.R. Durkin. 1985. Ranking chemicals based upon chronic
toxicity data. Toxicol. md. Health. 1(4): 177-92.
Toxicology Excellence for Risk Assessment 6-69 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietar Risk
DeRosa, C.T., M.L. Dourson, and R. Osborne. 1989. Risk assessment initiatives for noncancer
endpoints: Implications for risk characterization of chemical mixtures. Toxicol. md Health.
5(5): 805-824.
Dourson, M.L., R.C. Hertzberg, R. Hartung, eta]. 1985. Novel approaches for the estimation of
acceptable daily intake. Toxicol. md. Health. 1(4): 23-41.
Dewailly, E., P. Ayotte, C. Laliberte, et al. 1996. Polychiorinated biphenyl (PCB) and
dichlorodiphenyl dichioroethylene (DDE) concentrations in the breast milk of women in Quebec.
Am. J. Pub. Health. 86: 124 1-1246.
Dewailly, E., J-P Weber, S. Gingras, eta!. 1991. Coplanar PCBs in human milk in the province
of Quebec, Canada: are they more toxic than dioxin for breast fed infants 7 Bull Environ Contam
Toxicol. 47: 491-498.
Durkin, P. 1998. Personal communication with Michael Dourson. December.
Feely, M.M., and S.A. Jordan. 1998. Dietary and tissue residue analysis and contaminant intake
estimations in rats consuming diets composed of Great Lakes salmon: a multigeneration study.
Reg. Toxicol. Pharmacol. 27: S8-S17.
Fleming, L.E., S. Watkins, R. Kaderman, eta!. 1995. Mercury exposure in humans through
food consumption from the everglades of Florida. Water Air Soil Pollut. 80: 4 1-48.
Gillum, R.F. 1996. Fish consumption and stroke incidence. Stroke. 27(7): 1254
Grandjean, P., P. Weihe, R.F. White, eta]. 1997. Cognitive deficit in 7-year-old children with
prenatal exposure to methylmercury. Neurotoxicol. Teratol. 19: 417-428.
Hartung, R. and P.R. Durkin. 1986. Ranking the severity of toxic effects; potential applications
to risk assessment. Comments on Toxicology. 1(1): 49-63.
Hutchison, R. and C.E. Kraft. 1994. Hmong fishing activity and fish consumption. J. Great
Lakes Res. 20: 471-478.
Ip, H.N.H. 1990. Chlorinated pesticides in foodstuffs in Hong Kong. Arch. Environ. Contam.
Toxicol. 19: 291-296.
Jacobson, J.L. and S.W. Jacobson. 1996. Intellectual impairment in children exposed to
Polychlorinated biphenyls in utero. N. Engi. J. Med. 335: 783-789.
Jarabek, A.M. 1994. Inhalation RfC methodology: dosimetric adjustments and dose-response
estimation of noncancer toxicity in the upper respiratory tract. Inhal. Toxicol. 6(suppl): 301-
325.
Toxicology Excellence for Risk Assessment 6-70 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Kannan, K., S. Tanabe, H.T. Quynh, eta!. 1992. Residue pattern and dietary intake of persistent
organochiorine compounds in foodstuffs from Vietnam. Arch. Environ. Contam. Toxicol. 22:
367-374.
Krabbenhoft, D,P. 1996. Mercury studies in the Florida Everglades. U.S. Department of the
Interior, U.S. Geologic Survey. Fact Sheet FS-166-96.
Kromhout, D., E.B. Bosschieter and C.L. Coulander. 1985. The inverse relation between fish
consumption and 20-year mortality from coronary heart disease. N. Engi. J. Med. 312(19):
1205-1209.
Marsh, D.O., T.W. Clarkson, C. Cox, eta]. 1987. Fetal methylmercury poisoning: relationship
between concentration in single strands of maternal hair and child effects. Arch. Neurol. 44:
1017- 1022.
Mendola, P., G.M. Buck, L.E. Sever, eta]. 1997. Consumption of PCB-contaminated
freshwater fish and shortened menstrual cycle length. Am. J. Epid. 146: 955-960.
Meek, M.E., R. Newhook, R.G. Liteplo, eta!. 1994. Approach to assessment of risk to human
health for priority substances under the Canadian Environmental Protection Act. Environmental
Carcinogenesis and Ecotoxicology Reviews. C12(2): 105-134.
Mes J. and D. Weber. 1989. Non-orthochlorine substituted coplanar polychiorinated biphenyl
congeners in Canadian adipose tissue, breast milk, and fatty foods. Chemosphere. 19 :1357-
1365.
Miettinen, J.K., T. Rahola, T. Hattula, etal. 1971. Elimination of 203-Hg methylmercury in
man. Ann. Clin. Res. 3: 116-122. (As cited in U.S. EPA, 1998)
Minnesota Department of Health. 1998. Fish Consumption Advisory. Minnesota.
N.Y. DEC (New York State Department of Environmental Conservation). 1994. Memorandum
concerning chemical contaminants in fish from the Niagara River from Lawrence C. Skinner,
Section Head, Environmental Monitoring Section, New York State Department of
Environmental Conservation, Bureau of Habitat..
Newsome, WH, D.J. Davies, and W.F. Sun. 1998. Residues of polychlorinated biphenyls (PCB)
in fatty foods of the Canadian Diet. Food Additives and Contaminants. 15: 19-29.
Phelps, R.W., T.W. Clarkson, T.G. Kershaw, et a]. 1980. Interrelationships of blood and hair
mercury concentrations in a North American population exposed to methylmercury. Arch.
Environ. Health. 35: 161-168. (As cited in U.S. EPA, 1999)
Pohi, H. R. and H. G. Abadin. 1995. Utilizing uncertainty factors in minimal risk levels
derivation. Reg. Toxicol Pharmacol. 22: 180-188.
Toxicology Excellence for Risk Assessment 6-71 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Ponce, R.A., S.M. Bartell, D. LaFlamme, et al. 1998. Quantitative analysis of risks and benefits
for public health decisions applied to fish consumption. The Toxicologist. 42(1-S): 45.
Price, P., R. Keenan, J. Swartout, et al. 1997. An approach for modeling noncancer dose
responses with an emphasis on uncertainty. Risk Anal. 17: 427-437.
Science Subgroup. 1997. Ecologic and precursor success criteria for South Florida ecosystem
restoration: A Report to the Working Group of the South Florida Ecosystem Restoration Task
Force. Online: http://everglades.fiu.eduITASKFORCEIPRECURSORItoc.html
Shapiro, Jean A., etal. 1996. Diet and rheumatoid arthritis in women: A possible protective
effect of fish consumption. Epidemiology. 7(3): 256-263.
Sherlock, J.C., D.G. Lindsay, J. Hislop, eta!. 1982. Duplication diet study on mercury intake by
fish consumers in the United Kingdom. Arch. Environ. Health. 37(5): 271-278. (As cited in
U.S. EPA, 1999)
Siscovick, D.S., et al. 1995. Dietary intake and cell membrane levels of long-chain n-3
polyunsaturated fatty acids and the risk of primary cardiac arrest. J.A.M.A. 274 (17): 1363-
1367.
Suzuki, T., T. Hongo, J. Yoshinaga et al. 1993. The hair-organ relationship in mercury
concentration in contemporary Japanese. Arch. Environ. Health. 48: 221 -229. (As cited in U.S.
EPA, 1999)
Tsubaki, T.K. and K. Irukayama. 1977. Minamata Disease: Methylmercury Poisoning in
Minamata and Niigata, Japan. Elsevier Science Publishers, New York. p. 143-253. (As cited in
U.S. EPA, 1999)
U.S. EPA. 1988. Technical support document on risk assessment of chemical mixtures. Office
of Research and Development. EPAJ600/8-90/064.
U.S. EPA. 1989. Risk Assessment Guidance for Superfund. Volume I: Human Health
evaluation Manual (Part A). Interim Final, December 1989. EPAJ54OI1-891002.
U.S. EPA. 1992. National Study of Chemical Residues in Fish. Office of Science and
Technology. EPA 823-R-92-008b.
U.S. EPA. 1995. Use of the benchmark dose approach in health risk assessment.
EPAJ63OIR-94/007
U.S. EPA. 1996. Proposed Guidelines for Carcinogen Risk Assessment. Office of Research
and Development. Washington, DC. EPA/6001P-92/003C
U.S. EPA. 1998. Comparative Risk Framework: methodology and case study. SAB review
draft. National Center for Environmental Assessment. Cincinnati, OH. NCEA-C-0135
Toxicology Excellence for Risk Assessment 6-72 8/6/99
-------
Cootierative Agreement with U.S. EPA on Comparative Dietary Risk
U.S. EPA. 1999. Integrated Risk Information System (IRIS). Substance file for methyl
mercury. Online: http://www.epa. gov/ngispgm3/iris/substIOO73.htm .
Weihe, P., P. Grandjean, F. Debes, et al. 1996. Health implications for Faroe Islanders of heavy
metals and PCBs from pilot whales. Sci Total Environ. 186: 1-179.
WHO (World Health Organization). 1990. Environmental Health Criteria 101: Methylmercury.
Geneva. (As cited in U.S. EPA, 1998)
Toxicology Excellence for Risk Assessment 6-73 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
7 Using and Communicating the Comparative Dietary Risk Framework
This framework and approach could be used by state, tribal, and local risk managers who set fish
advisories to provide additional information on possible health benefits to those who fish and eat
fish. Because of the data intense process and results of the FCI, a solid risk communication
program is necessary to insure successful usage of the information generated. The risk
communication process associated with fish consumption health advisories has been described in
depth in U.S. EPA’s C uidance for Assessing Chemical Contaminant 11 ata for Use in Fish
Advisories, Volume 4 (U.S. EPA 1995). This chapter summarizes key elements of that process
applied to the comparative dietary risk framework, emphasizing that risk communication is a
process of information exchange between the target audience and the risk communicator.
Two cautions about communicating information from the framework should be reiterated. First,
instituting a risk communication program assumes the existence of quality information to
communicate. Developing a risk communication approach at this stage of evolution in the
Comparative Dietary Risk Framework is appropriate; however, implementing a risk
communication program is not appropriate until the data are available for calculating the actual
values that would be used in the framework and the FCI.
Second, although the framework provides a mechanism for comparing risks and benefits
associated with fish consumption, it is not a justification for accepting fish consumption risks as
long as there is a net benefit. Decisions about acceptable risks and distribution of risks and
benefits throughout society is a social decision, to be made collectively by the communities
affected (Shrader-Frechette, 1990; Knuth, 1995). That the FCI may demonstrate cases in which
fish consumption benefits appear to outweigh the risks is not a license to pollute. Rather, society
must determine policy about long-term goals for minimizing environmental pollution based on a
range of ethical, economic, and social criteria. Further, environmental justice and equity issues
are raised when certain communities are forced to assume more health risks than others. For
example, some communities consuming large amounts of fish may score high on the “benefits”
side of the equation. If hose fish are chemically contaminated, however, the same communities
also score high on the “risks” side. Use of the framework and FCI does not imply the proper
choice is simply achieving a situation in which the net risks and benefits are zero. Rather, the
framework helps make the tradeoffs between risks and benefits transparent, and should be used
to foster discussion on environmental equity and justice issues, and questions of who should bear
the costs of pollution vs. derive the benefits from the fisheries resource.
7.1 Overview of Risk Communication as a Process
Risk communication includes several stages: problem analysis; audience needs assessment;
communication strategy design; communication strategy implementation, and evaluation (Fig. 7-
1). Problem analysis includes examination of both external and internal factors that may or
should influence the risk communication program. This first phase also involves identifying the
specific objectives to be achieved through the risk communication process.
The second phase, audience identification and needs assessment, begins with target audience
identification. The risk communication objectives established in the problem analysis phase
Toxicology Excellence for Risk Assessment 7-1 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
provide insights about potential types of target audiences. In this phase, those audiences are
characterized in terms of demographics; awareness and knowledge about advisories,
contaminants, and fish consumption; beliefs and attitudes about related topics; and behaviors
related to fishing and fish consumption. Based on this information about target audiences, health
advisory information needs are identified.
Strategy design and implementation, phase three, reflects the communication objectives and the
target audience information needs identified earlier. Strategy design includes considerations
about the style of communication (e.g., format, tone, text vs. graphics, reading level), the content
(e.g., comparisons of sites, health effects, health benefits), and means for disseminating the
information (e.g., mass media, interpersonal contacts, specialized media). Strategy
implementation includes pre-testing the message, modifying the design as needed, creating a
timetable, and finalizing and disseminating the message(s).
Evaluation as a component of the risk communication process occurs at three stages of the
process. Formative evaluation occurs during problem analysis, audience needs assessment, and
the initial stage of communication strategy design. Process evaluation occurs during the
communication strategy implementation period. Summative evaluation occurs after the
communication strategy has been implemented, but refers back to information identified in the
problem analysis and audience needs assessment phases.
7.2 Designing, Implementing, and Evaluating a Communication Program for the
Comparative Dietary Risk Framework
This section examines each stage of the risk communication process in relation to using the
Comparative Dietary Risk Framework, indicating both strengths and challenges. The
Comparative Dietary Risk Framework responds to several risk communication needs identified
for fish consumption health advisory programs. Recent studies of angler and fish consumer
response to advisory communications suggest that potential fish consumers desire particular
types of information, although these information needs may differ among target audiences.
Information about how risks change with different levels of fish consumption has been identified
by anglers, fishery experts, and health care experts as very important for health advisory
communication programs (Velicer and Knuth, 1994). Studies of licensed anglers have indicated
the perceived importance of health advisory information about topics such’ as potential health
benefits and risks associated with fish consumption, how risks change as more or less fish is
eaten, and comparing the health risks of eating fish with the risks from other protein sources
(e.g., Connelly etal., 1992; Connelly and Knuth, 1993).
Other studies have demonstrated that anglers do respond to health risk information by changing
their fishing-related behavior. Changes include eating less sport-caught fish, changing fish-
cleaning methods, changing fishing locations, changing species eaten, changing the size of fish
eaten, and changing cooking methods (Connelly eta!. 1992). Connelly et aL (1996) provided
Toxicology Excellence for Risk Assessment 7-2 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
Figure 7-1. The risk communication process, adapted from Velicer and Knuth (1994).
RISK COMMUNICATION PROCESS
[ Problem Analysis
Audience Needs
Assessment
.
Communication
________ Communication Strategy
lmp rnentation 4
Toxicology Excellence for Risk Assessment 7-3 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
evidence that fish consumption suppression (anglers eating less fish than they would in the
absence of health advisories) was prevalent among Lake Ontario anglers. Montgomery and
Needelman (1997) outlined a method to quantify the economic impacts of fishing behavior
related to chemical contamination of fisheries.
Key among information needs identified is the desire of many potential fish consumers to
understand the impacts of the advisory, and of fish consumption, for them individually. The
framework enables risk communicators to facilitate this understanding, providing a mechanism
to better meet many risk communication goals.
The FCI is the key communication element of the framework. The FCI may be conveyed
graphically across a range of fish consumption rates (see Figure 6-5 Health Scale as a Function
of Fish Consumption Rate), or as a single value. The latter approach is similar to many current
health advisory communication programs, in which a determined Level of fish consumption is
recommended for a particular group (e.g., eat no more than 1 fish meal a month). The graphical
presentation of FCI, however, conveys a greater degree of information than does a single fish
consumption rate. The reader is able to visualize how benefits and risks, and hence net benefits,
change with changes in the fish consumption rate. Ideally, it would be possible to convey this
comparative information for different species of fish, because many fish consumers face the
question of ‘substitution” rather than “abstinence;” i.e., changing species, sizes, or locations of
fish caught and eaten, rather than reducing or eliminating overall fish consumption.
In theory, this graphical presentation increases the information available to the decision-maker (a
government agency or the potential fish consumer) and is thus more individualized. Used in this
way, however, the framework focuses on individual diets, increasing the risk communication
challenge particularly for agencies used to developing and disseminating a” one size fits all” type
of health advisory. Limited resources for health advisory and associated communication
programs, coupled with a mandate to address large, diverse groups in society, constrain the
abilities of agencies to target very specific audiences.
From a risk communication standpoint, the framework’s greatest strengths (i.e., use of the FCI
across varying fish consumption rates to estimate changes in net benefits) lie in application to
local areas, to situations in which individual consumer concerns can be identified, and to
internally homogenous groups with particular cultural or dietary concerns.. On larger (e.g.,
statewide, region wide) scales, using a single FCI value on which to base fish consumption
recommendations may be the best option. Statewide, portraying all the possible combinations of
exposure and benefits would be infeasible, but presenting summary information would be
possible to help anglers decide on appropriate “substitutions” - i.e., switching species, sizes, or
fishing locations to target those that have, on balance, greater benefits than risks. In local or
special audience cases, however, it may be feasible to present a range of figures (FCI graphs)
demonstrating a limited number of exposure and benefit scenarios to enhance local
understanding of the options and tradeoffs available.
Toxicology Excellence for Risk Assessment 7-4 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
7.2.1 Problem Analysis
The first stage in risk communication is problem analysis, a careful consideration of external and
internal factors that influence the selection of risk communication objectives and the likelihood
of meeting those objectives. Objectives indicate the outcomes that reflect the mandate of the
agency and the impact to be achieved through the risk communication program.
Analyzing external and internal factors improves understanding of the context in which the
health advisory risk communication program will occur. Without an understanding of the
context, it is difficult to establish realistic objectives. Contextual factors external to the health
advisory program include the characteristics of the community(ies) to be affected by the health
advisory, and the degree of certainty and completeness of information used to establish the
health advisory. Data needs for calculating the FCI within the framework have been discussed
earlier. Internal factors include staff, budget, and other resources available to or required by the
health advisory program.
A variety of objectives are often associated with health advisory programs. The involvement of
more than one agency in the development and dissemination of health advisories (Reinert et aL
1991) often complicate such programs. For example, environmental quality agencies may
conduct the chemical and fish tissue monitoring programs. Health agencies may conduct the risk
assessment (or calculate the FCI). Health agencies in cooperation with fishery management
agencies may be charged with communicating the health advice. Because these agencies have
differing mandates, they may have differing objectives they hope to accomplish through health
advisory programs.
In a study of Great Lakes agencies involved in health advisory programs, one of the most
frequently cited objectives (by all types of agencies) for health advisory communication
programs was to enable consumers to “make their own, informed decision” about fish
consumption (Knuth and Connelly 1991). Other objectives focused on reducing human health
risks, educating people about risk-reducing fish preparation methods, encouraging public support
for fisheries and facilitating positive fishery resource use, and following agency mandates.
Value-based risk management judgments are inherent in risk communication, particularly using
the FCI-based approach. Because the framework (through the graphic presentation of FCI)
compares potential health risks and benefits of eating fish affected by contaminants, risk
communicators have an increased ability to help individual fish consumers be informed in their
decision making, addressing the first objective noted above. Including a graph such as Fig. 4 in
health advisory communication programs would improve a potential fish consumer’ s ability to
make an informed decision based on his or her own choice about balancing fish consumption
benefits with risks. However, objectives focused on reducing human health risks might prompt
agencies to choose the FCI represented by zero or near-zero risk, the PCI represented by
maximum net benefits, or some other value. Objectives focused on encouraging public support
for and use of fishery resources might support choosing the FCI represented by zero net benefits
(i.e., that level at which net risks are offset by net benefits so that maximum fish consumption
opportunities at zero net risk are promoted) or some other value.
Toxicology Excellence for Risk Assessment 7-5 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Thus, the communication resources available to an agency, the context in which the
communication will occur, and the objectives inherent in the health advisory program will
influence the way in which the framework and FCI are implemented. As noted, the FCI may be
communicated graphically as a range of values across differing fish consumption levels, or as a
point value recommending one set fish consumption rate as a maximum allowable level. The
single point value selected would differ depending on the program objectives, reflecting an
agency s decision about the appropriate manner in which to balance risks and benefits associated
with fish consumption.
7.2.2 Audience Identification and Needs Assessment
The second stage in the risk communication process is audience identification and needs
assessment. A combination of” expert” input from those within the health advisory agency (ies)
and those knowledgeable about the potential target audiences, and direct input from the potential
audiences is usually required. u Expert” viewpoints about target audience information needs may
not agree with needs identified by the target audiences themselves (Velicer and Knuth 1994).
Judgments about factors such as the relative importance of risks and benefits to community
members, and about cultural importance of fish consumption in the diet or as part of local
tradition, are needed. Risk communication experts may not be informed enough or aware
enough to make such judgments without considerable local input.
Identifying potential target audiences is the first step in this phase. Ideally, selection of target
audiences would have been completed during development of the FCI. Calculations within the
framework require a variety of data about health status and impacts, dietary tradeoffs, etc., which
should be collected for the target audiences of concern. From a communication standpoint,
target audience segmentation should be based on identifying groups that are relatively
homogenous from the perspective of information content needs, and who can be reached through
similar information dissemination mechanisms.
Characterizing the information needs of target audiences includes assessing a variety of factors,
such as audience demographics (age, gender, education, language ability, income, residence,
race, family status); typical information sources used; fishing and fish consumption experience;
and prior awareness, knowledge, beliefs, and behaviors related to fishing and fish consumption,
including cultural forces. These information needs are similar to those in the risk
communication components of existing health advisory programs. Using the framework,
however, these information needs are relevant not only to the risk communication process, but
also to the preceding process of developing the basic health advisory recommendations.
The information requirements for an audience needs assessment using the framework occur
earlier in the health advisory development process than in traditional health advisory approaches.
The framework can consider different subgroups (e.g., adults, children, breast-feeding mothers
and infants), health benefits, cancer and non-cancer health endpoints, biological and perceived
severity of health endpoints, and cultural values. Because of this, more information about target
audience characteristics, behaviors, values, and information needs is required at the time the fish
consumption recommendations are generated - not just at the risk communication stage.
Toxicology Excellence for Risk Assessment 7-6 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Emphasis must be placed on incorporating information about target audiences early in the
process of developing data to insert in the framework.
Applying the framework to a particular audience requires knowledge of how the target
community (or individuals within the community) perceives the severity associated with
different health outcomes, what health outcomes are most important in the community given its
demographics, and the cultural values the community assigns to fish consumption. Other
important information includes understanding how dietary behavior changes in response to
reduced fish consumption (e.g., what alternative food sources would replace fish) so the
appropriate comparisons of health risks/benefits between current and modified (less fish
consumption) dietary patterns can be factored into the framework.
Early and in-depth assessment of target audience information needs is particularly important for
applications of the framework in which cultural risk factors will be incorporated. Rarely (if ever)
will a group of risk-management or risk-assessment experts be able to characterize adequately
the cultural risks associated with fish consumption and/or potential loss of fish consumption. If
fish consumption and associated activities are a key element of the local culture (e.g., see
Chapter 5 Socio-cuitural Considerations ofFish Consumption) decision-making methods are
needed that will allow the local community to help quantify or characterize the perceived
severity and cultural risk factors that will be incorporated into the calculations of the framework.
Techniques for determining target audience information needs, based on input from both
audience members and experts, are detailed in U.S. EPA (1995). These include personal
interviews and group discussions, mail and telephone surveys, and document review.
7.2.3 Communication Program Strategy Design and Implementation
The flexibility of the framework for designing and implementing risk communication programs
is both appealing and challenging. Because of the types of information used in operating the
framework, very specific risk communication messages can be developed that are responsive to
the special concerns of a given subpopulation, community, or individual. The converse,
however, is the challenge of providing all of the information needed for the framework to be
applied to its fullest - detailed descriptions of perceptions, cultural values, and behaviors within
the community of concern. Thus, this approach can be particularly cost-intensive in terms of
information and the staff resources needed to acquire it.
In many cases, decisions will have to be made about which set of perceptions and values to
apply, treating a known heterogeneous community as if it was homogenous, to simplify
calculation and communication of the FCI. In large, heterogeneous communities, therefore, the
full benefits of the framework may not be realized because of the number of assumptions (of
homogeneity) that will be necessary. However, in small communities that are homogeneous on
several parameters important in the framework (e.g., perceived severity of health outcomes,
cultural importance of fish in the diet), the FCI should improve the ability of individuals to make
their own “informed decisions” about an appropriate fish consumption rate, particularly if the
FCI is presented graphically relative to differing fish consumption levels and species or
locations.
Toxicology Excellence for Risk Assessment 7-7 8/6/99
-------
Cooperative Agreement with U.S EPA on Comparative Dietary Risk
8 Conclusions and Research Needs
8.1 Overall Conclusions and Research Needs
This document has ouflined an approach to evaluate the potential health benefits of consuming
fish against the potential health risks of eating contaminated fish. Some evidence exists for an
association between decreased risk of CHD or MI, and consumption of small amounts of fish,
including mainly lean (non-fatty) fish. Additional studies have seen some association between
eating fish and reduced risk of stroke and arthritis, and enhanced immunological and nervous
system development. These data, along with the superior nutritional value of fish, are strong
enough that public health officials routinely encourage the public to eat more fish.
Consuming uncontaminated fish (or at least fish that are smaller, younger, or in general less
contaminated) may provide health benefits as mentioned above, but without the potential health
risks associated with contamination. Before eating any contaminated fish, consumers should
consider fish supplies from cleaner water bodies, eating smaller, less contaminated fish, and
cooking and cleaning methods that reduce contaminants. The eating of such “cleaner” fish rather
than more contaminated fish would maximize the net benefit of fish consumption. This is shown
specifically in Figures 6-6 to 6-16 for low versus high concentrations of chemicals in fish, for
those chemicals that bioaccumulate, or for fish contaminated with more that one chemical.
When alternatives to the eating of contaminated fish are not available or desired, it may be
appropriate to weigh th€ risks of eating less of these contaminated fish with the benefits gained
from eating more of these same fish. The framework developed here can crudely compare these
risks and benefits. However, this framework has a number of significant data gaps. These gaps
are sufficiently large so as to prevent any definitive conclusions from this study or any overall
recommendations regarding existing fish consumption advisory programs of the U.S. or other
countries. Further study is needed to confirm and extend these preliminary findings.
This framework is an initial attempt to evaluate risks and benefits (both qualitatively and
quantitatively) on a common scale. Constructing this framework has identified numerous areas
that need further research and development. Two needs seem paramount. First, better
estimations of benefits are needed for the general population and its sensitive subgroups.
Although information ir this text is highly suggestive of the protective effects of eating fish and
allows some quantification, more definitive work is needed to support the quantitative values
shown in Table 6-1. Second, better risk information is needed on the chemicals that commonly
contaminate fish. Indeed, we already have sufficient knowledge on the toxicity of most of these
pollutants that quantifying risks above the RID can be done. This information is essential for
this framework, or any other construct, to be effective.
Specific conclusions and research needs on each technical chapter are summarized below.
8.2 Chapter 2
Some evidence exists for an association between decreased risk of CHD or MI, reduced risk of
stroke and arthritis, enhanced immunological and nervous system development, and consumption
Toxicology Excellence for Risk Assessment 8-1 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
of small amounts of fish, including mainly lean (non-fatty) fish. However, it seems unlikely that
decades-long intake of small amounts of fish protect, if fish is indeed etiologically protective, via
the very small amounts of omega-3 long-chain polyunsaturated fatty acids so ingested. The
resolution of this issue has important implications for public health and nutritional
recommendations. Thus, further studies -- observational and interventions, particularly clinical
trials -- are needed to resolve whether there is an etiologically significant protection against CHD
or MI afforded by regular ingestion of modest amounts of fish. Similarly, more research is
needed on the relationship of fish intake and health endpoints other than CHD or MI.
Data gaps and research needed on the benefits of fish consumption include:
• More understanding is needed on the benefits of consuming fish and why consuming fish
provides these benefits. For example, is it the n-3 FAs? Selenium or some other mineral?
Substitution for other less healthful foods? Or another mechanism or combination of factors
yet to be determined?
• Numerous epidemiological studies have been conducted which provide some evidence for an
association between consuming fish and reduced risk of coronary heart disease, stroke and
arthritis. More research is needed in this area and on the other possible beneficial effects of
fish consumption. More long-term studies and randomized controlled clinical trials are
needed.
• Studies are needed on groups of people who consume more fish than the national average.
These people are at most risk due to their high consumption, but the existing epidemiology
studies have not included groups with high rates of consumption. Do the potential benefits
increase with increases in consumption, or is there a point at which benefits plateau at some
consumption rate? Are there health detriments to even higher consumption of fish?
• The information used in this report on the change in specific health effects with consuming
fish was limited to studies primarily in adults and for only three health endpoints. Additional
studies on the benefits of fish consumption should be encouraged. Every effort should be
made to ascertain quantitative information on the benefits of fish consumption to pregnant
women, infants and young children, as well as health impacts of larger. consumption rates.
8.3 Chapter 3
Fish is high-quality protein that the public should be encouraged to eat. There are many
nutritional benefits associated with eating fish, regardless of the species type. Unlike red meats,
eggs and dairy products, fish provides very high quality protein anda “heart healthy”
combination of fatty acids. Further, fish (both lean and fatty) is one of the few foods that contain
n-3 (omega-3) fatty acids, a class of fatty acids that are essential for the development of the
nervous system and that may have other beneficial health effects. Calcium, iron, zinc, vitamin
A, niacin, vitamin B6, and vitamin D tend to be low in U.S. diets; fish supplies all of these
vitamins and minerals, in addition to others.
Toxicology Excellence for Risk Assessment 8-2 8/6/99
-------
Cooperative Agreement with U S EPA on Comparative Dietary Risk
Data gaps and research needed on the nutritional aspects of fish consumption include:
• Fish is known to be a good dietary source of selenium, but few reference data are published:
more research into the role of selenium in human health is also needed.
• Nutrient databases contain a wide range of fish species, but samples used to obtain nutrient
values are composites of cooked fish from various unknown locations.
• Nutrient values are generally expressed on the basis of a 100 gram cooked fish portion. This
limits the extent to which comparisons can be made with contaminant data, which are usually
based on raw tissue samples of wild fish gathered from specific geographic areas, and
expressed as concentrations rather than on a weight basis.
• Different methods of preparing and cooking fish will alter some of the organochlorine
contaminant levels. Ideally, the same samples of prepared and cooked fish would be sent for
both contaminant and nutrient analysis, and weighed records of amounts of the fish
consumed would be kept to enable researchers to better assess the physiological risks and
benefits to humans.
• A comparison of the nutritional and contaminant contents of protein sources other than fish
would be ideal, since it would give information on the benefits and risks of other protein
sources. This would allow one to make risk to risk comparisons with fish substitutions. See
Figure 6-3 for a hypothetical discussion of this issue.
8.4 Chapter 4
The risks of consuming fish with chemical contaminants are not completely understood.
However, information for the six chemicals selected for this document was available on EPA’s
IRIS (U.S. EPA, 1999). The majority of this information was of medium confidence, which
means that additional toxicity data may change the resulting risk values somewhat. For most
compounds, this risk information was based on data from laboratory animal studies
(methylmercury and chiorpyrifos were the exceptions). These results must be extrapolated to
humans with considerable uncertainties involved, but the methods used for this extrapolation are
widely accepted as health protective.
This framework requires an understanding of potential health risks at doses above those that are
considered “safe” or at a threshold for toxicity. Traditional cancer risk methods have provided
risk assessors with extrapolation to levels of environmental concern. While these estimates are
uncertain, they are geneally regarded as falling on the side of being health protective. The
method that was chosen to estimate risks above the RID is more recently developed and while is
designed to also be health protective, it has not been widely tested. It has the advantage of ease
of use with existing EPA information from IRIS.
Concordance of effects between laboratory animals and humans is not generally known.
Therefore, the critical effects driving the risk estimates derived from laboratory animal data may
not necessarily be the effects one would see in humans. However, the framework is flexible and
Toxicology Excellence for Risk Assessment 8-3 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
could be used with information on the non-critical effects of these same chemicals to further
refine the overall risk estimates.
Data gaps and research needed on the potential risks from fish consumption include:
• One of EPA’s methods for calculating risk above the RfD was used here (Price et al., 1997).
This method is new and needs further exploration. It has the advantages of being more
generally applicable than categorical regression or benchmark dose, and is less resource
intensive. It can be used directly from the existing data as on EPA’s IRIS. However, it is not
the only approach to the problem of risk above the RID, and as demonstrated, the method
does not work for all chemicals.
• For the framework to be most useful, noncancer risks above the RID must be estimated for
all significant critical effects of chemicals that contaminate fish, in particular, for the
contaminant PCBs. For example, the case study of the Vietnamese immigrant women
consuming Lake Ontario sportfIsh was severely hampered by the inability to estimate the
risks above the RfD for PCBs (Figure 6-24). Some exceedances of the PCB RID were as
much as 40-fold. Other chemicals need similar investigation.
• RfDs are designed to be protective for all adverse effects based on the data for the critical
effect. When doses exceed the RID, as the framework assumes they could, then the critical
effect may begin to manifest itself in the exposed population. The framework uses dose-
response information on the critical effect to predict the increased incidence of the critical
effect. But in addition to the critical effect, other effects may also be seen at higher doses.
Some of these non-critical effects may be more severe than the critical effect (e.g., reduced
body weight versus liver toxicity). At present, EPA has not developed dose-response
relationships for non-critical effects in humans. For the framework to fully characterize
potential risks, and the net health benefit of eating contaminated fish, dose-response
relationships for non-critical effects should also be developed.
8.5 Chapter 5
The benefits of catching and eating fish can go beyond the nutritional value and potential
reduced risk of certain diseases. For some subgroups such as tribes it may be important to
consider the social, religious and cultural importance of fish to that society. Economic impacts
might also be considered for this and other groups. Among isolated and/or lower-income groups,
fish may represent an important economic resource, and a source of needed high-quality protein,
that is not easily replaced. In such communities, advisories designed to limit consumption of
fish may have unforeseen detrimental socio-cultural impacts. These potential consequences need
to be considered when assessing the risk and benefits of fish consumption.
Data gaps and research needed on the cultural aspects of fish consumption-include:
• A scale to measure these impacts and benefits should be developed which can be directly
compared to those used to measure health risk and benefits. Several approaches might be
considered, including normalized scales being developed for use with tribal communities
Toxicology Excellence for Risk Assessment 8-4 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
(e.g., Harper, 1999). The affected individual or group should determine the magnitude of the
modification.
• More quantitative information needs to be amassed on specific consumption behaviors of
population groups for whom fish is important, with the aim of more productively combining
socio-cultural data with biological data in developing risk assessments and consequent risk
management strategies.
• More research is needed on environmental justice with fish consumption, and on the
relationship between fish consumption and group sovereignty, especially in regard to Native
American communities.
8.6 Chapter 6
The current version of the framework represents a significant step forward over the way risks
and benefits of eating fish have been addressed in the past. However, future work should further
explore several important aspects of the framework. A number of conclusions and
recommendations for adjitional work are listed below.
• Incorporate full range of benefits data. The examples of benefits that are presented in the
framework are representative based on the available data. However, they do not incorporate
the entire quantitative benefits data (see Table 2-1). At a minimum, all the data sets
supporting, or contradicting, the existence of a particular health benefit should be further
summarized and discussed, and data should be presented for any endpoint having
quantitative benefits information. A meta-analysis might be considered for each endpoint
supported by more than one data set. This might allow the development of a single dose-
response curve for each health endpoint. Such single dose-response benefit curves would
make the framework easier to use.
• Severity Schemes. For this framework, the severity approach of EPA and ATSDR for
estimating RfDsIRfCs and MRLs (Table 6-2) was used to modify the health risks associated
with chemical exposure. This approach has the advantages of simplicity, familiarity and
consistency with the se of information from EPA’s IRIS, and of ATSDR information found
in its toxicological profiles. One shortcoming of this approach is the implied equal spacing
between levels. There is no scientific or mathematical justification proposed for a FEL being
considered three times as ‘severe” as a less serious LOAEL. Other caveats were discussed in
chapter 6 (see Table 6-3).
In like fashion, a modifier to the magnitude of health benefits accrued from eating fish was
used to roughly compare with the risk of different health endpoints. This modifier of health
benefits (e.g., coronary heart disease avoided) was ranked as none, minimal, moderate or
maximum. This modifier has the advantages of simplicity and consistency with the use of
information for health risks. As for health risks, however, the scheme for health benefits is
being used in a quantitative fashion in the framework, and this results in several
shortcomings which were discussed in Chapter 6 (see Table 6-3).
Toxicology Excellence for Risk Assessment 8-5 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
Other severity schemes should be explored for comparing the health risks and benefits of fish
consumption. The results are likely to be more complex, however. Several of these schemes
will necessitate additional judgment regarding the appropriate severity level of both the
critical effect and benefit. At least one of these schemes (i.e., Ponce et a!., 1998) also
incorporates the concept of duration of the effect or benefit through the use of QALYs.
Every attempt should be made to see if these more complex severity schemes add value when
compared to the simpler one, which was used here.
Explicitly incorporate uncertainty. It is important to recognize that with the exception of
noncancer risks (see Figures 4-1 and 4-2), uncertainty in health benefits and risks is not dealt
with explicitly by the framework in its current version. Moreover, the uncertainty
surrounding the estimates of the different benefits and risks associated with eating fish are
unlikely to be the same. For example, the uncertainty surrounding estimated cancer risks
based on animal toxicity data is likely much greater than the uncertainty surrounding
estimated coronary heart disease benefits based on human data.
An important future refinement of the framework would be explicit consideration and
quantification of uncertainty surrounding estimates of potential health risk and benefit,
because both have the potential to alter the interpretation of the framework and the resulting
FCI.
• Conduct a sensitivity analysis. The current version of the framework uses fixed inputs for
most of the variables that determine potential risk, potential benefits and the FCI. Such fixed
information helped develop the framework and also allowed for exploration of a number of
issues associated with its use. However, many of these fixed parameters can and do vary,
and additional work is needed to investigate how the FCI changes when these parameters are
changed. Such a sensitivity analysis would greatly improve interpretation of the framework
results and perhaps help focus future work on the input variables that have the greatest
potential to affect the FCI.
• Evaluation of additional mixtures of chemicals. The framework and case studies used only a
few chemicals and concentrations to examine the relationship between potential risks and
benefits of eating contaminated fish. While the choice of these chemicals reflected the
frequency of residues and number of fish consumption advisories (Table 4-1), other
chemicals are also found in fish. While the analysis of a limited number of chemicals is
useful for the development of the framework and its application, the choice of concentrations
could perhaps better reflect those typically observed in waters of the U.S (the example
concentrations presented here were much higher than average). Based upon comments from
the Advisory Committee, methylmercury, PCBs and dioxin are the chemicals for which
advisories are most commonly needed and typical high concentrations might vary between
0.2 and 1 mg/kg for methylmercury and PCBs, and be around 1 ng/kg for dioxin toxic
equivalents.
• Risk curves for non-sensitive groups. For health risks, specific risk curves for non-sensitive
members of the population could also be developed. This would avoid matching the health
risk for the sensitive individual with the health benefit to the average individual. For
Toxicology Excellence for Risk Assessment 8-6 8/6/99
-------
Cooperative Agreement with U.S. EPA on Comparative Dietary Risk
example, with methylmercury the risk curve is based on risk to the infant and fetus, whereas
the benefit curve was for the adult. Use of an adult risk curve would have changed the
conclusions of the Florida Everglades case study.
8.7 Chapter 7
A strong communication program is needed to best implement use of the framework and
approach outlined in this document. This approach can provide individuals and groups
(communities, states or tribes) the ability to describe and analyze tradeoffs between benefits and
risks. Ultimately, however, no approach will be successful if it cannot be understood and applied
by the audiences for which it is intended. Therefore research is needed with at-risk populations
(e.g., tribes with potentially heavy fish consumption, women of childbearing age, fish-eating
families with children), to identify their information needs. This is an iterative development of
communication approaches and content, with communicators and target audiences working in
partnership.
A key to the approach proposed in this document is research-based evaluation. Since no
approach will be successful if it cannot be understood and applied by the audiences for which it
is intended, both formative and summative evaluation research efforts are needed.
• Formative evaluation research would include working with the target audiences to identify
their information needs. Ideally, formative evaluation begins with in-depth, qualitative
analysis of information needs and the range of potential responses to and concerns about
various types of information. Focus groups and other interactive forums often provide the
best mechanism for this stage of research. Formative evaluation continues with iterative
development of communication approaches and content, with communicators and target
audiences working in partnership.
• Summative evaluation, an empirical assessment of the impact of the communication process,
is a critical research need to assess the efficacy of the FCI approach. Summative evaluation
is often hypothesis-based. For example, possible hypotheses related to use of the FCI
include:
H 1 : Availability of health benefit/risk comparison information via the FCI will be related
to increased confidence of fish consumers that they are making an informed
decision about fish consumption:
H 2 : Increased information provided to fish consumers through the FCI will lead to
improved compliance with health advisory recommendations.
Summative evaluation assesses the extent to which program objectives were achieved. Thus,
achievement of the objectives of health advisory programs using the FCI should be evaluated
systematically, both before implementing FCI, and after. Collecting baseline data is critical to
evaluating the impact of new risk management and communication programs.
Toxicology Excellence for Risk Assessment 8-7 8/6/99
-------
Cooperative Agreement with U S. EPA on Comparative Dietary Risk
8.8 Final Comment
The need to consider the health and nutritional benefits of fish consumption has long been
recognized when crafting public policy regarding people eating fish contaminated with low
levels of chemicals. Due in part to this well-defmed need, the purpose of this research was to
develop an understanding and framework by which to evaluate the comparative risks posed by
dietary changes as a result of fish consumption advisories. We have been partially successful in
this endeavor. This research should lead to a better understanding of the health benefits and
health risks of fish consumption, although further work is needed before the framework that we
suggest can be generally useful. We anticipate that public health officials and consumers will
use this increased understanding to evaluate a broader range of dietary information before
making decisions about this important resource.
8.9 References
Harper, B.L. 1999. Personal communication with Michael Dourson, TE: ‘A. June.
Ponce, R.A., S.M. Bartell, D. LaFlamme, eta]. 1998. Quantitative analysis of risks and benefits
for public health decisions applied to fish consumption. The Toxicologist. 42(1-S): 45.
Price, P., R. Keenan, J. Swartout, et a!. 1997. An approach for modeling ioncancer dose
responses with an emphasis on uncertainty. Risk Anal. 17(4): 427-437.
U.S. EPA. 1999. Integrated Risk Information System (IRIS). National Center for
Environmental Assessment. Online: http:/Iwww.epa.gov/irisl
Toxicology Excellence for Risk Assessment 8-8 8/6/99
------- |