United States Science Advisory EPA-SAB-EC-98-001
Environmental Board
Protection Agency Washington DC October 1997
&EPA AN SAB REPORT: REVIEW
OF THE EPA DRAFT
MERCURY STUDY
REPORT TO CONGRESS
REVIEW OF THE ERA'S MERCURY
REPORT TO CONGRESS BY THE
MERCURY REVIEW SUBCOM-
MITTEE
-------
October 10, 1997
EPA-SAB-EC-98-001
Honorable Carol M. Browner
Administrator
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, DC 20460
Subject: Science Advisory Board's review of the EPA Draft Mercury Study
Report to Congress
Dear Ms. Browner:
The 1990 Clean Air Act Amendments directed the EPA to perform a study on the
impacts of mercury as an air pollutant and to provide a report to the Congress. The Act
required EPA to address several specific topics, including mercury emissions from
electric utility steam generating units, municipal waste combustion units, and "area"
sources; the rate and mass of these emissions; the associated health and environmen-
tal effects; technologies for controlling these emissions; and the costs of such control
technologies.
In response, the EPA developed a seven volume draft report (EPA-452/R-96-
001 a-h, June, 1996) which was stated by the Agency to provide data on types, sources,
and trends in mercury emissions; evaluated the atmospheric transport of mercury;
assessed the impacts of mercury emissions on organisms/ecosystems close to the
emitting source; identified major exposure pathways to humans and non-human biota;
identified mercury exposure levels likely to produce adverse effects in humans, and the
nature of those effects; evaluated mercury exposure effects for ecosystems and non-
human organisms; identified populations especially at risk from mercury exposure due
to special sensitivity or high exposure; and made estimates of the effectiveness of
control technologies and their costs.
After completing the draft report, the Agency stopped short of issuing a formal
final mercury report to Congress because of a growing consensus that such an analysis
should wait for a full assessment of several relevant studies now underway. These
-------
studies address the effects of methylmercury on neurological development of children
in fish-consuming populations in the Seychelles and Faeroe Islands, and were ex-
pected to be completed and published in the next year or two. Their results should be
considered before making a new assessment of health risks. In terms of currently
available data, however, the Agency decided to proceed with finalizing the existing
report by having the Science Advisory Board (SAB) review it (as is customary with
major scientific documents) before transmitting it to the Congress. The Agency's goal
was "... to receive SAB's view of the overall report as well as to seek SAB assistance in
prioritizing research needs and identifying any important weaknesses in the current
evaluation - other than those which are addressed by the on-going studies - with an
eye on improving the assessment" (memorandum from Dr. Robert Huggett, then
Assistant Administrator for Research and Development (ORD) and Ms. Mary Nichols,
then Assistant Administrator for the Office of Air and Radiation (OAR), June 23, 1996).
Following receipt of the above referenced memorandum from the ORD and the
OAR, the SAB created a special Subcommittee (the Mercury Review Subcommittee
(MRS)) of its Executive Committee. Composed of 34 scientists from the United States
and Canada, and reflecting a wide range of disciplines and expertise, the MRS
convened in public meeting on February 13/14, 1997 in Washington, DC. The Sub-
committee was organized into three Workgroups (Sources; Exposure, Doses, and Body
Burdens; and Human Health Effects), and addressed a Charge of 46 enumerated
questions developed via discussions between EPA, the Food and Drug Administration,
and SAB staff. This report, developed via mail in the several months following the
public meeting, reflects the discussions at that meeting.
The following discussion presents the Subcommittee's conclusions regarding the
major topic areas of the Charge; because of the unusual length and complexity of the
Charge (which is provided in its entirety in Enclosure A), this letter will not attempt to
address every specific enumerated issue.
First, and perhaps most importantly, the Subcommittee believes that the major
findings of the draft report are well supported by the scientific evidence, and that the
Agency has done a very creditable job in amassing, analyzing, and drawing conclu-
sions from a truly vast amount of data. Naturally, as with any such project of this scope
and depth, there are areas where improvement in the use of available scientific
information is possible, as well as in the organization and presentation of information.
Detailed suggestions for such improvements are noted in our report, and are noted
below.
Addressing specific findings, the MRS concluded that:
a) The majority of the human population is not experiencing methylmercury
exposures that are of concern from the standpoint of human health. It is
-------
not now possible to establish a quantitative relationship between sources
and actual exposures in humans or wildlife. The draft EPA report is seen
by the Subcommittee as a document which, in general, reflects the current
state of the art with regard to human health and atmospheric mercury
transport issues. However, the Subcommittee feels that the report does
not adequately present or model Hg fate and transport in ecosystems, Hg
bioaccumulation or wildlife exposure. It is also noted that the high end of
the distribution of methylmercury exposures is very uncertain with respect
to exposures, total number of people (and percent of the population) who
may be experiencing exposures high enough to cause adverse health
effects, and the actual sub-groups who are highly exposed. Conse-
quently, total population risk is not, and cannot be, fully characterized at
this time.
b) In general, from the standpoint of looking at human health effects and the
uncertainties in general, the draft report is a very good document and an
important step forward in terms of bringing the relevant information
together into one place for the first time. The current Reference Dose
(RfD), based on the Iraqi and New Zealand data, should be retained at
least until the on-going Faeroe and Seychelles Islands studies have
progressed much further and been subjected to the same scrutiny as has
the Iraqi data.
c) The Subcommittee identified some problems vis-a-vis human health
issues as discussed in the draft document - a lack of recognition and
emphasis on consistency of the data across multiple studies, and, most
seriously, a failure to link animal studies to the human risk assessment
process. Major problems also exist in interpreting the biology of the
outcome in children versus adults; interpreting potential susceptibility
differences between populations; and identifying factors which may
modify response, such as diet. Note however, that the Subcommittee
believes that there is sufficient data to conclude that the developing
organism is vulnerable during the entire period of development and that in
utero as well as early postnatal exposure to methylmercury is of concern.
d) With respect to modeling the linkage between emissions from anthropo-
genic sources and human exposures, the Subcommittee agrees that it is
plausible that current anthropogenic emissions are contributing to human
exposures. However, the relative contributions of current anthropogenic
emission sources, global backgrounds, recycling of old emissions and
natural background to human exposures of methylmercury are highly
uncertain and difficult to determine. Furthermore, the time constants for
intermedia transport are not known. Since mercury does not degrade in
the global environment, current anthropogenic emissions will add to the
-------
fraction of global mercury that is actively circulating in the biosphere,
surface water, and soils. In addition, emissions from both the U.S. and
from other countries can impact the U.S. environment and exposures of
humans and wildlife. The Subcommittee also notes, however, that this
section of Volume III of EPA's draft report is not clearly written and
confuses environmental transport and fate modeling with exposure
analysis.
e) The Subcommittee concurs with the conclusion that fish are the major
source of methylmercury exposures for the human population. The
exposure analysis based on fish consumption must be regarded as a
"snapshot" in time, however. In addition, there are two potential biases
affecting the exposure analysis based on fish consumption. First, the
consumption data from the Department of Agriculture's Continuing Sur-
veys of Individual Food Consumption (CSFII) should be checked against
production data. These will not agree exactly but should be reasonably
consistent. Second, the average concentrations of methylmercury in fish
used for the exposure assessment were provided by the National Marine
Fisheries Service. The Subcommittee recommends that EPA determine
how measurements of methylmercury in fish which were below the analyti-
cal limits of detection (BDL) were statistically treated in arriving at the
mean concentrations.
f) The Subcommittee strongly recommends that EPA edit and shorten
Volume III on Exposure to eliminate redundancies and clarify the logical
sequence of the exposure analysis.
g) The Subcommittee notes that in the process of assessing control costs,
the draft mercury report devotes most of the cost analysis to end-of-pipe
controls and gives little attention to other types of controls. The Subcom-
mittee recommends that the cost analysis also give consideration to other
approaches for controlling mercury emissions that might prove to be more
flexible and more cost-effective.
h) The Subcommittee generally agreed that the mercury wildlife criterion (as
currently presented) is overly conservative and is lower than appears
necessary to protect wildlife species. The criterion was derived from a
bioaccumulation factor (BAF) based on total Hg rather than methyl-
mercury (MeHg), which is the mercury species that bioaccumulates. As a
result, the magnitude of the error in the BAF is associated with, and
reflects, the wide variability in MeHg concentrations among ecosystems.
The Subcommittee suggests that EPA, using the best techniques now
available to recalculate BAF and wildlife risk criteria based on dissolved
-------
MeHg in water, and that BAF calculations be specific to ecosystem or
water body type.
The problems noted above notwithstanding, the Subcommittee believes
that piscivorous wildlife are at risk from elevated mercury exposures and
exhibit toxic effects in the areas of concern identified by the U.S. EPA. In
addition, there is good evidence showing effects of mercury on wildlife; a
summary of the population that appear to be affected; and a statement
that wildlife exposure can be a harbinger of hard-to-detect human effects
should be added to the final report
I) The Subcommittee finds that Volume V, in total, is not based on the best
available and sound science. To a large degree this chapter follows EPA
guidelines for ecological risk assessment. However, it deviated in terms
of quantifying the risk. This is a critical component of risk assessment.
j) The final document should emphasize the fact that there are significant
information gaps in the understanding of the biogeochemistry of mercury
species and that the absence of this information limits the reliability with
which any mercury control program can be evaluated.
k) The modeling of atmospheric mercury transport and deposition is largely
sound, but the modeling of the post-deposition fate of mercury in ecosys-
tems is oversimplified, neglects available information on speciation, and
does not reflect recent advances in the science. This deficiency is one of
the most serious problems in the draft report. This modeling should be
revised, with the goal of having the report reflect current scientific under-
standing. Also, the modeling, as used, contributed to sizeable errors in
the estimated wildlife criterion. It is important to bring out the importance
of MeHg production in ecosystems in the report. Because MeHg is the
species that bioaccumulates in human and wildlife food supplies, under-
standing the methylation process is critical to modeling Hg fate and
exposure. Mercury methylation and the variability in MeHg among
ecosystems were neglected in the report, because of the perceived
difficulties in modeling methylation. The report assumes that MeHg
constitutes the same fraction of total Hg in sediments, soils and waters
across ecosystems. This is not the case. Rather, the variability in MeHg
production and bioaccumulation among ecosystem types is many orders
of magnitude, and may be as large as the influence of Hg contamination.
Without consideration of this variability, the report's fate and transport
models, and the resultant exposure models and wildlife criteria models
cannot predict Wildlife Criteria with less than two to three orders of
magnitude associated error.
-------
The final EPA report should detail the state of our understanding of Hg
methylation process within Volume III; address MeHg production along
with fate and transport models in Volume III (which models should be
specific for a suite of ecosystem types); and the resultant information be
used to model and discuss wildlife exposure in Volume V. Because the
models for estimating the distributions of human exposure were based on
measured values for Hg in fish and other food, rather than modeled
values, these suggested changes will not affect the human exposure
analysis in the report.
We appreciate the opportunity to review this document, and look forward
to your response to the issues we have raised.
/signed/
Dr. Genevieve Matanoski, Chair
Science Advisory Board, and
Co-chair, Mercury Review Subcommittee
/signed/
Dr. Joan Daisey, Co-chair
Mercury Review Subcommittee
ENCLOSURE
-------
ENCLOSURE A
DETAILED CHARGE FOR THE MERCURY REVIEW SUBCOMMITTEE
GENERAL
1. The EPA has decided to defer evaluation of the dose-response relationship between
methylmercury and fetal and developmental effects until additional scientific data have
been evaluated. These issues will be addressed in Phase Two of the SAB review.
Putting these matters aside, are the overall assessment and conclusions based
on sound and appropriate evaluation of the relevant science?
SOURCES
2. In order to approximate future mercury emissions, the sources are simplified to
model plants placed to approximate current U.S. emission patterns. Is this approach
consistent with the best available scientific practice? Are the implications of this
simplifying assumption adequately presented?
3. US EPA limited the scope of its assessment to anthropogenic sources of mercury
emissions. Are these reasonable and defensible assumptions? The emissions
inventory estimates these emissions in the United States. There is a short discussion
of natural emissions and re-emitted emissions as well. The discussion of these
emissions was based on language suggested by the scientific peer review panel. The
modeling of local point sources did not include natural emissions, while the long-range
transport modeling did incorporate a background level of mercury (to account for
natural and re-emitted Hg) which has been measured over the open ocean. Are the
reasons for this approach adequately stated? Are the uncertainties and implica-
tions of this approach consistently and appropriately described?
4. For the atmospheric fate and transport modeling conducted using both the Regional
Lagrangian Model of Air Pollution (RELMAP) and the Complex Terrain and Deposition
Air Dispersion Model (COMDEP), best estimates from the peer-reviewed literature of
the species of mercury emitted from each source were utilized. Were these speci-
ation profiles appropriate?
5. Can the Committee comment on the evidence concerning the role of human
activity, and if possible, the role of U.S. sources to methylmercury in ocean
seafood?
-------
ENVIRONMENTAL FATE/TRANSPORT
6. US EPA relied on modeling to describe mercury fate and transport. In the report, the
inventory served as the building block and was linked to the models (which were
themselves linked). Both local and regional impacts were assessed together (additive)
and separately for both humans and wildlife. Model plants were used as surrogates for
actual facilities as well as hypothetical locations and populations. Would a case study
approach using actual facilities be more appropriate? Are the rationales for the
selected approach appropriately presented? Please comment on the usefulness
of measured data for estimating mercury exposures.
7. In its assessment, US EPA assumed that available data on the deposition of nitric
acid could be used to approximate the likely wet deposition of divalent mercury. Is the
rationale for this assumption adequate (see Volume 3, Section Chapter 6, Section
6.1.3, and Volume 3, Appendix D, Section D.1.3.2)? Classical Gaussian and
Lagrangian puff models were adapted to reflect mercury chemistry. Were these
modifications appropriate?
8. Further, the assessment neglects the potential conversion of elemental mercury to
divalent forms in cloud droplets in both the local scale and the regional analysis. The
Agency has assumed that any such conversions would happen on such a small scale
that they could have only a very nominal impact on the ultimate exposure characteriza-
tion. Is this rationale fully explained and adequate? Are the uncertainties of this
assessment explained?
9. The Agency recognizes that mercury fate and transport is an immature field;
capabilities to describe phenomena such as the temporal distribution of mercury
concentrations in soil continue to evolve. In its assessment, US EPA attempted to
modify existing approaches (the 1990 Methodology of Assessing Health Risks associ-
ated with Indirect Exposure to Combustor Emissions and the 1993 Addendum) to reflect
the more recent science. Were the assumptions regarding equilibrium among
mercury species present in soil appropriate?
10. Was the selection of the model input parameters (such as the soil-to-water
partition coefficients) appropriately justified and the uncertainties explained? Is
the estimation of the watershed and water body fluxes adequate and appropri-
ate? Should fluxes in large lakes, rivers and other large water bodies also have
been estimated? Is the rationale for not doing so adequate? Does the Agency
appropriately justify and explain the uncertainty surrounding assumptions about
plant uptake of mercury directly from the atmosphere? Is the likely contribution
of mercury from plants to soil contributions sufficient that such sources should
be added to future modeling efforts?
A-2
-------
11. A number of qualitative conclusions are drawn from the results of the atmospheric
modeling; given the uncertainty in the modeling and in the state of the science,
evaluate the scientific quality of these conclusions. Were the uncertainties ade-
quately described?
12. A key variable in estimating ultimate potential human exposure is the
bioaccumulation factor (BAF) for methylmercury through the aquatic chain. Was the
BAF developed by the Agency appropriate and its utilization in the models
scientifically sound? Was the uncertainty analysis adequate?
13. Were the models and modeling conducted for the terrestrial food chain
adequate and appropriate?
EXPOSURE
14. The exposure assessment is contained in Volume III. Please comment on the
scientific basis of this assessment, alternative approaches, and research
priorities. In particular, please advise the Agency on how Appendix H should be
used, improved, or expanded.
15. Because of uncertainties inherent in the emissions inventory and the model, the
exposure assessment was characterized as being a "qualitative assessment based on
quantitative modeling." Is this a reasonable characterization? Were the
uncertainties of and conclusions drawn from linking the models together
appropriately described?
16. US EPA has estimated methylmercury and fish consumption using both cross-
sectional and longitudinal data for the general US population. In these studies, EPA
consistently finds that a portion (albeit a small one) of the population consumes very
large quantities of fish. It is these individuals who face the largest risks from
methylmercury exposure and who pose the greatest concern to the Agency. Has EPA
adequately assessed the number and types of individuals who consume fish in
these relatively large quantities? The group at issue includes 5% of the fish-
consuming population. Are the uncertainties associated with characterization of
such small portions at the extreme end of the population distribution
appropriately considered and presented? Could the estimate of 5% of the fish-
consuming population be a methodologic artifact?
17. US EPA has relied on three nationally-based sources of data for mercury
concentrations in marine and fresh-water fish and shellfish. Some of these data were
obtained within the past three years, whereas other data were obtained as long ago as
20 years.
A-3
-------
Is the use of these older data appropriate? Does the document appropriately note
the limitations of this methodology?
18. The USDA has published data indicating that fish and shellfish consumption have
risen approximately 25% since the early 1970s . These consumption data were
employed without adjustment in this report to estimate current fish consumption
(Volume III, Appendix H; and Volume VI, Section 4.2.3.1). Should these data be
adjusted to reflect the overall trends in fish consumption? If so, how would the
estimates of very high consumption (above the 90th percentile) be adjusted?
19. In Volume III, Appendix H, calculations of methylmercury intake from crab were
grouped without specific consideration of the particular species. These data are then
linked to dietary survey data which typically do not list the individual species of many
fish and shellfish consumed. How might this approach bias the final assessment?
Are the limitations of this approach adequately characterized in the report? Are
these limitations going to materially alter the assessment of health risk
associated with methylmercury exposures?
20. In early versions of the draft report (and in Volume III, Appendix H), US EPA
characterized total risk by considering methylmercury exposure from marine fish and
shellfish together with consumption of methylmercury from fresh-water or estuarine fish
and shellfish. It has been recommended that the report focus exclusively on fish from
inland sources. Does the current version adequately characterize total risk? Are
the uncertainties resulting from this methodology appropriately presented?
21. Some data exist on hair mercury concentrations in U.S. residents. Is this data
base adequate to predict the distribution of hair mercury in the general U.S.
population?
22. The draft report identifies hair mercury concentration as the most appropriate
available index of methylmercury exposure. Is this assumption consistent with the
available data? Are the exposure estimate approaches used relevant and
appropriate? Are the predicted exposure ranges consistent with other published
exposure analyses? Are the uncertainties of this assumption appropriately
presented?
23. Although the modeling exposure assessment focused on anthropogenic emissions,
the fish consumption analysis considered measured mercury concentrations in fish
tissue regardless of the mercury's origin. Thus, there is considerable difficulty in
assessing or describing how much of the mercury in fish is attributable to current
anthropogenic emissions. Is the approach taken by the Agency in this assessment
appropriate given the available data? What is the advice of the SAB regarding
A-4
-------
the differentiation between current emissions and their impacts relative to the
"body burden" approach of the fish consumption analysis?
DOSES/BODY BURDENS
24. In its assessment, U.S. EPA assumes that the biokinetics of methylmercury from
contaminated grain approximate those resulting from methylmercury from fish. Is this
assumption consistent with the available science?
25. Are the uncertainties in the fish and grain consumption analyses adequately
and consistently presented in the draft report?
26. Similarly, US EPA assumes the biokinetic parameters for children are identical to
those of adults. Does such an assumption introduce bias into the assessment?
Are the uncertainties adequately and consistently presented in the draft report?
Is the conclusion that children's lower body weight results in higher exposure
than to adults? Could this be an artifact of exposure modeling using 3-day
consumption survey data?
27. What methodologies other than hair concentrations could be used to estimate
body burdens of mercury?
HEALTH ENDPOINTS and SUSCEPTIBLE SUBPOPULATIONS
28. Volume 4, Section 6.3, and Volume 6, Chapter 2, Sections 2.1 and 2.2.21 present
US EPA's interpretation of the data demonstrating increasing severity and frequency of
nervous system effects (particularly impairment in visual-motor integration) among
pediatric subjects with increasing levels of maternal hair mercury. These sections
present US EPA's interpretation of the data demonstrating increasing severity and
frequency of nervous system effects (particularly impairment in visual-motor
integration) among pediatric subjects with increasing levels of maternal hair mercury.
Recognizing that additional data may shortly be available, is EPA's assessment
methodology based on the best scientific practice? Are the uncertainties in the
characterization of potential effects accurately described in the draft text?
29. Data from experimental animals (including primates with long-term exposures to
methylmercury) show methylmercury-induced nervous system damage, particularly on
the visual system, although the animals appear clinically normal. The traditional RfD
methodology neglects such impairment. Are these data important endpoints? Are
they appropriately characterized in the draft report? How could such data be
better evaluated by the Agency?
A-5
-------
30. The available data present information on fish consumption drawn over relatively
short time periods (e.g., days) and are used to extrapolate consumption patterns over
longer periods (e.g.. a month). Although the exact developmental window affected by
methylmercury in humans is not precisely defined, it is thought to be less than three
months. Is such extrapolation appropriate in this case? Does the document
appropriately present the limitations of this methodology?
31. Volume 4, Section 6.3, and Volume 6, Chapter 2, Sections 2.1 and 2.2.21 present
US EPA's interpretation of the data demonstrating increasing severity and frequency of
nervous system effects (particularly impairment in visual-motor integration) among
pediatric subjects with increasing levels of maternal hair mercury. What biases are
introduced into the assessment by focusing on subtle endpoints of
neurobehavioral function in contrast with traditional metrics of child
development? Should these indicators be dismissed if traditional metrics (such
as age of first walking) are within normal ranges? Do the available data indicate
whether children are more sensitive than adults to the effects of methylmercury?
32. Are the uncertainties in the characterization of potential effects accurately
described in the draft text?
33. Traditional methods for estimating potential human health risks from environmental
hazards do not distinguish risks among subpopulations (e.g. racial or ethnic) groups. If
methylmercury exposures are comparable and some groups, but not all, show
impairment in traditional and/or specialized neurophysiological/neurobehavioral
tests, how should US EPA reflect these differences in its analyses? Do factors
such as nutritional status, life-styles (e.g., substance abuse), or economic status
play a role is mediating these differences?
34. Similarly, many assessments of methylmercury risk clump risk to fetuses and
children together with risk to adults. Is this approach scientifically valid for
methylmercury? Methods for estimating potential human health risks from
environmental hazards do not distinguish risks among subpopulations (e.g. racial or
ethnic) groups. If methylmercury exposures are comparable and some groups, but
not all, show impairment in traditional and/or specialized
neurophysiological/neurobehavioral tests, how should US EPA reflect these
differences in its analyses?
35a. What wildlife effects (based on what metric) can be interpreted as harbingers
of likely human health effects?
35b. Could the Committee provide any short-term advice on human health issues
not addressed in the coming epidemiologic studies such as toxicokinetics?
A-6
-------
ISSUES ON WILDLIFE ASSESSMENT
35c. Chapter VI presents the Agency's current mercury risk characterization. Please
advise the Agency on the appropriateness and environmental significance of the
characterization of wildlife effects.
36. Volume V describes fairly significant wildlife effects which are attributed to elevated
mercury concentrations in some ecosystems. Is this evaluation based on the best
available and sound science, and are they consistent with EPA eco risk
guidelines? How could this evaluation be improved?
37. This report makes the inference that mercury emissions are related to reproductive
effects and neurobehavioral changes in fish consuming birds and mammals. These
effects have been documented in the Great Lakes and in the Southeastern United
States. EPA's models predict that regional hot spots (relatively high concentrations of
methylmercury) would occur in these same areas. Is the analysis of the evidence
linking emissions and effects scientifically sound?
38. Have any "hot spots" not shown evidence of wildlife effects? Do these
predicted "hot spots" correlate with methylmercury levels in fishery products?
39. Which mercury-related effects in wildlife should be identified as warning
signs that analogous effects may occur in humans? Could "wildlife
epidemiology" be used as a surrogate for controlled lab studies?
RESEARCH NEEDS
40. Based on the research needs identified in the draft report, areas identified
through discussion of the report, or on other information, has EPA identified the
highest priority areas for research? If not, could the Committee suggest what
areas need to be addressed?
QUESTIONS RELATED TO SOCIAL COST
41. Chapter VII previously contained a section on the "Social Costs" of mercury
contamination (this section is attached to comments by the Council of Economic
Advisors in the supplementary information). This section described the value of the
fisheries in the U.S. as well as other values such as maintaining a healthy ecosystem.
The intent was to balance the discussion of mercury control costs. Some reviewers
objected to the inclusion of the Social Costs section partly because the impact of
anthropogenic sources could not be directly and quantitatively related to these impacts
(e.g., the declining Florida panther population). The section was consequently deleted.
A-7
-------
Is such a discussion appropriate for this study? How would the SAB advise the
EPA to describe benefits of mercury reductions if a) the impacts of such
reductions are not directly quantifiable, and b) the monetary value of such
benefits are not easily quantified?
A-8
-------
Distribution List
Administrator
Deputy Administrator
Assistant Administrators
Deputy Assistant Administrator for Pesticides and Toxic Substances
Deputy Assistant Administrator for Research and Development
Deputy Assistant Administrator for Water
EPA Regional Administrators
EPA Laboratory Directors
EPA Headquarters Library
EPA Regional Libraries
EPA Laboratory Libraries
Staff Director, Scientific Advisory Panel
Library of Congress
National Technical Information Service
-------
NOTICE
This report has been written as a part of the activities of the Science Advisory
Board, a public advisory group providing extramural scientific information and advice to
the Administrator and other officials of the Environmental Protection Agency. The
Board is structured to provide balanced, expert assessment of scientific matters
relating to problems facing the Agency. This report has not been reviewed for approval
by the Agency and, therefore, the contents of this report do not necessarily represent
the views and policies of the Environmental Protection Agency, nor of other agencies in
the Executive Branch of the Federal government, nor does mention of trade names or
commercial products constitute a recommendation for use.
-------
ABSTRACT
In response to the 1990 Clean Air Act Amendment's directive, the EPA
developed a draft report on mercury, and asked the Science Advisory Board to review
it. The Mercury Review Subcommittee convened on February 13/14, 1997 in
Washington, DC.
The Subcommittee believes that the major findings of the draft report are well
supported by the scientific evidence. There are areas where improvement in the use of
available scientific information is possible. Detailed suggestions for such
improvements are noted below:
a) The majority of the human population is not experiencing methylmercury
exposures that are of concern from the standpoint of human health. The
current Reference Dose, based on the Iraqi and New Zealand data,
should be retained at least until the on-going Faeroe and Seychelles
Islands studies have progressed much further and been subjected to the
same scrutiny as has the Iraqi data.
b) The Subcommittee identified some problems vis-a-vis human health
issues - a lack of recognition and emphasis on consistency of the animal
data across multiple studies.
c) It is plausible that current anthropogenic emissions are contributing to
human exposures, and that fish are the major source of methylmercury
exposures for the human population.
d) The Subcommittee recommends that the cost analysis also give
consideration to other approaches for controlling mercury emissions that
might prove to be more flexible and more cost-effective.
e) The mercury wildlife criterion is overly conservative and is lower than
appears necessary to protect wildlife species. However, piscivorous
wildlife are at risk from elevated mercury exposures.
f) Volume V, in total, is not based on the best available and sound science.
g) The final document should emphasize the fact that there are significant
information gaps in the understanding of the biogeochemistry of mercury
species.
h) The modeling of atmospheric mercury transport and deposition is largely
sound, but the modeling of the post-deposition fate of mercury in
ecosystems does not reflect recent advances in the science.
KEYWORDS: mercury; methylmercury; RfD; piscivorous wildlife; fish; seafood; mercury
biogeochemistry.
-------
U.S. Environmental Protection Agency
Science Advisory Board
Mercury Review Subcommittee
February 26-27, 1997
Co-Chairs
Dr. Genevieve M. Matanoski, Professor of Epidemiology, School of Hygiene and Public
Health, The John Hopkins University, Baltimore, MD
Dr. Joan Daisey, Program Head, Indoor Environment Program, Lawrence Berkeley
National Laboratory, Berkeley, CA
Members and Consultants
Dr. William J. Adams, Kennecott Utah Copper Corp., Magna, UT
Dr. David Bellinger, Neuroepidemiology Unit, Children's Hospital, Boston, MA
Dr. Nicolas Bloom, Chief Science Officer, Frontier Geosciences Inc., Seattle, WA
Dr. Dallas Burtraw, Resources for the Future, Washington, DC
Dr. Thomas Burbacher, Department of Environmental Health, School of Public Health
and Community Medicine, University of Washington, Seattle, WA
Dr. John A. Dellinger, Department of Preventive Medicine, Medical College of
Wisconsin, Milwaukee, Wl1
Dr. Kim N. Dietrich, University of Cincinnati, College of Medicine, Department of
Environmental Health, Division of Biostatistics and Epidemiology, Cincinnati, OH
Dr. Lawrence J. Fischer, Institute for Environmental Toxicology, Michigan State
University, East Lansing, Ml
Dr. Bruce A. Fowler, University of Maryland at Baltimore, Program in Toxicology,
Baltimore, MD
Now Director, Illinois Poison Control Center, Chicago, IL
-------
Dr. Steven G. Gilbert, Biosupport, Inc., Redmond, WA2
Dr. Cynthia C. Gilmour, The Academy of Natural Sciences, Estuarine Research Center,
St. Leonard, MD
Dr. Gary Heinz, Patuxent Wildlife Research Center, Laurel, MD
Dr. Robert E. Hueter, Center for Shark Research, Mote Marine Laboratory, Sarasota,
FL
Dr. Harold E. B. Humphrey, Michigan Department of Community Health, Division of
Health Risk Assessment, Lansing, Ml
Dr. James P. Hurley, Wisconsin Department of Natural Resources and University of
Wisconsin, Water Chemistry Program, Madison, Wl.
Dr. Joseph L. Jacobson, Department of Psychology, Wayne State University, Detroit,
Ml
Dr. Ronald J. Kendall, The Institute of Wildlife &Environmental Toxicology, Clemson
University, Pendleton, SC
Dr. Lynda P. Knobeloch, Wisconsin Division of Health, Bureau of Public Health,
Madison, Wl
Dr. Steven E. Lindberg, Environmental Sciences Division, Oak Ridge National
Laboratory, Oak Ridge, TN
Dr. Thomas McKone, School of Public Health, University of California, Berkeley, CA
Dr. Michael W. Meyer, Wisconsin Department of Natural Resources, Rhinelander, Wl
Dr. Maria Morandi, University of Texas Health Science Center at Houston, School of
Public Health, Houston, TX
Dr. Christopher Newland, Department of Psychology, Auburn University, Auburn, AL
Dr. Jerome 0. Nriagu, The University of Michigan, School of Public Health, Department
of Environmental and Industrial Health, Ann Arbor, Ml
Dr. W. Steven Otwell, Professor, Seafood Technology, Food Science and Human
Now at the Institute of Neurotoxicology and Neurological Disorders, Redmond WA
IV
-------
Nutrition, Aquatic Food Products Program, University of Florida, Gainesville, FL
Dr. Deborah C. Rice, Toxicology Research Center, Ottawa, Ontario, Canada
Dr. Ellen K. Silbergeld, University of Maryland, Dept. Of Epidemiology, Baltimore, MD
Mr. Howard A. Simonin, New York State Department of Environmental Conservation,
Rome, NY
Dr. Ann Spacie, Purdue University, Lafayette, IN
Mr. David G. Strimaitis, Earth Tech, Concord, MA
Dr. Valerie Thomas, Center for Energy and Environmental Studies, Princeton University
Princeton, NJ
Designated Federal Officials
Samuel R. Rondberg, U.S. Environmental Protection Agency, Science Advisory Board,
401 M Street, SW, Washington, DC 20460
A. Robert Flaak, U.S. Environmental Protection Agency, Science Advisory Board,
401 M Street, SW, Washington, DC 20460
Dr. Donald G. Barnes, U.S. Environmental Protection Agency, Science Advisory Board,
401 M Street, SW, Washington, DC 20460
Staff Secretary
Mary L. Winston, U.S. Environmental Protection Agency, Science Advisory Board,
401 M Street, SW, Washington, DC 20460
-------
TABLE OF CONTENTS
1. EXECUTIVE SUMMARY 1
2. BACKGROUND AND CHARGE 6
2.1 Background 6
2.2 Detailed Charge 7
3. SOURCES 15
3.1 Source Simplification in Modeling Mercury Emission Patterns 15
3.2 Modeling Sources of Mercury Emissions 15
3.3 Speciation Profiles in Atmospheric Fate and Transport Modeling 16
3.4 Anthropogenic Sources and Methylmercury in Ocean Seafood 17
3.5 Modeling Mercury Fate and Transport 17
3.6 Nitric Acid as a Surrogate for Wet Deposition of Divalent Mercury 19
3.7 The Impact Of the Conversion of Elemental Mercury to Divalent Forms on
Exposure Characterization 23
3.8 Assumptions Regarding Equilibrium among Mercury Species in Soil and
Sediments 24
3.9 Biogeochemical Issues in Modeling 26
3.10 Uncertainties in Atmospheric Modeling 29
3.11 Estimating the Bioaccumulation Factor (BAF) for the Aquatic Chain ... 30
3.12 Modeling the Terrestrial Food Chain 31
3.13 Characterization of Wildlife Effects 32
3.14 Evaluation of Wildlife Effects 33
3.15 Linking Emissions and Wildlife Effects 37
3.16 "Hot Spots" and Wildlife Effects 38
3.17 Wildlife Effects as Warning Signs of Humans Effects 39
3.18 "Social Costs" of Mercury Contamination 40
3.19 Research Needs 44
3.20 Conclusions 44
4. Exposure, Doses, and Body Burdens 47
4.1 The Scientific Basis of the Volume 3 Exposure Assessment 47
4.2 Characterization of the Exposure Assessment 51
4.3 Methylmercury and Fish Consumption 52
4.4 Sources of Data for Mercury Concentrations in Marine and Fresh-Water
Fish and Shellfish 53
4.5 Fish and Shellfish Consumption 55
4.6 Calculations of Methylmercury Intake from Crab 55
4.7 Characterizing Total Population Exposure and Risk 56
4.8 Hair Mercury Concentrations 57
VI
-------
4.9 Hair Mercury Concentration as an Index of Methylmercury Exposure ... 58
4.10 Linking Mercury in Fish and Current Anthropogenic Emissions 61
4.11 The Biokinetics of Methylmercury in Grain and Fish 62
4.12 Uncertainties in the Fish and Grain Consumption Analyses 63
4.13 Biokinetic Parameters for Children vs. Adults 66
4.14 Methodologies for Estimating Body Burdens of Mercury 66
4.15 Priority Areas for Research 67
4.16 Conclusions 68
EXPOSURE APPENDIX 1 71
General Comments 71
Specific Comments 73
5. HEALTH ENDPOINTS AND SUSCEPTIBLE SUBPOPULATIONS 75
5.1 EPA's Assessment Methodology 75
5.2 Alternative Neurological Endpoints 77
5.3 Fish Consumption Patterns and the "Developmental Window" 81
5.4 Subtle Endpoints Versus Traditional Metrics of Child Development 82
5.5 Uncertainties in the Characterization of Potential Effects 86
5.6 Risks among Subpopulations 88
5.7 Estimating Risks to Fetuses, Children, and Adults 91
5.8 Wildlife Effects as Harbingers of Human Health Effects 92
5.9 Unaddressed Human Health Issues 93
5.10 Research Needs 94
5.11 Conclusions 94
HEALTH APPENDIX 96
REFERENCES R-1
VII
-------
1. EXECUTIVE SUMMARY
The EPA undertook a "Herculean" effort developing its draft report on mercury.
Members of the Mercury Review Subcommittee were impressed by the robust and
extensive data amassed for the report, and by its generally correct analysis and
interpretation. As with any such large and complex task, however, modifications and
clarifications can be identified. The Subcommittee's major conclusions, based on the
existing data as presented in the EPA's draft report and the analysis of these data
(based on the scientific background of the Subcommittee's Members) are outlined
below. Such a listing tends to highlight problem areas rather than areas of strength,
and readers of this summary should keep this fact in mind. Specific findings of the
Subcommittee are:
a) That the majority of the human population is not experiencing
methylmercury exposures that are of concern from the standpoint of
human health. The current state of the science, however, doesn't allow
the establishment of a quantitative relationship between the sources and
actual exposures in humans or wildlife. The draft report is seen by the
Subcommittee as a document which, in general, reflects the current state
of the art with regard to human health and atmospheric mercury transport
issues. However, the Subcommittee feels that the report does not
adequately present or model Hg fate and transport in ecosystems, Hg
bioaccumulation or wildlife exposure. It is also noted, however, that the
high end of the distribution of methylmercury exposures is very
uncertain with respect to exposures, total number of people (and
percent of the population) who may be experiencing exposures high
enough to cause adverse health effects, and the actual sub-groups
who are highly exposed. Consequently, total population risk is not,
and cannot be, fully characterized at this time.
b) In general, from the standpoint of looking at human health effects
and the uncertainties in general, the draft report is a very good
document and an important step forward in terms of bringing the
relevant information together into one place for the first time. The
current Reference Dose, based on the Iraqi and New Zealand data,
should be retained at least until the on-going Faeroe and Seychelles
Islands studies have progressed much further and been subjected to
the same scrutiny as has the Iraqi data.
c) The Subcommittee identified some problems vis-a-vis human health
issues as discussed in the draft document - a lack of recognition
and emphasis on consistency of the data across multiple studies,
-------
and, most seriously, a failure to link animal studies to the human risk
assessment process. As pointed out at the public meeting, the
animal studies are poorly integrated into the general document In
addition, major problems exist in interpreting the biology of the outcome in
children versus adults; interpreting potential susceptibility differences
between populations; and identifying factors which may modify response,
such as diet. Note however, that the Subcommittee believes that there is
sufficient data to conclude that the developing organism is vulnerable
during the entire period of development and that in utero as well as early
postnatal exposure to methylmercury is of concern. Also, very little is
known about long-term chronic exposure in adults, because of the issue
of possible heightened sensitivity to mercury in older people, and whether
long-term exposure causes some different neurological responses.
d) With respect to the modeling linkage between emissions from
anthropogenic sources and human exposures, the Subcommittee
agrees that it is plausible that current anthropogenic emissions are
contributing to human exposures. However, the relative
contributions of current anthropogenic emission sources, global
backgrounds, recycling of old emissions and natural background to
human exposures of methylmercury are highly uncertain and difficult
to determine. Furthermore, the time constants for intermedia transport
are not known. Since mercury does not degrade in the global
environment, current anthropogenic emissions will add to the
fraction of global mercury that is actively circulating in the
biosphere, surface water, and soils. In addition, emissions from
both the U.S. and from other countries can impact the U.S.
environment and exposures of humans and wildlife. However, this
section of Volume III of EPA's draft report is not clearly written and
confuses environmental transport and fate modeling with exposure
analysis.
e) The Subcommittee concurs with the conclusion that fish are the
major source of methylmercury exposures for the human population.
The exposure analysis based on fish consumption must be regarded as a
"snapshot" in time. The species offish that are being consumed and their
sources are both changing rapidly due to over fishing of some species,
changes in the global market in marine fish and more domestic fish
supplied by fish farms. The impact of local anthropogenic sources of
mercury is likely to be greatest on freshwater and estuarine fish. There
are, however, two potential biases affecting the exposure analysis based
on fish consumption which should be checked in order to provide greater
confidence in the analysis. First, the consumption data from the
-------
Department of Agriculture's Continuing Surveys of Individual Food
Consumption (CSFII) should be checked against production data. These
will not agree exactly but should be reasonably consistent. Second, the
average concentrations of methylmercury in fish used for the exposure
assessment were provided by the National Marine Fisheries Service. The
Subcommittee recommends that EPA determine how measurements of
methylmercury in fish which were below the analytical limits of detection
(BDL) were statistically treated in arriving at the mean concentrations. If
the BDL values were excluded, then the means are biased high. If BDL
values were included in the estimation of the mean using one of the
standard methods for this, then the means can be considered good
estimators. However, if a very large percentage of the measurements for
any fish species were BDL, then the estimates of the means are much
less reliable.
f) The Subcommittee strongly recommends that EPA edit and shorten
Volume III on Exposure to eliminate redundancies and clarify the logical
sequence of the exposure analysis. In particular, the definitions of dose
and exposure presented in this volume should be used correctly and
consistently throughout the Volume, and transport and fate modeling be
clearly distinguished from exposure analysis. In general, the earlier EPA
Dioxin Risk Assessment provides a good model for the organization and
analysis of information on mercury for the entire draft report.
g) The Subcommittee notes that in the process of assessing control costs,
the draft mercury report devotes most of the cost analysis to end-of-pipe
controls and gives little attention to other types of controls. The
Subcommittee recommends that the cost analysis also give consideration
to other approaches for controlling mercury emissions that might prove to
be more flexible and more cost-effective. For example, new FDA
regulations (FDA 21 CFR Parts 123 and 1240-12/18/95) could have an
impact on exposures of the U.S. population to methylmercury. Some
analysis of their potential impact would be useful to EPA in evaluating
possible approaches for control.
h) The Subcommittee generally agreed that the mercury wildlife criterion as
currently presented ) is overly conservative and is lower than appears
necessary to protect wildlife species. The criterion was derived from a
bioaccumulation factor (BAF) based on total Hg rather than
methylmercury (MeHg), which is the mercury species that
bioaccumulates. As a result, the magnitude of the error in the BAF is
associated with, and reflects, the wide variability in MeHg concentrations
among ecosystems. Further, statements in the report (page ES-5) infer
-------
risk to various species without presenting risk calculations, a serious
shortcoming.
The Subcommittee suggests that EPA, using the best techniques
available in 1997, calculate BAF and wildlife risk criteria based on
dissolved
MeHg in water, and that BAF calculations be specific to ecosystem or
water body type.
The problems noted above notwithstanding, the Subcommittee
believes that piscivorous wildlife are at risk from elevated mercury
exposures and exhibit toxic effects in the areas of concern identified
by the U.S. EPA. In addition, there is good evidence showing effects of
mercury on wildlife; a summary of the population that appear to be
affected; and a statement that wildlife exposure can be a harbinger of
hard-to-detect human effects should be added to the final report
I) The Subcommittee finds that Volume V, in total, is not based on the
best available and sound science. To a large degree this chapter
follows EPA guidelines for ecological risk assessment. However, it
deviated in terms of quantifying the risk. This is a critical
component of risk assessment.
j) The final document should emphasize the fact that there are significant
information gaps in the understanding of the biogeochemistry of mercury
species and that the absence of this information limits the reliability with
which any mercury control program can be evaluated.
k) The modeling of atmospheric mercury transport and deposition is
largely sound, but the modeling of the post-deposition fate of
mercury in ecosystems is oversimplified, neglects available
information on speciation, and does not reflect recent advances in
the science. This deficiency is one of the most serious problems in
the draft report. This modeling should be revised, with the goal of
having the report reflect current scientific understanding. Also, the
modeling, as used, contributed to sizeable errors in the estimated wildlife
criterion. It is important to bring out the importance of MeHg production in
ecosystems in the report. Because MeHg is the species that
bioaccumulates in human and wildlife food supplies, understanding the
methylation process is critical to modeling Hg fate and exposure. Mercury
methylation and the variability in MeHg among ecosystems were
neglected in the report, because of the perceived difficulties in modeling
methylation. The report assumes that MeHg constitutes the same fraction
-------
of total Hg in sediments, soils and waters across ecosystems. This is not
the case. Rather, the variability in MeHg production and bioaccumulation
among ecosystem types is many orders of magnitude, and may be as
large as the influence of Hg contamination. Without consideration of this
variability, the report's fate and transport models, and the resultant
exposure models and wildlife criteria models cannot predict Wildlife
Criteria with less than two to three orders of magnitude associated error.
In addition, the proposed use of nitric acid as a surrogate for the wet
deposition of mercury may not be justified, given the differences in the
chemical and physical properties of the two species.
The final EPA report should detail the state of our understanding of Hg
methylation process within Volume III; address MeHg production along
with fate and transport models in Volume III (which models should be
specific for a suite of ecosystem types); and the resultant information be
used to model and discuss wildlife exposure in Volume V. Because the
models for estimating the distributions of human exposure were
based on measured values for Hg in fish and other food, rather than
modeled values, these suggested changes will not affect the human
exposure analysis in the report.
-------
2. BACKGROUND AND CHARGE
2.1 Background
The 1990 Clean Air Act (CAA) Amendments (section 112(n)(1)(B)) directed the
EPA to perform a study on the impacts of mercury as an air pollutant and provide a
report to the Congress. The subject section required EPA to address several specific
topics, including mercury emissions from electric utility steam generating units,
municipal waste combustion units, and "area" sources; the rate and mass of these
emissions; the associated health and environmental effects; technologies for controlling
these emissions; and the costs of such control technologies.
In response, the EPA developed a seven volume (draft) report which:
a) provides data on types, sources, and trends in mercury emissions
b) evaluates the atmospheric transport of mercury
c) assesses the impacts of mercury emissions on organisms/ecosystems
close to the emitting source
d) identifies major exposure pathways to humans and non-human biota
e) identifies mercury exposure levels likely to produce adverse effects in
humans, and the nature of those effects.
g) evaluates mercury exposure effects for ecosystems and non-human
organisms
h) identifies populations especially at risk from mercury exposure due to
special sensitivity or high exposure
I) estimates effectiveness of control technologies and their costs
As noted above, the EPA report is still in draft. The Agency initially stopped
short of issuing a formal final mercury report to Congress because of a growing
consensus that such an analysis should wait for a full assessment of several relevant
studies now underway. These studies address the effects of methylmercury on
neurological development of children in fish-consuming populations in the Seychelles
and Faeroe Islands, and were expected to be completed and published in the next year
or two. Their results should be considered before making a new assessment of health
risks. In terms of current data, however, the Agency decided to proceed with finalizing
-------
the existing report by having the Science Advisory Board (SAB) review it (as is
customary with major scientific documents) before transmitting it to the Congress.
Following a request from the Office of Air Programs and the Office of Research
and Development, the SAB created a special Subcommittee (the Mercury Review
Subcommittee (MRS)) of its Executive Committee. Composed of 34 experts on mercury
from the United States and Canada, the MRS convened in public meeting on February
26/27, 1997 in Washington, DC. The Subcommittee was organized into three
Workgroups, and addressed a Charge of 43 enumerated questions (see section 2.2)
developed via discussions between EPA and SAB staff. This report, developed via
mail in the several months following the public meeting, reflects the discussions at that
meeting.
2.2 Detailed Charge
The detailed Charge follows. Specific questions are set in boldface type to
differentiate them from background/expository material.
GENERAL
1. The EPA has decided to defer evaluation of the dose-response relationship between
methylmercury and fetal and developmental effects until additional scientific data have
been evaluated. These issues will be addressed in Phase Two of the SAB review.
Putting these matters aside, are the overall assessment and conclusions based
on sound and appropriate evaluation of the relevant science?
SOURCES
2. In order to approximate future mercury emissions, the sources are simplified to
model plants placed to approximate current U.S. emission patterns. Is this approach
consistent with the best available scientific practice? Are the implications of this
simplifying assumption adequately presented?
3. U.S. EPA limited the scope of its assessment to anthropogenic sources of mercury
emissions. Are these reasonable and defensible assumptions? The emissions
inventory estimates these emissions in the United States. There is a short discussion
of natural emissions and re-emitted emissions as well. The discussion of these
emissions was based on language suggested by the scientific peer review panel. The
modeling of local point sources did not include natural emissions, while the long-range
transport modeling did incorporate a background level of mercury (to account for
natural and re-emitted Hg) which has been measured over the open ocean. Are the
reasons
-------
for this approach adequately stated? Are the uncertainties and implications of
this approach consistently and appropriately described?
4. For the atmospheric fate and transport modeling conducted using both the Regional
Lagrangian Model of Air Pollution (RELMAP) and the Complex Terrain and Deposition
Air Dispersion Model (COMDEP), best estimates from the peer-reviewed literature of
the species of mercury emitted from each source were utilized. Were these
speciation profiles appropriate?
5. Can the Committee comment on the evidence concerning the role of human
activity, and if possible, the role of U.S. sources to methylmercury in ocean
seafood?
ENVIRONMENTAL FATE/TRANSPORT
6. U.S. EPA relied on modeling to describe mercury fate and transport. In the report,
the inventory served as the building block and was linked to the models (which were
themselves linked). Both local and regional impacts were assessed together (additive)
and separately for both humans and wildlife. Model plants were used as surrogates for
actual facilities as well as hypothetical locations and populations. Would a case study
approach using actual facilities be more appropriate? Are the rationales for the
selected approach appropriately presented? Please comment on the usefulness
of measured data for estimating mercury exposures.
7. In its assessment, U.S. EPA assumed that available data on the deposition of nitric
acid could be used to approximate the likely wet deposition of divalent mercury. Is the
rationale for this assumption adequate (see Volume 3, Section Chapter 6, Section
6.1.3, and Volume 3, Appendix D, Section D.1.3.2)? Classical Gaussian and
Lagrangian puff models were adapted to reflect mercury chemistry. Were these
modifications appropriate?
8. Further, the assessment neglects the potential conversion of elemental mercury to
divalent forms in cloud droplets in both the local scale and the regional analysis. The
Agency has assumed that any such conversions would happen on such a small scale
that they could have only a very nominal impact on the ultimate exposure
characterization. Is this rationale fully explained and adequate? Are the
uncertainties of this assessment explained?
9. The Agency recognizes that mercury fate and transport is an immature field;
capabilities to describe phenomena such as the temporal distribution of mercury
concentrations in soil continue to evolve. In its assessment, U.S. EPA attempted to
modify existing approaches (the 1990 Methodology of Assessing Health Risks
8
-------
associated with Indirect Exposure to Combustor Emissions and the 1993 Addendum) to
reflect the more recent science. Were the assumptions regarding equilibrium
among mercury species present in soil appropriate?
10. Was the selection of the model input parameters (such as the soil-to-water
partition coefficients) appropriately justified and the uncertainties explained? Is
the estimation of the watershed and water body fluxes adequate and
appropriate? Should fluxes in large lakes, rivers and other large water bodies
also have been estimated? Is the rationale for not doing so adequate? Does the
Agency appropriately justify and explain the uncertainty surrounding
assumptions about plant uptake of mercury directly from the atmosphere? Is the
likely contribution of mercury from plants to soil contributions sufficient that
such sources should be added to future modeling efforts?
11. A number of qualitative conclusions are drawn from the results of the atmospheric
modeling; given the uncertainty in the modeling and in the state of the science,
evaluate the scientific quality of these conclusions. Were the uncertainties
adequately described?
12. A key variable in estimating ultimate potential human exposure is the
bioaccumulation factor (BAF) for methylmercury through the aquatic chain. Was the
BAF developed by the Agency appropriate and its utilization in the models
scientifically sound? Was the uncertainty analysis adequate?
13. Were the models and modeling conducted for the terrestrial food chain
adequate and appropriate?
EXPOSURE
14. The exposure assessment is contained in Volume III. Please comment on the
scientific basis of this assessment, alternative approaches, and research
priorities. In particular, please advise the Agency on how Appendix H should be
used, improved, or expanded.
15. Because of uncertainties inherent in the emissions inventory and the model, the
exposure assessment was characterized as being a "qualitative assessment based on
quantitative modeling." Is this a reasonable characterization? Were the
uncertainties of and conclusions drawn from linking the models together
appropriately described?
16. U.S. EPA has estimated methylmercury and fish consumption using both cross-
sectional and longitudinal data for the general U.S. population. In these studies, EPA
-------
consistently finds that a portion (albeit a small one) of the population consumes very
large quantities of fish. It is these individuals who face the largest risks from
methylmercury exposure and who pose the greatest concern to the Agency. Has EPA
adequately assessed the number and types of individuals who consume fish in
these relatively large quantities? The group at issue includes 5% of the fish-
consuming population. Are the uncertainties associated with characterization of
such small portions at the extreme end of the population distribution
appropriately considered and presented? Could the estimate of 5% of the fish-
consuming population be a methodologic artifact?
17. U.S. EPA has relied on three nationally-based sources of data for mercury
concentrations in marine and fresh-water fish and shellfish. Some of these data were
obtained within the past three years, whereas other data were obtained as long ago as
20 years. Is the use of these older data appropriate? Does the document
appropriately note the limitations of this methodology?
18. The USDA has published data indicating that fish and shellfish consumption have
risen approximately 25% since the early 1970s . These consumption data were
employed without adjustment in this report to estimate current fish consumption
(Volume III, Appendix H; and Volume VI, Section 4.2.3.1). Should these data be
adjusted to reflect the overall trends in fish consumption? If so, how would the
estimates of very high consumption (above the 90th percentile) be adjusted?
19. In Volume III, Appendix H, calculations of methylmercury intake from crab were
grouped without specific consideration of the particular species. These data are then
linked to dietary survey data which typically do not list the individual species of many
fish and shellfish consumed. How might this approach bias the final assessment?
Are the limitations of this approach adequately characterized in the report? Are
these limitations going to materially alter the assessment of health risk
associated with methylmercury exposures?
20. In early versions of the draft report (and in Volume III, Appendix H), U.S. EPA
characterized total risk by considering methylmercury exposure from marine fish and
shellfish together with consumption of methylmercury from fresh-water or estuarine fish
and shellfish. It has been recommended that the report focus exclusively on fish from
inland sources. Does the current version adequately characterize total risk? Are
the uncertainties resulting from this methodology appropriately presented?
21. Some data exist on hair mercury concentrations in U.S. residents. Is this data
base adequate to predict the distribution of hair mercury in the general U.S.
population?
22. The draft report identifies hair mercury concentration as the most appropriate
10
-------
available index of methylmercury exposure. Is this assumption consistent with the
available data? Are the exposure estimate approaches used relevant and
appropriate? Are the predicted exposure ranges consistent with other published
exposure analyses? Are the uncertainties of this assumption appropriately
presented?
23. Although the modeling exposure assessment focused on anthropogenic emissions,
the fish consumption analysis considered measured mercury concentrations in fish
tissue regardless of the mercury's origin. Thus, there is considerable difficulty in
assessing or describing how much of the mercury in fish is attributable to current
anthropogenic emissions. Is the approach taken by the Agency in this assessment
appropriate given the available data? What is the advice of the SAB regarding
the differentiation between current emissions and their impacts relative to the
"body burden" approach of the fish consumption analysis?
DOSES/BODY BURDENS
24. In its assessment, U.S. EPA assumes that the biokinetics of methylmercury from
contaminated grain approximate those resulting from methylmercury from fish. Is this
assumption consistent with the available science?
25. Are the uncertainties in the fish and grain consumption analyses adequately
and consistently presented in the draft report?
26. Similarly, U.S. EPA assumes the biokinetic parameters for children are identical to
those of adults. Does such an assumption introduce bias into the assessment?
Are the uncertainties adequately and consistently presented in the draft report?
Is the conclusion that children's lower body weight results in higher exposure
than to adults? Could this be an artifact of exposure modeling using 3-day
consumption survey data?
27. What methodologies other than hair concentrations could be used to estimate
body burdens of mercury?
HEALTH ENDPOINTS and SUSCEPTIBLE SUBPOPULATIONS
28. Volume 4, Section 6.3, and Volume 6, Chapter 2, Sections 2.1 and 2.2.21 present
U.S. EPA's interpretation of the data demonstrating increasing severity and frequency
of nervous system effects (particularly impairment in visual-motor integration) among
pediatric subjects with increasing levels of maternal hair mercury. These sections
present U.S. EPA's interpretation of the data demonstrating increasing severity and
frequency of nervous system effects (particularly impairment in visual-motor
11
-------
integration) among pediatric subjects with increasing levels of maternal hair mercury.
Recognizing that additional data may shortly be available, is EPA's assessment
methodology based on the best scientific practice? Are the uncertainties in the
characterization of potential effects accurately described in the draft text?
29. Data from experimental animals (including primates with long-term exposures to
methylmercury) show methylmercury-induced nervous system damage, particularly on
the visual system, although the animals appear clinically normal. The traditional RfD
methodology neglects such impairment. Are these data important endpoints? Are
they appropriately characterized in the draft report? How could such data be
better evaluated by the Agency?
30. The available data present information on fish consumption drawn over relatively
short time periods (e.g., days) and are used to extrapolate consumption patterns over
longer periods (e.g.. a month). Although the exact developmental window affected by
methylmercury in humans is not precisely defined, it is thought to be less than three
months. Is such extrapolation appropriate in this case? Does the document
appropriately present the limitations of this methodology?
31. Volume 4, Section 6.3, and Volume 6, Chapter 2, Sections 2.1 and 2.2.21 present
U.S. EPA's interpretation of the data demonstrating increasing severity and frequency
of nervous system effects (particularly impairment in visual-motor integration) among
pediatric subjects with increasing levels of maternal hair mercury. What biases are
introduced into the assessment by focusing on subtle endpoints of
neurobehavioral function in contrast with traditional metrics of child
development? Should these indicators be dismissed if traditional metrics (such
as age of first walking) are within normal ranges? Do the available data indicate
whether children are more sensitive than adults to the effects of methylmercury?
32. Are the uncertainties in the characterization of potential effects accurately
described in the draft text?
33. Traditional methods for estimating potential human health risks from environmental
hazards do not distinguish risks among subpopulations (e.g. racial or ethnic) groups. If
methylmercury exposures are comparable and some groups, but not all, show
impairment in traditional and/or specialized neurophysiological/neurobehavioral
tests, how should U.S. EPA reflect these differences in its analyses? Do factors
such as nutritional status, life-styles (e.g., substance abuse), or economic status
play a role is mediating these differences?
34. Similarly, many assessments of methylmercury risk clump risk to fetuses and
children together with risk to adults. Is this approach scientifically valid for
12
-------
methylmercury? Methods for estimating potential human health risks from
environmental hazards do not distinguish risks among subpopulations (e.g. racial or
ethnic) groups. If methylmercury exposures are comparable and some groups, but
not all, show impairment in traditional and/or specialized
neurophysiological/neurobehavioral tests, how should U.S. EPA reflect these
differences in its analyses?
35a. What wildlife effects (based on what metric) can be interpreted as harbingers
of likely human health effects?
35b. Could the Committee provide any short-term advice on human health issues
not addressed in the coming epidemiologic studies such as toxicokinetics?
ISSUES ON WILDLIFE ASSESSMENT
35c. Chapter VI presents the Agency's current mercury risk characterization. Please
advise the Agency on the appropriateness and environmental significance of the
characterization of wildlife effects.
36. Volume V describes fairly significant wildlife effects which are attributed to elevated
mercury concentrations in some ecosystems. Is this evaluation based on the best
available and sound science, and are they consistent with EPA eco risk
guidelines? How could this evaluation be improved?
37. This report makes the inference that mercury emissions are related to reproductive
effects and neurobehavioral changes in fish consuming birds and mammals. These
effects have been documented in the Great Lakes and in the Southeastern United
States. EPA's models predict that regional hot spots (relatively high concentrations of
methylmercury) would occur in these same areas. Is the analysis of the evidence
linking emissions and effects scientifically sound?
38. Have any "hot spots" not shown evidence of wildlife effects? Do these
predicted "hot spots" correlate with methylmercury levels in fishery products?
39. Which mercury-related effects in wildlife should be identified as warning
signs that analogous effects may occur in humans? Could "wildlife
epidemiology" be used as a surrogate for controlled lab studies?
RESEARCH NEEDS
40. Based on the research needs identified in the draft report, areas identified
through discussion of the report, or on other information, has EPA identified the
13
-------
highest priority areas for research? If not, could the Committee suggest what
areas need to be addressed?
QUESTIONS RELATED TO SOCIAL COST
41. Chapter VII previously contained a section on the "Social Costs" of mercury
contamination (this section is attached to comments by the Council of Economic
Advisors in the supplementary information). This section described the value of the
fisheries in the U.S. as well as other values such as maintaining a healthy ecosystem.
The intent was to balance the discussion of mercury control costs. Some reviewers
objected to the inclusion of the Social Costs section partly because the impact of
anthropogenic sources could not be directly and quantitatively related to these impacts
(e.g., the declining Florida panther population). The section was consequently deleted.
Is such a discussion appropriate for this study? How would the SAB advise the EPA to
describe benefits of mercury reductions if a) the impacts of such reductions are not
directly quantifiable, and b) the monetary value of such benefits are not easily
quantified?
14
-------
3. SOURCES
3.1 Source Simplification in Modeling Mercury Emission Patterns
In order to approximate future mercury emissions, the sources of mercury were
simplified to model plants located to approximate current U.S. emission patterns.
Question 2 of the Charge asked the Subcommittee to assess this approach for
consistency with the best available scientific practice, and to consider the possible
implications of this simplifying assumption.
The issue of the use of "source simplification" is best addressed by considering
the current (existing) sources and the potential for new sources. According to EPA's
Table ES-3 (Volume I), current principal sources include fossil fuel combustors,
incinerators, utility boilers, commercial/industrial boilers, and industrial sources. Will
the regional distribution of incinerators in the country change significantly in the near
future? Probably not. Will there be significant changes in the number and locations of
utility and commercial/industrial boilers in the near future? Not very likely.
With the reduction of mercury use in existing commercial products, it seems
unlikely that many major new uses will be discovered that will drastically change the
emission scenario in the near future; consequently the Subcommittee does not note
any particular problems arising from the simplification approach, nor its
presentation in the draft report.
3.2 Modeling Sources of Mercury Emissions
The EPA limited the scope of its assessment to anthropogenic sources of
mercury emissions. The modeling of local point sources did not include natural
emissions, while the long-range transport modeling did incorporate a background level
of mercury (to account for natural and re-emitted Hg) which has been measured over
the open ocean. Charge question 3 asked for the Subcommittee's findings on the
major assumptions made, the statement of the supporting rationales, and the
implications of using such an approach.
The role of natural sources of Hg and re-emission of "old" anthropogenic
sources of Hg is recognized as being important on a global scale (Expert Panel, 1994).
The EPA draft report acknowledges this, but with a lack of regional data and models to
treat these sources specifically, was unable to include them in the overall analysis.
Many caveats to this effect are stated, but the impacts of ignoring such sources on
model results and interpretation are not clearly defined. In particular, it must be
acknowledged that such sources cannot be readily controlled (if at all), and that the
results of reducing U.S. anthropogenic emissions will be less than expected (with
15
-------
respect to resulting exposure levels) because of these other sources. In other words,
there will not be a 1:1 relationship between decreases in U.S. industrial emissions and
changes in receptor exposure, with the possible exception of a site very close to a
strong point source of reactive gaseous mercury. The Expert Panel (1994) estimated
that 95% of the 200,000 Mg of Hg emitted since 1890 remains in surface soils and lake
and oceanic sediments, so it only requires a re-emission of a fraction of a percent each
year for this source to rival U.S. industrial emissions (see Lindberg etal., accepted;
Carpi and Lindberg, accepted A; and Kim etal., 1995).
3.3 Speciation Profiles in Atmospheric Fate and Transport Modeling
Charge question 4 addressed the atmospheric fate and transport modeling
conducted using both the Regional Lagrangian Model of Air Pollution (RELMAP) and
the Complex Terrain and Deposition Air Dispersion Model (COMDEP), with specific
attention to the speciation profiles used.
For its atmospheric fate and transport modeling, EPA used the speciation
profiles based on European data from Peterson etal. (1995) for its baseline scenario.
The Electric Power Research Institute and the U.S. Department of Energy have been
working to improve methods for measuring the speciation of mercury emissions from
coal-fired utilities. Data from these new studies have not yet appeared in the
peer-reviewed literature and EPA cannot be faulted for not having included these
findings in its draft. To our knowledge, there are no U.S. data on the species of
mercury emitted from municipal or hospital waste combustors. The EPA also used an
alternate speciation profile, which had a high percentage of mercury in the particulate
form. Although it is useful to explore the sensitivity of the model to changes in the
speciation profiles, the plausibility of this alternate speciation profile should be clarified.
Consequently, the speciation profiles used by EPA in its baseline scenario were
not inappropriate, given the information available at the time. However, EPA
should evaluate new speciation data as it becomes available, and revise its
speciation profiles if necessary.
Speciation is critical to the modeling of mercury fate and transport. It controls the
predictions of the amount and location of deposition. It also controls the total loading at
a given location, and affects the predictions of the eventual loading of methylmercury in
ecosystems. Thus improvements in the understanding of mercury speciation are of the
highest importance for improving the entire mercury risk assessment.
3.4 Anthropogenic Sources and Methylmercury in Ocean Seafood
16
-------
In Charge question 5, EPA asked for an assessment of the evidence concerning
the global relationship of human activity to methylmercury levels in ocean seafood, and
if possible, the contribution of U.S. sources to methylmercury levels.
The upper layer of the ocean is estimated to have roughly a factor of 3 times the
mercury content of pre-industrial times (Mason et. al, 1994). Total anthropogenic
mercury emissions are estimated to be about twice global natural emissions. Both of
these estimates suggest that the anthropogenic contribution to ocean seafood
may be substantial. However, because there is currently no detailed
understanding of the quantitative relationship between the amount of
methylmercury in ocean seafood and anthropogenic mercury emissions, we
cannot quantify the anthropogenic contribution to methylmercury in ocean
seafood.
Seafood from coastal areas may be a special case. Seafood from U.S. coastal
areas can be expected to be exposed to anthropogenic mercury from U.S. sources in
much the same way as freshwater fish. Although U.S. emissions of mercury represent
only about 5% of global anthropogenic emissions as of 1995, the impact of U.S.
sources on U.S. coastal seafood can be expected to be larger than this ratio, due to
the influence of local sources. However, the EPA's draft report did not attempt to model
this pathway. Such an assessment should be considered in the EPA's revised report.
3.5 Modeling Mercury Fate and Transport
U.S. EPA relied on modeling to describe mercury fate and transport. In the Draft
Report, the known Hg inventory served as the building block and was linked to the
prediction models (which were themselves linked). Both local and regional impacts
were assessed together (additive) and separately for both humans and wildlife.
Modeled industrial plants were used as surrogates for actual facilities as well as
hypothetical locations and populations. In Charge question 6, the Agency asked for
guidance on the issue of using a case study approach using actual facilities rather than
modeling; the Subcommittee's appraisal of the rationales for the selected modeling
approach; and comments on the usefulness of measured data for estimating mercury
exposures.
Clearly, a case study approach would not have been more appropriate than
the modeling approach chosen by EPA. To the Subcommittee's knowledge, there
are no case studies linking air emissions from a single specific point source to
exposure in a specific water body. The Aquatic Cycling of Mercury in the Everglades
(ACME) study, sponsored by a broad consortium of federal, state, and corporate
entities, together with several other studies monitoring atmospheric deposition patterns
in South
17
-------
Florida, may be able to link the suite of South Florida sources to bioaccumulation and
effects in the Everglades, but the study is not yet complete.
Case studies limited to relating emissions to deposition around individual
facilities may have been useful in the report (Carpi etal., accepted B; Stratton and
Lindberg, 1995). However, we do not know if adequate case studies linking emissions,
deposition, and exposure exist.
Considering the limitations in knowledge of the Hg cycle and Hg exposure
modeling, the Subcommittee believes that it is important to use what limited measured
data are available for estimating Hg exposures to both people and wildlife.
Although the Subcommittee agrees with EPA that it is necessary to rely on
modeling to describe mercury fate and transport, the rationales for the selected
modeling approach were not appropriately presented. The Subcommittee divided
its consideration of the fate and transport modeling into two parts: air transport and
deposition, and fate and transport after deposition and bioaccumulation within
ecosystems.
The draft report's rationale and presentation of air transport and deposition
modeling was reasonable overall, although it has some shortcomings. These issues
are discussed in sections 3.7 and 3.8 of this report. However, the Subcommittee was
disappointed with the models downstream from the deposition models, particularly the
fate and transport in ecosystems and the bioaccumulation models. The consensus of
the Subcommittee was that the rationale for the selection of these post-deposition
models was poorly justified and inappropriate. The Subcommittee felt that these
weaknesses were the most serious in the report overall. Specifically, EPA's
decision to model %MeHg (methylmercury) as constant across ecosystems, and
to model fish bioaccumulation factors (BAF) based on total Hg (rather than
MeHg) were most problematic.
It is not possible at this time to produce highly quantitative models for MeHg
production among ecosystems, starting with emissions or deposition. However,
"quantitative modeling yielding qualitative responses" (an approach used in Vol. Ill and
V to provide rough ideas of how deposition responds to source strength) is possible for
modeling MeHg production and exposure in aquatic systems. The Indirect Exposure
Methodology (IEM2) model discussed in Vol. Ill, which is mainly a watershed transport
model, provides estimates for the total Hg concentration in water that is protective of
wildlife (the Wildlife Criteria, or "WC") of about 0.5 ng Hg/L. This value is well below
that of most natural inland waters at this time. The WC provided by the model has an
extremely large associated error - 3 orders of magnitude between the 5 and 95%
confidence intervals. The state of the science in 1997 provides an opportunity for EPA
to make WC evaluations more ecosystem specific by explicitly including ecosystem
18
-------
type and variability in the analysis (St. Louis et al., 1996; Watras et al., 1995; Driscoll et
al., 1994; Hurley et al., 1995). By assessing %MeHg in surface water by ecosystem
type, the EPA model could better capture variability among ecosystems. In addition,
basing BAF on MeHg in water rather than on total Hg, and normalizing BAF by fish
species, can also improve the credibility of the WC models. Even though many
uncertainties in the transport and fate analysis will remain, these changes will improve
the ability of the models to appropriately account for variations in WC. By making use
of existing models external to EPA, or by applying literature values of %MeHg
among ecosystems to the existing EPA model, the report could be revised
adequately, and fairly quickly.
3.6 Nitric Acid as a Surrogate for Wet Deposition of Divalent Mercury
The Agency modified existing models to describe mercury atmospheric fate and
transport, and, in the case of wet deposition of divalent mercury, used data on the
deposition of nitric acid as a surrogate for mercury. In addition, classical Gaussian and
Lagrangian puff models were adapted to reflect mercury chemistry as part of the overall
modeling approach. In Charge 7, the Subcommittee was asked to comment on the
adequacy of the modifications, as well as on the use of nitric acid as a surrogate for
mercury.
In brief, the Subcommittee noted that:
a) Modifications made to RELMAP to model mercury transport,
transformation, and deposition appear to be appropriate, but several
areas (identified below) should be clarified in the report.
b) Modifications made to COMPDEP are not completely satisfactory. The
dry deposition velocities for Hg(ll)(9) should be amended to include the full
specification found in Table D-17 of the draft report. Furthermore, a
similar table should be developed and implemented for Hg(0). The
scavenging coefficients for wet removal should be reformulated
specifically to simulate below-cloud processes.
c) Application of COMPDEP to the model plants must be revisited, in view of
the recommended changes to COMPDEP. In particular, the present study
produces dry Hg(ll)(9) fluxes that are most appropriate for land-use
categories of water and barren land (see Table D-17). Dry deposition
velocities to mixed forest/wetland areas can be 3 to 4 times larger!
Variation in land use between the source and "critical receptors" in the
simulations should also be considered. For example, depletion of the
plume during transport due to dry deposition to forests (say at a velocity
of 3 cm/s) will reduce the
19
-------
near-surface Hg concentration available to dry-deposit to the surface of a
distant watershed (at a velocity of 1 cm/s).
d) Although other studies have made this assumption, the use of nitric acid
as a surrogate for the wet deposition of mercury may not be justified,
given the chemical and physical properties of the two species.
The only justification given for EPA's assumption that available data on the
deposition of nitric acid could be used to approximate the likely wet deposition of
divalent mercury is that the water solubility of these two species is presumed to be
"comparably high," in the words of Petersen et al. (1995), the paper which forms the
basis for much of the deposition formulation used in the draft report. This high
solubility implies that both vapors will rapidly dry-deposit to surfaces which they
contact. Furthermore, for wet deposition, high solubility is expected to increase the
washout efficiency. However, the presumption ignores the influence of associated
anions on mercury solubility. Does it matter whether mercuric sulfate, oxide, hydroxide,
or chloride nitrate is present in the aqueous medium? Most likely. The possible effects
of complex formation on Hg solubility are also ignored. Differences between the ionic
Hg(ll) and molecular properties (N02) would also argue against the assumption: (a)
HN03 is a bigger molecule than Hg(ll); (b) the washout coefficient for the two species
differ; c) the mass median diameter for the two species differ; and (d) the densities of
the two species are different. Consequently, although other studies have made this
assumption, it may not be justified on the basis of the chemical and physical
properties of the two species.
Given the importance of the substantial estimates of wet and dry deposition fluxes
of divalent mercury, the report should state the uncertainties involved with using the
properties of nitric acid as a surrogate for divalent mercury, it should cite other studies
which make this assumption when estimating fluxes of divalent mercury, and it should
provide a discussion of the mechanisms by which high solubility enhances removal
processes.
The report makes reference to the use of nitric acid as a surrogate for the
deposition of divalent mercury vapor in discussing the Local Impact analysis (Chapter 6
of Volume III), and in the modeling details presented in Appendix D, but not when
discussing the results of the Long Range Impact analysis (Chapter 5). Because the
base-case versus alternate speciation simulations illustrate the sensitivity of deposition
rates to speciation assumptions involving the presence of divalent mercury vapor, the
basis for the deposition estimates should also be stated here (e.g. page 5-19).
EPA adapted classical Gaussian and Lagrangian puff models to reflect mercury
chemistry in estimating deposition. As detailed below, the Subcommittee noted some
problems in EPA's application of these models, particularly with respect to the use of
20
-------
certain adjustment factors for local dry deposition. In the RELMAP model, injection of
emissions into layer 1 (for "area sources") or layer 2 (for point sources) appears
appropriate, but is only effective at night because during the day, RELMAP mixes all
advected species throughout its 4 layers. At the same time, a local dry deposition
factor is applied to particulate mercury emissions released into layer 1. The draft report
states that half of the emissions into layer 1 are assumed to be dry-deposited within the
local cell (nominally 40 km), based on a similar adjustment in Petersen etal. (1995) for
their 1-layer puff model. This approach is designed to compensate for underestimating
dry deposition fluxes for near-surface releases when near-surface concentrations are
underestimated due to complete mixing in the vertical. At the public meeting, EPA staff
stated that the final RELMAP simulations do not use this adjustment. The report
should be updated to remove the present discussion, and to replace it with the
rationale for not including such an adjustment. In particular, the discussion should
address the effectiveness of the adjustment during the day compared to the treatment
at night, and it should address uncertainties in selecting an appropriate removal factor,
if it were used. Petersen et al. (1995) estimate this factor to be 0.5 for Hg(ll)(g), and
0.025 for Hg(part). They obtain these estimates by assuming that the factor is
proportional to the dry deposition velocity for each species, and use a deposition factor
of 0.1 for S02 as a basis. With a dry deposition velocity of 0.8 cm/s for S02, and 0.2
cm/s for Hg(part), the factor is 0.025. The deposition velocity for Hg(part) used in
RELMAP varies from 0.11 cm/s to 0.02 cm/s, so that deposition factors of 0.014 to
0.0025 would be estimated from this method. (Differences in the grid system will
influence the value for this factor as well.)
Ozone reacts with elemental mercury, changing it to the divalent form (Hg++),
which is reactive with the environment, constituting a loss mechanism for atmospheric
mercury. The question which must be posed is "How strong are spatial gradients in
ozone aloft (as provided to RELMAP) during precipitation events, when RELMAP uses
ozone in its wet scavenging algorithm for Hg(0)?" That is, how important is ozone
advection aloft on the aqueous phase chemical transformations?
Ozone concentrations aloft (layers 3 and 4) are specified for each cell, as
computed from observed hourly surface measurements from the EPA's Aerometric
Information Retrieval System (AIRS). The two midday 3-hour time steps use the
corresponding 3-hr averaged concentration measurements. Recognizing that surface
ozone measurements are not representative of ozone concentrations aloft at night, the
average concentration over the period between 1000 and 1600 is persisted throughout
the 18 hour period from 1600 to 1000 the next day. This procedure does not simulate
ozone transport aloft. The report should include a discussion/analysis of the
assumptions concerning ozone gradients to the computed aqueous phase
conversion of Hg(0) to Hg(ll)
21
-------
Wind data used for advection in RELMAP are stated to be the wind field
initialization data for the Nested Grid Model (NGM). The NGM uses a nominal 90 km
grid. Results of NGM simulations were reportedly stored at 2-hr intervals for the 1989
year used in this study, but the Subcommittee understands that the NGM is initialized
at 12-hour intervals. At the public meeting, EPA staff stated that 3-hour average wind
data were used in the study. The report should include a discussion that clarifies
the source and characteristics of the meteorological data.
Dry deposition velocities for soot were computed for a matrix of stability
class/land use types using the California Air Resources Board (GARB) subroutine.
This required an assignment of near-surface wind speed (nominally 10 cm/sec), and a
roughness length. Wind speeds were assigned on the basis of the stability class, but
the wind speeds associated with classes A and D (Table D-3) are substantially different
from those typically used in the stability class typing procedure (see Table D-7).
Speeds for class A are typically of order 2 m/s or less, not 10 m/s. And those for class
D are typically greater the 6 m/s, not the 2.5 m/s reported in this study. Regarding the
roughness length assigned to each land use category, several of the values in Table D-
4 appear to be smaller than those typically used. Urban areas and forests are typically
given a roughness length greater than 1.0m, but a length of 0.5 is used here. Also,
agricultural uses are given smaller roughness in the spring-summer than in the autumn-
fall. Typically, the larger roughness is assigned for summer. The curves on which the
GARB model is based do not include roughness lengths greater than 0.1 m, so
deposition velocities obtained from this model may not fully characterize dry deposition
to rougher surfaces. The report should identify uncertainties associated with this
use of the CARB subroutine. Also, the rationale for the pollutant depletion rate of 5%
per 3-hr time step chosen to represent the diffusive mass exchange out of the top of the
model should be provided.
In the COMDEP, dry deposition velocities for Hg(ll)(g) are said to be taken from
the RELMAP implementation, and specified by stability class (COMPDEP was modified
to allow such user-supplied deposition velocities as a function of stability class).
However, a single value of 1.0 cm/s is used for classes A, B, and C, and a single value
of 0.3 cm/s is used for classes D, E, and F, in spite of the presence of a full set of
values given in Table D-17. The rationale for this scheme is not presented in detail,
but the neutral/stable deposition velocity of 0.3 cm/s is said to represent nighttime
periods when RELMAP uses 0.3 cm/s. Class D (neutral) occurs frequently both day
and night, as it applies for the larger wind speed classes, so assigning all class D hours
a "nighttime" deposition velocity will underestimate some daytime fluxes by as much as
a factor of 3 to 9.The values listed in Table D-17 should be used directly in the
COMPDEP simulations, for the dominant land use prescribed for each site. It
would be preferable to adopt an existing gas phase dry deposition module, and place it
into COMPDEP or other Gaussian plume model (GPM) for this study.
22
-------
During the public meeting, one Member of the Subcommittee pointed out that dry
deposition of Hg(0) is not zero, as assumed in the study. The Lindberg etal. (1992)
report modeled dry deposition velocities (weekly averaged) to the deciduous forest at
the Walker Branch Watershed that range from about 0.006 cm/s during the winter to
about 0.1 cm/s during a few weeks in the summer. This increase is the result of leaf
stomatal control of dry deposition during the growing season. Another effect of this
control is the strong diurnal pattern in their computed deposition velocities, which vary
from about 0.04 cm/s at night to 0.09 to 0.18 cm/s during the day for a simulated period
in mid-June. Hg(0> dry deposition should be included in the near-field modeling
assessment.
The precipitation scavenging treatment adopts the washout parameters of
Petersen etal. (1995) for use in COMPDEP. However, the highly parameterized
scavenging coefficient approach of this and similar models does not differentiate
individual wet removal processes, such as in-cloud nucleation and aerosol growth, and
below-cloud interception. Computed fluxes depend on vertically-integrated pollutant
concentration, which increase rapidly as one approaches the source of a Gaussian
plume. The efficiency of the scavenging process is assumed to be independent of the
vertical scale of the plume. It is not apparent that the washout ratios used to estimate
wet removal are appropriate for estimating the below-cloud removal of Hg(ll)(9) and
Hg(part) from highly localized plumes, in which case the algorithm will likely
overestimate wet deposition fluxes in the near-field. The scavenging coefficients
should be modified to reflect below-cloud processes of importance near point
sources.
3.7 The Impact Of the Conversion of Elemental Mercury to Divalent Forms on
Exposure Characterization
The mercury assessment neglects the potential conversion of elemental mercury
to divalent forms in cloud droplets in both the local scale and the regional analysis; this
decision was made because it was estimated that any such conversions would happen
on a small scale and would have nominal impact on the ultimate exposure
characterization. Charge 8 asks the Subcommittee to evaluate the rationale for this
assumption, and to comment on the relevant uncertainties.
In estimating deposition, EPA states that it has chosen to neglect the potential
conversion of elemental mercury to divalent forms in cloud droplets in both the local
scale and the regional analysis. Many caveats re this approach are stated, but few
uncertainties are shown ( e.g., some modeled fluxes are given to 4 significant figures.).
The models used do include a form of Hg°- Hg2+ conversion, but ignore the
back reaction. This is acceptable for clouds with sufficient soot, and the model is
probably reasonable for most of the U.S. The model ignores cloud aqueous chemistry
23
-------
for non-precipitating clouds. However, the background levels of aerosol Hg are very
low and not highly significant in Hg dry deposition. Use of background levels of aerosol
Hg in the wet deposition model is appropriate.
In addition, dry deposition of Hg° is ignored in both near and far field models,
and publications on dry deposition of Hg° were misinterpreted making the application
of the alternate Vd in Table 6-39 invalid (Volume III p. 5-26). The model of annual
mean air concentration is inadequate to estimate plant exposure and uptake because
brief spikes control deposition. The compensation point is confused with no uptake,
also as being a fixed value. The modeled internal plant Hg concentrations (Fig. 6-8)
are low compared to measured data. There are 3 detailed field studies published on
the dry deposition of Hg, (Sweden, Lake Champlain, Walker Branch) and all show dry
greater than wet (Rea etal., 1996, Iverfeldt etal., 1996, Lindberg, 1996).
3.8 Assumptions Regarding Equilibrium among Mercury Species in Soil and
Sediments
The Agency attempted to modify existing approaches to reflect the most recent
science on estimating the temporal distribution of mercury concentrations in soil.. In
Charge 9, the Subcommittee was asked to comment on the assumptions regarding
equilibrium among mercury species present in soil.
The Subcommittee does not find the draft report's assumptions about the
speciation of Hg in soils, and more importantly sediments of lakes and wetlands, to be
appropriate. The major species are inorganic Hg", Hg° and MeHg. Production of
MeHg is the key step in bioaccumulation of Hg in food chains; production of Hg° from
Hg(ll) is an important loss mechanism to soils and source term to the atmosphere.
Rates of both reactions vary substantially among ecosystems, as functions of both
ecosystem biogeochemistry and Hg loading (Watras etal., 1994; Henry etal., 1995; St.
Louis etal., 1996; Krabbenhoff etal., 1992; Gilmour etal., in press).
However, EPA chose to model MeHg as a constant percent of total mercury
across all soil types and across all surface waters. The fraction of Hg as Hg° was also
set to a constant for all soils and natural waters. The report states clearly that this
approach was chosen because of the complexity of the methylation processes, and the
idea that MeHg production could not be adequately modeled. However, this simplistic
approach is not appropriate, given the current level of understanding of this process.
The predictive abilities of the IEM2 fate and transport, and the human and wildlife
exposure models are weakened to the point where they have little predictive capability
because the interecosystem variability in MeHg production is not considered.
The natural differences in MeHg production among ecosystems are at least as
great as the differences due to Hg deposition. Considerable information is available in
24
-------
the literature and should be reflected in the EPA's report (Benoit et al., submitted;
Watras etal., 1994; Driscoll etal., 1994; Hurley etal., 1995). The lack of use of
available information on the Hg speciation in soils and waters is thus a major
shortcoming of this report. During the public review meeting, EPA stated that fate
and transport and exposure modeling in the Hg report was done to provide information
on Hg exposure that is "not available in the data." However, because the factors that
affect Hg methylation and reduction were neglected in the models, the models actually
provide much less information on MeHg production, bioaccumulation exposure than
does the current literature.
A more sophisticated and up-to-date approach to modeling Hg methylation and
reduction in soils and sediments should be taken, making use of the available
information on the control of methylation and reduction processes. In fact, very
significant progress has been made in understanding MeHg production in watersheds
and water bodies in the last few years (much of it funded by EPA and the Electric
Power Research Institute), and that information should be used by EPA in this report to
provide improved estimates of Hg exposure and the potential to reduce that exposure
through source regulation Gilmour et al., 1992; Gilmour et al., in press; Driscoll et al.,
1994; Lee etal.,
1995). Nevertheless, many uncertainties will remain and will, for the most part, prevent
accurate assessment of source-receptor relationships.
Certain types of aquatic ecosystems are sensitive to high levels of MeHg
production and bioaccumulation including wetlands (e.g. St. Louis etal., 1994, 1996;
Hurley etal., 1995; Krabbenhoft, etal. 1995; Lee etal., 1995; Branfireun etal., 1996;
Gilmour etal., in press; Heyes etal., 1996); new reservoirs (Bodaly etal., 1984;
Jackson, 1988; Kelly, etal. 1997) lakes impacted by acid deposition (Weiner, 1988;
Weiner etal., 1990; Gilmour etal., 1992; Driscoll etal., 1994; Watras etal., 1995),
lakes with high levels of disolved carbon (Watras et al., 1995) and lakes with anoxic
hypolimnia (Henry etal., 1995). Fish in smaller lakes may accumulate more Hg than
fish in larger lakes of the same type (Bodaly et al., 1993), possibly because warmer
average sediment temperatures favor higher rates of MeHg production.
Among ecosystems not impacted by point sources of Hg, total Hg concentration
is not a good predictor of MeHg in either water or sediment (Kelly etal., 1995).
However, in lakes located near aquatic Hg point sources, MeHg concentrations in
water and sediments, and Hg in fish, are generally elevated. Elevations in MeHg in
water, sediment and fish are not proportional to the increase in total Hg concentration
in these systems (Parks etal., 1986, 1989; Henry, etal. 1995).
A number of recent studies have examined concentration patterns in estuaries
(Coquery and Martin, 1995; Cosa and Noel, 1987 ). Estuaries and coastal waters
appear less sensitive to MeHg production and bioaccumulation, probably due to sulfide
25
-------
inhibition of MeHg production (Compeau and Bartha 1983, 1985; Craig and Moreton,
1983; Gilmour etal. 1992; Henry, 1992; Choi and Bartha, 1994; Benoit ef a/in press).
The location of MeHg production within ecosystems has been fairly well defined.
The interface between oxic and anoxic conditions (0/A boundary) in sediments
(Gilmour et al., 1992; Gilmour et al. in press; Krabbenhoft et al. in press) or in the water
column (Watras etal., 1995) is a region of intense methylation activity because of the
high availability of substrates (both organic matter and sulfate) for sulfate reduction in
this zone. Within lakes, the 0/A interface may move throughout the year as oxygen is
depleted and replenished. Areas where S is rapidly reduced and reoxidized seem to
be particularly important. For example, the top of the hypolimnion, where green and
purple S bacteria supported reoxidation of sulfide, can be a very active area of
methylation (Watras et al., 1995). Littoral and epilimnetic sediments are often the
favored site for methylation within lakes (Gilmour et al., 1992; Ramlal et al., 1993;
Krabbenhoft et al. in
press) because of higher average temperatures and more available substrate than
deeper sediments.
The first mass balance for Hg in was completed in the early 1990's using a
hydrologically constrained seepage lake, Little Rock Lake (LRL), Wisconsin (Weiner et
al., 1990; Watras etal., 1994). The study included production of a sophisticated
biogeochemical model (with minimal hydrology) parameterized from the experimental
studies in the lake. The LRL study was also important in demonstrating the importance
of acid deposition on Hg methylation and bioaccumulation, a factor that has not been
considered in the EPA report. Work in LRL was later extended to other lakes and
watersheds in the region - Pallette Lake (Hurley et al., 1994; Krabbenhoft et al. in
press); Allequash Creek (Krabbenhoft etal., 1995); and interlake comparisons (Watras
et al. b and c).
It is thus critically important to bring out the importance and inter-system
variability of methylation in this report. Specifically, it is important to discuss and
emphasize the methylation process, including intersystem variability in the report's
executive summary; to detail the state of our understanding of methylation in Vol. Ill; to
model (as well as possible) MeHg production and exposure among different
ecosystems types using state-of-the-art transport and biogeochemical models in Vol.
Ill; and to use the resultant information to model and discuss MeHg exposure to wildlife
in Vol V.
3.9 Biogeochemical Issues in Modeling
EPA's Charge question 10 comprised a wide range of six separate questions on
modeling the behavior of mercury in the environment and its uptake by plants. These
questions addressed the selection of the model input parameters (such as the soil-to-
26
-------
water partition coefficients); the estimation of the watershed and water body fluxes;
estimation effluxes in large lakes, rivers and other large water bodies also have been
estimated; the treatment of the uncertainty surrounding assumptions about plant uptake
of mercury directly from the atmosphere; and the likely contribution of mercury from
plants to soil.
The following discussion addresses the various specific questions in logical
combinations.
The Subcommittee has significant reservations about the selection of model
input parameters for the soil-to-water partition coefficients. Watershed type is
extremely important in regulating metals transport to lakes. Land use/land cover
characteristics can greatly affect fluxes and yields of Hg to a lake (St. Louis, et al.,
1994, 1996; Hurley etal., 1995). The IEM-2 model presented is a high-quality
watershed model, attempting to evaluate the effects of Hg deposition on partitioning
and transport to a lake. The model is mainly soil-based and represents a watershed
with untilled soils. While this approach can be used for upland soils, there is
increasing awareness in the literature of the influence of wetlands on Hg cycling. In
several regions of the U.S. and Canada where fish advisories exist for mercury,
wetlands constitute a significant portion of the watershed. Recent evidence has clearly
indicated that wetlands are zones of MeHg production(St. Louis, et al., 1994, 1996;
Hurley etal., 1995; Kelly etal., 1997). The modeling of watershed effects on Hg
bioaccumulation, therefore, should evaluate contrasting watershed types. The model in
the report only compares an eastern and western lake system that is dependent upon
soil-water partitioning in the watershed. The IEM-2 fate and transport model used in
the draft report is not designed to model transformations among Hg species (e.g.,
methylation and reduction). As suggested above, a more rigorous comparison
using simulations with different watershed characteristics, would be a better
approach to assessing impacts of deposition on aquatic systems.
The lack of published studies prior to 1994 with respect to soil-water partitioning
led the authors to use a similar coefficient for Hg(ll) and MeHg. The K,, used for soil-
water is appropriate for the soil setting but is insensitive to any biogeochemical
processes that may occur (for example, soil flooding or anoxia). Similarly, the
sediment-water partition coefficient reflects similar values for Hg(ll) and MeHg. A
strength of the modeling effort was the calibration of chosen partition coefficients to
existing databases.
The review also raised some issues regarding the estimation of the watershed
and water body fluxes. The important partition coefficient for bioaccumulation is the
K^, the suspended sediment-water coefficient. The test simulations indicated that this
particular coefficient was insensitive to calibration test. Discussions with the EPA
27
-------
modeling group at the public meeting revealed that the low suspended solids
concentrations used for the simulations was the main cause for insensitivity. This
further strengthens the argument that a more rigorous water column model needs
to be used to assess aquatic effects. Assessment of particle types (i.e., soil vs.
phytoplankton) and amounts would provide additional important bioaccumlation
simulation comparisons.
It is unfortunate that soil and water loss degradation constants were not
incorporated in the model. Several recent studies have shown that Hg° production and
evasion are common processes in soils and surface waters. Similarly,
photodegradation has been shown to affect MeHg concentrations in surface waters.
The approach of using
a set fraction of Hg species in lake waters (i.e., 15% as MeHg) is better replaced
by a valid geochemical model, as mentioned above.
EPA did not estimate fluxes in large lakes, rivers and other large water bodies.
The Subcommittee does not, however, consider this to be a major problem. Rivers can
be extremely important vectors for trace metal transport to lakes but modeling such
efforts is undoubtedly beyond the scope of this report. Incorporation of large lakes and
rivers complicates the assessment of atmospheric inputs to bioaccumulation because
of the confounding effects of point-source discharges. Instead, tables in volume III
should be expanded to include Hg and MeHg concentrations from selected river
studies.
The Lake Michigan Mass Balance Model, funded by the EPA, should provide the
first detailed large-lake Hg cycling model. The study quantified atmospheric and
tributary inputs to the system and coupled this information with in-lake assessment of
aqueous, sedimentary and biotic pools. This model, when completed, cannot treat
riverine watersheds in the way that the IEM-2 model can, due to the effects of
contaminated zones in rivers and harbors of Lake Michigan.
Sections of the draft report addressing assumptions about plant uptake of
mercury directly from the atmosphere are adequate for the report's purposes, but could
be improved. Most of the studies cited were conducted prior to 1990. The air-plant
BCF (bioconcentration factors) was taken as the midpoint of the Mosbeak et al. (1988)
study for a variety of plant types. However, several studies have been conducted more
recently to assess accurately the effects of foliar uptake and release of Hg (Cocking, et
al., 1995; Hanson et al., 1995; Munthe et al., 1995). Detailed studies in specific foliar
settings are being conducted to assess dry deposition and through fall of Hg. Since the
model presented did not evaluate the effects of plant types on watershed processing of
Hg, the effort is probably adequate.
Finally, the Charge asks if the likely contribution of mercury from plants to soil is
28
-------
great enough to require adding such sources to the input for future models. The
Subcommittee recommends that a statement be incorporated into the report's
text that specifies that both plant-to-soil and soil-to-plant pathways can represent
important Hg transport processes. Mass-balance studies that incorporate "through
fall" in forest canopies have indicated that dry deposition onto plant surfaces can
account for significant inputs of Hg to terrestrial systems. Similarly, evasion of Hg
through the canopy can be an important loss mechanism and that decaying vegetation
can be a significant source of Hg to soils.
3.10 Uncertainties in Atmospheric Modeling
Charge 11 asked for comments on a number of qualitative conclusions drawn
from the results of the atmospheric modeling; in particular, the Agency was concerned
about the effects of uncertainty.
A number of significant qualitative conclusions are drawn from the results of the
draft report's atmospheric modeling. Chief among these are:
a) mercury species from combustion sources are transported over relatively
great distances
b) mercury is deposited to soil and terrestrial vegetation, but at levels that do
not result in human exposures likely to be detrimental to health though
terrestrial exposure pathways
c) significant inputs of mercury in lakes and ponds result from both direct
deposition of mercury compounds into lakes and ponds and run-off inputs
of mercury compounds from the watershed land surface following land
deposition
The Subcommittee considered three issues relevant to the question of whether
uncertainties are adequately described: a) were the important uncertainties listed; b) is
consideration given to how the qualitative conclusions could change as a result of the
uncertainties; and c) what questions cannot be addressed because of the
uncertainties?
As far as the Subcommittee could determine, the draft report does provide a
fairly complete list of the important sources of uncertainty in the atmospheric model.
However, the significance of these uncertainties and strategies for confronting these
uncertainties in reaching conclusions are not addressed and should be. Although
uncertainties and lack of data are discussed throughout the report, both the Executive
Summary and other parts of the document need an emphasis on the fact that significant
information is missing regarding the biogeochemistry of mercury species and that the
29
-------
absence of this information limits the reliability with which any mercury control program
can be evaluated.
It should be recognized that, in its current form, this document contains
quantitative modeling for making order of magnitude estimates, but provides only
qualitative treatment of uncertainty. There is no explicit quantification of uncertainties.
We recognized that it may not be possible at this point in time to carry out a detailed
uncertainty analysis. However, even without a quantitative uncertainty analysis, an
effective way to communicate uncertainties is to state clearly: a) all the key
assumptions made in the analysis; b) missing information; and c) how robust are the
conclusions relative to the assumptions. Perhaps the most important issue in this
list is whether the conclusions are invariant with respect to parameter values that
are missing or to the value ranges of parameters that are uncertain or have
incomplete data. The uncertainties should be ranked in terms of their relative
impact on the conclusions.
3.11 Estimating the Bioaccumulation Factor (BAF) for the Aquatic Chain
Charge 12 sought comment on the BAF developed by the Agency, and its
utilization in the models.
The Subcommittee agreed that the generation of the reference dose as currently
presented for wildlife is flawed. The Wildlife Criteria (WC) was derived from a
bioaccumulation factor (BAF) based on total Hg rather than methylmercury (MeHg),
which is the species that bioaccumulates. As a result, the magnitude of the error in the
BAF is associated with, and reflects, the wide variability in MeHg concentrations among
ecosystems. This is inadequate to make predictions about the effects of Hg on wildlife,
or the exposure of humans to non-farmed fish. The Subcommittee feels that EPA
should calculate BAF and WC based on dissolved MeHg in water, and that BAF
calculations be specific to ecosystem or water body type.
The draft report describes three approaches for calculating bioaccumulation
factors (BAF). Each of these approaches represent state-of-the-art techniques for
deriving a BAF, and the Subcommittee supports the preference of BAFs over BCF.
The report clearly (and correctly) points out that field derived BAFs are preferred
because they reduce the uncertainty in the number. BAF3 x PPF4 was chosen as the
preferred method because it was based on field data for the BAF3 and because the
results showed less variability. The disadvantage to this approach is that the PPF
multiplier is a fixed value. This value can vary from site to site and therefore utilizing a
single PPF introduces uncertainty.
The uncertainty analysis conducted by the Agency was adequate although
it did not include uncertainty that is due to analytical error. Understanding the
30
-------
extent to which mercury is accumulated by various aquatic species is central to both
the human health and ecological risk assessment. Unfortunately, both calculated and
field measured BAFs show a great deal of variability in the numbers for the same
species. The data shown in the draft report's Table 4-1 show that the field-derived
BAF3 and BAF4 can vary by as much two orders of magnitude when comparing 5th and
95th percentiles. The variability is attributed to site specific parameters which control
the extent of mercury methylation as well as sorption of mercury to suspended particles
and dissolved organics. One additional factor which should be included in
deriving BAFs is fish age or size. Calculation of BAFs for methylmercury as well as
for total mercury would be useful and might help reduce the uncertainty.
For purposes of aquatic risk assessment, geometric mean BAF values
were used, which is appropriate. However, it should be pointed out that the
potential to overestimate or underestimate the BAF is significant and that the
error may be an order of magnitude or more. This amount of error is sufficient to
make a difference between risk and no risk. This amount of variability, in part, helps
explain why some populations of sensitive aquatic species do not appear to be affected
in the field while others may be affected.
3.12 Modeling the Terrestrial Food Chain
In Charge question 13, EPA asked for an evaluation of the models and modeling
conducted for the terrestrial food chain.
The EPA's model does not appear to provide WC for mercury that
approximate threshold (effect) values and hence the utility and accuracy of the
value is questioned. The Subcommittee remains supportive of the approach
used to derive the WC, but has reservations concerning the value derived for
mercury and its subsequent universal application. Use of site-specific data in
WC models, however, is supported. The SAB has, in previous reviews, supported
the development of criteria for wildlife and continues to do so. However,
reservations about model formulations have been expressed and still exist. The
values provided by the model for several species are in the range of 0.2-0.4 ng/L (ppt)
HgT. These values are a factor of 2-5 below levels in many lakes (1.0 ng/L) that have
healthy populations of wildlife species including those selected as most at risk for
mercury exposure. Derivation of WC as shown in the Report is overly conservative and
not particularly useful for risk assessment. The value may be thought of in the same
manner as a reference dose, that is, a concentration that is calculated to provide an
adequate margin of safety for all species potentially at risk. Ecological risk
assessments are based on comparison of threshold values with exposure values. The
WC do not appear to represent a reasonable threshold value. Therefore, it is
recommended that they not be used for risk assessment purposes. The WC
values appear to be overly conservative due to the two uncertainty factors and
31
-------
the large BAFs used in the model. The model input parameters, including those for
methylmercury, need to be further evaluated.
3.13 Characterization of Wildlife Effects
Chapter VI of the draft report presented the Agency's current mercury risk
characterization of wildlife effects. In Charge 35c, the Subcommittee was asked to
comment on the appropriateness and environmental significance of the
characterization.
The Agency identified the most significant laboratory and field toxicology studies
which evaluated the effects of mercury exposure on wildlife. The findings are clearly
presented in Volumes V and VI. These data were summarized to produce dose/
response relationships to methylmercury for wildlife species. The establishment of an
avian Hg Lowest Observed Adverse Effects Level (LOAEL) of 0.064 mg/kg-day (0.5 //g
Hg/g diet) is justified on the basis of Heinz's three generation study with mallard ducks
(Heinz, 1979). However, an avian No Observed Adverse Effects Level (NOAEL) was
not established. The Agency correctly established a mammalian LOAEL of 0.16
mg/kg-day and NOAEL of 0.05 mg/kg-day on the basis of Wobeser's findings of
nervous tissues lesions in mink(Wobeser, 1973; Wobeser etal., 1976a; Wobeser etal.,
1976b).
The methods used to calculate a wildlife criteria (WC) for mercury, using a
wildlife reference dose and methods described in the Proposed Great Lakes Water
Quality Guidance, are compromised by data deficiencies and numerous uncertainty
factors in the model. The uncertainty factors, questionable application of BAF
ratios, and incomplete data on diet habits of target species produce WC for
mercury which are not scientifically defensible. The Agency should recognize that
wildlife Hg exposure (as measured by egg, brain, muscle tissue Hg concentrations)
differs by a factor of ten between most aquatic systems in North America and that the
dose-response curve for methyl-Hg is unusually steep. Applying uncertainty factors
of 9-10 times to the test dose (TD) produces a reference dose which is so high
that it is not likely to be encountered by most free-ranging piscivorous wildlife in
North America. Furthermore, other Members of the Sources Workgroup noted at the
public meeting that water column total Hg concentrations are not predictive offish
methyl-Hg concentrations, an assumption made by the WC model. Consequently, we
recommend a BAF based on water column methyl-Hg concentrations and site
specific BAF characterizations. Additional study of target species dietary habits
is also needed (especially mink) and the inclusion of additional target species
(i.e. common loon) is recommended. The bottom line is that the Agency's own
32
-------
estimate of WC for mink (Mercury Report to Congress, Vol. VI and Great Lakes Water
Quality Initiative) ranges from 400 - 4000 pg/L. The same data sets were used to make
the estimates, only the assumptions differ. This range of water column total Hg spans
nearly the entire range currently measured in North American aquatic systems.
However, the Subcommittee believes that piscivorous wildlife are at risk to
elevated mercury exposure and toxic effects in the areas of concern identified by
the U.S. EPA. For instance, Hg concentrations in piscivorous wildlife tissue are greater
in the areas of concern than in other regions of North America. Also, egg Hg
concentrations of avian piscivores in the upper Great Lakes, mid Atlantic coast, New
England, and Florida exceed levels associated with reproductive impairment. MeHg
intake rates of common loons, osprey, and otter likely exceed the avian and
mammalian LOAEL in many aquatic systems within these regions. Fish contaminant
databases (developed for human consumption advisories) can be accessed to estimate
Hg concentrations in many prey species consumed by wildlife. These estimates should
be used to establish risk levels (in terms of probabalistic estimates incorporating
confidence limits) for target wildlife species in the areas of concern.
The Agency should be cautious when using feather and liver total Hg
concentrations as toxicological benchmarks. Interpretation of feather Hg
concentrations as an index of exposure requires knowledge of feather type, molt
pattern, and individual age. Interpretation of feather Hg concentrations as a
toxicological benchmark for exposure is complicated by the fact that feathers are sites
of methyl-Hg elimination and sequestration. Thus, feather Hg concentrations of
piscivorous wildlife are expected to be greater than non-piscivores, even in the
absence of toxic effects. Also, liver total Hg concentrations need to be interpreted with
care as it has been shown that most liver Hg in piscivores has been demethylated and
poses less risk than does methyl-Hg. Conversely, methyl-Hg is the primary form in
egg, brain, muscle, and blood, which likely provide more meaningful toxicological
benchmarks of exposure.
3.14 Evaluation of Wildlife Effects
Charge 36 asks the Subcommittee to evaluate the section of Volume V
describes fairly significant wildlife effects (which are attributed to elevated mercury
concentrations in some ecosystems). In particular, they sought comment on the
scientific underpinnings of the evaluation, including its consistency with EPA's
ecological risk guidelines.
The Subcommittee does not believe Volume V, in total, is based on the
best available and sound science. To a large degree this chapter follows EPA
guidelines for ecological risk assessment. However, it deviated in terms of
33
-------
quantifying the risk. This is a critical component of risk assessment.
Risk assessment is the process of coupling exposure (or more explicitly, dose)
and response (effect). It is the imputed relationship between biological effects and
exposure to the agent(s) of concern. The draft report uses predicted exposures from
computer models as the primary bases for exposure estimates (dietary mercury). This
is not adequate for state-of-the-art risk assessment. Site specific monitoring data
are needed (and as noted in the draft report, are available for many sites). These
data could be used to provide more scientifically defensible examples of exposures at a
few selected sites, for which risk quotients could be calculated to quantify the
exposure/effects relationship (risk) at specific sites. Using this approach, one can
define the potential for effects to occur at various sites and can perform probabilistic
assessment of the potential for exposure to exceed designated threshold effect levels
for species most art risk. These techniques are described in detail in SETAC (1994)
and Solomon etal. (1996).
Volume V could be greatly improved through the use of calculated risk quotients
for specific sites and species most at risk. There are sufficient data to do this for
several areas and species. This requires calculating a threshold values from chronic
laboratory feeding studies for the species of interest (e.g., threshold value is calculated
as the geometric mean of the LOEL and NOEL). The threshold value can then be
divided by the level of mercury assumed to be in the diet of a given species and a risk
quotient for a given area is derived. An example follows for loons using the chronic
data from a three generation feeding study with mallard ducks (Heinz, 1979) to derive a
threshold value. The LOEL in this study was 0.5 ppm; a NOEL was not obtained, but it
could be estimated by dividing the NOEL by 3 providing a value of 0.17 ppm. The
geometric mean of the two values is 0.29 ppm. This value could then be used to
calculate risk quotients for areas where dietary mercury is calculated from site specific
mercury concentrations in fishes eaten by loons.
Care has to be taken in comparing mercury levels in tissues of animals that died
or were collected in the 70s and 80s with samples taken in the 90s. Atmospheric
releases of mercury appear to have declined somewhat over the past five years.
Additionally, site specific characteristics are important in determining mercury
methylation, fate, and transport; care should be taken in comparing site specific data
and drawing conclusions relative to levels resulting from atmospheric deposition based
on estimates derived from RELMAP.
In addition to these general comments on the wildlife sections of the draft report,
the Subcommittee has identified some specific issues worth noting. These include:
a) Table ES-2 shows that <1 % of the panther's range overlaps mercury
areas of concern. This suggests that the potential for exposure to exceed
34
-------
the effect level (i.e., risk >1.0) is low. This should be stated clearly.
Conclusions drawn relative to the panther appeared to be based on two
unpublished (non-peer reviewed reports). Although inclusion of these
reports is appropriate, there are insufficient data to judge dose-response
and significance of the field residue panther data. The data are anecdotal
in nature, although indicative of the potential for increased exposure and
increased risk to panthers from methyl mercury. Over reliance on the
raccoon pathway for exposure is placed in this report. Multiple avenues
exist for the panther, one of which is the raccoon. Concentrations of
mercury in raccoons, may not always exceed those in other species. Risk
statements could be drawn for panthers through the calculation of risk
quotients for panthers using site specific data for the diet and a threshold
effect level for cats based on a chronic feeding study (summarized
elsewhere in this report).
b) Page ES-5 lists the five species identified as having the greatest potential
to be at risk , along with the percent of the species' range that overlaps
mercury areas of concern. This is a good start towards a risk statement.
The report should complete the presentation by providing risk
calculations, i.e., effect levels divided by exposure levels. The
statements on this page infer risk to the various species without
presenting risk calculations. This is a serious short-coming to this
report.
c) Consistency as to comments on observed effects in the report is needed.
Page ES-5 reports that Bald Eagle populations are improving in five
states, yet the draft report later also infers risk to bald eagles from
mercury and indicates that the Great Lakes is one of the areas of highest
concern. This is a key area where eagle populations have been
improving over the past decade in spite of mercury levels in fish. Page
ES-11, however, indicates that "field data suggests that bald eagles have
not suffered adverse effects toxic effects due to airborne mercury
emissions." The report paints a biased picture without taking
advantage of the opportunity to test or verify the assessment with
the analyses of field populations in the extant literature. Once again,
risk is inferred without a real risk calculation.
d) Page ES-8 indicates that "otter population declines do not overlap to a
large extent with regions of concern; however, the area of decline does
coincide with RELMAP predictions of high mercury deposition rate." It is
apparent from these statements that the reasons for otter declines are not
clear, yet risk is inferred without definitive risk calculations. A critical
factor in risk assessment is the co-occurrence of the exposure and
35
-------
effects. The data from RELMAP suggest that there may be increased
exposure in areas where otters live, but this is strictly an inference and
site specific/region specific monitoring data needs to be used. The
Agency should point out the weakness and lack of information to
draw definitive conclusions about risk to otters or make the case
based on risk calculations. The inference of risk based on exposure
estimates overstates the case.
e) Page ES-10 (second paragraph) states "...although causal links have not
been established , mercury originating from airborne deposition may be a
contributing factor to population effects on bald eagles, river otters, and
mink. " Without a causal relationship, the above statement is
unsubstantiated and does not suggest state-of-the-art ecological risk
assessment. There are many factors affecting otter, eagle and mink
populations, especially, the expanding human population in the U. S. (and
trapping in the case of mink). The single biggest factor resulting in
species loss in the U.S. is now recognized as habitat loss and habitat
conversion due to human intervention.
f) Page ES-11, in several places, infers risk due to elevated mercury
exposure levels (levels in the diet). Exposure by itself does not infer risk,
it must be accompanied by an equally rigorous effects assessment and
risk characterization. State-of-the-art risk assessment is not based on
inferences. It is acceptable to draw conclusions, but they must be
qualified as to the strength of the data on which they are based. Also,
note that on page ES-11, lines 17-19, it is stated that "...field data are
inconclusive to conclude whether the mink, river otter, or kingfisher have
suffered adverse toxic effects due to airborne mercury emissions." When
this statement is taken together with a similar statement on ES-10
relative to bald eagles, one can conclude that field data
demonstrating effects on the species most at risk are lacking. This
is a key statement and should be brought forward for Congress to
evaluate.
g) Page ES-13 (last line) incorrectly states that"... the designation of an
area as a region of concern implies an increased risk of mercury toxicity
to wildlife." More correctly, designation of an area as a region of concern
implies an increased potential for exposure to mercury for species that
co-occur with the exposure. The consistent theme here is that exposure
does not equate to risk!
h) Page 2-26 (2.3.3, second sentence) states "...the species are at high risk
of mercury exposure and effects because they either are piscivores or eat
36
-------
piscivores." This statement again infers that exposure = risk.
Additionally, it states that the species are at high risk of both exposure
and effects. This is a definitive conclusion without a risk calculation (risk
quotient). On the other hand, one might more appropriately conclude, on
the basis of the data presented, that elevated tissue residues can result in
an increase in exposure for the species of interest and that this exposure
could lead to effects if effect thresholds are exceeded. Better yet, if risk
quotients were calculated, one could make a definitive statement about
the apparent risk
relative to other species and other areas where mercury exposure levels
and risk quotients are lower.
I) Page 2-28 (lines 5 and 9), page 2-27 (lines 24, 25, and 27) report units
for Barr (1986) as ug/Kg (ppb), whereas they should be ug/g (ppm).
j) Page 2-28 (third paragraph) states "The viability of loon populations
within their traditional habitats in the United States is unclear. None of
the studies reviewed was able to demonstrate clear population declines
on a regional or national level." This sentence should be brought forward
to the Executive Summary for Volume V. It makes a clear summary
statement and indicates that population trends are uncertain. Put in
context with the sensitivity of loons and their potential for elevated
exposure, further field evaluation appears warranted. It also appears that
more definitive risk statements could be made which would show risk to
loons if risk quotients were calculated (chronic threshold bird values for
fish in the diet/dietary mercury; e.g., mallard duck chronic effects
threshold value divided by mercury in the diet of loons at specific sites).
3.15 Linking Emissions and Wildlife Effects
The draft report makes the inference that exposure to mercury emissions is
related to reproductive effects and neurobehavioral changes in fish consuming birds
and mammals. These effects have been documented in the Great Lakes and in the
Southeastern United States. EPA's models predict that regional hot spots (relatively
high concentrations of methylmercury) would occur in these same areas. In question
37, the Subcommittee is asked to comment on the soundness of the analysis of the
evidence linking emissions and effects.
The effort to link emission inventory to effects in piscivorous birds and mammals
is a leap of faith with little actual scientific evidence to substantiate any correlations.
The draft report recognizes that there is still considerable uncertainty as to the transfer
mechanisms and rates of mercury from source to the target species. The following
37
-------
major gaps in knowledge and/or model assumptions make it difficult to draw meaningful
inference between source and effect:
a) Most of the mercury is emitted in inorganic form and there are large gaps
in our understanding of the environmental movement and deposition of
airborne mercury into aquatic ecosystems. The model ignores any
re-emission processes which can amount to a large fraction of the
deposited mercury.
b) Processes and rates involved in converting the inorganic mercury to
methylated forms are only partially understood, although the current level
of understanding is not reflected in the draft report.
c) The use of BAF to quantify the relationship between the dissolved
mercury concentration in the water column and methylmercury
concentration in fish does not seem to have much scientific justification.
The BAF data used have a number of major limitations (nature of Hg in
water, detection limit, and contamination artifacts).
3.16 "Hot Spots" and Wildlife Effects
Charge 38 deals with mercury concentration "hot spots" and associated wildlife
effect (or lack of effects). EPA wanted to know if any "hot spots" have not shown
evidence of wildlife effects, and if predicted "hot spots" correlated with methylmercury
levels in fishery products.
The wording of the question on "hot spots" and wildlife effects pre-supposes that
deposition hot spots are, in fact, reflected in wildlife Hg concentrations and effects.
This also implies that in areas of low Hg deposition, we would not see wildlife
impacts/or high fish concentrations. The actual situation is much more complex for
several reasons. First, there is a small range (3x) in Hg deposition between the least
and most impacted regions, while the effects on methyl Hg production caused by in situ
factors (pH, SO/2, watershed area, DOC, etc.) are much greater (>10x). Additionally,
although it has not been demonstrated, anecdotal evidence suggests that the
relationship between inorganic Hg input and MeHg production is not linear: i.e., as the
total Hg increases by a factor of 10, methyl Hg increases perhaps only by a factor of
two. This can be seen by looking at contaminated sites. For example, in the Lahonton
Reservoir (NV) system, total Hg is 1000 ng/L, while piscivorous fish have 5 ppm MeHg
by comparison; lakes in Wisconsin may have 1 ng/L Hg, but the piscivorous fish have
0.5 ppm MeHg. Onondaga Lake (NY) is in-between, having 25 ng/L total Hg, and fish
with 1 ppm MeHg. Or, looking at deposition, we see that in Elephant Butte Reservoir
(NM), an area with < 5 ug/m2 deposition, piscivorous fish have 0.5 ppm MeHg, similar
38
-------
to Northern Wisconsin, which has 10 ug/m2, and upstate New York, which has 15-20
ug/m2 (Driscoll etal., 1994; Watras etal., 1994; Henry etal., 1995)
This is not to say that deposition "hot spots" have no impact on wildlife
(the Subcommittee agrees that they might), but that a myriad of biogeochemical
factors are at work in real systems which have not been incorporated into the
model. It is not unlikely that each of the important factors (pH, DOC, SO/2, HgT,
watershed area, mean temperature, surface area to volume ratio, trophic state, etc.)
has a temporally different, and non-linear effect upon the methyl Hg production in a
given lake. Thus, aside from saying that MeHg will eventually go up if Hg(ll) deposition
is increased, we cannot say how much, or when. And, if two or more factors change at
the same time (say, Hg inputs go up, but SO/2 goes down), predictions become very
difficult. However, EPA has not taken advantage of the state of the science in the draft
report.
3.17 Wildlife Effects as Warning Signs of Humans Effects
EPA asked the Subcommittee, in Charge 39, to advise it as to what wildlife
effects (based on what metric) could be interpreted as harbingers of likely human
health effects.
The Subcommittee's consensus was that wildlife epidemiology cannot serve to
totally replace well defined laboratory studies (Kendall etal., 1995). In the laboratory
we are able to demonstrate, under controlled conditions, a true dose-range of exposure
to the contaminant, in this case mercury, in a variety of wildlife species and generate
critically needed dose-response curves. With appropriate laboratory species and well
developed dose-response curves an opportunity then exists to utilize this information,
even as developed on surrogate species, to better understand the potential exposure
and subsequent tissue concentration that may exist in wild species. It is possible that
surveillance of key wildlife sentinels, if appropriately chosen, may be extremely useful
in the future to better understand impacts from environmental contaminants (Sheffield
and Kendall, 1997). The generation of well designed laboratory studies to demonstrate
cause and effect under a range of appropriate doses integrated with field exposure and
perhaps effects measurements can be extremely important in the development of a
wildlife ecological risk assessment (Kendall etal., 1996).
Looking at mercury's effects in general, the classic case of "Minimata Disease"
in Japan demonstrated that mehylmercury can have profound effects on humans as a
teratogen and reproductive toxicant, as well as causing death in human and animals.
Other forms of mercury have also been found to be toxic in humans and animals. For
example, wildlife have also been killed by exposures to mercury compounds used to
treat seeds used in agriculture. .
39
-------
With sublethal exposure to mercury, a variety of toxicological endpoints may be
manifested in wildlife that also could occur in human beings (Kendall et al., 1995). First
of all, mercury can act as a reproductive toxicant, particularly as a teratogen which has
been demonstrated with laboratory studies in wildlife. There is also concern as verified
with Minimata Disease of teratogenicity with mercury exposure in humans. Mercury
may act as an endocrine modulator in wildlife species. (Lower and Kendall, 1990;
Kendall and Dickerson, 1996).
With regard to mercury acting as a reproductive toxicant in wildlife, of critical
interest is its influence on population statistics. It is quite difficult in the field to
measure the numbers of wildlife in their natural environment to understand their true
population ecology (Kendall and Lacher, 1994). It appears that several wildlife
sentinels, including wild mink and loons, may have experienced population declines
correlated with mercury exposure. Future field studies of other species may identify
additional species which could also serve as sentinels to identify reproductive and
population impacts related to exposure to environmental mercury.
Mercury is also a potent neurotoxicant. The "Madhatters Syndrome" has been
demonstrated in humans with severe neurotoxicological symptoms. In addition,
sublethally exposed wildlife have demonstrated neurotoxicological responses to
mercury.
Mercury existing as a strongly positive cation in vivo manifests its toxic action to
a large degree through the disturbance of various enzyme functions, particularly those
having sulfhydryl groups. In this regard, a variety of types of tissue damage,
particularly in the liver and kidney, may occur with enhanced mercury exposure.
3.18 "Social Costs" of Mercury Contamination
The draft EPA report (Chapter VII) previously contained a section on the "Social
Costs" of mercury contamination assessing the value of the fisheries in the U.S. as well
as other values, such as maintaining a healthy ecosystem. The intent was to balance
the discussion of mercury control costs. Some peer reviewers objected to the inclusion
of the Social Costs section partly because the impact of anthropogenic sources could
not be directly and quantitatively related to these impacts (e.g., the declining Florida
panther population). The section was consequently deleted. In Charge question 41,
the Agency asked the Subcommittee to comment on the advisability of including such a
study, and to advise it the EPA how to describe best the benefits of mercury reductions
if a) the impacts of such reductions are not directly quantifiable, and b) the monetary
value of such benefits are not easily quantified.
The term "social costs," as used in the context of this Charge question (41)
addressing the economics of Hg controls refers to the benefits of reducing mercury
40
-------
emissions from a societal point of view. The supplementary material provided to the
Subcommittee includes several pages deleted from a previous version of the mercury
report that described social costs. This section was deleted following criticism received
through an inter-agency review of the study. We agree that the deleted discussion
of social costs was inadequate.
Nonetheless, the report is sorely lacking for a discussion of social costs
(benefits). By including a discussion of the costs of reducing mercury
emissions, but omitting discussion of the benefits of emissions reductions, the
Agency fails to provide a basis for weighing costs and benefits of any measures
to reduce mercury emissions.
The control of mercury emissions is inter-twined with the control of other harmful
pollutants. The benefits of controlling other pollutants should not be forgotten in the
consideration of the benefits of mercury control. Even where the quantification and
monetization of some potential benefits associated with reductions in mercury
emissions is difficult, the description of benefits from the ancillary reduction in other
pollutants that is achieved through mercury control may be readily available. These
ancillary benefits should be described for consideration along with other benefits and
with the costs of mercury control (McConnell, 1990).
Environmental benefits are often difficult to quantify, and indeed, the physical
effects of pollution changes are often difficult to state precisely. Consequently,
wherever possible, scientific analysis should characterize the effects of potential policy
in probabilistic terms, rather than as point estimates. The state of the art in integrated
assessment calls for the propagation of these probabilities through an entire model
linking the source of emissions, their various environmental pathways, and endpoints
that are of concern to humans and to the policy process. The result will be probability
distributions describing the impacts of changes in emissions that can be combined with
available information about willingness-to-pay for these changes, to provide a bound
and range of potential benefits. Such analysis also is of tremendous use in exploring
the value of additional information by identifying research needs that will yield the
greatest "bang for the buck" through reducing uncertainties about the benefits of
emission reductions.
The organization of a comprehensive integrated assessment along these
lines is not achievable in a time frame that is available for revisions to the
Mercury Study. However, as discussed below, a framework for such analysis
should be included in the Study to guide further research efforts by the Agency.
It is not possible at the present time to describe the benefits of reductions in
mercury emissions in monetary terms in a reliable fashion. In place of monetary
41
-------
estimates, a rigorous framework for the consideration of benefits could be articulated.
It should take advantage of the best scientific information available for the description
of those benefits, and as importantly, describe how such benefit estimates could be
achieved in the future. These benefit comparisons occur at the margin. Hence a
discussion of the total value of an environmental asset or resource (such as was
contained in the deleted material from a previous version of the Mercury Study) may be
useful background but it is not relevant to the analysis of small changes in mercury
exposure. Also, the framework should identify the various endpoints that in principle
could and should be valued and it should identify the proper methodology for doing so
in each case. The benefits discussion should be accomplished in the time frame
available before release of the final Mercury report.
For the case of mercury, the evaluation of social costs is particularly challenging
because there is not a single peer- reviewed study that has attempted to evaluate
specifically the social costs of mercury pollution. There is likely to be disagreement
about the degree to which related valuation studies provide useful information pertinent
to understanding issues regarding mercury. However, the existing literature on the
valuation of related benefit endpoints should be organized and brought to the fore.
There are other pollutants that pose generally similar threats to the environment or
public health as mercury, yielding effects such as delayed child development,
neurological effects in adults, and threats to species diversity and habitat. In
supporting analysis for the regulation of lead, for example, these types of subtle effects
have been studied for the purpose of economic valuation and contributed to the
promulgation of regulations. Although lead is a very different toxicant than mercury,
valuation estimates for these subtle lead effects may provide some useful information
to describe the likely magnitude of economic benefits of reduced exposure to mercury.
More importantly, however, they provide immediately useful information for an
evaluation of the methodologies available for estimation of the benefits for these
type of effects, and they would help to establish priorities for research aimed
specifically at valuation of benefits for mercury.
Recent surveys that provide helpful background include Lee etal., 1995;
Haggler Billy Consulting, Inc. 1995; and the European Commission, 1995.
The review of costs of mercury control in Volume VII appears to have been done
carefully, as far as it goes. However, it is flawed in two important ways. One serves to
potentially overstate the private (out-of-pocket) cost of emission reductions by a large
amount. The other serves to understate the social cost of government regulations.
Interestingly, these two omissions are closely related, from the standpoint of options
that might be considered by a policy maker.
The Study overstates the private cost of emissions reductions because it does
42
-------
not comprehensively explore the range of management strategies available to the
Agency for mercury emission reductions. The Study limits its review exclusively to a
set of post-combustion technologies of proven capability. It should be updated
to consider emerging technologies for post-combustion control (Feeler and Ruth,
1996) and expanded to explore the opportunities for pollution prevention, such as
strategies aimed at reducing the mercury content of fuels. (PERI, 1997; EPA, 1996;
Michigan Mercury Pollution Task Force, 1996). Although mercury differs from sulfur in
that there is greater variation within a deposit, the average mercury content differs
importantly among coal basins, providing a potentially low cost way of achieving
sizable emission reductions from electric utilities.
More important, the analysis provides information that is organized for only a
single management strategy—the implementation of Maximum Attainable Control
Technology regulations. This approach is not one which is likely to achieve cost
effectiveness in emission reductions. The Study only mentions in passing, and
provides no analysis of, incentive-based approaches to achieving environmental goals.
Incentive-based approaches that should be considered are deposit-refund systems
(batteries, industrial uses of mercury), emission fees and tradable permits. Numerous
studies of potential control of various pollutants have found that the costs of control can
be decreased by 20 to 80% through the use of incentive based approaches in place of
technology standards or inflexible emission rate standards (that is, emission rate
standards that are calibrated to be achievable with a specific technology). (Tietenberg,
1985). The Study should put forward information about options in addition to the
specification of emission guidelines (limits). However, in Volume VII the draft report
states (p.4-15): "Other nontraditional approaches such as emissions trading or
application of a use tax, or other market-based approaches may also prove feasible for
mercury control. However, these options are not presented in detail in this Report as
the control technology analyses focused on what might be achievable under the
statutory language of sections 112 and 129 of the CAA."
A second flaw that leads to the cost estimates in Volume VII understating
(potentially) the social costs of regulatory policies is the failure to acknowledge
changes expected to occur in a general equilibrium setting. In brief, one can expect the
existence of preexisting regulation and tax policy to cause the incremental social cost
of additional policy to be greater than the private financial cost, due to its interaction
with preexisting policy.
One can think of a new regulation as a "virtual tax" in that it depresses the real
wage of workers in the economy by raising the cost of goods and services. Almost any
tax has the unfortunate property, from an efficiency perspective, that it is expected to
distort economic behavior. However, to an important degree, this problem can be
remedied if a new regulation raises revenue (through a mechanism such as auctioned
permits, or emission fees). For further background, see Goulder, etal. 1996; Goulder,
1995; Oates, 1995; and Parry, 1995.
43
-------
The two ways in which the cost estimates are flawed invite the consideration of
incentive based approaches to environmental policies. Cost effectiveness in attaining
an environmental goal can be enhanced through such approaches. Further, the tax
interaction effect and social costs in a general equilibrium setting are reduced as the
cost of policies are reduced, and even further if such policy can be a source of revenue
for government that might displace other taxes. Analysis that takes these issues into
account, as well as consideration of the performance of alternative management
strategies with respect to the costs of control, is essential to improving the scientific
breadth and rigor of the Study.
3.19 Research Needs
In Charge 40, the Agency asked for suggestions to identify needed research.
With regard to the relation of anthropogenic mercury releases to human mercury
exposure, the major research priority is to quantify the relationships between human
mercury exposure and current and past anthropogenic sources of mercury. Quant-
ification of these relationships would allow a quantitative prediction of the effect of
mercury emissions control strategies on human mercury exposures.
Although there are many areas of uncertainty in the modeling of mercury fate
and transport, lack of information about the methylation and bioaccumulation of
mercury in various ecosystems is the area which contributes the greatest uncertainty to
the overall modeling effort. Studies determining rate constants and factors affecting
rate constants would lead to a more sound, defensible source-effect model. In
addition, research is needed to understand better the link between lake acidity and
mercury concentrations in fish, as well as watershed studies to evaluate the relative
contribution of historical, versus current, deposition and the residence time of mercury
in the watershed.
Additional study of target species' dietary habits is needed in order to better
characterize wildlife mercury exposure in at-risk species. In order to better understand
inter-relation of mercury sources to effects of mercury on wildlife, further research is
needed to determine NOAELs and LOAELs, particularly in piscivorous avian species
such as the cormorant, common loon, heron, osprey, etc.
3.20 Conclusions
The Sources Workgroup, after considering its discussions at the public meeting
and developing the preceding report, reached the following conclusions about EPA's
draft mercury report:
a) The emissions inventory of anthropogenic sources of mercury is largely
sound.
44
-------
b) The report does not address adequately the contribution of natural
sources of mercury or of past anthropogenic mercury emissions, both of
which may be important contributors to exposures.
c) The chemical form of the emitted mercury may have a significant effect on
its fate. The fraction of mercury emitted in various forms is uncertain,
although current research efforts may provide clarification. EPA should
incorporate new speciation results as they become available.
d) The modeling of atmospheric mercury transport and deposition is largely
sound. However, the treatment of dry deposition velocities should be
carefully reviewed.
e) The modeling of the post-deposition fate of mercury in ecosystems is
oversimplified and does not reflect recent advances in the science. This
modeling should be revised, with the goal of having the report reflect
current scientific understanding.
It is important to bring out the importance of MeHg production in
ecosystems in the draft report. Because MeHg is the species that
bioaccumulates in human and wildlife food supplies, understanding the
methylation process is critical to modeling Hg fate and exposure. Mercury
methylation and the variability in MeHg among ecosystems were
neglected in the report, because of the perceived difficulties in modeling
methylation. The report assumes that MeHg constitutes the same fraction
of total Hg in sediments, soils and waters across ecosystems. This is not
the case. Rather, the variability in MeHg production and bioaccumulation
among ecosystem types is many orders of magnitude, and may be as
large as the influence of Hg contamination. Without consideration of this
variability, the report's fate and transport models, and the resultant
exposure models and wildlife criteria models cannot predict WC with less
than two to three orders of magnitude associated error.
The report should detail current understanding of Hg methylation process;
MeHg production should be specifically included within fate and transport
models; and these models should be specific for a suite of ecosystem
types.
f) The derived wildlife criterion for mercury is a value that is below
concentrations found in many lakes with healthy biota and the utility of the
value is questioned. The risks to wildlife are not well characterized and
should be revised.
45
-------
g) Regarding the costs of controlling mercury emissions, the Report limits its
assessment to end-of-pipe, post-combustion pollution control systems.
Pollution prevention and market-based approaches should also be
considered. In addition, the benefits from the simultaneous reduction of
other pollutant emissions, which could result from some strategies, should
be included in the cost/benefit analysis. Although the benefits of reduced
mercury emissions can not, with the current-state-of-the-art, be quantified
at this time, at least a framework to evaluate the benefits of reduced
mercury emissions should be provided.
h) Even with revisions, the EPA should clarify that it is not currently possible
to quantify the relation between anthropogenic sources of mercury and
resulting mercury concentrations in fish or other biota. The EPA's
approach was to use state-of-the-art models to provide a
semi-quantitative link between mercury sources and mercury in biota.
Given the limitations in the ability of fate and transport models to quantify
mercury concentration in biota, the Subcommittee found the EPA's
semi-quantitative approach to be acceptable, although critical
biogeochemical processes were neglected. Further, by using better
models, including existing models such as MCM or R-MCM, a respectable
effort could be made at predicting effects. A more sound approach could
have been taken in contrasting watersheds testing different ancillary
parameters and loading rates.
46
-------
4. Exposure, Doses, and Body Burdens
4.1 The Scientific Basis of the Volume 3 Exposure Assessment
The Subcommittee, in Charge 14, was asked to comment on the scientific basis
of the exposure assessment, alternative approaches, and research priorities, and, as
well, advise the Agency on how Appendix H of Volume 3 should be used, improved, or
expanded.
The scientific methods used in the draft report's exposure assessment included
estimates of population exposures based on fish consumption survey data and
methylmercury concentration measurements in fish, scenario exposure estimates
based on environmental transport and fate modeling to estimate concentrations in
freshwater fish, and existing (albeit limited) data on mercury concentrations in hair.
Based on these factors, and market basket survey data on methylmercury
concentrations in food presented at the public review meeting, but included in the draft
report, it was concluded that ingestion of mercury-contaminated fish is the major source
of human exposure to methylmercury. Appendix H summarizes fish consumption
information from several surveys of the general population and of special "high intake"
populations. In addition, this appendix summarizes data on mercury levels in marine
and freshwater fish and seafood. This methodology is scientifically sound and is
consistent with approaches that have been used by EPA and other federal agencies for
assessing the health risks posed by environmental contamination.
The Subcommittee's review of this portion of the document identified several
areas where the exposure assessment could be strengthened or clarified. These areas
are:
a) The Mercury Study Report concludes that there is "a plausible link
between mercury emissions from anthropogenic combustion and
industrial sources and methylmercury concentrations and freshwater fish."
A critical question that still needs to be answered is "Do anthropogenic
atmospheric mercury emissions significantly affect mercury levels in fish
and seafood?"
Answering this question requires an explicit multimedia
transport/transformation assessment for mercury species. Only some of
the relevant transport and source media are currently considered, i.e., air
sources with transfer to surface waters by deposition. Other sources and
cross-media transfers need to be considered.
The overall question is only partially addressed in the draft report through
the use of the RELMAP and COMPDEP models which were used to
47
-------
predict the impact on of atmospheric releases on inland surface water.
However, the models were not used to evaluate the impact of U.S.
atmospheric releases on marine and estuarine environments, or to
evaluate the impact of foreign emissions on U.S. waters.
Information about other sources of mercury to the aquatic environment
were omitted from the report. Exclusion of this information creates the
impression that atmospheric releases are the only source of mercury to
the aquatic environment. This is not accurate, however, since municipal
and industrial wastewaters which are directly discharged to surface water
often contain traces of mercury. In addition, waste sludge from industrial
processes or municipal sewage plants, which is often used to amend
surface soil, can contribute to surface water contamination via runoff or
following volatilization to air. Many readers of the final report will be
aware of these sources and may wonder about their "relative contribution"
to surface water. It would be wise to include some information about this
in the final report. If it is impossible to estimate relative inputs from
atmospheric releases vs. land disposal and surface water discharges,
explain that to the reader. Perhaps this could be a future research issue.
b) Fish consumption rates were based on the Department of Agriculture's
Continuing Surveys of Individual Food Consumption (CSFII) 3-day dietary
history survey. This survey provides important information about fish
consumption patterns within the general population. However, the 3-day
survey has several limitations which are listed below:
1) The survey cannot be used to evaluate the frequency of fish
consumption by an individual consumer.
2) Daily intake rates for participants who reported eating at least one
fish meal during the survey period may be higher than actual since
the weight of fish ingested was divided by 3 days to calculate a
grams/day estimate.
3) More than half of those who normally ingest one fish meal per
week would have reported no fish consumption during a 3-day
period. Thus, the survey may not accurately reflect the percentage
of U.S. residents who consume fish and seafood on a weekly
basis. The survey provides very little information about patterns of
fish consumption by individuals, i.e. how frequently a single
consumer eats a particular species of fish; or about the geographic
source of the fish he/she consumes. Therefore, the survey cannot
be used to assess the randomness of a consumer's selection of
48
-------
fish and seafood. This is problematic since the risk assessment for
methylmercury in fish appears to be based on an underlying
assumption that consumers eat a random selection of
commercially-available and sport-caught freshwater and marine
species - an assumption the Subcommittee believes is incorrect.
Members of the Exposure Workgroup have studied consumption
patterns and found individuals eating very repetitive diets that
include daily or weekly meals of the same species of fish that they
obtained from the same commercial source or local fishery
(Knobeloch et al., 1995). This weakness in the survey data should
be mentioned in the report and the assumption of randomness
should be evaluated by future research. These inherent
weaknesses in the survey design should be discussed in Appendix
H. In addition, the authors should explain how the report
compensated for these weaknesses; for example, by including
information from other, longer dietary surveys.
c) Occupational, intentional, and accidental exposure to mercury are only
partially covered in this report. Limited information on occupational
exposure and exposure to mercury from dental amalgams is included,
however the report does not cover other exposures, such as household
spills, use of mercury for ritualistic or medicinal purposes, etc. The
uneven and incomplete coverage of non-environmental exposure sources
weakens the report. We recommend expansion of this section of the
report to include a more comprehensive literature review. Several cases
of mercury poisoning have been reported in the Center for Disease
Control's (CDC) Morbidity and Mortality Weekly Report over the past two
years. National databases could be used to determine the number of
mercury poisoning cases that are reported to CDC or to Poison Control
Centers annually. This information serves to confirm the importance of
mercury as a cause of human illness.
d) The process of averaging mercury levels for several different subspecies
offish and shellfish to develop a mean level for each species may, in
some cases, lead to inaccurate exposure estimates. When a small
number of sub-species dominate the commercial marketplace the average
mercury level for the species should be weighted accordingly.
In future surveys, it may also be useful to record the geographic origin of
the marine and freshwater fish that are sampled for mercury analysis.
This information could be used to weight mean estimates based on the
percentage of the total catch that are harvested from each location (e.g.,
N. Atlantic vs Pacific). Correction for differences between mercury levels
49
-------
in fish that are harvested from inland waters, estuaries, marine waters,
and commercial fish farms would further refine the exposure estimates.
Additional information on these topics should be identified as a research
priority.
e) The average methylmercury intake estimates that are included in the
Executive Summary do not provide the reader with a sense of the extreme
variability of exposure that occurs within the U.S. population. A more
accurate picture of exposure would be provided by citing a range of
values such as the 10th, 50th and 90th percentile estimates.
f) Is information available that could be used to estimate mercury intakes
among Native Alaskans that result from ingestion of marine mammals
such as whales and seals? If so, this information could be included in
Appendix H. If not, this could be identified as a data gap.
g) Humans and wildlife most at risk are those who repeatedly consume fish,
aquatic mammals or birds from a single location that is contaminated with
mercury. When sport-caught fish are involved, advisories are used to
provide information on the safety of frequent or long-term consumption.
Advisories are not used for commercially-sold species such as swordfish
or shark. This report could address the appropriateness of labeling
commercial fish species that are suspected of being high in mercury. The
report might also mention the FDA's new Hazard Analysis and Critical
Control Point Program (HACCP) designed to prevent control food safety
problems. The HACCP, which goes into effect in December, 1997,
monitors "Critical Control Points," which could reduce exposure to
methylmercury in fishery products.
Appendix H provides an overview offish consumption rates for the general U.S.
population and for several "high risk" populations. Fish intake rates broken down by
age-group and gender were used to estimate methylmercury intake rates for each
subgroup based on mean mercury levels for freshwater fish. Mercury levels for
freshwater fish were taken from Bahnick et al. (1994) and Lowe et al. (1985).
This appendix is a valuable addition to the report. It is best used as a summary
of U.S. fish consumption data and tissue mercury levels. It could be improved by
updating the fish sampling results and expanded by including information about fish
consumption patterns by certain "at risk" populations, such as pregnant and nursing
women, pre-school aged children, and the elderly; and by adding a section that
addresses the randomness offish consumption.
50
-------
4.2 Characterization of the Exposure Assessment
Because of uncertainties inherent in the emissions inventory and the model, the
exposure assessment was characterized as being a "qualitative assessment based on
quantitative modeling." In Charge 15, the Agency asked if that was a reasonable
characterization, and if the uncertainties and conclusions drawn from linking the
models together were appropriately described?
In the view of the Subcommittee, EPA's characterization is not only reasonable
characterization, but, given the quality and reliability of the information available very
appropriate characterization. However, the Subcommittee noted that, even though
quantitative modeling is used for making order of magnitude estimates, there remains
only qualitative treatment of uncertainty. Since the qualitative conclusions are based
on the quantitative results from models, it would be useful to put some quantitative
bounds on the results presented. In addition, the Subcommittee observed that, in
making qualitative conclusions, models are used more than data, and the data and
models are not effectively compared or reconciled.
One area in which we believe the report needs to provide a better
communication of uncertainties is in the use of an appropriate number of significant
figures. Qualitative results do not have three significant figures! This is a problem both
in the Overall Executive Summary, the Executive Summary for Volume II, and
throughout Volume II. Too much precision is implied by the way the numbers are
presented with so many significant figures. Most of the results in this report are only
good to one significant figure, if that. For example in Tables listing sources of mercury
a better qualitative presentation would be to list sources as major (i.e. 20 to 50%),
significant (1 to 10%) and minor (<1%).
There should be more effort to make sense of model predictions relative to
multiple observations of mercury deposition over land and water. These observations
could be used to put some type of bounds on the model reliability. For example, the
EPA could pose and test the premise that Hg emissions are essentially 80 to 90%
accounted for and then test models and deposition data against this premise to see if it
makes sense.
Comparing human and natural emissions of mercury carries with it the implicit
assumption that human emissions have the same impact on human health and
ecosystems as natural emissions. This assumption needs to be given more
consideration. Do human emissions have the same biogeochemical cycle as natural
emissions? Do they spread over the same spatial range? What is the relative
persistence of human versus natural emissions?
51
-------
One issue that is not well-characterized is how to distinguish between global
average concentrations and local and regionally elevated environmental levels. This
will be an important issue for interpretation of deposition data over land, lakes, and
ocean waters.
Instead of simply explaining how the models work and showing the results, as is
done in this report, it would be more useful to work through the logic of the models and
identify where important uncertainties exist and how these uncertainties impact the
source/receptor relationships derived from the models. The modeling and exposure
evaluation process has an implicit tiered approach that should be made more explicit.
In the report there are several modeling and evaluation scales, such as local models
(within 10 to 20 km of the source), regional airshed/watershed models, continental
scale models, and global scale models. In the absence of more data, simple bounding
mass balance assessments are probably all that can be done given the current
scientific understanding of the biogeochemistry of mercury and its compounds.
4.3 Methyl mercury and Fish Consumption
The EPA has estimated methylmercury and fish consumption using both cross-
sectional and longitudinal data for the general U.S. population. In these studies, EPA
consistently finds that a portion (albeit a small one) of the population consumes very
large quantities of fish. It is these individuals who face the largest risks from
methylmercury exposure and who pose the greatest concern to the Agency. Charge 16
sought guidance to determine if EPA had adequately assessed the number and types
of individuals who consume fish in these relatively large quantities. Related questions
addressed the uncertainties associated with characterization of such small portions at
the extreme end of the population distribution, and the possibility of a methodologic
artifact.
The Subcommittee believes that the agency has adequately assessed the
number and types of individuals who consume fish in these large amounts, given the
state of existing information. As stated on page 3-35 of Volume I, the EPA estimates
that between 1 and 5% of the population consumes more than 100 grams offish per
day. This estimate is supported by the 1973-1974 National Purchase Diary survey that
identified the 99th percentile intake rate of 112 grams/day and by the 1989-91 CSFII
survey that identified 95th percentile estimates of 118.9 and 134 grams/day for adult
females and males, respectively; and 99th percentile estimates of 178-224 grams/day.
In addition, several studies cited in Figure 3-6 (3-34) of special populations such as
sport fishermen, Native Americans and Native Alaskans identified 90th and 95th
percentile estimates that exceeded 100 grams/day. In fact, two independent studies of
Alaskan Natives reported mean daily intake rates of 109 and 452 grams/day.
52
-------
In addition, the Subcommittee finds that the uncertainties associated with
characterization of such small portions at the extreme end of the population distribution
are appropriately considered and presented.
The Subcommittee found the intent of Charge item questioning the possibility of
a methodologic artifact affecting the estimate of 5% of the fish-consuming population as
being high-end consumers to be unclear. All population distributions have an upper
5% tail. Therefore, any survey offish consumption can be expected to define an upper
95th percentile subpopulation that consumes fish at a higher rate than the rest of the
respondents.
The real issue seems to be, "Is the fish intake estimate for the upper 95th
percentile population accurate?" This is a more difficult question to answer. Certainly
the 3-day survey method could have overestimated the daily intake rate, however,
several earlier surveys did find people with intake rates of 100 g/day or more
(Humphrey, 1974, 1988).
4.4 Sources of Data for Mercury Concentrations in Marine and Fresh-Water Fish
and Shellfish
The Agency relied on three nationally-based sources of data for mercury
concentrations in marine and fresh-water fish and shellfish. Some of these data were
obtained within the past three years, whereas other data were obtained as long ago as
20 years. Question 17 poses issue of the use of these older data, and the report's
documentation of the limitations of this methodology.
There are several reasons to question older fish data, including a) concerns over
analytical quality and documentability; b) representativeness of the sampling protocols;
and c) whether, due to changes in Hg inputs to the systems, average values may have
increased or decreased between the time of the survey and 1997. Overall, the data at
hand can be used with confidence. Of course, if more recent survey data are available,
it would be prudent to utilize those where possible.
It is fairly unlikely that significant analytical bias exists in any of the major fish
surveys of total Hg collected in the past several decades. This is because techniques
necessary to quantify total Hg in fish have long been available. Because fish tissue is
relatively high in Hg (ppb to ppm), the risk of contamination is small (unlike the case for
water samples, where contamination at the parts-per-trillion level is commonplace).
Factors that could bias analytical data include the following: (a) how were the below the
detection limits (BDL) values were handled in constructing mean values? (If less than
detects were discarded, then an over-estimate of mean values will occur); (b) what
were the BDLs of the earlier surveys? (The poorer the BDLs, the greater the risk of
bias - most likely in the direction of over-estimation of mean values); c) how were fish
53
-------
processed prior to analysis-i.e., whole, filleted, skin on or skin off? (Inclusion of tissue
other than muscle will result in an under-estimation of consumption of methyl Hg); (d)
were the fish analyzed truly representative of the size, region, and species normally
eaten by the American population?
It would be wise to look over the original documentation for these surveys to be
sure that protocols were both appropriate and equivalent between surveys. Finally, it
should be noted that if any of the data were of methyl Hg, rather than total Hg, there is
a significant risk of bias too low, as earlier methyl Hg quantification techniques (solvent
extraction/GC-ECD detection) are notorious for poor recoveries and reproducibility.
Low recoveries for methyl Hg combined with the potential for contamination with Hg(ll)
has likely been responsible for the low bias in estimates of % methyl Hg in fish (70-
90%) from the 70s and '80s, compared to estimates from the '90s (95-100%).
With respect to the representativeness of the sampling protocols, the Committee
notes that Bahnick et al. (1994) sampled only 5 bottom feeders and 5 game fish from
314 sample sites selected based on proximity to either point or non-point pollution
sources. Only 35 of the sites were considered remote. Data for the bottom feeders
was based on analysis of the whole fish rather than edible portions. Although this
information is indeed useful in monitoring site specific concerns, its application to a
national consumption guideline including marine species is questionable.
Having made these comments, the Subcommittee regards the mean
concentrations from the Lowe data set as too low, and suggests that the EPA/Bahnick
data sets are more representative of fresh water fish levels. This position is based
upon comparisons of species for which there are many additional studies (usually from
the Midwestern states where these fish are mostly consumed), such as walleye, pike,
and yellow perch. From the summary data, there is no indication as to why these data
set means differ, but we suspect that looking into the primary data will show differences
in the representativeness of the fish analyzed (e.g., regions, lake types, fish size). If no
such assessment can be made for the report, we suggest dropping the Lowe data set in
favor of the EPA/Bahnick data sets, and/or other more recent surveys. The mean Hg
for marine species in the National Marine Fishery Service (NMFS) database appears to
be appropriate, although a look at the underlying assumptions and analytical figures of
merit would be worthwhile. It is of concern that the NMFS database is mostly
unpublished data (although the report stressed that only published data were used), for
which the level of quality control through time is not established. Of particular
importance is verification that the named fish in the data base do indeed correspond to
the same fish that people in consumption surveys think they are eating. This is
important at least for mackerel, where a big discrepancy between the observed
concentrations between surveys may be attributed to different species called mackerel
being analyzed.
54
-------
4.5 Fish and Shellfish Consumption
The U.S. Department of Agriculture (USDA) has published data indicating that
fish and shellfish consumption have risen approximately 25% since the early 1970s .
These consumption data were employed without adjustment in this report to estimate
current fish consumption. Question 18 asks if these data should be adjusted to reflect
the overall trends in fish consumption, if so, how would the estimates of very high
consumption (above the 90th percentile) be adjusted.
As the Subcommittee interprets this question, the issue is whether the
consumption data from the 1970s should be used without adjustment, even though
recent data indicate that consumption offish and shellfish has risen approximately 25%
since the 1970s. Based on comparisons to other types of market information, the
Subcommittee is of the opinion that the numbers used by EPA for fish consumption are
representative and perhaps somewhat high for representing the mid-range fish
consumer (this is discussed elsewhere in this review). The USDA report of increasing
fish consumption is difficult to interpret in terms of its impact on the very-high fish
consuming individual. It is typical in food-consumption surveys, that high-end
consumers who make food choices based on economic and taste preferences are less
likely to move up and down with market trends and diet trends as much as the
mid-range population. Thus, high-end consumers are not likely to increase their
consumption proportional to the mid-range fish consumer. The Subcommittee
expressed the view that the fish consumption data for the high end consumers should
not be adjusted to reflect the trend reported by USDA.
4.6 Calculations of Methylmercury Intake from Crab
Calculations of methylmercury intake from crab were grouped without specific
consideration of the particular species. These data are then linked to dietary survey
data which typically do not list the individual species of many fish and shellfish
consumed. Question 19 asks if this approach might bias the final assessment, if the
limitations of this approach are adequately characterized in the report, and if these
limitations going to materially alter the assessment of health risk associated with
methylmercury exposures.
EPA's calculations of methylmercury intake from crabs were made by grouping
data without specific consideration of the particular crab species consumed. These
data are then linked to dietary survey data which typically do not list the individual
species of many fish and shellfish consumed. The Subcommittee agreed that this was
not an issue of concern for two reasons. First, individual crab species values did not
differ drastically from one to another (confirmed by data for Dungeness crabs collected
by a Subcommittee Member), meaning that the mean value is relatively robust. More
55
-------
importantly however, is that crabs make up a very small fraction of American fish and
shellfish consumption. Thus, for any reasonable value for Hg in crab, there would be no
discernible change in the overall human methylmercury consumption values. The
methylmercury consumption for average Americans is overwhelmingly dominated by
tuna, pollack, and shrimp. In the case of high end consumption of crabs, this can only
realistically be assessed by identification of the exposed population and the particular
crabs that they are eating.
A concern was raised, however, over the more general issue of averaging
different species, or fish of the same species but from different locations, under a single
common name. The risk inherent in this was noted above for mackerel, where an order
of magnitude difference is seen in mackerel of different species -and it was pointed out
that the species lower in Hg are not the species that are commonly consumed by
mackerel eaters. As noted above however, for the purposes of national population
exposure estimates, these issues will not affect a mean exposure that is so heavily
weighted by three well-documented species (tuna, pollack, shrimp). In assessing the
high end consumer groups, it is again critical to know clearly the exact species and
lakes from which the fish are harvested, as order of magnitude differences can be seen
over a few mile radius due to differences in water chemistry (in situ methylation rates).
4.7 Characterizing Total Population Exposure and Risk
In Volume III (Appendix H), the EPA characterized total exposure (and by
association) total risk by considering methylmercury exposure from marine fish and
shellfish together with consumption of methylmercury from fresh-water or estuarine fish
and shellfish. The Agency, in Charge 20, wanted input as to how well the current
approach characterized total exposure and dealt with associated uncertainties vis-a-vis
sources offish (inland freshwater versus marine).
The Subcommittee agrees that inland lake fish appear to represent the greatest
acute source of exposure, but also notes that the contribution made by the marine
environment is important and should not be ignored because it represents background
or chronic exposure. Certain near-shore marine waters may also represent acute
exposure sources.
The draft report correctly concludes that the exposure levels at which one infers
risk for the general population are relatively low. However, total population exposure
(all routes and pathways) and the related risk is not and cannot be fully characterized
because of the large variability in human consumption offish and interspecies and
interregional mercury concentrations. Total risk cannot be characterized at this time
because of the uncertainty of upper bounds exposure due to the variability found in
existing data bases. For freshwater fish, there are only two surveys of "national" scope
and the mean mercury body burden for all fish differed between them by more than a
56
-------
factor of two. This large difference was also seen in the highest mercury levels
recorded in the surveys (Lowe versus Bahnick). Although more data on marine fish are
available, there appears to be limited information on near shore marine fish. These
areas receive freshwater discharge and may well represent sources of acute exposure
similar to certain inland freshwater lakes. Total population exposure/risk is difficult to
characterize given the variability and gaps in existing data sets. These impact most
significantly on the ability to accurately model and assess upper bound risk.
The draft report does provide a thorough and complete record of the existing
data available. Many studies and data bases were found and reported in Appendix H.
The uncertainties in and variability of existing information are also well presented
(pages H-65-67).
Given the uncertainty in exposure assessment, especially for the upper
bound of human exposure, we are not confident that accurate modeling of risk
for this group can be performed. The assumptions used in Volume VI to model
human risk may be overstated. Given the variability of mercury in fish and information
gaps on "subsistence" or high consumers offish, one must question use of the
assumptions that these people eat trophic level 4 fish exclusively, that all trophic level 4
fish have 5-fold higher mercury levels than trophic level 3 fish, what fish mercury level
should be used in the model and what is the actual size of the population at greatest
risk. Overestimates in assumptions can show estimated human intake exposure in
excess of that actually occurring in the field. The conclusions section of Volume VI
(pages 6-1 to 6-3) acknowledges this problem.
4.8 Hair Mercury Concentrations
Some data exist on hair mercury concentrations in U.S. residents. Question 21
asks if this data base is adequate to predict the distribution of hair mercury in the
general U.S. population.
Hair total mercury levels provide an index of blood total mercury when the major
mercury exposure is in the form of organic mercury, particularly methylmercury. Hair
mercury levels do not correlate well with blood mercury when exposure involves
primarily inorganic or elemental mercury. There is a relatively small body of data
describing hair mercury levels in U.S. residents. Most of the available data comes from
research studies containing small numbers of individuals whose hair was measured in
the course of studies on methylmercury exposure and disposition. There are many
more studies involving hair analysis for mercury that have been conducted outside of
the U.S. These also are research oriented and do not constitute population monitoring
efforts. One exception is an unpublished study supported by the Food and Drug
Administration which contained approximately 1500 U.S. residents from across the
country. The draft report mentions this study but suggests that it does not have much
57
-------
scientific weight because the variability in the data was not given. An expanded
explanation should be given because this apparently is the largest attempt to measure
mercury levels in the population and the data would appear to be useful.
The current draft mercury report correctly indicates that statistically
representative population monitoring studies are not available at this time. The
Subcommittee's Exposure Workgroup discussed the need to monitor the U.S.
population for mercury exposure. Some Members felt that it would be beneficial and
others thought that it was sufficient to monitor more highly exposed subpopulations
such as subsistence fish eaters.
4.9 Hair Mercury Concentration as an Index of Methylmercury Exposure
The draft report identifies hair mercury concentration as the most appropriate
available index of methylmercury exposure. Question 22 asks if this assumption is
consistent with the available data, if the exposure estimate approaches used are
relevant and appropriate, if the predicted exposure ranges are consistent with other
published exposure analyses, and if the uncertainties of this assumption appropriately
presented.
The preponderance of studies reviewed that are relevant to the developmental
health endpoint, upon which the Reference Dose (RfD) for methyl mercury is based,
have used total hair Hg concentrations as the most appropriate metric of exposure. The
use of hair has practical advantages over utilization of body fluids or tissues given the
non-invasive nature of such samples. In addition, Hg concentrations in hair provide an
integrated measure of dose over time, as compared to concentrations measured in
whole blood or some fraction thereof, which only provides levels existing at the time of
sampling. In addition, there is a general correlation between total mercury
concentrations in hair and blood, which support the assumption that hair analysis may
be comparable in its validity as a bioindicator to blood, with the additional practical
advantages described above. Finally, the current and proposed RfD is based on
results from the Iran Hg poisoning episode, from which only maternal hair analysis data
are available as an indicator of in-utero exposures. To this extent, the draft report is
generally consistent with the preponderance of the studies, and presents some of the
limitations regarding the RfD estimate. In addition, the authors have made an
extraordinary effort at compiling and evaluating the available literature. They should be
commended for this effort, given the difficulty and controversial aspects of this task.
The comments that follow are not intended as a criticism but are meant to provide some
guidelines to strengthen the review with respect to exposure-dose issues.
Most of the discussion on the appropriate metric of dose is presented in Chapter
2 of Volume IV: Health Effects of Mercury and Mercury Compounds, with further
reference in Chapter VI which presents the risk characterization for the U.S. population.
58
-------
Estimated exposures for the U.S. population based on emissions and multimedia
transport models are presented in Volume III. The document could benefit from a
section dedicated exclusively to exposure and exposure-dose relationships, either as
an additional Chapter in Volume III, or as the first Chapter in Volume IV. This section
should address specifically some of the issues that are not discussed in the present
format of the document, or that are presented in a diffuse manner throughout Chapter
IV. These issues include:
a) clear definitions of what metric constitutes an indicator of exposure, dose,
or effect and how these metrics relate to each other. In its current
version, the draft document (and the preponderance of the past and
present population-based studies) uses the number offish meals as the
main indicator of exposure while exposure should be the concentration of
methyl mercury present in the fish times the amount offish consumed
during a day, week, etc. The number offish meals is a surrogate for
exposure and, as such, could lead to misclassification. The most
appropriate exposure indicator would be the amount of methyl mercury
ingested via the food chain over a relevant period of time. Since the
concentration of methyl mercury varies by type offish, where it is caught,
and what parts of the fish are eaten, the number of fish meals consumed
over time may or may not be a good indicator of the exposure depending
on the other determinants of exposure, as mentioned above.
The use of number offish meals could lead to individual misclassification
of exposure. For example, inspection of the data presented in Table 2 of
Grandjean et al. (1992) shows that if the only information available for
exposure classification were the number offish dinners consumed per
week, we would be likely to place individuals in the wrong "exposure"
category according to the blood or hair levels because the range of the
latter measurements is relatively broad. In part, individual variability in
compartmental distribution and disposition of methyl mercury could
account for the lack of better predictive validity. However, since the
number of fish dinners consumed per week has not been demonstrated to
be a good indicator of exposure (based on actual Hg concentration and
fish consumption data), we cannot determine if only individual variability
is important in explaining the potential for misclassification. Mercury
concentrations in blood, urine or hair could be indicators of exposure if it
can be shown that they are correlated with exposures as determined
above. On the other hand, mercury concentrations in biological fluids or
tissues could be used as markers of effect if they were shown to be
associated with particular effects in a consistent manner. The current
document uses mercury concentrations in hair for dual purposes, that is,
as a marker of exposure (as indicated by number offish meals consumed)
59
-------
and developmental effects produced by in-utero exposure. EPA has an
opportunity in the future to evaluate the association between exposure
and the appropriateness of using the concentration of Hg in biofluids and
tissues as indicators of exposure by better estimating actual exposures
(that is, beyond number offish meals consumed) using the recently
completed and on-going population studies. The investigators in these
studies should be able to derive those estimates which can then be
related to the bioindicator data. Perhaps it would also be useful to
combine the data from the various studies in New Zealand, Peru, the
Faeroe and Seychelles Islands using a meta-analysis type approach, if
possible, so the exposure-bioindicator relationships could be evaluated in
more detail as a function other population variables.
b) broaden the discussion of limitations on the usefulness of hair mercury
concentrations to include practices among the U.S. population that may
affect these concentrations. Besides the limitations discussed in the
document (e.g., acute vs. chronic exposures, genetic differences, etc.)
that may affect the usefulness of this bioindicator in U.S. populations,
there are cultural practices that could also be important (and for which
data are probably lacking). Populations in developed countries,
particularly women in the U.S., engage in chemical treatment of their hair
for cosmetic purposes far more that those in developing countries. Do
treatments such as permanents, straightening, bleaching, etc. alter the
concentration of metals present in hair, including Hg? If this is the case,
the usefulness of mercury concentrations in hair would have limited
applicability in the U.S.
c) related to b) above, discuss the usefulness of hair analysis given multiple
exposure pathways to different forms of Hg. Exposure to Hg in the U.S.
may be dominated by other than methyl mercury in foods, for example
exposures to elemental Hg emissions from Ag/Hg dental amalgams.
Would this affect the applicability of analysis of mercury in hair as an
indicator of in-utero, methyl mercury exposures?
Other issues that the document should address include:
a) since brain development does not end at birth but continues well into
childhood, there could be a potential for subtle developmental effects due
to ex-utero exposures, for example via breast milk. There is no RfD for
children. Is there evidence that exposures after birth could affect
development and, therefore, should there be a guideline for children's
exposure?
60
-------
b) a more careful attempt at trying to identify U.S. sub-populations that may
be at increased risk of exposures, particularly since the "average"
individual is not at risk.
4.10 Linking Mercury in Fish and Current Anthropogenic Emissions
Although the modeling exposure assessment focused on anthropogenic
emissions, the fish consumption analysis considered measured mercury concentrations
in fish tissue regardless of the mercury's origin. Thus, there is considerable difficulty in
assessing or describing how much of the mercury in fish is attributable to current
anthropogenic emissions. Questions 23 addresses the approach taken by the Agency
in this assessment, asks the advice of the Subcommittee regarding the differentiation
between current emissions and their impacts relative to the "body burden" approach of
the fish consumption analysis.
This Charge issue, addressing the exposure assessment, goes to the very heart
of the prime objective of the EPA report: Can a quantitative linkage be established
between patterns of anthropogenic atmospheric emissions and mercury concentrations
ultimately manifested in fish?
As the various specific questions in this issue point out, it is extremely difficult at
this time to determine such a relationship in quantitative terms, although the report
concludes a "plausible link" exists between this proposed cause and effect. Therefore,
a quantitative linkage was not established in the EPA study in a real sense based on
comprehensive field data, but a qualitative linkage was postulated based on theoretical
models and limited field data.
Why was this not done? At least two reasons are responsible for this shortfall.
The first of these is the protracted time course of mercury effects in the context of the
global mercury cycle. There are many dynamics in the relationship between
atmospheric mercury and methylmercury levels in fish, and the various steps may span
long time periods, making the quantitative description of the linkages between steps all
the more difficult. This problem is understandable and perhaps deserves a bit more
emphasis in the EPA report.
The second reason for the shortfall is the lack of field data, covering the full
spectrum from emission through deposition to methylation and finally appearance in
fish, near U.S. anthropogenic sources of concern. Few experimental studies exist that
precisely document the cascade of changes that occur when mercury emissions from a
given source are varied. A 100-year mercury emission elimination experiment would
be a revealing, albeit completely impractical, research project to address this question.
Perhaps more feasible would be an historical analysis of mercury deposition changes,
61
-------
using core samples in lake beds and ocean bottoms. Many such studies, based on
sediment cores from lakes, oceans, and ombrotrophic bogs are available (Hurley et al.,
1994; Rada etal., 1993; Rada etal., 1989; Compeau etal., 1985; Choi, S.-C. and R.
Bartha, 1994; Gagnon et al., 1996) These data could be used in
geochemical/ecological models to predict the down-the-line changes that should have
occurred given what we do know about the linkages. The model outputs could then be
compared to data on historically relevant changes, such as fish methylmercury levels
(where known), and adjusted accordingly. But this is intended as a research
recommendation, not a currently feasible analysis.
Given these problems, therefore, it is reasonable and appropriate that EPA took
the approach that they did. Advice on dealing with this issue follows:
a) Although EPA should not necessarily try to quantitatively estimate the
contribution of current air emissions to mercury in consumed fish, it
should nevertheless provide some analysis of natural and old
anthropogenic emissions, using the above mentioned sediment studies,
so as to indicate the effect that might be expected in a reduction of
current emissions.
b) In doing so, EPA should showcase localized studies, especially on point-
source emissions in freshwater systems (and perhaps marine/estuarine
cases like Minamata), where a direct linkage between field data of
mercury emission and human effects is demonstrable because the
temporal and spatial scales are limited.
c) EPA should stress in the report the global picture: U.S. emissions
contribute to total global atmospheric mercury and the resulting
deposition in the oceans and on land. This ultimately must contribute to
total mercury loads in fish, since elemental mercury in the environment is
neither created nor destroyed.
Because methylation is complex and not well understood, however, it
cannot be assumed that a change in total mercury emissions will be
linearly related to any resulting change in methylmercury in fish, even
taking into account the role of natural and old anthropogenic sources. It
should be made clear that the variability in MeHg production among
ecosystems may be as large as the effect of Hg contamination Therefore,
EPA should emphasize to decision-makers the difficulties of quantifying
the linkages, while underscoring the plausibility of the qualitative
connections on a local scale and the certainty of a linkage on a global
scale.
62
-------
4.11 The Biokinetics of Methylmercury in Grain and Fish
In its assessment, U.S. EPA assumes that the biokinetics of methylmercury from
contaminated grain approximate those resulting from methylmercury from fish. In
Question 24, it asks if this assumption consistent with the available science.
There is no compelling evidence to suggest that the toxicokinetics of
methylmercury (MeHg) ingested in grain (as in the Iraqi poisoning episode) is different
from that resulting from ingestion offish (the typical exposure route of humans). The
best evidence for this is a study in which cats were fed contaminated fish, control fish,
or MeHg in a non-fish diet (Charbonneau et al., 1974). No differences were observed
in degree of MeHg neurotoxicity, latency to toxicity (ataxia), tissue levels or distribution
of Hg. Some discussion in the EPA document focused on potential for Selenium (Se)
to protect against MeHg neurotoxicity, since Se levels may be higher in fish than most
other foods. While co-administration of MeHg and Se apparently results in decreased
MeHg concentrations in kidney, Hg levels in brain and liver are increased (Suzuki and
Yamamoto, 1984; Brzenicka and Chmielnicka, 1985; Komsta-Szumska etal., 1983).
Se also increased methylmercury staining in spinal cord and nerve cell bodies (M0ller-
Madsen and Danscher, 1991). A positive correlation between brain Hg and Se levels
was observed in monkeys exposed to MeHg with no exposure to Se other than regular
diet (Bjorkman etal., 1995); the mechanism for this is unknown. The apparent
protective effect of Se against overt high-dose MeHg toxicity has been attributed to the
decreased accumulation of MeHg in kidney in the presence of Se (Stillings et al.,
1974). It is doubtful that this effect on kidney is relevant at environmental levels of
MeHg. It has also been suggested that the formation bis (methylmercury) selinide may
render MeHg less toxic (Naganuma and Imura, 1980), but there is no direct evidence
for this. In addition, although increased fish consumption was associated with a very
modest increase in (cord) blood Se levels in a fish-eating population, the blood Hg
levels increased much more dramatically (Grandjean etal., 1992). Grain may also
contain substantial levels of Se, depending on the soil in which it is grown. Based on
the questionable relevance of any protective effect of Se against high-dose MeHg
nephrotoxicity, the fact that increased Se intake results in increased brain Hg levels
following MeHg ingestion, and a complete lack of data on the Se status of the human
(Iraqi) population exposed to MeHg via grain, there is no reason to postulate that
ingestion of MeHg in a fish matrix would result in decreased toxicity compared to
ingestion in a non-fish matrix. Indeed, the study that addressed this directly found
absolutely no difference in toxicity or tissue levels when MeHg was administered as
contaminated fish or added to a non-fish meal (Charbonneau et al., 1974).
4.12 Uncertainties in the Fish and Grain Consumption Analyses
63
-------
Charge question 25 asks if the uncertainties in the fish and grain consumption
analyses are adequately and consistently presented in the draft report.
A major uncertainty in the fish consumption data are that the current supply of
'fish' for U.S. consumption has significantly changed through the last decade and
projected trends and predictions indicate continuing changes which must be accounted
for in any national projection to address methyl mercury consumption in the United
States. The EPA report relies on methyl mercury assessments focused on fish from
United States environments, rather than existing markets. This approach is
understandable in a report also attempting to link mercury sources with contaminated
foods, but it only accounts for foods from domestic sources. Since the 1980's, nearly
70% of the seafood, including fish, consumed in the U.S. are imported and this supply
trend is predicted to increase in response to steady-state domestic sources and
increasing restrictions on both commercial and recreational harvest. Likewise,
aquaculture sources, both domestic and international, have grown to represent nearly
20% of U.S. seafood, including fish, consumption. Aquaculture remains the most
plausible source for future fish supply to meet domestic demands. The EPA report
does not account for this segment of consumption, which is based on cultured products
grown in shorter duration in more controlled conditions for water quality and feed
composition. Although there is no evidence to suggest that MeHg levels are
substantially higher or lower in imported fish than in fish from domestic environments,
this should be noted in the report as an additional source of uncertainty.
In addition, the EPA cautions against consumption of certain aquatic predators,
i.e. swordfish, shark, and barracudas. Recent federal and state-based fishery
management plans since 1990 are restricting the production of such fish due to
conservation concerns, recreational plans, and separate food safety issues. For
example, harvesting of sharks about Florida is soon to be prohibited; many recreational
billfish practices have adopted, and enforced catch-and-release programs, and
concerns for ciguatera have banned harvest and commercial sales of barracuda in
regions of southern Florida, Puerto Rico, etc. These supply changes, which favor the
reduction of risk from consumption offish from the United States, should be noted in
the final report.
The Committee particularly wishes to emphasize and alert EPA to all of
these ongoing changes in the sources offish in American diets and to
recommend that EPA be prepared to address such changes in future risk
assessments.
EPA's methods for determining the methyl mercury load in "fish," as documented
in the draft report are confusing and difficult to understand in several places. Section
2.1 in Appendix H (p. H-45) exemplifies the confusion. This is a critical paragraph in
the
64
-------
report and warrants correction and clarification. Part of the problem seems to stem
from typographic errors:
a) Line 9,...from Table H-36 (Consumption of freshwater fish.. .) does not
correspond with Table H-36, page no. H-51.
b) Use of methyl mercury data from limited freshwater fish alone (p. H-45,
section 2.1) to estimate the range of [national] exposure from total fish
consumption is not appropriate. The vast majority offish consumed is
from marine sources. In fact, the EPA reports that the CSFII 89/91 survey
reported less than 31% of the identified 'fish eaters' ate freshwater fish
(page. no. H-8). In conflict, the Executive Summary (page no. 3-33)
states that the CSFII 89/91 survey found 33% of the 'fish and shellfish'
eaten came from freshwater and estuarine species. It is not clear how
shellfish and estuarine species only boost consumption by 2%
(31%+2%=33%).
The report offers no definitions for marine, estuarine, river and freshwater fish.
This distinction is important, particularly for certain subpopulations considered to be at
higher risk. For example, would the Alaskan salmon be marine, estuarine or river fish?
Would commercial at sea harvest differ in concern for the same fish taken in river
migration? Simple reference to 'fish' complicates interpretation of some parts of the
report. Addition of some definitions is recommended; the text should also be reviewed
and edited to reflect the added definitions.
It is also not completely clear how the NMFS methyl mercury data for marine
species was used. The Executive Summary (p. 3-33) mentions a "mixture of
fish/shellfish reported in CSFII 89/91 to be consumed by persons surveyed is 0.134
ppm." How the 'mixture' was calculated is not clear in Vol. Ill on Section H. There are
also concerns that the NMFS data are not easily accessible and lack peer review.
Portions of the report and subsequent discussions by EPA staff during the public
meeting mention use of weighted intakes for methyl mercury based on matching the
'fish eaters' reported fish vs. the respective reported methyl mercury levels for the
respective fish. No such comparisons are offered in the report and if available would
be subject to critique for a) proper fish identity, and b) appropriate source/comparisons.
It is highly probable that some of the recall survey data on the type of 'fish' consumed is
inaccurate. The proper identification is limited by the persons' knowledge, use of local
vernacular, and/or reliance of package labels or restaurant menus that are also
suspect. Product mislabeling in retail and restaurant settings has been measured as
high as one-third of the commercial events in certain states. Increasing use of
imported products further complicates this problem. Thus, aligning 'recalled' fish
consumption by fish type with estimated methyl mercury concentration is compromised
65
-------
by proper product identity. The report should note this source of uncertainty explicitly.
A concern that the EPA report placed too much reliance on CSFII 89/91 and
failed to incorporate data from other consumption surveys has been repeatedly
expressed by numerous reviewers. The survey, by virtue of its short (3-day) sample
and division by 3 days per amount eaten, tends to overstate consumption for an
infrequently consumed product. It also tends to understate the total number of
consumers. This survey also extrapolated data on 'recalled' species and amount of fish
generally consumed in a single meal to an estimate of daily exposure over an extended
period. There is no data to support such an extension. The only way such an estimate
should be valid is by assuming a consumer eats the same species and amount in each
and every three day period throughout the year. Data from CSFII 89/90 does provide
good information on average amounts consumed in an eating occasion. Utilizing this
information with other surveys which indicate frequency offish eating occasions over a
longer period of time would produce a better estimate offish consumption and methyl
mercury intake.
4.13 Biokinetic Parameters for Children vs. Adults
The EPA assumes the biokinetic parameters for children are identical to those of
adults. The Subcommittee was asked, in Question 26, if such an assumption
introduces bias into the assessment, if the uncertainties were adequately and
consistently presented in the draft report, if the conclusion that children's lower body
weight results in higher exposure than to adults, and if using 3-day consumption survey
data could introduce an artifact affecting the findings.
There are no published data which indicate that the biokinetics of mercury differ
between children and adults. On the other hand, there are data which indicate that the
blood mercury concentrations of newborn humans or non-human primates are
approximately 1.5-1.8 times higher than those of their mothers, suggesting that the
fetus may become a sink for mercury following maternal exposure. This is also
consistent with the concept that the fetus and new born probably experience higher
exposures due in part to lower body weight and given the above data, this concept is
probably not an artifact of the 3 day consumption data. The uncertainties regarding
this concept are adequately addressed in the EPA's draft document. It should also be
noted that fetuses and newborns may be at greater risk for mercurial toxicity since
there are a number of mercury sensitive cellular and molecular processes which are
highly activated during this period. For example, methylmercury interferes with cell
migration in the brain, a process which occurs during fetal development in humans.
4.14 Methodologies for Estimating Body Burdens of Mercury
Charge question 27 asks if methodologies other than hair concentrations could
66
-------
be used to estimate body burdens of mercury.
At steady-state exposure to methylmercury, hair and blood mercury levels are
highly correlated. Both of these biological materials can be used to estimate an
individual's mercury exposure. The principal advantage of using hair as a biomarker is
the ability of segmented samples to provide an exposure history. The major
disadvantage is that hair is not routinely collected or analyzed. Therefore, researchers
who limit their analyses to hair will be unable to use existing biological specimens that
were collected for another purpose such as the National Health And Nutritional
Examination Survey study for mercury exposure assessment. In addition, since hair
mercury
analyses are less commonly done than blood and urine mercury levels, many
laboratories are unable to perform the test and costs may be higher than for blood.
Concerns have also been raised about the ability of hair to adsorb mercury from
the air. This is most likely to occur in an occupational setting. Unless the hair is
carefully washed to remove surface-deposited mercury, measured mercury levels are
likely to overestimate dietary intake rates for some individuals. The Subcommittee's
preference is to use hair and blood mercury levels in combination to assess an
individual's exposure since neither is a perfect predictor by itself.
4.15 Priority Areas for Research
In response to Question 40, the Subcommittee recommends the following
research priorities in exposure assessment:
a) Exposure measurements to better delineate the upper or high end of the
distribution of population exposures should be a high research priority.
b) Exposure measurements of pregnant women should also be a high
priority since this will enable EPA to address exposures to the most
vulnerable population - fetuses. It should be possible to obtain
statistically selected samples of blood and/or hair from pre-natal clinics
and analyze these for mercury. Complementary data on fish eating and
health of the baby should also be obtainable for this population. This will
also provide information on potential high end exposures in the
population.
c) Research is needed to estimate exposures of the population to inorganic
mercury since these exposures can add to the risks.
d) Research is needed to better delineate biological susceptibility in the
population. Even among highly exposed populations, not all subjects
67
-------
experience adverse health effects, suggesting that there are differences
in susceptibility.
e) Given the rapid changes in the types of fish being consumed and their
sources, e.g., changes in international sources of marine fish, over fishing
of some species, and increasing supplies of farm-grown fish, the
Committee recommends that EPA and FDA cooperate to provide
surveillance offish for environmental contaminants such as mercury.
f) More long-term research is needed to better understand the dynamic
behavior of mercury in the environment. Such research should included
integrated and iterative development and use of models and
environmental measurements. Opportunities for international cooperation
and coordination in such research should be sought since there is a
global component to environmental mercury contamination.
4.16 Conclusions
The Subcommittee's major conclusion, based on the existing data and the
analysis of these data, is that the majority of the population is not experiencing
methylmercury exposures that are of concern. The Subcommittee also agrees that
the high end of the distribution of methylmercury exposures is very uncertain with
respect to exposures, total number of people (and percent of the population) who may
be experiencing exposures high enough to cause adverse health effects, and the actual
sub-groups who are highly exposed. Additional comments on the human health
aspects of the EPA's draft report will be found in Section 5 of this report.
With respect to the modeling linkage between emissions from anthropogenic
sources and human exposures, the Subcommittee agrees that it is plausible that
current anthropogenic emissions are contributing to human exposures. However, the
relative contributions of current anthropogenic emission sources, global backgrounds,
recycling of old emissions and natural background to human exposures of
methylmercury are highly uncertain. Furthermore, the time constants for intermedia
transport are not known. Nonetheless, since mercury does not degrade in the
environment, current anthropogenic emissions will add to various sinks as well
as to the fraction of global mercury that is actively circulating in the biosphere,
surface waters, soils. In addition, emissions from both the U.S. and from other
countries can impact the U.S. environment and exposures of humans and
wildlife. The estimated exposures (based on transport and fate environmental
modeling) are consistent with other lines of evidence regarding population exposures.
However, this section of Volume III of EPA's draft report is not clearly written and
confuses environmental transport and fate modeling with exposure analysis.
68
-------
The Subcommittee concurs with the conclusion that fish are the major
source of methylmercury exposures for the human population. This is supported
by the exposure analysis based on methylmercury concentrations in fish and fish
consumption data. The market basket survey data of methylmercury in various foods
consumed by the population also provide strong support for this conclusion.
The exposure analysis based on fish consumption must be regarded as a
"snapshot" in time. The species offish that are being consumed and their sources are
both changing rapidly due to over fishing of some species, changes in the global
market in marine fish and more domestic fish supplied by fish farms. The impact of
local anthropogenic sources of mercury is likely to be greatest on freshwater and
estuarine fish. There are, however, two potential biases to the exposure analysis
based on fish consumption which should be checked in order to provide greater
confidence in the analysis. First, we recommend that the consumption data from the
CSFII survey should be checked against production data. These will not agree exactly
but should be reasonably consistent. Second, the average concentrations of
methylmercury in fish used for the exposure assessment were provided by the National
Marine Fisheries Service. The Subcommittee recommends that EPA determine how
measurements of methylmercury in fish which were below the analytical limits of
detection (BDL) were statistically treated in arriving at the mean concentrations. If the
BDL values were excluded, than the means are biased high. If BDL values were
included in the estimation of the mean using one of the standard methods for this, then
the means can be considered good estimators. However, if a very large percentage of
the measurements for any fish species were BDL, then the estimates of the means are
much less reliable.
The Subcommittee strongly recommends that EPA edit and shorten
Volume III on Exposure to eliminate redundancies and clarify the logical
sequence of the exposure analysis. The basic questions that should be addressed
in the volume on exposure are:
a) What is the distribution of exposures of the U.S. population to
methylmercury?
b) What are the exposures of wildlife to methylmercury?
The introduction of this volume should more clearly lay out the three approaches
that are used to estimate exposures to methylmercury - fish consumption data
combined with methylmercury concentration measurements in fish; scenario exposure
estimates based on environmental transport and fate modeling to estimate
concentrations in freshwater fish; and existing, albeit limited data on mercury
concentrations in hair.
69
-------
Environmental transport and fate modeling is used in two ways in this volume,
and it is presented and discussed in a confusing way. First, the volume should be
revised to clearly distinguish between environmental transport and fate modeling and
exposure analysis modeling (which are confused in this document). These two
techniques are not the same thing. Secondly, transport and fate modeling is also used
in two ways in this volume. It is used to estimate of methylmercury concentrations in
fish for human and wildlife exposure estimation. It is used as well to provide a basis for
linking domestic anthropogenic emissions of mercury to methylmercury exposures of
the U.S. population and wildlife. These two different applications need to be clearly
distinguished. Finally, the materials on mercury measurements in hair should be in
Volume III since they provide a biomarker of exposure.
In the Subcommittee's opinion, EPA's Dioxin Risk Assessment report (EPA,
1995) provides a very good model for the organization and analysis of
information on mercury.
70
-------
EXPOSURE APPENDIX 1
General Comments
Volume III should be rigorously edited for greater clarity, improved logical flow and
elimination of redundancies. This volume is now written with an emphasis on the
transport and fate modeling of emissions, rather than an objective scientific analysis of
human (and wildlife) exposures to MeHg. The central question that this Volume should
be addressing is "What is the distribution of exposures of the U.S. population?" Some
suggestions for reorganizing and editing this volume are as follows:
a) The Introduction should present an overview of the three different
approaches used to estimate exposures of the population to MeHg and
explain that each had a somewhat different purpose. The introduction should
also present a clear road map of the overall flow of the arguments to be
made. This would help the reader enormously. As now written, the reader
must figure this out the hard way - read through the entire volume and then
go back and re-read sections to make sure he/she understands what it's all
about. These three approaches and their purposes are:
First, transport and fate modeling is used to estimate the possible
contributions of anthropogenic sources of Hg to MeHg in fish. Long-range
transport and fate modeling is used to estimate the "average" or background
concentrations of MeHg in fish, etc., from emissions that are dispersed over
long distances. The contributions of local sources are added to the
"background" contributions to estimate totals. Exposure scenarios are then
used to evaluate the potential contributions of anthropogenic sources to
exposure through the food chain. This should be explained succinctly in the
introduction.
The second approach is to combined diet consumption data with the
concentrations of MeHg measured in food in order to estimate the average
exposure for the population. This is probably the most important and
extensive analysis of exposure presented. lt"s the only one that gives an
estimate (and it is reasonably robust) of the average exposure of the
population. This should not be relegated to an Appendix (H).
The third approach is to examine the concentrations of Hg in hair to see if
they are consistent with the estimates of average exposures from the dietary
analyses. The Hg in hair data should be included in Volume III. These data,
while not as statistically representative as the dietary exposure assessment,
provide important supporting evidence that the dietary exposure assessment
is correct.
71
-------
b) The overall organization of this volume is very confusing. We suggest that a
more logical organization would be as follows:
Present the environmental measurements first, as Chapter 2. The purpose of
presenting these data is to show the levels in different media and conclude
that air, water and soil are NOT major exposure concerns. However, the food
measurements do indicate that fish is the major source of exposure. The
tables that are presented should be only those need to provide the evidence
that these are scientifically sound conclusions. It may also be possible to
eliminate some of the tables in this section, particularly those that are
redundant with tables in Appendix H.
Also, the section on the chemical properties would be more logical in Volume
2, before the discussion of source emissions. (See the Dioxin Report).
Chapters 3 & 4 should present the transport and fate modeling, with an
explanation of the purposes of the modeling. The environmental
concentration predicted by the modeling can be compared to the existing
measurement data, where appropriate, to demonstrate that the models and
measurements are consistent (one purpose). The second purpose is to use
the modeled concentrations in scenario estimates of potential human
exposures.
Chapter 5 should then be the exposure analyses for the scenarios. This
chapter also provides some upper bound estimates of exposures.
Chapter 6 should present the population exposure assessment based on
dietary data. lt"s very surprising that this ends up in an appendix instead of
being a main feature of the exposure assessment volume. This and the hair
Hg are much more immediate estimates of human exposure than multi-media
models that depend on many input parameters that are not well known or not
known at all.
Chapter 7 should present the Hg in hair data and compare these data to the
data from Chapter 6 and from the scenario modeling (Chapter 5).
Transport and fate modeling is used to understand the processes by which
the pollutant is transported through the environment and finally comes into
contact with humans or wildlife. But it is not exposure assessment perse!
Exposure analysis requires the additional step of estimating contact between
the pollutant in an environmental medium with a human (or animal).
It is strongly recommended that the definitions of exposure and dose, which
are given in this volume, be used correctly and consistently throughout the
72
-------
volume and that transport and fate modeling be clearly distinguished form
exposure analysis.
Specific Comments
The term "exposure" is frequently confused with (and used interchangeably with)
transport and fate modeling in this volume. These are not the same. Exposure is
defined correctly on p. 1-9 and 1-13 but then not always used correctly throughout this
volume. For example, p. ES-2 - "Exposure Assessment of Local Deposition of Mercury"
- delete "Exposure." This section discusses deposition of Hg, not exposure. Another
example, on p. 1-10, paragraph 2 - direct measurements of mercury concentrations in
source emissions DO NOT provide an estimate of exposure - unless the human is
inhaling or ingesting the emissions.
H-4 First full paragraph is confusing. Why is the default value 6.5 g/day used by EPA if
the overall consumption rate from the NPD survey was 14.3 g/day? Which rate is the
last sentence referring to?
Table H-1 Do men and women over age 75 really report serving sizes that are 12%
larger than those consumed by working aged adults? Has this pattern of consumption
changed over the past 20 yrs?
Table H-2 Title could be improved to better describe the table contents. The no. and
(%) reporting would be useful in second column, e.g. 70 (17%).
Table H-3 Shows grams per day in each column. Why aren't children aged 2-11
included?
Table H-4 Why are 2-9 yr olds omitted? It would be helpful to include lower
consumption rates in table so that the columns total to 100%.
Tables H-6 and H-8. The footnote indicates that the data were weighted to be
representative of U.S. population. How was this weighting accomplished? Some
explanation of the weighting procedure should be included in the text.
Table H-10 Includes the Peterson et al. study. The percentage of males and females
that reported eating 3 or more fish meals in this study are inaccurately listed as 0.15
and 0.26%. The percentages were actually 15 and 26%.
Pages H23 and H24 discuss the dependence of Alaskans from subsistence economies
on marine wildlife including marine mammals. However, there is no attempt to
quantitate dietary mercury intake for this susceptible population and no mention of the
73
-------
source of mercury to the habitat of these animals. Such information would be a
valuable addition to this section of the report.
Page H-24. The Peterson etal. study is again cited. Again, the percentages are listed
as 0.15 and 0.26%. These should be corrected to 15 and 26% as in Table H-10 (see
comment above).
Page H-26 First sentence - the word anadromous should be defined since it is likely to
be unfamiliar to many readers. Second full paragraph - something appears to be wrong
with the units in this paragraph and in Table H-19. The passage states that the finfish
consumption rate for male members of these tribes was 14 times the default rate of 6.5
g/day. [6.5 x 14 = 91 g/day]. The next sentence indicates that the 50th percentile for
these tribes was 32 g/kg bodyweight/day. [32 g/kg x 70-kg = 2240 g/day] Later in the
paragraph the author states that the daily fish consumption rate for both tribes was 73
g/day. None of these values appear in Table H-19. Instead most of the values in this
table are well below 1 gram/day.
Table H-20. Suggest changing the title to "The percentage of nursing women who
reported consuming locally-caught fish."
Table H-21. Fish purchases are reported in pounds per year. All other tables in this
appendix have used metric weights of kilograms per year or given both units.
Tables H-22, 23, 24, 26 & 27 The listing of a mean values would be more meaningful if
the number of samples tested and the standard deviation were included. The title
should be revised to indicate that the values are averages.
Page H-42. The discussion of the "National study of chemical residues in fish" appears
to have focused on mercury levels in fish caught at sites that were selected based on
their proximity to contamination sources. Most of these contamination sources appear
to have posed a threat via surface water discharges or groundwater seepage. This
should be discussed in the report since the mercury levels measured in these fish are
unlikely to be related to atmospheric deposition.
Tables 29 and 30. Please include an explanation of how the methylmercury exposure
levels were calculated. It might also be helpful to explain why the estimated mercury
intakes are essentially the same when calculated using data sets compiled by Bahnick
and Lowe, and to explain how the data were weighted as mentioned in the footnote.
Page H-65 second paragraph. This paragraph was confusing and difficult to follow.
74
-------
5. HEALTH ENDPOINTS AND SUSCEPTIBLE SUBPOPULATIONS
5.1 EPA's Assessment Methodology
The EPA reports available data demonstrating increasing severity and frequency
of nervous system effects (particularly impairment in visual-motor integration) among
pediatric subjects with increasing levels of maternal hair mercury. In question 28, the
Agency requested comment on the soundness of its assessment methodology, and on
the presentation of the uncertainties in the characterization of potential effects in the
draft text.
The U.S. EPA has established an RfD for ingested methylmercury (MeHg) at 0.1
ug/kg-day based on the presence of the developmental delay or neurological deficits in
infants exposed prenatally to MeHg in Iraq. The RfD in the draft report was based on
the calculation of a benchmark dose approach that does not assume or require a No
Observed Adverse Effects Level (NOAEL). A benchmark dose was estimated using a
Weibull model on grouped data with a 95% lower limit on a 10% response level. These
calculations yielded a benchmark dose of 11 ppm concentration of mercury in maternal
hair. This was then converted to a mercury blood level assuming a ratio of 250:1. An
uncertainty factor of 10 was used to account for variability in the human population,
variation in the biological half-life and variation in the hair to blood ratio. All these
assumptions were extensively documented. Additional analysis was performed using
different groupings of the data which yielded similar benchmark dose levels. Basing the
RfD on the Iraqi study is not ideal. This study involved an acute high level exposure
from contaminated seed grain, rather than the chronic low level exposure from fish
consumption that is the principal risk the RfD is designed to protect against. The report
cites several studies involving chronic exposure from fish, including the New Zealand
and Cree studies (see Table 2, below), from which benchmark doses have been
derived, which are very similar to that derived from the Iraqi data. Moreover, recent risk
assessments derived from experimental animal data (Gilbert and Grant-Webster (1995;
Rice, 1992; 1996) supported a similar RfD. The conclusions of the draft report
would be strengthened considerably if the authors were to emphasize the
convergence of these data from multiple studies based on different ethnic
populations and species, exposures and developmental endpoints, all
suggesting similar RfDs. The Subcommittee has prepared a summary of the study
data that is overwhelmingly supportive of the EPA RfD (see Table 2 - prepared in part
from a table presented by EPA at the meeting).
Investigators conducting two new major prospective longitudinal studies- one in
the Seychelles Islands the other in the Faeroe Islands- have recently begun to publish
findings in the literature and are expected to continue releasing their findings during the
next 2-3 years. These studies have advantages over those cited in the previous
75
-------
Table 2 Summary of Estimates for the Effects of MeHg in Humans and Animals
Study
Iraq
New
Zealand
Duration of N Exposure Endpoint
Exposure
ST- seed grain 81 in utero all developmental
all developmental -
"timing" endpoints
all developmental
delayed walking
all developmental
delayed walking
delayed walking
delayed walking
LT-fish 237 in utero IQ (WISC-R),
language
Analysis
BMD
BMD
NOAEL
LOAEL
LOAEL
threshold
threshold
threshold
threshold
Maternal Hair
Level (ppm)
11
15
7-10
14
11?(44ug/L
blood)
10
80
>100?
5-15
Hair:
Blood
250
250
250
250
250
Dos
1.1
1.5
1.0
1.2
0.7'
0.8
used a PBPK model
BMD
10-31 (17)
Cree Indians
Amazon
natives
Minamata
Peru
Data review
Data review
LT - fish
LT
LT?
LT
Animal & Human
Animal & Human
247
29
986
131
in utero tendon reflex
adult visual discrimination
adult Minamata disease
in utero
in utero
LOAEL
LOAEL
threshold
NOAEL
LOAEL
LOAEL
10-20
20
20
30
10-20
10
250
250
250
250
250
1.0
1.0
1.0
Adapted in part from slide presented by EPA at the Subcommittee's public meeting.
LT - Long-term; ST Short-term
paragraph in that they have much larger samples sizes, a larger number of
developmental endpoints, potentially more sensitive developmental endpoints, and
control a more extensive set of potential confounding influences. On the other hand,
the studies have some limitations in terms of low exposures (to PCBs in the Faeroes)
and ethnically homogenous societies. Since only a small portion of these new data
sets have been published to date and because questions have been raised about the
sensitivity and appropriateness of the several statistical procedures used in the
analyses, the Subcommittee concluded that it would be premature to include any data
from these studies in this report until they are subjected to appropriate peer review.
Because these data are so much more comprehensive and relevant to
76
-------
contemporary regulatory issues than the data heretofore available, once there
has been adequate opportunity for peer review and debate within the scientific
community, the RfD may need to be reassessed in terms of the most sensitive
endpoints from these new studies.
The EPA draft document provided an extensive review of the uncertainties
associated with the studies cited. The reanalyses of the Iraqi data using various
grouping factors is commendable. However, we recommend that the EPA consider
information that suggests that the uncertainty factor be increased. It is striking
that no new data are available for risk assessment. For example, the Faeroe Islands
data (also animal data) indicates that the fetal exposure may be greater than maternal
exposure. In this study fetal cord blood mercury levels averaged 80.2 ppb while
maternal blood levels were only 38.1 ppb. Animal data supports that the fetus may act
as a sink for mercury. The report extensively reviews blood mercury kinetics but has
little to say about fetal brain mercury levels. Although the data are slight there are
indications from recent monkey studies that the brain mercury half-life is very long
(Vahter et al., 1995). The RfD for MeHg is based on results from an acute exposure
study while most MeHg exposure is thought to be long term. This may be an additional
reason to increase the uncertainty factor. There are also indications of age related
changes where MeHg may accelerate neurodegeneration associated with aging from
human data (Igata, 1993) and animal data (Rice, 1989a; 1989b). In evaluating
neurotoxic effects from low exposures such as with methyl mercury, it must be
remembered that few individuals may actually demonstrate clinical signs of disease but
many individuals may suffer subtle changes which can produce total population effects.
5.2 Alternative Neurological Endpoints
Data from experimental animals (including primates with long-term exposures to
methylmercury) show methylmercury-induced nervous system damage, particularly on
the visual system, although the animals appear clinically normal. The traditional RfD
methodology neglects such impairment. In question 29, the Subcommittee was asked
to determine if these data are important endpoints, and if they are appropriately
characterized in the draft report.
The reviewers were impressed by the draft report's thoroughness of the review
of the animal literature as it existed in 1994. The document appropriately distinguished
among the different forms of mercury (elemental, organic, and inorganic) and it
distinguished between adult and developmental exposure. The tables identifying
LOAELS/NOAELS in animal studies were most helpful, although a few errors are noted
below. We are generally in agreement that the comments below can and should be
addressed quickly, but we also agreed that timely publication is a high priority.
Answers to the specific questions, and critiques of the discussion of the animal
literature are provided below.
77
-------
The endpoints described in the animal literature are important and they
have been induced by dosing protocols that are relevant to human exposures. In
experiments using nonhuman primates, sensory (visual, somatosensory, auditory),
cognitive (learning under concurrent schedules, recognition effaces), social play, and
schedule-controlled operant behavior (not referenced in the tables) are all identified as
having been affected by methylmercury. The sensory, cognitive, and motor deficits
appear reliably over a consistent range of doses in nonhuman primates exposed to
methylmercury during development. Subtle, but believable and important deficits
appear in several functional domains. The draft document calls these 'subclinical,'
which is true only in the sense that they are not overtly obvious upon casual
observation. These are identifiable signs when appropriate testing conditions are
applied, conditions that could be applied in clinical settings.
The rodent studies buttress these conclusions in general, though the results are
not always consistent. The rat is the most common rodent species used and the
endpoints identified to date have usually been less specific than those examined in the
primate literature, but schedule-controlled operant behavior (Bornhausen etal., 1980;
Schreiner etal, 1986) and subtle characteristics of motor function (Eisner, 1991) have
been examined. The latter two approaches represent some of the more sensitive
endpoints in the rodent literature, but these papers are not without their difficulties.
The Bornhausen experiment used such low doses that replication is called for before
too much weight is given to that study. Eisner's study used sophisticated endpoints on
rats that could be trained to perform the task with extensive analysis conducted on the
best performing subjects. In that study, performance on a task in which the
methylmercury-impaired rat was required to produce a carefully defined force with its
forepaw was affected. This procedure was sensitive, the effects are consistent with
some effects identified with humans and nonhuman primates, and are consistent with
the accumulation of mercury in areas of the nervous system important to motor
function. Many of the other rodent studies used less sensitive screening measures, but
nevertheless deficits have been identified, and in a range of doses consistent with the
different blood-brain ratios usually assumed for the rat (0.06) as compared with the
nonhuman primate (2-4).
Where studied, there is some overlap in the effects of mercury vapor and
methylmercury, but there are important differences, too. Appropriately, the draft
document keeps the different forms of mercury (organic, elemental, and inorganic)
separate.
The sensory and motor deficits imply directly that the exposed individual is
missing the full complement of important capabilities. Moreover, recognition that
forms of learning and reading dysfunction in people can be traced to subtle
alterations in sensory systems is growing, so these findings raise concerns
about deficits in functional domains not traditionally linked directly to sensory
78
-------
function. The motor deficits are consistent with neural systems that are affected
by methylmercury and therefore indicate important deficits. Whether the cognitive
endpoints are traceable to this sensory loss remains to be determined but some, such
as the learning deficits under concurrent schedules or alterations in fixed-interval
schedule performance may be independent of such loss.
The description of the relevant endpoints misses, in places, the importance of
the endpoints studied. In this sense the coverage is more encyclopedic than
interpretative. Similarly, describing the visual contrast sensitivity function as simply a
measure of spatial vision misses the point of the investigation. This approach to visual
psychophysics provides a rich lode of information about function and dysfunction in the
visual system. Contrast sensitivity functions enable links between important features of
visual function as expressed in behavior and the neural mechanisms underlying vision.
The learning impairments observed in behavior under concurrent schedules not only
raise concerns about cognitive effects of methylmercury exposure but also point to
behavioral mechanisms by which these effects occur. The primate studies out of the
University of Washington can make contact with cognitive and visual impairment.
Deficits in visual recognition effaces make contact with well-established areas in the
primate visual cortex that are tuned to identify complex features such as faces. Results
suggesting deficits in the visual recognition effaces make contact with well-established
areas of neuroscience that show how higher-order functioning is accomplished in the
primate sensory (including visual) cortex. Although somewhat preliminary, these
results could point to links between the integration of complex visual information and
higher order cognitive abilities.
The rodent studies generally find effects at doses that might be predictable
based upon the kinetics of methylmercury in these species and the sensitivity of the
procedures used. Some of the studies, e.g., those by Eisner, are sensitive, consistent
with other known effects of methylmercury, and should be taken quite seriously. The
authors of the draft report have provided a comprehensive review of all studies, both
positive and negative. However, it would be helpful if they provided further
interpretation of how the studies on rodents relate to human risk.
The most disappointing aspect of the section on the health effects of
mercury (elemental, inorganic, and organic) is that it fails to link the animal
studies to the risk assessment process. The effects of methylmercury in these
studies are important ones and the doses at which they occur are consistent across
studies. Where the human epidemiological studies are non-experimental in nature and
therefore contain confounding variables and unknown terms which may make it difficult
to determine causality; the animal studies are experimental in nature and provide better
control over potential confounding factors. As pointed out at the public meeting,
however, the animal studies are poorly integrated into the general document.
This is unfortunate because the animal studies have much to offer to the risk
79
-------
assessment process.
First, laboratory studies can be drawn upon to examine critically the absence of
effects in human studies. For example, if contrast sensitivity functions, high-frequency
hearing, or learning are especially sensitive to methylmercury, then the absence of an
effect in an epidemiological study can be interpreted only if these functions are
examined directly. Human studies that rely on relatively insensitive screening
measures rather than refined measures of functions known to be affected by mercury
should be interpreted cautiously if refined measures are absent. The primate studies,
and even some of the rodent studies are clear in showing that subtle features of
behavior are affected in animals that show no gross impairment.
Second, these studies can be used to resolve ambiguities in threshold
estimates. Some estimates from epidemiological studies are so heavily dependent on
the statistical model that order-of-magnitude differences in the estimated threshold
arise. These disagreements can be resolved by a review of the animal literature. For
example, many pages are devoted to detailed discussions of statistical models used to
characterize four cases of delayed walking in the Iraqi study but not a single line draws
from the large and growing animal literature to support the different arguments, even
though empirically grounded interpretations from well-controlled experiments are
available.
Perhaps most important, an independent risk assessment can be conducted with
the animal studies. This has been accomplished recently (Rice, 1996, Gilbert and
Grant-Webster, 1995) and the RfD's recommended there are lower (but within an order
of magnitude) than those recommended by the U.S. EPA. This lower estimate coupled
with the long half-life of mercury elimination from the brain (Vahter etal., 1995) raise
the
possibility that an additional uncertainty factor might be considered to account for
accumulation and possible neurodegenerative effects.
As an additional note, the summaries of the animal studies, and especially some
of the primate studies, are accompanied by criticisms about the small sample size
used. A small sample size is a problem if one is trying to determine a benchmark dose,
but in many regards the small sample size is a strength. This experimental tactic
permits the use of refined endpoints to identify effects of low exposure levels, can point
to mechanisms of toxicant-induced behavioral disruptions, and facilitates links to other
studies, such as kinetic or in vitro experiments. Because of the refined analysis
permitted by smaller studies, these studies, especially the primate ones, identify
LOAEL's that can form the basis for a comparative risk assessment between species.
The Subcommittee offers the following specific recommendations, while
urging that timely publication of the document be given the highest priority:
80
-------
a) Integrate a discussion of risk based estimates upon the animal studies
that have been published. Do so by drawing on papers published since
this document was written (Stern, 1993; Rice, 1996; Gilbert and Grant-
Webster, 1995).
b) Update tables and other descriptive information with papers that have
been published describing animal studies since the document was
written. We limit this recommendation to the animal literature because of
uncertainties in how to interpret recent human epidemiological studies.
Some studies are listed below.
c) Implement suggested corrections and clarifications found in the Appendix
of this report.
d) Where it is possible to correct interpretations of the animal literature do
so quickly. Don't delay the document too long by rewriting entire sections
5.3 Fish Consumption Patterns and the "Developmental Window"
The draft report provides information on fish consumption drawn over relatively
short time periods (e.g., days). These data are used to extrapolate consumption
patterns over longer periods (e.g. a month). Although the exact developmental window
affected by methylmercury in humans is not precisely defined, it is thought to be less
than three months. The Subcommittee was asked, in question 30, to advise if such an
extrapolation was appropriate in this case, and if the draft document appropriately
presented the limitations of this methodology.
The Subcommittee noted that data from studies of humans and animals
presented in the draft report do not support the notion of a precisely defined 3 month
window of vulnerability for the effects of methylmercury on the developing fetus or child.
In contrast, the Subcommittee indicated that there is sufficient data to conclude
that the developing organism is vulnerable during the entire period of
development and that in utero as well as early postnatal exposure to
methylmercury is of concern. The Subcommittee also indicated that intermittent or
short-term exposure to methylmercury at a critical period in development should also
be considered. Exposures prior to pregnancy may also be of concern given the half-life
of methylmercury.
The draft report presents some but not all of the limitations of the methodology
used to assess Hg exposure from fish consumption. The report notes that available
data are frequently based on samples gathered up to 20 years ago; it should also state
that, although Hg levels in marine species appear to have remained fairly stable over
the past 20 years, levels in freshwater fish have fluctuated considerably. Information
81
-------
on recent changes in marketing that have led to increased consumption of non
contaminated farm-raised fish should also be incorporated. More emphasis should be
placed on the paucity of data on high level fish eaters who are likely to be the only
populations at risk and particularly on the dearth of data on exposure of women of
child-bearing age. One study of Mohawk Indians in upstate New York indicates a
dramatic decrease in fish consumption among women of child-bearing age upon
learning of its potential health hazards.
5.4 Subtle Endpoints Versus Traditional Metrics of Child Development
Volume 4, Section 6.3, and Volume 6, Chapter 2, Sections 2.1 and 2.2.21 of the
draft report present EPA's interpretation of the data demonstrating increasing severity
and frequency of nervous system effects (particularly impairment in visual-motor
integration) among pediatric subjects with increasing levels of maternal hair mercury.
In question 31, the Agency sought the Subcommittee's comments on the possible
biases introduced into the assessment by focusing on subtle endpoints of
neurobehavioral function, rather than the traditional metrics of child development and
on establishing the conditions under which such subtle measures should be employed.
Advice was also requested as to whether the available data indicate that children are
more sensitive than adults to the effects of methylmercury.
The Subcommittee's findings on these issues are herein summarized (Detailed
discussion follows below):
a) Subtle neuralbehavorial measurements should be used even if traditional
metrics are within normal ranges.
b) No specific biases are introduced by the use of subtle neuralbehavorial
endpoints. However, given the questions about the Iraqi study, the EPA
report should clearly acknowledge the potential significance of these
issues.
c) The data are unequivocal in indicating that the fetus is more susceptible
to mercury toxicity than is the adult
The Subcommittee interprets the term "subtle endpoints" to mean scores yielded
by individually-administered sensory-motor or neuropsychological tests. Focusing on
subtle endpoints versus ".... traditional metrics of child development" such as age at
achievement of milestones (e.g., walking, talking) will not introduce bias in the usual
sense of distorting the estimate of the magnitude of an association because of
selection bias or confounding bias.
Insofar as the wording of the question posed to the Subcommittee implies the
82
-------
equation of "subtle" with "more sensitive to performance variation within the subclinical
range," a statistical test involving a more subtle endpoint may have greater statistical
power if the gain in sensitivity is not matched by a loss in specificity and thus be more
likely to detect a non-zero association with the index of exposure, should one actually
exist. The key issue would seem not to be one of bias, but of interpretation, namely
whether the performance differences found using more sensitive endpoints are
important enough to act on. The lead and PCB literatures provide useful lessons for
thinking about this issue. Neither compound would probably be viewed as a
developmental neurotoxicant, at least at low levels of exposure, if reliance had been
placed solely on "traditional metrics" such as developmental milestones. This
experience with other neurotoxicants indicates that the answer to the question "Should
these indicators be dismissed if traditional metrics are within normal ranges?" is clearly
"No."
Use of traditional metrics such as milestones of development requires
consideration of several measurement issues. Ages at milestone achievement are
necessarily based on the reports of an observer in close contact with a child, and often
a considerable amount of time may separate the achievement from solicitation of the
report. As
such, these reports are subject to a number of errors, both non-differential (random)
and differential (systematic), that are attributable to characteristics of the reporters.
The possibility of non-differential misclassification arises because the reports
depend on the intensity of the reporter's "diagnostic surveillance" for the target
behavior, the acuity of the reporter's observational skills, the criteria used by the
reporter to identify when the milestone was achieved (e.g., when is a verbal
approximation accepted as a "word"?), and the accuracy of the reporter's memory of
the milestone achievement. The distributions of responses provided by the Iraqi
parents about the ages at which their children walked and talked suggest that some of
these factors may have influenced these reports. The most notable feature is a clear
digit preference for both judgments. For "age at walking," 70 of 78 values are even
numbered months, and 50 or nearly 3 of 4 responses are multiples of 6 months (i.e.,
12, 18, 24, 36, 60, and 72). This is true for "age at talking" responses as well (70/73
are even numbered months). These patterns are unlikely to have occurred by chance
alone and suggests that the respondents were giving "ball park" responses.
Depending on the way Iraqi nomads mark the passage of time, it may also indicate that
the interviewers had to suggest response categories for the respondents (The report
does indicate that no particular significance is ascribed either to the onset of walking or
to date of birth in this culture). This may not be a big problem if the data are analyzed
categorically3 (since distinctions within the groups below the cut point or above the cut
point are irrelevant). As noted in the EPA report, however, response misclassification
Especially if the category includes a preference point as the middle of a category.
83
-------
will be a problem when it occurs in the neighborhood of the cut point chosen as
"abnormal". Various analyses summarized indicate the extreme sensitivity of the
threshold of effect to the choice of cut point. Without having Iraqi norms for ages at
walking and talking, it may be inappropriate to designate any cut-offs for identifying an
"abnormal" response (particularly if that "point" represents a preference number).
Treatment of age at milestone achievement as a continuously distributed variable
would reduce the problem of misclassification in the neighborhood of the cut-point. It
would be interesting to learn about the estimated ages of the Iraqi children at the time
they were identified. Was the age of the child at the time a respondent was interviewed
associated, either positively or negatively, with age at which milestones were reported
to have been achieved? If so, the next issue to address would be whether child's age
at time of interview was associated with estimated mercury burden?
The draft report says that the data were collected 30 months after the episode,
although this is difficult to reconcile with the fact that one child was reported to have
been 6 years old before he or she walked or talked and another 2 to have been 5 years
old. If the interval really was 30 months, one would expect some censoring of the data,
i.e., missing data because some children had not yet achieved the milestones at the
time of the interview. Some data are missing, but these are not identified as due to
censoring.
Differential misclassification is a type of systematic bias arising when a
respondent's judgment about whether or when the target behavior occurred is affected
by knowledge of exposure status, i.e., a mother who knows that she consumed more
contaminated food grain than most other women tends to overestimate the age when
her child began to walk or talk. It is not possible to evaluate the extent of this type of
misclassification in the Iraqi data set.
Aspects of the Iraqi data raise doubts about its quality, but do not clearly
suggest whether the problem is non-differential or differential misclassification. For
"age at walking," 46% (36/78) reportedly did not walk until 18 months of later. In the
standardization sample of the 1969 version of the Bayley Scales, 95% of children met
the criterion for "walks alone" (at least 3 steps without support) by age 17 months
(Bayley, 1989). In the Iraqi sample, 37% of the children (27/73) did not produce 2 or 3
words until 24 months or later. In the U.S. standardization sample for the MacArthur
Communicative Development Inventory, a parent-completed checklist, the median age
at which children were speaking 2 or 3 words was 10-12 months(Fenson et al., 1993).
At age 24 months, the 5th percentile for Vocabulary Production was 70 words for girls
and 48 words for boys, and by this age, 89 of girls and 83% of boys were combining
single words into rudimentary syntactic structures.
The data presented by Marsh et al. (1987) may accurately mirror Iraqi norms, or
may reflect a sample in which children with developmental delays are greatly over-
84
-------
represented (even among infants with low maternal hair mercury levels).
Besides being more sensitive than caretaker memories of ages at milestone
achievement, assessments of more subtle sensory-motor or neuropsychological
endpoints are generally less subject to these sources of classification error because
they typically involve standardized administration of items, clear definitions of the target
behavior and criteria for response classification, and can be completed by an individual
blinded to exposure status and whose performance can be monitored by some quality
control procedure.
The issue of endpoint sensitivity also bears importantly on the question of
whether one is interested in drawing inferences about the individual that is exposed or
the population that is exposed. An individual-centered approach would frame the
public health impact of exposure in terms of clinically significant impact on a particular
child (e.g., a delay in walking or talking that would warrant a referral for further
evaluation). A population-centered approach would frame the public health impact in
population terms. A 5 point decline in the IQ of an individual child, roughly 1.5 times
the standard error of measurement of full-scale IQ, is unlikely to have clinical import for
that child, i.e., to move the child's score out of the normal range. Weiss' example of the
implications of a shift of 5 points in mean IQ on the population distribution of IQ scores
shows that this might result in a doubling of the numbers of the children with scores in
the clinically meaningful range (Weiss, 1988).
The Subcommittee recognizes that there is little EPA can do at this point to
address these concerns about data quality in the Iraqi studies and, specifically, to
reduce the uncertainties attending their interpretation. The EPA report should,
however, more clearly acknowledge the potential significance of these issues and their
potential impact on the results of the risk assessments conducted.
We assume that use of the term "childhood exposure" here excludes fetal
exposure, referring only to postnatal exposure. The data in humans and in several
animal models are unequivocal in indicating that the fetus is more susceptible to
mercury toxicity than is the adult. One of the most important lessons learned from the
Minamata and Iraqi episodes of congenital methyl mercury poisoning is that fetotoxicity
can occur in the absence of clinical signs or symptoms in the mother. In contrast, little
evidence is available on the relative susceptibility of children versus adults. In most
human studies, it is impossible to discriminate the effects of fetal exposure from those
of lactational exposure or to exposure incurred by the postnatal child by virtue of his or
her own patterns of consumption. Animal data appear to suggest age-related changes
in mercury metabolism that would likely be expressed as increased vulnerability in the
early postnatal period (e.g., greater retention of ingested mercury by suckling rats
compared to adults, especially in brain) (Null, et al., 1973). In addition, data from
animal models and human neuropathology studies clearly indicate that the immature
85
-------
brain is more sensitive than the mature brain to methyl mercury Rheul, et al., 1979). In
the adult, selective focal damage to the visual cortex, granule layer of the cerebellum,
and somatosensory cortex may be seen following high dose exposures.. In contrast,
damage to the developing brain of the fetus and infant is diffuse and widespread,
resulting in microcephaly, broad gyri, shallow sulci, ectopic neurons and enlarged
ventricles. Animal models confirm that methyl mercury can cause late mitotic arrest in
proliferating neurons (Sager et al., 1984). Examination of the brains of adult, infant,
and fetal victims of Minamata disease clearly show a strong relationship between
developmental stage at the time of exposure and the extent of cortical and subcortical
damage (Takeuchi, 1968; Takeuchi et al., 1979). Another reason to assume an age-
associated decline in vulnerability is the fact that many of the developmental processes
presumed to underlie neurotoxicity in the fetus continue into the postnatal period,
including myelination and organization of the neuronal cytoarchitecture involving
processes of cell proliferation, migration, and organization of cellular layers (Rodier,
1994). A similar inference is supported by the available evidence regarding the
molecular bases of mercury neurotoxicity, which include interference with Ca2+
homeostasis and ion channel function (particularly the PKC pathway), with
neurotransmitter function, and with microtubule function, all of which may ultimately
affect the fine structure of the nervous system (Rodier, 1994). It should also be pointed
out that these effects are clearly less severe than lead encephalopathy and may not
require the conservative safety factors that are applied when such severe sequelae
follow exposure to other neurotoxicants.
5.5 Uncertainties in the Characterization of Potential Effects
Charge question 32 asked if the uncertainties in the characterization of potential
effects were accurately described in the draft text.
The Subcommittee found, that the draft report included a very good discussion
of the uncertainties in the characterization of the potential effects of MeHg. There are
some specific issues requiring discussion, however.
Appendix D (volume 4) presents a comprehensive set of complex highly
quantitative simulations and sensitivity analyses to assess the impact of differing levels
of uncertainty (e.g..producing a series of cumulative bootstrap threshold distributions
multiplied by values of the dose conversion distributions). In that Appendix, the report
states "The principal uncertainties in calculating the reference dose arise from the
following sources: the variability of susceptibilities within the Iraqi cohort; population
variability in the pharmacokinetics processes reflected in the dose conversion;
response classification error; and exposure classification error."
The text seems to assume (p. D-2, D-3) that the only type of exposure
misclassification of concern is classifying an unexposed child as exposed because of
86
-------
uncertainty as to whether the period of exposure overlapped with the developmental^
critical
period. The same uncertainty could result in the inverse classification error of
assuming that an exposed child was unexposed.
Considerable attention is focused on the assumptions underlying the bootstrap
method used to generate the threshold distributions (e.g., p. D-5: assumes that "...the
observed sample was a random sampling of a larger population and that each
observation was equally likely to occur in additional samples." In the Subcommittee's
view, greater consideration should be given to the sampling mechanism that generated
the Iraqi mother-infant pairs, to issue of whether this data set can support analyses of
the sophistication and precision of the bootstrap models.
The draft report does a better job identifying the uncertainties in the purely
toxicologic factors than it does the uncertainties associated with epidemiologic design
issues.
One uncertainty mentioned but dismissed is selection bias in the Iraqi sample.
The sample of mother-child pairs studied by Marsh et al. (1987) appears to be a
convenience sample insofar as no sampling frame or referral mechanism is described
to indicate how the pairs contributing data came to be included in the study. One
concern is that the characteristics of those who were selected (or self-selected) for
study differed systematically from those who were not. One uncertainty, for instance, is
whether the likelihood of participating was associated with the severity of outcome in
the child. If so, this would distort an estimate of the association between mercury
exposure and the endpoints of interest. On the other hand, if the mother-child pairs
who make up the data set were identified by some specified mechanism, the data
would permit estimation of the population distribution of both hair mercury levels, ages
at which children achieve developmental milestones, and the prevalence of abnormal
findings on neurological examination. These can be known only if all members of the
base population have an equal probability of being selected (rarely achieved).
Some of the problem stems from the absence of a denominator. The draft report
states that, "...there are no records of the size of the population who consumed grain
treated with methylmercury fungicide. Likewise, there are no reliable estimates of the
numbers of people who consumed methylmercury-treated grain and developed signs
and symptoms of mercury toxicity, but did not obtain medical attention or become
identified as part of the epidemic" and "Whether or not those who obtained medical
care represented a more sensitive subpopulation is not known."
Is it possible that mothers who elected to participate in the study had particular
concerns about their children's developmental delays? The sample would suffer from
87
-------
selection bias if there were women in the population who did not participate, who had
equally high hair mercury levels, and whose children did not manifest developmental
delays, and so did not seek to participate in the study. By the same token, were there
women who did not participate, perhaps because they did not eat much contaminated
grain and knew that was the investigators' focus, and whose children had significant
developmental delay?
This is important because calculation of a threshold, benchmark dose, or
reference dose depends on having a denominator for the size of the exposed
population, and a background prevalence of the adverse health impact of interest. The
concept of "added risk" also requires an accurate estimate of background prevalence
insofar as it represents "... the added incidence of observing an effect above the
background rate relative to the proportion of the population of interest that is not
expected to exhibit such an effect" (vol, 6, 2-6). On what basis can a 10% incidence in
response be hypothesized if the background prevalence is not known? Any
quantitative estimates will be biased in ways that are not known. From the data
presented by Marsh etal. (1987), the background prevalence of abnormal
development appears exceedingly high, at least using U.S. data as a reference. The
prevalence of delayed walking (>18 months) among children whose mothers had hair
mercury levels <10 ppm) was 11/31 (36%) (versus 5% in the U.S. population), and the
prevalence of delayed talking (>24 months) was 6/27 (22%). In the U.S.
standardization sample of the MacArthur Communicative Development Inventory, more
than 95% of 24 month old children were producing 48 words or more (Fenson, et
a/., 1993). By 24 months of age, 89% of girls and 83% of boys were combining single
words into rudimentary syntactic structures (Fenson, etal., 1993).
5.6 Risks among Subpopulations
Traditional methods for estimating potential human health risks from
environmental hazards do not distinguish risks among subpopulations (e.g. racial or
ethnic) groups. Question 33 asks for advice on dealing differential response across
population groups, and how this factor could be addressed in the report. The
Subcommittee was also asked to determine if factors such as nutritional status, life-
styles (e.g., substance abuse), or economic status play a role is mediating these
differences between groups.
Effect modification occurs when, at a given dose of a neurotoxicant, an adverse
outcome is observed in some members of a population but not others. Epidemiologists
seek to define the characteristics of sensitive and insensitive subjects. Those
individuals who are affected may possess one or more distinctive characteristics such
as age (stage of development), gender, social class, or certain premorbid health factors
(e.g., diabetes, liver disease, pulmonary dysfunction) or genetic predisposing factors
that are not well represented in unaffected members of the population.
88
-------
Unfortunately, most studies of environmental chemical influences on child
development have treated potential effect modifiers as covariates or confounders in
multiple linear regression models without interaction terms models, or as matching
variables in comparing so-called "exposed" and "unexposed" groups. Interactions are
infrequently explored by pediatric neurotoxicologists and, when they are found, their
significance is often dismissed as a symptom of "data dredging."
However, the animal model literature is brimming with examples of neurotoxicity
enhancement or buffering as a result of species, strain, drug, and physical and social
environmental interactions. This is the "experimental system" that has been largely
ignored in environmental neuroepidemiology (Bellinger, 1995).
An example taken directly from the methylmercury literature is the phenomenon
of male vulnerability. Several epidemiologic studies have shown that the relative risk of
perinatal morbidity and mortality is higher in males (Abramowizc and Barnett, 1970;
Naeye et al., 1971), including the risk of poor reproductive outcomes and postnatal
development due to fetal exposure to industrial pollutants (Scragg etal., 1977;
McKeown-Eyssen et al., 1983). Males also have a higher rate of mental developmental
disability in the general population (Gross and Wilson, 1974; Schaffer etal., 1985) and
display more profound intellectual deficits as a result of cortical lesions (Bornstein and
Matazarro, 1984; Inglis and Lawson, 1981).
Gender-related differences in susceptibility to perinatal methyl mercury exposure
were originally reported in a study by McKeown-Eyssen in the early 1980's (McKeown-
Eyssen etal., 1983). An interesting sexual dimorphism was observed with male infants
in particular presenting with dose-related deficits in sensorimotor behaviors as
assessed by the Bayley Scales of Infant Development (Bayley, 1989). Furthermore,
the prevalence of abnormal muscle tone and deep tendon reflexes was positively
associated with methyl mercury dose in males but not females. An examination of
cases from the Iraqi episode confirmed that more severe neurological effects were
observed in males (Marsh etal., 1987). Animal experiments have also observed sex
differences in neurodevelopmental vulnerability. Thus, for example, Sager et al. (1984)
administered a single dose of methyl mercury to neonatal mice. At the lower doses,
only males evidenced mitotic arrest in cells of the granule layer of the cerebellum.
The episode of pediatric elemental mercury poisoning via medicinals earlier in
this century supports the concept that organismic and environmental factors may
modify the severity of this metal's developmental neurotoxicity. Thus, while many were
exposed to these iatrogenic preparations, only a relatively small number of children
developed the complex of cutaneous, neurologic, and psychiatric symptoms that
clinicians have called "acrodynia" (Warkany and Hubbard, 1951).
One of the important lessons of mercury neurotoxicology is that a substantial
89
-------
amount of interindividual variability in the response of fetuses and children to equal
doses of an environmental contaminant can be expected. These differences in
response may be due to genetic, nutritional, maternal, metabolic or other premorbid
factors. The difficult task is identifying these factors in rigorous epidemiologic studies
and animal model experiments.
There is evidence that lifestyle factors such as the quality of the home
environment and nutrition play a role in the expression of developmental neurotoxicity.
Although social interactions have not been extensively investigated in the mercury
literature, lead studies have found deeper neurocognitive deficits in exposed
individuals from the poorest families (e.g., Bellinger et al., 1989; Dietrich etal., 1987;
Harvey et al., 1984; Lansdown et al., 1986; Winneke and Kraemer, 1984).
Positive nutritional factors associated with a seafood diet may be one reason for
the greater delay in the onset of the Minamata as compared to the Iraqi outbreaks.
Early results from the Faeroe islands studies have shown an unexpected positive
association between cord blood methyl mercury concentrations and birth weight
(Grandjean, et al., 1992; Grandjean et al., 1995). The authors attribute the finding to
the benefits of n-3 polyunsaturated fatty acids in a high seafood diet. Selenium may
also play a protective role against methyl mercury neurotoxicity. Infant hair methyl
mercury concentrations at 12 months were also positively associated with the
attainment of motor developmental milestones. The authors attributed this unexpected
correlation to the benefits of breast-feeding which, in itself, can lead to higher methyl
mercury intake by the infant.
The protective effects of genetic and environmental factors may be evident in the
most recent results of the studies performed by the Rochester/Seychelles group
(Davidson, et al, 1995). One of the most intriguing aspects of this study is the overall
performance of the Seychelles infants on the Bayley Scales and other measures of
sensorimotor development. Thus, for example, at 19 and 29 months scores on the
Psychomotor Development Index (PDI) were very negatively skewed. The mean PDI
scores at 19 and 29 months were 1.7 and 1.3 standard deviations above the United
States means of 100 +/-16 points respectively. That is, the Seychelles means for the
PDI are well above what would be classified as "accelerated performance" on these
scales. Furthermore, the inexplicable developmental health of this sample is also
reflected in the incredibly small number of subjects attaining "abnormal" scores on the
Denver Developmental Screening Test-Revised when compared to samples in the
United States. Only three out of 737 individual examinations or 0.4% were rated as
abnormal (i.e., below the 10th percentile for U.S. norms). Developmental^, this
appears to be an extremely robust sample of infants. Accelerated motor development
has been noted in previous studies African cultures and is also observed in African-
American infants under two years of age.
90
-------
The apparent neuromotor developmental precocity of this sample leads one to
question the negative results of the initial studies. On the older version of the Bayley
Scales, both the MDI and PDI are highly influenced by the motoric skills of the infant. It
is conceivable that, by virtue of culture, nutrition, genetic factors, or some combination
of these variables Seychelles infants are buffered from the adverse neuromotor
consequences of low level in utero methyl mercury exposure. For now, this remains a
matter of conjecture and debate.
All of these factors notwithstanding, the data regarding effect modification
in human epidemiologic studies of mercury poisoning are currently too meager to
base separate estimates of human health risks or establish different RFD's for
various subpopulations.
5.7 Estimating Risks to Fetuses, Children, and Adults
Many assessments of methylmercury risk clump risk to fetuses and children
together with risk to adults. Charge question 34 asks if this approach is scientifically
valid for methylmercury.
The RfD has been set for neurodevelopmental effects on the fetus, which is
currently believed to be the most sensitive endpoint in the most sensitive
subpopulation. This approach is in accordance with current RfD methodology. Both
human epidemiological and animal experimental data suggest, however, that it may be
appropriate to consider a second, higher RfD for adult males and adult females beyond
child-bearing age. In other words, although there is strong scientific justification for the
current RfD for the fetus, there are no data justifying such a low RfD for adults. The
fetal RfD would, of course, continue to be a criteria for risk management pertaining to
Hg levels in air, but a two-tier RfD may be more appropriate for issuing fish
consumption advisories. Dual-criteria fish advisories are commonly issued by state
agencies, and these advisories would benefit from rigorous EPA risk assessments for
both the more and less vulnerable populations. The original EPA RfD of 0.3 ug/kg/day,
which was based on adult paresthesia in Iraq, provides a basis for deriving an RfD for
the less vulnerable adult population, but other new studies, including those on effects
of long-term Hg exposure on dentists and dental hygienists, should also be considered.
Unfortunately, there is a paucity of data on long-term chronic effects of lifetime
Hg exposure, particularly as the individual ages. The concern with Hg effects on older
adults is heightened by the recognition that over time MeHg may be demethylated in
the body and the Hg may be preferentially stored in the nervous system. Because the
current RfD is based primarily on the chronic short-term Iraqi exposure, we recommend
noting in the Report that future risk assessments might consider an additional
uncertainty factor to take into account possible effects of lifetime exposure. With
91
-------
regard to possible differences in sensitivity to MeHg among different ethnic groups, the
committee concluded that such differences should not be considered in assessing
human health risks unless clear evidence indicating racial/ethnic differences in
sensitivity emerges in the future.
5.8 Wildlife Effects as Harbingers of Human Health Effects
In question 35a, the Agency asked the Subcommittee to recommend what
wildlife effects (based on what metric) could be interpreted as harbingers of likely
human health effects.
Beyond the obvious behavioral, reproductive, and immunologic effects that
should be monitored in both sentinel wildlife and laboratory studies, there is
considerable evidence to support a stronger use of wildlife data. In 1991, the National
Research Council (NRC) published Animals as Sentinels of Environmental Health
Hazards. That monograph clearly outlines the use of wildlife, food animal, and
companion animal data for human epidemiology.
First in the introduction, the history of the use of wildlife in human toxicology is
described beginning with the canary in mines and then concluding with Table 1-1 listing
the environmental toxicants first identified in animals as human hazards. In Table 1-1
mercury epidemics in animals included cats in Minamata Japan developing "dancing
cat disease" in the 1950s prior to general recognition of the human epidemic; birds in
Sweden (1950s) becoming ill from eating mercurial treated grain; and (1954-1971) pigs
developing neurologic disease and death following ingestion of mercury treated grain in
the U.S. These reports clearly depict that animals can provide important indications of
human health risks. Specifically, Chapter 5 provides an in depth discussion of "Fish
and Other Wildlife as Sentinels." Table 7-1 depicts the advantages and
disadvantages of animal sentinel systems for risk assessment. The EPA should
consider including the sections of the NRC monograph listed here possibly
incorporating the section as an integral part of the EPA report to Congress.
However, we must recognize that wildlife studies will provide little if any insight
into mechanisms of action and cause/effect relationships. Laboratory animal studies
will be imperative to answer those types of questions. The data from both wildlife and
laboratory studies should be used to calculate comparable RfDs to give us more
confidence in our human data. Such calculations were published earlier by other
Members of this Subcommittee.
As a final note, we believe that EPA should be cautious in assuming that wildlife
studies are inherently less sensitive than human studies. Wildlife populations are
constantly operating at the limit of their resources and therefore stressors such as
mercury on those systems may display population effects at levels less than those for
92
-------
either laboratory or human studies. As demonstrated at Minamata, stray cats were
overtly affected long before the problem was identified in humans. Conversely,
protective mechanisms such as concurrent uptake in selenium may serve as a natural
buffer against adverse health effects of mercury in wildlife or in humans. Complex
dietary interactions are difficult to test in the laboratory setting. Therefore,
generalizations should be avoided, and all three types of data must be considered
(wildlife, laboratory, and human).
5.9 Unaddressed Human Health Issues
In Charge 35b, the Agency asked for the Subcommittee's advice on human
health issues not addressed in the coming epidemiologic studies (such as
toxicokinetics).
The Subcommittee identified several significant human health issues not being
addressed in ongoing epidemiological studies (i.e..Seychelles, Faeroes, the Agency for
Toxic Substances and Disease Registry, Canada, and Brazil):
a) Studies of the long term effects of neonatal exposures to methylmercury
b) reproductive toxicity, with an emphasis on germ cell effects and
transgenerational effects
c) immunological effects
We see the highest priority for research in human populations as:
a) Research on the long term neurobehavioral effects of mercury exposure,
including measures of neurosystem stability, recovery of delay, and the
appearance of progressive or long latency effects.
b) Measurements of neurobehavioral development, based on a presumptive
biological mechanism to assess relative vulnerability depending on age of
exposure (especially from prenatal to puberty)
c) Effects on germ cells and other genotoxic markers (lymphocytes) and
relevance to reproduction (not just fetal development)
d) Immunologic effects (T- and B-cell functions; cytokines,
immunoglobulines) and relevance to host immune status and resistance
e) Understanding basic Hg kinetics and blood:hair ratios in diverse
populations (e.g. differences between Seychelles and Faeroes)
93
-------
5.10 Research Needs
In terms of overall research needs, the Subcommittee notes the following areas
in response to Question 40 of the Charge:
a) Biomarkers of integrated and/or past exposure to methyl mercury
b) Understanding variability in response
c) Long term (even latent) neurotoxic effects; interactions of aging with
earlier mercury exposure
d) Reproductive toxicity (not developmental), especially related to published
data on germ cells, pre implantation embryo, and distribution to the
pituitary
e) Low dose dose/response studies in models appropriate to the research
design and issue under study.
f) Nature and significance of immunologic effects
g) Effects of pregnancy and lactation on distribution of mercury
h) Distribution and accumulation of mercury in brain/cellular and regional),
de-methylation in brain, role of glia in neurotoxicity
I) Improve linkage between animal and human neurobehavioral measures.
5.11 Conclusions
The U.S. EPA has established a RfD for MeHg of 0.1 microgram from the
International Agency for Research on Cancer Monographs Programme on the
Evaluation of Carcinogenic Risks to Humans (Preamble, pp7-8, Volume 67, 1996). The
RfD is based on neurological deficits and developmental delay in Iraqi children who
ingested contaminated seed grain. A benchmark dose was estimated for this
population based on the level of 11 ppm of mercury in maternal hair which was
extrapolated to an equivalent dose by an ingestion route. The Subcommittee noted
that the Iraqi population had some limitations for purposes of setting standards for
several reasons. Possible selective factors may have been involved in determining
94
-------
who participated in the original study which could bias the data. The exposure in this
instance represented an acute toxicity rather than a chronic exposure which
populations have from chronic exposure to fish with a mercury burden. Other flaws in
the epidemiologic study have been reported such as the fact that there is little
information on the background performance scores for the normal Iraqi population, a
concern which has arisen because of the very low performance of the children in this
population versus children in the U.S. However, the Subcommittee generated a table
reflecting the maternal hair level in other studies which was partially based on the data
provided by the EPA at the meeting and found that the Iraqi data and those from other
sources are compatible. This reassured the group that the use of this database is
adequate for setting the RfD in the absence of newer data.
The data from the animal literature was appropriately characterized in regard to
the available published data at the time of the writing of the report. The major deficit in
this review was that the report's authors did not adequately tie the animal data to the
human data. The animal data have identified sensory, cognitive, social play and
schedule-controlled operant behavior deficits related to mercury exposure. The data
also indicate that there are changes in the brains of animals exposed in the fetal and
infant periods which suggest that there are wide spread changes in brain including
microcephaly, broad gyre, shallow sulci, and other abnormalities which indicate that
there may be differences in effects based on age at exposure.
The data, although scanty , suggest that there may be differences in response
of human populations based on various biologic factors other than age. For example
the male of most species appears to be more susceptible than the female. It would be
important in further studies to examine those factors that, like sex, appear to influence
the risk of demonstrating neurologic deficits.
The Subcommittee was briefed on the current status of the studies in two
populations which have high MeHg exposure by virtue offish ingestion and which have
been under study for several years. Because the current results from these studies
have not been published the populations have not been used for setting the RfD. The
studies raise issues about the important influence of other factors such as genetics,
nutrition, and metabolism in determining the detectible effects of mercury on fetuses
and infants. The Faeroe Island population have demonstrated a positive relationship
between levels of MeHg and both birth weight and motor development. The unexpected
result may be related to breast feeding. Additionally, data from the Seychelles have
shown a very low rate of developmental abnormalities compared to the normal U.S. and
an accelerated Psychomotor Development Index which may be attributed to the
precocious motor development of children with African heritage. These influences of
other factors on the neurologic performance of infants exposed to MeHg will need to be
considered further in any future analysis.
95
-------
96
-------
HEALTH APPENDIX.
Volume I. Executive summary.
The units are confusing in places. Units of ppm and ug/g, are used
interchangeably, for example, as though they represent different measures. To
complicate further the reading of the summary, units of 10"4 are used instead of one of
these measures. In other places abrupt changes among biomarkers such as blood
and hair make it difficult to follow the argument. It would help to summarize the
argument using a single biomarker, even if it requires a calculated estimate. If such an
estimate is called for then, of course, it should be clearly stated that such an estimate
has been made. Attending to the inconsistencies in units of measure could make the
discussion on pages 3-19 to 3-23 more readable.
Volume IV: Health Effects of Mercury and Mercury Compounds.
Page 2-12. Table 2-2. The detection limit for atomic absorption spectrophotometry is
in the ppb range, not 0.5 ppm.
Page 3-31. The reference to the Fredriksson et al (1992) study incorrectly lists brain
levels as 1700-63000 ug/g. The true values are 0.017 to 0.063 ug/g.
Page 3-31. Newland etal., 1996 can be added to Table 3-27 describing developmental
toxicity of elemental mercury.
Page 3-64. First paragraph. Bornhausen(1980) gave doses of 0.01 and 0.05 mg/kg 5
times, not once as implied in this paragraph.
Page 3-91, Table 3-68. Omits Rice, 1992. Also, consider adding Rice and Gilbert,
1995 to this table.
Page D-4 Only three cases of delayed walking are listed below 100 ppm hair mercury.
Four cases are shown in the figure from Cox et al. that has been published from this
study.
97
-------
REFERENCES
Abramowicz, M. and H.L. Barnett. 1970. Sex ratio of infant mortality: Trends of
change. Amer J Dis Child. 119:314-315.
Bahnick, D., Sauer, C., and D.W. Kuehl. 1994. A national study of mercury
contamination in fish. 4, Analytical Methods and Results. Chemosphere\/2Q
N3:537-546.
Bayley, N. 1969. The Bayley Scales of Infant Development. San Antonio, TX: The
Psychological Corporation.
Bellinger, D., Leviton, A., Waternaux, C., Needleman, H., and M. Rabinowitz. 1989.
Low-level lead exposure, social class, and infant development. Neurotoxicology
and Teratology, 10: 497-503.
Bellinger, D.C. 1995. Interpreting the literature on lead and child development: The
neglected role of the "experimental system." Neurotoxicology and Teratology,
17:201-212.
Bjorkman, L, Mottet, K., Nylander, M., Vahter, M., Lind, B., and L. Friberg. 1995
Selenium concentrations in brain after exposure to methylmercury: relations
between the inorganic mercury fraction and selenium. Arch. Toxicol., 69:228-
234.
Bodaly, R.A., J.M.W. Rudd, R.J.P. Fudge and C.A. Kelly. 1993. Mercury
concentrations in fish related to the size of remote Canadian Shield Lakes. Can.
J. Fish. Aquat. Sci. 50:980-987.
Bodaly, R.A., R.E. Hecky and R.J.P. Fudge. 1984. Increases in fish mercury levels in
lakes flooded by the Churchill River diversion, northern Manitoba. Can. J. fish.
Aquat. Sci. 41:682-691.
Bornhausen, M., M.R. Musch and H. Greim. 1980. Operant behavior performance
changes in rats after prenatal methyl mercury exposure. Toxicol. Appl.
Pharmacol. 56:305-316.
Bornstein, R.A., and J.D. Matazarro. 1984. Relationship of sex and the effects of
unilateral lesions on the Wechsler Scales: Further considerations. J Nerv Ment
Dis. 172:707-710.
R-1
-------
Brianfireun, B.A., A. Heyes and NT. Roulet. 1996. The hydrology and methylmercury
dynamics of a Precambrian Sheild headwater peatland. Water Res. Res.
32:1785-1794.
Brzeznicka, E. and J. Chmielnicka. 1985. Interaction of alkylmercuric compounds with
sodium selenite II. Metabolism of methylmercuric chloride administered alone
and in combination with sodium selenite in rats. Env. Health Persp., 60:411-421.
Callister, S.M. and M.R. Winfrey. 1986. Microbial methylation of mercury in Upper
Wisconsin River Sediments. Water Air Soil Poll. 29:453-465.
Carpi, A. and S.E. Lindberg. Application of a teflon dynamic flux chamber for
quantifying soil mercury fluxes: tests and results over background soils.
Atmos.Envir. (In press A).
Carpi, A., Lindberg, S.E., Prestbo, E.M., and N.S. Bloom. Global and regional impacts
of elemental and methyl mercury emitted by soils to the atmosphere. J.Env.
Qual. (In press B).
Charbonneau, S.M., Munro, I.C., Nera, E.A., Willes, R.F., Kuiper-Goodman, T., Iverson,
F., Moodie, C.A., Stoltz, D.R., Armstrong, F.A.J., Uthe, J.F., and H.C. Grice.
1974. Subacute toxicity of methylmercury in the adult cat. Toxicol. Appl.
Pharmacol. 27:569-581.
Choi, S.-C. and R. Bartha. 1994. Environmental factors affecting mercury methylation
in estuarine sediments. Bull. Environ. Contain. Tox. 53:805-812.
Cocking, D., M. Rohrer, R. Thomas, J. Walker and D. Ward. 1995. Effects of root
morphology and Hg concentration in the soil on uptake by terrestrial vascular
plants. Wat. Air Soil. Pollut. 80:1113-1116.
Compeau, G. and R. Bartha. 1985. Sulfate-reducing bacteria: principle methylators of
mercury in anoxic estuarine sediment. Appl. Environ. Microbiol. 50:498-502.
Davidson, P.W., Meyers, G.J., and C. Cox. 1995. Longitudinal neurodevelopmental
study of Seychellois children following in utero exposure to methylmercury from
maternal fish ingestion: Outcomes at 19 and 29 months. Neurotoxicol. 16: 677-
688.
Dietrich, K.N., Krafft, K.M., Bornschein, R.L., Hammond, P.B., Berger, 0., Succop, P.A.,
and M. Bier 1987. Low level fetal lead exposure effect on neurobehavioral
development in early infancy. Pediatrics 80: 721-730.
Driscoll, C.T., C Yan, C.L. Schofield, R. Munson, and H. Holsapple. 1994. The
R-2
-------
chemistry and bioavailability of mercury in Adirondack lakes. Environ. Sci. and
Technol. 28:136A-143A
Electric Power Research Institute. 1997. Mercury in the Environment.
Eisner, J. 1991. Tactile-kinesthetic system of rats as an animal model for minimal brain
dysfunction. Arch Toxicol. 65(6):465-473.
EPA. 1996. Analyzing Electric Power Generation Under the CAAA, (July), Office of Air
and Radiation.
European Commission. 1995. Externalities of Energy: ExternE Project. For the
Directorate General XII. Prepared by Metroeconomica, CEPN, IER, Eyre
Energy-Environment, ETSU, Ecole des Mines.
Expert Panel on Mercury Atmospheric Processes. 1994. Mercury Atmospheric
Processes: A Synthesis Report. Report No. TR-104214.
Feeley, Thomas, and Lawrence Ruth. 1996. "the U.S. Department of Energy's
Advanced Environmental Control Technology Program", U.S. DOE, Federal
Energy Technology Center, Pittsburgh, PA.
Fredriksson, A.L., Dahlgren, Danielsson, etal. 1992. Behavioral effects of neonatal
metallic mercury exposure in rats. Toxicology. 74(2-3): 151-160.
Gagnon, C., E. Pelletier, A. Mucci and W.F. Fitzgerald. 1996. Diagenetic behavior of
methylmercury in organic-rich coastal sediments. Limnol. Oceangr. 41:428-434.
Gilbert, S.G., and K.S. Grant-Webster. 1995. Neurobehavioral effects of
developmental methylmercury exposure. Environmental Health Perspectives
103 (Suppl6), 135-142.
Gilmour, C.C., G.S. Riedel, M.C. Ederington, J.T. Bell, J.M. Benoit, G.A. Gill and M.C.
Stordal. In press. Mercury methylation and sulfur cycling across a trophic
gradient in the northern Everglades. Biogeochemistry.
Gilmour, C.C., E.A. Henry and R. Mitchell. 1992. Sulfate stimulation of mercury
methylation in freshwater sediments. Environ. Sci. Technol. 26:2281-2287.
Goulder, Lawrence H. 1995. "Environmental Taxation and the 'Double Dividend': A
Reader's Guide," International Tax and Public Finance, 2 (2): 157-183.
Goulder, L., Parry I., and Dallas Burtraw. 1996. "Revenue-Raising vs. Other
Approaches to Environmental Protection: The Critical Significance of Pre-
R-3
-------
Existing Tax Distortions", Resources for the Future, Washington DC, November
1996.
Grandjean, P., Weihe, P., and R.F. White. 1995. Milestone development in infants
exposed to methylmercury from human milk. Neurotoxicol. 16: 27-34.
Grandjean, P., Weihe, P., J0rgensen, P.J., Clarkson, T., Cernichiari, E., and T. Vider0.
1992. Impact of maternal seafood diet on fetal exposure to mercury, selenium,
and lead. Arch. Environ. Health, 47:185-195.
Grandjean etal. (Archives of Environmental Health, 47(3), 1992, page 192).
Gross, M.B., and W.C. Wilson. 1974. Minimal Brain Dysfunction: A Clinical Study of
Incidence, Diagnosis, and Treatment in over 1000 Children. New York:
Brunner/Mazel.
Hagler Bailly Consulting, Inc. 1995. The New York State Externalities Cost Study.
Oceana Publications, Dobbs Ferry, N.Y.
Hanson, P. J., S. E. Lindberg, T. A. Tabberer, J. G. Owens and K-H. Kim. 1995. Foliar
exchange of mercury vapor: Evidence for a compensation point. Wat. air Soil.
Pollut. 80:373-382.
Heinz, G.H. 1979. Methylmercury: Reproductive and behavioral effects on three
generations of mallard ducks. J. Wildl. Mgmt. 43:394-401.
Henry, E.A., L.J. Dodge-Murphy, G.N. Bigham, S.M. Klein and C.C. Gilmour. 1995.
Total mercury and methylmercury mass balance in an alkaline, hypereutrophic
urban lake (Onondaga Lake, N.Y.) Wat. Air Soil Poll. 80: 509-518
Heyes, A. 1996. Methylmercury in disturbed and undisturbed wetlands. Ph.D.
dissertation. McGill Univ.
Humphrey, H.E.B. 1974. Mercury Concentrations in Humans and Consumptiom of
Fish Containing Methylmercury, Heavy Metals in the Environment, International
Symposium Proceedings in Progress In Water Technology. 7: 33.
Humphrey, H.E.B. 1988. Chemical Contaminants in the Great Lakes:The Human
Health Aspect, in M.L. Evans (Ed.) Toxic Contaminants and Ecosystem Health:
A Great Lakes Focus. Wiley, New York, Chapter 7.
Harvey, P., Hamlin, M., and R. Kumar. 1984. Blood lead, behavior and intelligence
R-4
-------
test performance in pre-school children. Sci Total Environ, 40: 45-60.
Hurley, J. P., J. M. Benoit, C. L. Babiarz, M. M. Schafer, A. W. Andren, J. R. Sullivan,
R. Hammond and d. a. Webb. 1995. Influences of watershed characteristics on
mercury levels in Wisconsin rivers. Environ. Sci. Technol. 29:1867-1875.
Hurley, J.P., D.P. Krabbenhoft, C.L. Babiarz and A.W. Andren. 1994. Cycling of
mercury across the sediment-water interface in seepage lakes. In: L.A. Baker
(editor) Environmental Chemistry of Lakes and Resrvoirs. American Chemical
Society, pp. 425-449.
Igata, A. 1993. Epidemiological and clinical features of Minimata disease. Environ.
Res. 63:157-169.
Inglis, J., and J.S. Lawson. 1981. Sex differences in the effects of unilateral brain
damage on intelligence. Science 212: 693-695.
lverfeldt,A., Munthe,J., and H. Hultberg. 1996. Terrestrial Mercury and Methylmercury
Budgets for Scandinavia. IN: Baeyens, W. Ebinghaus, R., Vasiliev, 0. (eds):
Global and Regional Mercury Cycles: Sources, Fluxes and Mass Balances.
NATO-ASI-Series, Vol. 21, Kluwer Academic Publishers, Dordrecht, The
Netherlands, 381-401.
Jackson, T.A. 1988. The mercury problem in recently formed reservoirs of northern
Manitoba (Canada): effects of impoundment and other factors on the production
of methyl mercury by microorganisms in sediments. Can. J. Fish. Aqua. Sci.
45:1744-1757.
Kelly, C.A., J.W. M. Rudd, R. A. Bodaly, N.P. Roulet, V. L. st. Louis, A. Heyes, T. R.
Moore, S. Schiff, R. Aravena, K. J. Scott, B. Dyck, r. Harris, B. Waner, and G.
Edwards. 1997. Increases in fluxes of greenhouse gasses and methyl mercury
following flooding of an experimental reservoir. Environ. Sci. Technol.
331:1334-1344.
Kelly, C.A., J.W.M. Rudd, V.L. St. Louis and A. Heyes. 1995. Is total mercury
concentration a good predictor of methylmercury concentration in aquatic
systems? Wat. Air Soil Poll. 80:715-724.
Kendall, R.J., C.M. Bens, G.P. Cobb III, R.L. Dickerson, K.R. Dixon, M.J. Hooper, S.J
R-5
-------
Klaine, I.E. Lacher, Jr., T.W. LaPoint, ST. McMurry, R. Noblet and E.E. Smith.
1995. Aquatic and Terrestrial Ecotoxicology. IN; Cassarett and Doull's
Toxicology: The Basic Science of Poisons. 5th Ed. C.D. Klassen, (ed).
McGraw-Hill, New York. Chap. 29 pp. 883-905.
Kendall, R.J. and R.L. Dickerson. 1969. Pinciples and processes for evaluating
endocrine disruption in wildlife. Environ. Toxicol. Chem. 15:1253-1254.
Kendall, R. J. and T.E. Lacher, Jr., editors. 1994. Wildlife Toxicology and Population
Modeling: Integrated Studies of Agroecosystems. CRC Press/Lewis Publishers,
Chelsea, Michigan. 576 pp.
Kendall, R.J., T.E. Lacher, Jr., C.Bunck, F.B. Daniel, C. Driver, G.E. Grue, F. Leighton,
W. Stansley, P.G Watanabe and M.Whitworth. 1996. An ecological risk
assessment of lead shot exposure in upland game birds and raptors. Environ.
Toxicol. Chem. 15:4.20.
Kim, K.H., Lindberg, S.E., andT.P. Meyers. 995. Micrometeorological measurements
of mercury fluxes over background forest soils in eastern Tennessee. Atmos.
Envir. 27:267-282.
Knobeloch, L.M., Ziarnik, M., Anderson, H.A. and V.N. Dodson. 1995. Imported
seabass as a source of mercury exposure: A Wisconsin Case Study.
Environmental Health Perspectives, 103:6, 604-606.
Komstra-Szumska, E., Reuhl, K.R., and D.R. Miller. 1983. Effect of selenium on
distribution, demethylation, and excretion of methylmercury by the Guinea pig.
J. Toxicol. Env. Health, 12:775-785.
Krabbenhoft, D., C.C. Gilmour, J. Benoit, J. Hurley and A. Andren. In press.
Methylmercury production and flux in a groundwater inflow zone of a pristine
seepage lake. Can. J. Fish. Aquat. Sci.
Lansdown, R., Yule, W., Urbanowicz, M., and J. Hunter. 1986. The relationship
between blood-lead concentrations, intelligence, attainment and behavior in a
school population: the second London study. IntArch Occup Environ Hlth 57:
225-235.
Lee, Russell, etal.. 1995. Estimating Externalities of Electric Fuel Cycles: Analytical
Methods and Issues, Estimating Externalities of Coal Fuel Cycles, and additional
volumes for other fuel cycles. McGraw-Hill/Utility Data Institute, Washington,
D.C.
R-6
-------
Lindberg, S.E., Hanson, P.J., Myers, T.P., and K-Y Kim. Micrometeorological studies of
air/surface exchange of mercury over forest vegetation and a reassessment of
continental biogenic mercury emissions. Atmos, Envir. (In press).
Lindberg, S.E. 1996. Forests and the Global Biogeochemical Cycle of Mercury: The
Importance of Understanding Air/vegetation Exchange Processes. IN: Baeyens,
W., Ebinghaus, R., Vasiliev, 0. (Eds.): Global and Regional Mercury Cycles:
Sources, Fluxes and Mass Balances. NATO-ASI Series, Vol.21, Kluwer
Academic Publishers, Dordrecht, The Netherlands, 359-380.
Lowe, T.P., May, T.W., Brumbaugh, W.G., and D.A. Kane. 1985. National
Contaminant Biomonitoring Program: Concentration of seven elements in fresh-
water fish, 1978-1981. Arch. Environ. Contamin. Toxicol. 14:363-388.
Lower, W.R. and R.J. Kendall. 1990. Sentinel Species and Sentinel Bioassay. IN;
Biomarkers of Environmental Contamination. John F. McCarthy and Lee R.
Shugart, (eds.) Lewis Publishers, pp. 309-331.
Marsh, D.O., Clarkson, T.W., Cox, C., Meyer, G.J., Amin-Zaki, and S. L. AI-Tikriti.
1987. Fetal methylmercury poisoning. Relationship between concentration in
single strands of maternal hair and child effects. Arch Neurol 44: 1017-1022.
Mason R.P., Fitzgerald, W.F., and F.M.M. Morel. 1994. The Biogeochemical Cycling
of Elemental Mercury: Anthropogenic Influences. Geo. Cosmo. Act.
58(15):3191-3198.
McConnell, Virginia D. 1990. Costs and Benefits of Vehicle Inspection: a Case Study
of the Maryland Region," Journal of Environmental Management, 30, 1-15.
McKeown-Eyssen, G.E., Reudy, J., and A. Neims. 1983. Methylmercury exposure in
Northern Quebec 2: Neurologic findings in children. Am J Epidemiol 118: 470-
479.
Michigan Mercury Pollution Prevention Task Force. 1996. "Mercury Pollution
Prevention in Michigan," Michigan Department of Environmental Quality.
M0ller-Madsen, B. and G. Danscher. 1991 Localization of mercury in CNS of the rat.
Toxicol. Appl. Pharmacol., 108:457-473
Mosbaek. H., J.C. Tjell and T. Sevel. 1988. Plant uptake of airborne mercury in
background areas. Chemosphere. 17:1227-1236
R-7
-------
Munthe, J., H. Hultberg and A. Iverfeldt. 1995. Mechanisms of deposition of
methylmercury and mercury to coniferous forests. Wat. Air Soil. Pollut. 80:363-
371.
Naeye, R.L., Burt, L.S., Wright, D.L., et al. 1971. Neonatal mortality: The male
disadvantage. Pediatrics 48: 902-906.
Naganuma, A., and N. Imura. 1980. Bis(methylmercuric) selenide as reaction product
from methylmercury and selenite in rabbit blood. Res. Commun. Chem. Pathol.
P/7armaco/.,27:163-173.
National Research Council. 1991, Animals as Sentinels of Environmental Health
Hazards. Committee on Animals as Monitors of Environmental Hazards, Board
on Environmental Studies and Toxicology, Commission on Life Sciences.
National Academy Press, Washington DC, 160 pp.
Newland, M.C., Warfvinge, K., and M. Berlin. 1996. Behavioral consequences of in
utero exposure to mercury vapor: alterations in lever-press durations and
learning in squirrel monkeys. Toxicology and Applied Pharmacology, 139, 374-
386.
Gates, Wallace E. 1995. "Green Taxes: Can We Protect the Environment and Improve
the Tax system at the Same Time?" Southern Economic Journal, 61 (4), 914-
922.
Parks, J.W., A. Lutz, J.A. Sutton, and B.E. Townsend. 1989. Water column
methylmercury in the Wabigoon/English River-Lake System:Factors controlling
concentrations, speciation and net production. Can. J. FishAq. Sci.
Parks, J.W. and A.L. Hamilton. 1987. Accelerating recovery of the mercury
contaminated Wabigoon/English River System. Hydrobiologia 149:159-188.
Parks, J.W., J.A. Sutton and A. Lutz. 1986. Effect of point and diffuse source loadings
on mercury concentrations in the Wabigoon River: evidence of a seasonally
varying sediment-water partition. Can. J. Fish. Aquat. Sci. 43:1426-1444.
Parry, Ian W. H. 1995. "Pollution Taxes and Revenue Recycling," Journal of
Environmental Economics and Management, 29, S64-S77.
R-8
-------
Peterson, G.A., Iverfeldt, A., and J. Munthe. 1995. Atmospheric mercury species over
central and northern Europe, model calculations and comparison with
observations from the Nordic Air and Percipitation Network for 1987 and 1988.
GKSS Research Center, Institute of Physics, Max Planck-Str. 1, D-21502,
Geesthact, Germany.
Rada, R.G., D.E. Powell and J.G. Wiener. 1993. Whole-lake burdens and spatial
distribution of mercury in surficial sediments in Wisconsin seepage lakes. Can.
J. Fish. Aquat. Sci. 50:865-873.
Rada, R.G. J.G. Wiener, M.R. Winfrey and D.E. Powell. 1989. Recent increases in
atmospheric deposition of mercury in north-central Wisconsin lakes inferred
from sediment analysis. Arch. Environ. Contam. Toxicol. 18:175-181.
Rea, A.W., Keeler,G.J., and T. Scherbatskoy. 1996. The deposition of mercury in
throughfall and litterfall in the Lake Champlain Watershed: a short-term study.
Atmos. Environ. 80:1269-1278.
Rice, D.C. 1989a. Delayed neurotoxicity in monkeys exposed developmental^ to
methyl mercury. Neurotoxicology. 10(4):645-650.
Rice, D.C. 1989b. Brain and tissue levels of mercury after chronic methyl mercury
exposure in the monkey. J. Toxicol. Environ. Health. 27(2): 189-198.
Rice, D.C. 1996. Sensory and cognitive effects of experimental methylmercury
exposure in monkeys, and a comparison to effects in rodents. Neurotoxicology,
17, 139-154, 1996.
Rice, D.C., and S.G. Gilbert. 1995. Effects of developmental methylmercury exposure
or lifetime lead exposure on vibration sensitivity function in monkeys.
Toxicology and Applied Pharmacology, 134, 161 -169.
Rice, D.C. 1992. Effects of pre- plus postnatal exposure to methylmercury in the
monkey on fixed interval and discrimination reversal performance.
Neurotoxicology, 13, 443-452.
Schreiner, G., Ulbrich., B., and R. Bass. 1986. Testing strategies in behavioral
tertology: II. Discriminatory Learning. Neurobehav. Toxicol. Tertol. 8:567-572.
St. Louis, V. L, J. W. M. Rudd, C. A. Kelly, K. G. Beatty, R. J. Flett and N. T. Roulet.
1996. Production and loss of methyl mercury and loss of total mercury from
boreal forest catchments containing different types of wetlands. Environ. Sci.
Technol. 30:2719-2729.
R-9
-------
St. Louis, V. L, J. W. M. Rudd, C. A. Kelly, K. G. Beatty, N. S. Bloom and R. J. flett.
1994. Importance of wetlands as sources of methyl mercury to boreal forest
ecosystems. Can J. Fish Aquat. Sci. 51:1065-1076.
Sager, P.R., Aschner, M., and P.M. Rodier. 1984. Persistent differential alterations in
developing cerebellar cortex of male and female mice after methylmercury
exposure. Brain Research and Developmental Brain Research, 12, 1-11.
Scragg, R.K.R., McMichael, A.J., and P.O. Clark. 1977. Pollution and sex ratio of
births. MedJAust2: 68.
SETAC. 1994. Final Report: Aquatic Risk Assessment and Mitigation Dialogue Group.
Society of Environmental Toxicology and Chemistry, Pensacola, Florida.
Sheffield, S.R. and R.J. Kendall. 1997. Wildlife species as sentinels of environmental
health hazards. Proceedings of SGOMSEC 13 Meeting, Ispra, Italy (January
26-31, 1997).
Solomon, K. R. Et al. 1996. An ecological risk assessment of atrizine in North
American surface waters. Vol. 15, No. 1: 31-76.
Stern, A.M. 1993. Reevaluation of the reference dose for methylmercury and
assessment of current exposure levels. Risk Anal. 13(3):355-364.
Stillings, B.R., Lagally, H., Bauerfeld, P., and J. Scares. 1974. Effect of cystine,
selenium and fish protein on the toxicity and metabolism of methylmercury in
rats. Toxicol. Appl. Pharmacol. 3:243-254.
Stratton,W.J. and S.E. Lindberg. 1995. Use of a refluxing mist chamber for
measurement of gas-phase water-soluble mercury(ll) species in the atmosphere.
Water, Air, Soil, Pollut. 80:1269-1278.
Suzuki, T. and R. Yamamoto. 1984. Alteration of inorganic mercury accumulation due
to selenite in organs of mice fed methylmercury. Jpn. J. Ind. Health, 26:125-
129.
Tietenberg, Thomas. 1985. Emissions Trading, Resources for the Future.
Vahter, M.E., Mottet, N. K., Friberg, LT.., Lind, S.B., Charleston, J.S., and T.M.
Burbacher. 1995. Retention of inorganic mercury in different brain sites of
Macaca fascicularis monkeys during long-term subclinical exposure to
methylmercury. Toxicol. Appl. Pharmacol. 134:273-284.
R-10
-------
Warkany, J., and D.M. Hubbard. 1951. Adverse mercurial reactions in the form of
acrodynia and related conditions. Amer. J. Dis. Child. 81:335.
Watras, C.J., K.A. Morrison and J.S. Host. 1995. Concentration of mercury species in
relationship to other site-speific factors in the surface waters of northern
Wisconsin Lakes. Limnol. Oceanogr. 40:556-565.
Weiss, B. 1988. Neurobehavioral toxicity as a basis for risk assessment. Trends in
Pharmacological Sciences. 9:59-62.
Wobeser, G.A. 1973. Ph.D Dissertation. Aquatic Mercury Pollution: Studies of its
occurrence and pathologic effect on fish and mink. University of Saskatchewan
(Canada). Dissertation Number 73-24, 819. University Microfilms, Ann Arbor,
Ml.
Wobeser, G., N.D. Nielsen and B. Schiefer. 1976a. Mercury and mink I: The use of
mercury contaminated fish as a food for ranch mink. Can. J. Comp. Med.
40:30-33.
Wobeser, G., N.D. Nielsen and B. Schiefer. 1976b. Mercury and mink II: Experimental
methylmercury intoxication. Can. J. Comp. Med. 40:34-45.
R-11
------- |