EPA-823-R-01-001
January 2001
of Human Health: Methyl mercury
Final
Office of Science and Technology
Office of Water
U.S. Environmental Protection Agency
Washington, DC 20460
-------
-------
NOTICE
This document has been reviewed in accordance with U.S. Environmental Protection Agency
policy and approved for publication. Mention of trade names or commercial products does not constitute
endorsement or recommendation for use.
11
Methylmercury Water Quality Criterion 1/3/01
-------
-------
Authors
Denis Borum
ACKNOWLEDGMENTS
U.S. EPA Office of Science and Technology, Office of Water
Mary Ko Manibusan, M.P.H. U.S. EPA Office of Science and Technology, Office of Water
Rita Schoeny, Ph.D.
U.S. EPA Office of Science and Technology, Office of Water
Erik L. Winchester, M.S. U.S. EPA Office of Science and Technology, Office of Water
Contributors
Helen Jacobs, M.S.
Kate Mahaffey, Ph.D.
Debra Rice, Ph.D.
Keith Sappington, M.S.
EPA Reviewers
Larry Hall, Ph.D.
John Nichols, Ph.D.
Glen Rice, M.S.
Jeff Swartout, M.S.
U.S. EPA Office of Science and Technology, Office of Water
U.S. EPA Office of Science Co-ordination and Policy, Office of
Prevention, Pesticides and Toxic Substancest
U.S. EPA National Center for Environmental Assessment, Office of
Research and Development
U.S. EPA National Center for Environmental Assessment, Office of
Research and Development
U.S. EPA National Health and Environmental Effects Research
Laboratory, Office of Research and Development
U.S. EPA National Health and Environmental Effects Research
Laboratory, Office of Research and Development
U.S. EPA National Center for Environmental Assessment, Office of
Research and Development
U.S. EPA National Center for Environmental Assessment, Office of
Research and Development
Portions of this document were developed under contract with Great Lakes Environmental Center
(GLEC), Information Systems Solutions International (ISSI), Inc., and ICF Consulting, Inc.
Methylmercury Water Quality Criterion 1/3/01
111
-------
-------
EXTERNAL PEER REVIEWERS
The following individuals provided technical and scientific reviews of the content and scientific
information in the criterion document as part of a formal peer review process.
Methylmercury Reference Dose
Kim N. Dietrich, Ph.D., University of Cincinnati
Bruce A. Fowler, Ph.D., University of Maryland
Gary Ginsberg, Ph.D. (workshop chair), Connecticut Department of Public Health
Martha Keating, M.S., Keating Environmental ,
Chris Newland, Ph.D., Auburn University
Pam Shubat, Ph.D., Minnesota Department of Health
Andrew Smith, S.M., Sc.D., Maine Department of Human Services
Bioaccumulation of Methylmercury
Nicolas S. Bloom, M.S., Frontier Geosciences Inc.
James P. Hurley, Ph.D., University of Wisconsin Water Resources Institute
David P. Krabbenhoft, Ph.D., U.S. Geological Survey
David Maschwitz, Ph.D., Minnesota Pollution Control Agency
Darell G. Slotton, Ph.D., University of California
Edward Swain, Ph.D., Minnesota Pollution Control Agency
Potential areas for conflict of interest were investigated via direct inquiry with the peer reviewers
and review of their current affiliations. No conflicts of interest were identified.
IV
Methylmercury Water Quality Criterion 1/3/01
-------
-------
CONTENTS
Executive Summary jx
1.0 Introduction j.j
1.1 Purpose of this Document \.\
1.2 Primary Data Source 1_2
1.3 Chemical and Physical Properties 1-2
2.0 Toxicokinetics 2-1
2.1 Absorption 2-1
2.1.1 Oral Absorption 2-1
2.1.2 Absorption via Other Routes 2-1
2.2 Distribution 2-2
2.3 Metabolism 2-3
2.4 Excretion 2-4
2.5 Biological Monitoring 2-6
2.5.1 Blood 2-6
2.5.2 Hair 2-7
2.5.3 Methods of Analyzing Mercury Concentrations in Biological Samples 2-7
2.6 Pharmacokinetic Models 2-8
3.0 Toxicological Basis for Criteria 3-1
3.1 Introduction 3_1
3.2 Neurotoxicity 3-2
3.2.1 Human Studies 3-2
3.2.2 Animal Studies 3-36
3.3 Cardiovascular Toxicity 3-41
3.3.1 Human Studies 3-41
3.3.2 Animal Studies 3.43
3.4 Immunotoxicity 3.43
3.4.1 Human Studies 3.43
3.4.2 Animal Studies 3.44
3.5 Reproductive Toxicity 3.45
3.5.1 Human Studies 3-46
3.5.2 Animal Studies 3-46
3.6 Genotoxicity 3.45
3.6.1 Human Studies 3-46
3.6.2 Animal Studies 3.47
3.7 Carcinogenicity 3-48
3.7.1 Human Studies 3-48
3.7.2 Animal Studies 3.49
4.0 Risk Assessment for Methylmercury 4-1
4.1 Background 4-1
4.1.1 Other RfDs Published by EPA 4-2
4.1.2 Risk Assessments Done by Other Groups 4-7
4.1.3 SAB review of the Mercury Study Report to Congress 4-9
4.1.4 Interagency Consensus Process 4-10
4.1.5 National Academy of Sciences Review 4-12
4.1.6 External Peer Review of Draft RfD 4-14
Methylmercury Water Quality Criterion 1/3/01 V
-------
4.1.7 Revised RfD 4-14
4.2. Choice of Critical Study and Endpoint 4-15
4.2.1 Summary of Available Data * 4-15
4.2.2 Choice of Study 4-32
4.2.3 Choice of Critical Effect (endpoint) 4-48
4.3 Choice of Dose-Response Approach 4-62
4.3.1 Benchmark Versus NOAEL 4-62
4.3.2 Choice of Exposure Metric 4-63
4.3.3 Choice of BMD 4-64
4.3.4 Choice of Model 4-66
4.3.6 Selection of the Point of Departure for the RfD 4-68
4.4 Dose Conversion 4-68
4.4.1 PBPK Models Versus One-Compartment Model 4-69
4.4.2 One-Compartment Model for Methylmercury 4-69
4.5 Choice of Uncertainty Factor 4-77
4.5.1 Background 4-77
4.5.2 Toxicodynamics 4-78
4.5.3 Exposure Estimation as an Area of Uncertainty 4-79
4.5.4 Pharmacokinetic Variability 4-79
4.5.5 Uncertainty in Choice of Critical Effect 4-82
4.5.6 Choice of Uncertainty Factor 4-86
4.6 Calculation of the RfD 4-87
5.0 Exposure Assessment 5-1
5.1 Overview of Relative Source Contribution Analysis 5-1
5.2 Population of Concern 5-1
5.3 Overview of Potential for Exposure 5-2
5.4 Estimates of Occurrence and Exposure from Environmental Media 5-3
5.4.1 Exposure Intake Parameters 5-4
5.4.2 Intake from Drinking Water/ambient Water 5-5
5.4.3 Nonfish Dietary Exposures 5-12
5.4.4 Fish Consumption Estimates 5-18
5.4.5 Respiratory Exposures 5-30
5.4.6 Soil/Sediment Exposures 5-35
5.4.7 Occupational and Other Exposures 5-40
5.5 Exposure Data Adequacy and Estimate Uncertainties 5-42
5.5.1 Adequacy of Intake Estimate for Drinking Water 5-42
5.5.2 Intake from Nonfish Dietary Sources 5-43
5.5.3 Intake fromFish 5-44
5.5.4 Intake from Air 5-46
5.5.5 Intake from Soil 5-46
5.6 Total Exposure Estimates 5-47
5.7 Relative Source Contribution (RSC) Estimates 5-56
5.7.1 RSC Policy Summary 5-56
5.7.2 Target Population for RSC/rationale for Approach to Methylmercury 5-56
5.7.3 Data Adequacy for RSC Estimate 5-57
5.7.4 RSC Estimate/apportionment of the RfD 5-57
VI
Methylmercury Water Quality Criterion 1/3/01
-------
6.0 Mercury Bioaccumulation 6-1
6.1 Introduction 6-1
6.2 Issues in Developing Methylmercury BAFs 6-1
6.3 Consideration of Fish Tissue Residue Criterion 6-4
7.0 Water Quality Criterion Calculation 7-1
7.1 Equation for Tissue Residue Concentration and Parameters Used 7-1
7.2 Site-Specific or Regional Adjustments to Criteria 7-1
8.0 References R-l
Appendix A A-l
Section I: Draft National Methylmercury Bioaccumulation Factors A-l
Section H: Chemical Translators for Mercury and Methylmercury A-19
Methylmercury Water Quality Criterion 1/3/01
Vll
-------
Vlll
Methylmercury Water Quality Criterion 1/3/01
-------
EXECUTIVE SUMMARY
About This Document
This document is the basis for a human health Ambient Water Quality Criterion (AWQC) for
methylmercury. This AWQC replaces the AWQC for total mercury in published in 1980 and partially
updated in 1997. Under Section 304(a) of the Clean Water Act, EPA must periodically revise criteria for
water quality to accurately reflect the latest scientific knowledge on the kind and extent of all identifiable
effects of pollutants on human health.
This document uses new methods and information described in the Methodology for Deriving
Ambient Water Quality Criteria for the Protection of Human Health (2000) (2000 Human Health
Methodology) (U.S. EPA, 2000a,b). These new methods include updated approaches to determine
toxicity dose-response relationships for both carcinogenic and noncarcinogenic effects, updated
information for determining exposure factors, and new procedures to determine bioaccumulation factors.
The Mercury Study Report to Congress (MSRC) (U.S. EPA, 1997), an eight-volume report
prepared by the U.S. Environmental Protection Agency (EPA) and submitted to Congress in 1997, serves
as a primary information source on methylmercury. However, as the state of the science for
methylmercury is continuously and rapidly evolving, the information from the MSRC has been
supplemented by inclusion of published information since 1997.
Exposure to Methylmercury
The major pathway for human exposure to methylmercury is consumption of contaminated fish.
Dietary methylmercury is almost completely absorbed into the blood and is distributed to all tissues
including the brain; it also readily passes through the placenta to the fetus and fetal brain.
Major Health Effects of Methylmercury
Methylmercury is a highly toxic substance with a number of adverse health effects associated with
its exposure in humans and animals. Epidemics of mercury poisoning following high-dose exposures to
methylmercury in Japan and Iraq demonstrated that neurotoxicity is the health effect of greatest concern.
These epidemics led to observation of methylmercury effects on the fetal nervous system. High-dose
Methylmercury Water Quality Criterion 1/3/01
IX
-------
human exposure results in mental retardation, cerebral palsy, deafness, blindness, and dysarthria in utero
and in sensory and motor impairment in adults. Although developmental neurotoxicity is currently
considered the most sensitive health endpoint, data on cardiovascular and immunological effects are
beginning to be reported and provide more evidence for toxicity from low-dose methylmercury exposure.
Three large prospective epidemiology studies in the Seychelles Islands, New Zealand, and the
Faroe Islands were designed to evaluate childhood development and neurotoxicity in relation to fetal
exposures to methylmercury in fish-consuming populations. Prenatal methylmercury exposures in these
three populations were within the range of some U.S. population exposures. No adverse effects were
reported from the Seychelles Islands study, but children in the Faroe Islands exhibited subtle
developmental dose-related deficits at 7 years of age. These effects include abnormalities in memory,
attention, and language. In the New Zealand prospective study, children at 4 and 6 years of age exhibited
deficiencies in a number of neuropsychological tests.
In addition to the three large epidemiological studies, studies on both adults and children were
conducted in the Amazon; Ecuador; French Guiana; Madeira; Mancora, Peru; northern Quebec; and
Germany. Effects of methylmercury on the nervous system were reported in all but the Peruvian
population.
Other Health Effects of Methylmercury
Methylmercury causes chromosomal effects but does not induce point mutations. The MSRC
concluded that because there are data for mammalian germ-cell chromosome aberration and limited data
from a heritable mutation study, methylmercury is placed in a group of high concern for potential human
germ-cell mutagenicity. There is no two-generation study of reproductive effects, but shorter term
studies in rodents, guinea pigs and monkeys have reported observations consistent with reproductive
deficits. There are no data to indicate that methylmercury is carcinogenic in humans, and it induces
tumors in animals only at highly toxic doses. Application of the proposed revisions to the Guidelines for
Cancer Risk Assessment (EPA 1999)leads to a judgment that methylmercury is not likely to be
carcinogenic for humans under conditions of exposure generally encountered in the environment.
Methylmercury Water Quality Criterion 1/3/01
-------
Quantitative Risk Estimate for Methylmercury
The quantitative health risk assessment for a noncarcinogen relies on a reference dose (RfD). This
is an estimate (with uncertainty spanning perhaps an order of magnitude) of a daily exposure to the
human population (including sensitive subgroups) that is likely to be without an appreciable risk of
deleterious health effects during a lifetime. To derive an RfD, one first establishes a no adverse effect
level (NOAEL) for a particular endpoint. This can be done by inspection of the available data or by
using a mathematical modeling procedure to estimate the NOAEL; the latter approach was used for
methylmercury. Next the NOAEL is divided by a numerical uncertainty factor to account for areas of
variability and uncertainty in the risk estimate.
There has been considerable discussion within the scientific community regarding the level of
exposure to methylmercury that is likely to be without an appreciable risk of deleterious health effects
during a lifetime. In 1999, the Congress directed EPA to contract with the National Research Council
(NRC) of the National Academy of Sciences to evaluate the body of data on the health effects of
methylmercury. NRC was to concentrate on new data since the 1997 MSRC, and to provide
recommendations regarding issues relevant to the derivation of an appropriate RfD for methylmercury.
NRC published their report, Toxicological Effects of Methylmercury, in 2000. EPA generally concurred
with the NRC findings and recommendations. The NRC document was used as a resource in determining
the EPA RfD for methylmercury documented here.
Choice of Study
The adverse effect of methylmercury observed at lowest dose is neurotoxicity, particularly in
developing organisms. The brain is considered the most sensitive target organ for which there are data
suitable for derivation of an RfD. There is an extensive array of peer-reviewed, well-analyzed data from
human studies of low-dose exposure to methylmercury. NRC and EPA considered three epidemiologic
longitudinal developmental studies suitable for quantitative risk assessment: the Seychelles Child
Development Study (SCDS); the ongoing studies of children in the Faroe Islands; and the study of
children in New Zealand. All cohorts consisted of children exposed in utero through maternal
consumption of mercury-contaminated fish or marine mammals. In all studies there were
biomarkers of maternal exposure (hair), and in the Faroes study cord blood was also used as an additional
measure of fetal exposure. The SCDS yielded no evidence of impairment related to methylmercury
exposure, but the two other studies have found dose-related adverse effects on a number of
Methylmercury Water Quality Criterion 1/3/01
XI
-------
neuropsychological endpoints. EPA chose to base the RfD on data from the Faroes study. The SCDS has
no findings of effects associated with methylmercury exposure, and thus is not the best choice for a
public health protective risk estimate. While the New Zealand study does show mercury-related effects it
relatively small by comparison to the other two. Advantages of the Faroes study include these:
• Large sample size (n > 900 for some measures)
• Good statistical power as calculated by conventional means
• Use of two different biomarkers of exposure
• Comprehensive and focused neuropsychological assessment
• Assessment at an age and state of development when effects on complex neuropsychological
functions are most likely to be detectable
• Statistically significant observations which remain after adjusting for potential PCB effects
• Extensive scrutiny in the epidemiological literature
The Faroe Islands study was used for derivation of the RfD.
Estimation of the No Adverse Effect Level
A benchmark dose analysis was chosen as the most appropriate method of quantifying the dose-
effect relationship. The level chosen was a Benchmark Dose Lower Limit (BMDL); this was the lower
95% limit on a 5% effect level obtained by applying a K power model (K ^ 1) to dose-response data
based on mercury in cord blood. The BMDL was chosen as the functional equivalent of a no-adverse-
effect level for calculation of the RfD.
Choice ofEndpoint
Several endpoints are sensitive measures of methylmercury effects in the Faroese children. EPA
considered the recommendations of the NRC and EPA's external scientific peer review panel in coming
to a decision as to the appropriate endpoint. The NRC recommended the use of a BMDL of 58 ppb
mercury in cord blood from the Boston Naming Test (BNT). The external peer panel felt that the BNT
scores showed an effect of concomitant PCB exposure in some analyses. They preferred a PCB-adjusted
BMDL of 71 ppb mercury in cord blood for the BNT. A difficulty with this choice is that this BMDL is
based on scores from only about one-half of the total cohort. The peer panel further suggested using a
composite index across several measures in the Faroes data set. EPA prepared a comparison of the
Xll
Methylmercury Water Quality Criterion 1/3/01
-------
endpoints recommended by NRC and peer reviewers; this also included the BMDLs from the NRC
integrative analysis and geometric means of four scores from the Faroes. These BMDLs and
corresponding estimates of ingested methylmercury are within a very small range. Rather than choosing
a single measure for the RfD critical endpoint, EPA considers that this RfD is based on several scores.
These test scores are all indications of neuropsychological processes related to the ability of a child to
learn and process information.
Calculation of Ingested Methylmercury Dose
In the risk assessment discussion EPA uses the NRC-recommended BMDL of 58 ppb mercury in
cord blood as an example in the dose conversion and RfD calculation. The BMDL in terms of mercury
in cord blood was converted to an estimate of ingested methylmercury. This was done by use of a one-
compartment model similar to that used in the MSRC. Single-parameter estimates were used rather than
a distributional approach. It was assumed that the cord blood methylmercury level was equal to
maternal blood level. The ingested dose of methylmercury that corresponds to a cord blood level of 58
ppb is 1.081 |ig/kg bw/day.
Uncertainty Factor
Several sources of variability and uncertainty were considered in the application of a composite
uncertainty factor of 10. This included a factor of 3 for pharmacokinetic variability and uncertainty; one
area of pharmacokinetic uncertainty was introduced with the assumption of equivalent cord blood and
maternal blood mercury levels. An additional factor of 3 addressed pharmacokinetic variability and
uncertainty. Other areas of concern include inability to quantify possible long-term sequelae for
neurotoxic effects, questions as to the possibility of observing adverse impacts (such as cardiovascular
effects) below the BMDL, and lack of a two-generation reproductive effects assay.
Methylmercury Reference Dose
The RfD derived in this assessment is 0.1 p.g/kg bw/day or IxlO"4 mg/kg bw/day. The RfD for
methylmercury was not calculated to be a developmental RfD only. It is intended to serve as a level of
exposure without expectation of adverse effects when that exposure is encountered on a daily basis for a
lifetime. In the studies so far published on subtle neuropsychological effects in children, there has been
no definitive separation of prenatal and postnatal exposure that would permit dose-response modeling.
Methylmercury Water Quality Criterion 1/3/01
Xlll
-------
That is, there are currently no data that would support the derivation of a child (vs. general population)
RfD.
Relative Source Contribution
The assessment of methylmercury exposure from common media sources (e.g., diet, air) and
relative source contribution (RSC) estimates follows the 2000 Human Health Methodology. The RSC is
used to adjust the RfD to ensure that the water quality criterion is protective, given other anticipated
sources of exposure. The exposure assessment characterizes the sources of methylmercury exposure in
environmental media, providing estimates of intake from the relevant sources for children, women of
childbearing age, and adults in the general population. Based on available data, human exposures to
methylmercury from all media sources except freshwater/estuarine and marine fish are negligible, both in
comparison with exposures from fish and compared with the RfD. Estimated exposure from ambient
water, drinking water, nonfish dietary foods, air, and soil are all, on average, at least several orders of
magnitude less than those from freshwater/estuarine fish intakes. Therefore, these exposures were not
factored into the RSC. However, ingestion of marine fish is a significant contributor to total
methylmercury exposure. For the methylmercury criterion, the RSC is the estimated exposure from
marine fish intake. This is subtracted from the RfD when calculating the water quality criterion. One
hundred percent of the mercury in marine fish was assumed to be present as methylmercury. The
estimated average exposure to methylmercury from marine fish is 2.7 x 10-5 mg/kg-day. This exposure
represents almost 30% of the RfD.
Methylmercury Bioaccumulation
Methylmercury is a chemical that bioaccumulates and biomagnifies in aquatic food webs. The
fates of mercury and methylmercury in the environment are complex processes affected by numerous
biotic and abiotic factors that are subjects of ongoing research. Methylation of mercury is a key step in
the entrance of mercury into food chains. The biotransformation of inorganic mercury forms to
methylated organic forms in water bodies can occur in the sediment and the water column. Inorganic
mercury can be absorbed by aquatic organisms but is generally taken up at a slower rate and with lower
efficiency than is methylmercury. Methylmercury continues to accumulate in fish as they age. Predatory
organisms at the top of aquatic and terrestrial food webs generally have higher methylmercury
concentrations because methylmercury is typically not completely eliminated by organisms and is
xiv
Methylmercury Water Quality Criterion 1/3/01
-------
transferred up the food chain. Nearly 100% of the mercury that bioaccumulates in upper-trophic-level
fish (predator) tissue is methylmercury.
Numerous factors can influence the bioaccumulation of mercury in aquatic biota. These include,
but are not limited to, the acidity (pH) of the water, length of the aquatic food chain, temperature, and
dissolved organic material. Physical and chemical characteristics of a watershed, such as soil type and
erosion or proportion of area that is wetlands, can affect the amount of mercury that is transported from
soils to water bodies. Interrelationships among these factors are poorly understood and are likely to be
site-specific. No single factor (including pH) has been correlated with extent of mercury
bioaccumulation in all cases examined. Two lakes that are similar biologically, physically, and
chemically can have different methylmercury concentrations in water, fish, and other aquatic organisms.
The Methylmercury Criterion is a Fish Tissue Residue Criterion
EPA concluded that it is more appropriate at this time to derive a fish tissue (including shellfish)
residue water quality criterion for methylmercury rather than a water column-based water quality
criterion. This decision considered issues of mercury fate in the environment, the NRC report on the
toxicological effects of mercury, and in particular the methylmercury peer review comments. EPA
believes a fish tissue residue water quality criterion is appropriate for many reasons. Such a criterion
integrates spatial and temporal complexity that occurs in aquatic systems and that affects methylmercury
bioaccumulation. A fish tissue residue water quality criterion is more closely tied to the CWA goal of
protecting the public health because it is based directly on the dominant human exposure route for
methylmercury. The concentration of methylmercury is also generally easier to quantify in fish tissue
than in water and is less variable over the time periods in which water quality standards are typically
implemented in water quality-based. Thus, the data used in permitting activities can be based on a more
consistent and measurable endpoint. A fish tissue residue criterion is also consistent with how fish
advisories are issued. Fish advisories for mercury are based on the amount of methylmercury in fish
tissue that is considered acceptable, although they are usually issued for a certain fish or shellfish species
in terms of a meal size. A fish tissue residue water quality criterion should enhance harmonization
between these two approaches for protecting the public health.
The methylmercury water quality criterion is, thus, a concentration in fish tissue. It was calculated
using the criterion equation hi the 2000 Human Health Methodology rearranged to solve for a protective
concentration in fish tissue rather than in water.
Methylmercury Water Quality Criterion 1/3/01
XV
-------
BWx(Rfl) - RSC)
= 4 -
Where:
TRC = Fish tissue residue criterion (mg methylmercury/kg fish) for freshwater and estuarine fish
RfD = Reference dose (based on noncancer human health effects) of 0.0001 mg
methylmercury/kg body weight-day
RSC = Relative source contribution (subtracted from the RfD to account for marine fish
consumption) estimated to be 2.7 x 10"5 mg methylmercury/kg body weight-day
BW = Human body weight default value of 70 kg (for adults)
FI = Fish intake at trophic level (TL) i (i = 2, 3,4); total default intake is 0.0175 kg fish/day
for general adult population. Trophic level breakouts for the general population are: TL2
= 0.0038 kg fish/day; TL3 = 0.0080 kg fish/day; and TL4 = 0.0057 kg fish/day.
The resulting Tissue Residue Criterion is 0.3 mg methylmercury/kg fish. This is the concentration in fish
tissue that should not be exceeded based on a total fish and shellfish consumption-weighted rate of
0.0175 kg fish/day. EPA strongly encourages States and authorized Tribes to develop a water quality
criterion for methylmercury using local or regional data rather than the default values if they believe that
such a water quality criterion would be more appropriate for their target population.
xvi
Methylmercury Water Quality Criterion 1/3/01
-------
1.0 INTRODUCTION
1.1 PURPOSE OF TfflS DOCUMENT
This document provides guidance to States and Tribes authorized to establish water quality
standards under the Clean Water Act (CWA) to protect human health, pursuant to Section 304(a) of the
CWA. Under the CWA, States and authorized Tribes are to establish water quality criteria to protect
designated uses. While this document constitutes the U.S. Environmental Protection Agency's (EPA's)
scientific recommendations regarding concentrations of methylmercury in fish and shellfish that protect
human health, this document does not substitute for the CWA or EPA's regulations, nor is it a regulation
itself. Thus, it cannot impose legally binding requirements on EPA, States, Tribes, or the regulated
community, and may not apply to a particular situation based upon the circumstances. State and Tribal
decision makers retain the discretion to adopt approaches on a case-by-case basis that differ from this
guidance when appropriate. EPA may change this guidance in the future.
This document establishes a water quality criterion for methylmercury. The U.S. Environmental
Protection Agency (EPA) originally published an Ambient Water Quality Criterion (AWQC) for total
mercury in 1980. That AWQC was partially updated in 1997 to incorporate a change in the reference
dose (RfD). As required under Section 304(a) of the Clean Water Act, EPA must periodically revise
criteria for water quality to accurately reflect the latest scientific knowledge on the kind and extent of all
identifiable effects on human health from the presence of pollutants in any body of water. The criterion
uses new methods and information described in the Methodology for Deriving Ambient Water Quality
Criteria for the Protection of Human Health (2000) (2000 Human Health Methodology) and in the
Methodology's accompanying Federal Register Notice (U.S. EPA, 2000a,b). These new methods
include updated approaches to determine toxicity dose-response relationships for both carcinogenic and
noncarcinogenic effects, updated information for determining exposure factors, and new procedures to
determine bioaccumulation factors.
Development of a methylmercury criterion involves some unique considerations compared with
many of EPA's past efforts in the water quality criteria program. Traditionally, EPA has established
recommended 304(a) criteria to protect human health as ambient concentrations in water. For those
pollutants that bioaccumulate, such as methylmercury, exposure through the food pathway is estimated
by using a bioaccumulation factor (BAF). However, following review of available data and
Methylmercury Water Quality Criterion 1/3/01
1-1
-------
recommendations made by external peer reviewers (U.S. EPA, 2000c), EPA determined that it is more
appropriate to base the methylmercury criterion on a fish tissue residue concentration than on an ambient
water concentration. This determination was partly based on the current scientific understanding of the
fate of mercury and methylmercury in aquatic ecosystems. Another factor was the limited information on
sources of mercury and the conversion to methylmercury (and its bioavailability). Additional
considerations were the difficulty in measuring methylmercury in the water column and relating it to
concentrations in aquatic organisms. EPA believes that the latest data and science on methylmercury
exposure, effects, and environmental fate support the derivation of a fish tissue residue criterion.
1.2 PRIMARY DATA SOURCE
Much of the information in this document has been taken from the Mercury Study Report to
Congress (MSRC) (U.S. EPA, 1997b-h). This comprehensive, eight-volume study was prepared by EPA
and submitted to Congress in 1997 to fulfill the requirements of section 112(n)(l)(B) of the Clean Air
Act, as amended in 1990. The MSRC provides an assessment of the magnitude of U.S. mercury
emissions by source, the health and environmental implications of those emissions, and the availability
and cost of control technologies. As the state of the science for methylmercury continues to evolve,
information from the MSRC has been supplemented by data and analyses published since 1997. The
health effects information used in the derivation of the reference dose (RfD) for the fish tissue residue
concentration is based on the recommendations of the National Academy of Sciences National Research
Council report, Toxicological Effects of Methylmercury (NRC, 2000). For additional discussion on the
NRC recommendations, see Section 4 of this criteria document. The comments of the methylmercury
RfD scientific peer review panel also guided the risk assessment.
1.3 CHEMICAL AND PHYSICAL PROPERTIES
The water quality criterion is being derived for methylmercury (CAS No. 22967-92-6). Synonyms
for methylmercury include MeHg, methylmercury ion, methylmercury ion (1+), methylmercury (1+),
methyl mercury, and methylmercury(I) cation (Prager, 1997). A commonly occurring form of
methylmercury is methylmercuric chloride (CH3Hg+Cr), a stable salt form that exists as a white crystal.
This compound is often used in laboratory dosing experiments investigating the toxicological properties
of methylmercury. Because methylmercury exists as a free ion only in minute quantities (Prager, 1997),
the chemical and physical data provided below are for the chloride salt.
1-2
Methylmercury Water Quality Criterion 1/3/01
-------
The table below presents available chemical and physical data for methylmercuric chloride
(ATSDR, 1999; Kaufman, 1969).
Chemical formula
Chemical structure
Molecular weight
Physical state (25°C)
Boiling point (at 25 mm Hg)
Melting point
Density (25°C)
Vapor pressure (25°C)
Water solubility (2FC)
Log octanol/Water partition coeff.
Odor threshold (air)
Conversion factors (air)
CH3HgCl
CH3—Hg+Cl-
251.10 (g/mol)
White crystal
No data
170°C
4.06 g/mL
0.0085 mm Hg
<100 mg/L
No data
No data
1 ppm = 10.27 mg/m3;
lmg/m3 = 0.0974 ppm
Methylmercury Water Quality Criterion 1/3/01
1-3
-------
-------
2.0 TOXICOKINETTCS
This section presents information on the absorption, distribution, metabolism, and excretion of
methylmercury in humans and animals. This information is summarized from Volume V, Chapter 2 of
the Mercury Study Report to Congress (MSRC) (U.S. EPA, 1997e).
2.1 ABSORPTION
2.1.1 Oral Absorption
Methylmercury is efficiently absorbed from the gastrointestinal tract following ingestion.
Approximately 94%-95% of methylmercury in fish ingested by volunteers was absorbed from the
gastrointestinal tract (Aberg et al., 1969; Miettinen, 1973). Aberg et al. (1969) found uptake of greater
than 95% of radiolabeled methylmercuric nitrate administered in water to human volunteers.
Data from studies on rats, cats, and monkeys support these absorption estimates (ATSDR, 1999).
Studies on rats indicate rapid and complete absorption of inhaled methylmercury vapor into the
bloodstream (Fang, 1980). Female cynomolgus monkeys administered 0.5 mg mercury per kilogram of
methylmercuric chloride by oral gavage experienced complete absorption within 6 hours (Rice, 1989).
2.1.2 Absorption via Other Routes
Limited information is available on absorption via inhalation and dermal routes. There is one
reported human dermal exposure when a 48-year-old chemistry professor inadvertently spilled drops
(0.4-0.5 mL) of dimethylmercury from her pipette into her latex gloves. Penetration of dimethylmercury
through the gloves occurred instantaneously. Mercury hair level was elevated to almost 1,100 ppm, with
a half life of 74.6 days. Five months after exposure, the woman experienced severe neurotoxicity and
died 9 months later (Blayney et al., 1997; Nierenberg et al., 1998).
Skog and Wahlberg (1964) evaluated the dermal absorption of the methylmercuric cation in guinea
pigs. The test material was applied as the dicyandiamide salt. Absorption was estimated by
disappearance of the applied compound and by appearance of mercury in kidney, liver, urine, and blood.
Approximately 3% to 5% of the applied dose was absorbed during a 5-hour period.
Methylmercury Water Quality Criterion 1/3/01
2-1
-------
Indirect evidence in animals indicates that inhaled methylmercury vapor is absorbed readily
through the lungs. Fang (1980) showed a correlation between tissue mercury levels and both exposure
level and exposure duration in rats exposed to radioactively labeled methylmercury vapor. The percent
absorbed was not quantified.
2.2 DISTRIBUTION
After absorption from the gastrointestinal tract, methylmercury is readily absorbed into the blood
and distributes to all tissues, including the brain and fetus. The fraction of the absorbed dose that is
found in the blood has been estimated in three studies. Kershaw et al. (1980) reported an average
fraction of 0.059 of the absorbed dose in total blood volume, based on a study of five adult male subjects
who ingested methylmercury-contaminated tuna. In a group of nine male and six female volunteers who
had received 203Hg-methylmercury in fish, approximately 10% of the total mercury body burden was
present in 1 L of blood in the first few days after exposure; this dropped to approximately 5% over the
first 100 days (Miettinen et al., 1971). Sherlock et al. (1984) derived an average value of 1.14% for the
percentage of absorbed dose in 1 kg of blood from data on subjects who consumed a known amount of
methylmercury in fish over a 3-month period. Average daily intake in the study ranged from 43 to 233
Hg/day. There was a dose-related effect on percentage of absorbed dose that ranged from 1.03% to
1.26% in 1 L of blood. Each of these values was multiplied by 5 to yield the total amount in the blood
compartment, as there are approximately 5 L of blood in an adult human body.
Methylmercury in the blood is found predominantly in the red cells (Kershaw et al., 1980; Thomas
et al., 1986). It is distributed throughout the body following absorption from the gastrointestinal tract
into the blood (Clarkson, 1972; Hansen, 1988; Hansen et al., 1989; Nielsen and Andersen, 1992; Soria et
al., 1992; Suzuki et al., 1984). Although the distribution of methylmercury in the body is generally
uniform, at least one animal study indicates that high levels can be found in the kidney. Rice (1989b)
administered 0.025 or 0.05 mg mercury/kg-day as methylmercuric chloride in apple juice to cynomolgus
monkeys for approximately 2 years. Kidney tissue concentrations of mercury ranged from 10 to 28 ppm
in the cortex and 1 to 10 ppm in the medulla when assessed more than 200 days after cessation of
treatment. In contrast, mercury concentration was less than 2 ppm in the other tissues evaluated.
Methylmercury easily penetrates the placental barrier in humans and animals (Hansen, 1988;
Hansen et al., 1989; Nielsen and Andersen, 1992; Soria et al., 1992; Suzuki et al., 1984). Several studies
2-2
Methylmercury Water Quality Criterion 1/3/01
-------
have demonstrated mercury in newborn cord blood. The relationship to maternal blood is variable
(Grandjean et al., 1999). Information on this relationship is discussed in Section 4.5.4.1.
The distribution of methylmercury in animals may vary by age and sex (Thomas et al., 1982,, 1986,
1988). Female rats exposed to methylmercury had higher peak levels of mercury in the kidney (primarily
as methylmercury) than males; inorganic mercury levels did not differ significantly between the sexes
(Thomas et al., 1986). Accumulation of mercury was found to be higher in the bodies of neonatal rats
(Thomas et al., 1988) than in adult rats (Thomas et al., 1982). Ten days after administration of
methylmercury, 94% of the dose was still detected in neonates while approximately 60% was retained in
adults (Thomas et al., 1988). The longer retention of mercury in neonates may result from multiple
factors, including the high levels of mercury accumulated in the pelt of neonates owing to lack of
clearance (Thomas et al., 1988) and the lack of a fully developed biliary transport system in neonates
(Ballatori and Clarkson, 1982).
2.3 METABOLISM
The time required for methylmercury metabolism to inorganic mercury may account for the latent
or silent period observed in epidemiological studies from methylmercury poisoning incidents in Japan
and Iraq. During the latent period (both during and after the cessation of exposure) the patient feels no
untoward effects. It is possible that a number of biochemical changes may take place in parallel during
this period, and some may not be causatively related to the clinical outcome. Ganther (1978)
hypothesized that the carbon-mercury bond in methylmercury undergoes homolytic cleavage to release
methyl free radicals. The free radicals are expected to initiate a chain of events involving peroxidation of
lipid constituents of the neuronal cells. The onset of symptoms is delayed for the period of time that
cellular systems are able to prevent or repair effects of lipid peroxidation. When the cellular defense
mechanisms are overwhelmed, rapid and progressive degeneration of the tissue results. In the Iraqi
poisoning incident, the latent period before toxic signs were noted varied from a matter of weeks to
months. In contrast, the latency observed in the Japanese poisoning incident was as long as a year or
more. The difference in duration may in part be due to the presence of selenium in the fish ingested by
the Japanese population.
Rat liver microsomes can metabolize methylmercury into inorganic mercury via the NADPH-
cytochrome P-450 reductase, also known to control hydroxyl radical production in liver microsomes
(Suda and Takahashi). To a lesser degree, an oral dose of methylmercuric chloride may also be
Methylmercury Water Quality Criterion 1/3/01
2-3
-------
converted into inorganic mercury via the intestinal flora (Nakamura et al, 1977; Rowland et al., 1980).
The intestinal wall is poor in absorbing the inorganic mercury, thus almost all of it is excreted. Studies in
mice appear to indicate that toxicity from exposure to dimethylmercury results from the
biotransformation of dimethylmercury to methylmercury (Ostland, 1969). Following acute exposure to
methylmercury, most of the mercury in the brain is in the organic form; however, with chronic exposures,
a greater amount is in the inorganic form, suggesting that the rate of demethylation increases with long-
term exposure (Aschner and Aschner, 1990). Rice (1989a, 1989) demonstrated that tissue half-life of
methylmercury in the brain may be significantly longer than the blood half-life.
In rats, methylmercury in the body is relatively stable and is only slowly demethylated to form
mercuric ion (Norseth and Clarkson, 1970). The demethylation appears to occur in tissue macrophages
(Suda and Takahashi, 1986), intestinal microflora (Nakamura et al., 1977; Rowland et al., 1980), and
fetal liver (Suzuki et al., 1984).
2.4 EXCRETION
In humans, approximately 90% of the absorbed dose of methylmercury is excreted in the feces
(U.S. EPA, 1997e). Excretion via the urine is relatively minor but slowly increases with time; at 100 days
after dosing, urinary excretion of mercury accounted for 20% of the daily amount excreted. The urinary
excretion of mercury may reflect the deposition of demethylated mercury in the kidneys and its
subsequent excretion. In humans the major routes of excretion are via the bile and feces.
Feces are also the predominant route of methylmercury elimination in adult animals (Farris et al.,
1993; Hollins et al., 1975; Thomas et al., 1987). Biliary excretion of methylmercury and its
demethylation in gastrointestinal flora have been reported in rats (Farris et al., 1993). After a single oral
dose of methylmercury, the major elimination route was the feces (65% of the administered dose as
inorganic mercury and 15% of the administered dose as methylmercury) and the minor route was urine
(1% of the administered dose as inorganic mercury and 4% of the administered dose as methylmercury)
(Farris et al., 1993). Following administration of methylmercuric nitrate, 33% of the administered dose
was excreted in 49 days; 0.18% to 0.27% excretion in the urine in 10 days and 3.3% urinary excretion in
49 days. This continued for up to 71 days postingestion (Miettinen, 1973). Forty to 50 days
postingestion, <0.12% of the administered dose of mercury was found per gram of hair. The half-life for
methylmercury appeared to be 70-74 days. In humans the whole body half-life of methylmercury was
estimated to be between 70 and 80 days (Aberg et al., 1969; Miettinen, 1973; Bernard and Purdue, 1984).
2-4
Methylmercury Water Quality Criterion 1/3/01
-------
Mercury is excreted into the hair of methylmercury-exposed humans and animals. Incorporation of
mercury into hair is irreversible, and hair analysis is thus a useful tool for monitoring exposure to
methylmercury. Segmental analysis of hair may be used to provide a historical record of exposure
patterns.
Methylmercury is excreted in breast milk (Bakir et al., 1973; Sundberg and Oskarsson, 1992). The
ratio of mercury in breast milk to mercury in whole blood was approximately 1:20 in women exposed to
methylmercury via contaminated grain in Iraq between 1971 and 1972 (Bakir et al., 1973). Evidence
from the Iraqi poisoning incident also showed that lactation decreased blood mercury clearance half-
times from 75 days in males and nonlactating females to 42 days in lactating females; the faster clearance
due to lactation was confirmed in mice (Greenwood et al., 1978). In mice, of the total mercury in the
breast milk, approximately 60% was estimated to be methylmercury. Skerfving (1988) has found that
16% of mercury in human breast milk is methylmercury. Studies in animals indicate that the mercury
content of breast milk is proportional to the mercury content of plasma (Sundberg and Oskarsson, 1992;
Skerfving, 1988).
In rat and monkey neonates, excretion of methylmercury is severely limited (Lok, 1983; Thomas et
al., 1982). In rats dosed prior to 17 days of age, essentially no mercury was excreted (Thomas et al.,
1982). By the time of weaning, the rate of excretion had increased to adult levels. The failure of
neonates to excrete methylmercury may be associated with the inability of suckling infants to secrete bile
(Ballatori and Clarkson, 1982) and the decreased ability of intestinal microflora to demethylate
methylmercury during suckling (Rowland et al., 1977).
Currently, five studies report clearance half-lives for methylmercury. Three studies suggest a half-
life of approximately 70 to 80 days (Aberg et al., 1969; Bernard and Purdue, 1984; Miettinen, 1973).
Smith et al. (1994) reported a half-life of 44 days in a study of seven adult males treated intravenously
with methylmercury. In this study, methylmercury and inorganic mercury concentrations in blood and
excreta were determined separately based on differential extractability into benzene. The predominant
species in the blood was methylmercury; there was no detectable methylmercury in the urine. Al-
Shahristani and Shihab (1974) calculated a "biological half-life" of methylmercury in a study of 48 male
and female subjects who had ingested seed grain contaminated by organic mercurials. The half-life,
determined from distribution of mercury along head hair, ranged from 35 to 189 days with a mean of 72
days.
Methylmercury Water Quality Criterion 1/3/01
2-5
-------
The relatively long half-life of methylmercury in the body results partly from reabsorption of
methylmercury secreted into the bile (hepatobiliary cycling) (Norseth and Clarkson, 1971). In this cycle,
methylmercury forms a complex with glutathione in the hepatocyte and the complex is secreted into the
bile via a glutathione carrier protein (Clarkson, 1993). The methylmercury-glutathione complex in the
bile may be reabsorbed from the gallbladder and intestines into the blood. This cycle is terminated when
intestinal microorganisms demethylate methylmercury to form mercuric ion (Rowland et al., 1980).
Mercuric mercury is poorly absorbed from the-intestines and the fraction that is not reabsorbed is
excreted in the feces. As noted above, approximately 90% of the absorbed dose of methylmercury is
ultimately excreted in the feces as mercuric mercury.
2.5 BIOLOGICAL MONITORING
Distribution of methylmercury to hair and blood provides a means for biological monitoring of
methylmercury exposure. This section provides an overview of the use of hair and blood for assessing
exposure and outlines the available methods for quantitation.
2.5.1 Blood
Methylmercury distributes freely throughout the body, and thus blood is a good medium for
estimating short-term exposure. Blood levels may not necessarily reflect methylmercury intake over
longer periods, as an individual's intake may fluctuate (Sherlock et al., 1982; Sherlock and Quinn, 1988).
The characteristic partitioning of mercury in the blood permits identification of the form of
mercury to which an individual has been exposed. Measurements of blood hematocrit and mercury
concentrations in both whole blood and plasma can be used to calculate the red blood cell to plasma
mercury ratio. In the case of methylmercury, examination of this ratio enables estimation of interference
from exposure to high levels of elemental or inorganic mercury (Clarkson et al., 1988).
2.5.2 Hair
Scalp hair is a useful indicator for estimating methylmercury exposure (Phelps et al., 1980).
Mercury is incorporated into scalp hair at the hair follicle in proportion to its content in blood. The hair-
to-blood ratio in humans has been estimated as approximately 250:1 expressed as |ig mercury/g hair to
mg mercury/1 blood. Uncertainty in measurements, interindividual variation in body burden, differences
2-6
Methylmercury Water Quality Criterion 1/3/01
-------
in hair growth rates, and variations in fresh and saltwater fish intake have led to estimates ranging from
190:1 to 370:1 and higher (Birke et al., 1972; Skerfving, 1974; Phelps et ah, 1980; Turner et ah, 1980;
Sherlock et al., 1984). Once incorporated into the hair, the mercury is stable, and can give a longitudinal
history of blood rnethylmercury levels (Phelps et al., 1980; WHO, 1990). The identity of the
predominate chemical species (inorganic or methylmercury) depends on exposure patterns and the extent
of methylmercury demethylation.
Chemical analyses to determine mercury content of hair assay total mercury rather than chemical
species of mercury. As a result, the fraction of hair mercury that is methylmercury is an estimate based
on knowledge of environmental and occupational exposure patterns (U.S, EPA, 1997f). Analysis of hair
mercury levels may be confounded by several factors, including adsorption of mercury vapor onto the
hair strands, natural hair color, hair treatment, and growth rate (Francis et al., 1982; Suzuki, 1988).
Analysis of mercury in maternal hair has been utilized to estimate the fetal burden. This approach
has been validated by Cernichiari et al. (1995), who collected blood samples and autopsy brains from
terminally ill neonates in a population exposed to methylmercury via fish consumption. Maternal blood
and hair samples were also obtained. The concentrations of total mercury in six major brain regions of
the neonates were highly correlated with the concentration of mercury in a 1-cm segment of maternal hair
next to the scalp (correlation coefficients 0.6 to 0.8, p<0.01). These correlations were confirmed by a
series of comparisons utilizing maternal hair, maternal blood, neonate blood, and neonate brain tissue.
2.5.3 Methods of Analyzing Mercury Concentrations in Biological Samples
The most common methods used to determine mercury levels in biological media include atomic
absorption spectrometry, neutron activation analysis, X-ray fluorescence, and gas chromatography.
Another method is anodic stripping voltammetry (Liu et al., 1990). Gas chromotography-electron
capture is the only method capable of differentiating methylmercury from other species, whereas cold
vapor atomic absorption spectrometry will detect mercury at parts per billion in both urine (Magos and
Cernik, 1969) and blood samples (Magos and Clarkson, 1972). Mercury content in hair has been
measured by cold vapor atomic absorption spectrometry, atomic fluorescence spectrometry, X-ray
fluorescence, and neutron activation analysis (Zhuang et al., 1989).
Another method for analyzing biological samples containing methylmercury is with the use of
Pseudomonasputida strain FBI. The method is considered very reliable and specific for methylmercury
Methylmercury Water Quality Criterion 1/3/01 2-7
-------
quantification because chemical inference is negligible. The Pseudomonas putida bacteria is capable of
converting methylmercury to methane gas and elemental mercury (Baldi and Filippelli, 1991), thus
allowing the detection of 15 ng of methylmercury in 1 g of biological tissue with a coefficient of
variation of 1.9%.
New methods, such as inductively coupled plasma-mass spectrometry (Kalamegham and Ash,
1992) for analyzing mercury in biological samples are being developed, but are considered very costly
and unaffordable by many laboratories. For additional detail on other methods, please refer to the
Toxicological Profile for Mercury (Update) (ATSDR, 1999) and in the World Health Organization
(WHO) report Methylmercury (IPCS, 1990).
2.6 PHARMACOKINETTC MODELS
A number of extrapolations are generally required in risk assessments, including high-dose to low-
dose extrapolations, route-to-route extrapolations, cross-species extrapolations, and extrapolations for
varying exposure durations. Physiologically based pharmacokinetic (PBPK) modeling can increase the
accuracy of these extrapolations if one has data to use in the model parameters. (Clewell and Andersen,
1985, 1989; Clewell, 1995a; Andersen et al., 1995).
For methylmercury, PBPK modeling in the risk assessment process is used to estimate the
relationship between the measure of exposure used in epidemiological studies (mercury in hair and
blood) and the daily ingested dose used to determine a reference dose. Several human PBPK models
have been developed (Luecke et al., 1994, 1997; Smith et al., 1994; Gearhart et al., 1995; Clewell et al.,
1999) to address this issue. Two animal models (Farris et al., 1993; Gray, 1995) were also developed to
describe the disposition and metabolism of methylmercury and its major metabolite, mercuric mercury, in
rats. A brief description of the pharmacokinetic models developed for methylmercury is presented here.
A PBPK model was developed by Farris et al. (1993) to simulate the disposition of methylmercury
and its primary metabolite, inorganic or mercuric mercury, in the adult rat. Farris et al. (1993) also
conducted metabolism and distribution studies in rats to collect the data needed to understand the
processes that influence the pharmacokinetics of both methylmercury and mercuric mercury. This
model incorporated time-dependent compartment volume changes, compartment volume-dependent
clearance rates, and the recycling of mercury as a result of hair ingestion during grooming. The Farris
model served as the foundation for several subsequent models developed for methylmercury.
2-8
Methylmercury Water Quality Criterion 1/3/01
-------
On the basis of the modeling results reported by Farris et al. (1993), Smith et al. (1994) developed a
simple human PBPK model. Smith et al. (1994) assumed that methylmercury behaved as a single pool
while the behavior of its metabolite (inorganic mercury) varied in different tissues. Smith et al. (1994)
also conducted experimental studies in human volunteers to monitor levels of methylmercury and
inorganic mercury in the blood, urine, and feces following a single intravenous injection of a tracer dose
of methylmercury. The modeling results indicated that inorganic mercury accumulated in the body and
was the predominant form of mercury present at longer times following administration. The biological
half-life of methylmercury in the body was estimated to be 44 days, with an estimated 1.6% of the body
burden excreted each day.
Gray (1995) developed a PBPK model for methylmercury in the rat that could be used to evaluate
the developmental toxicity observed following in utero exposure to methylmercury. The model consists
of a maternal model with a fetal submodel. This model can be used to obtain fetal and maternal organ
methylmercury concentration-time profiles for any maternal dosing regimen, including the dosing
patterns used in rat developmental neurobehavioral studies.
Luecke et al. (1994) developed a generic PBPK model for human pregnancy that was applied
(Luecke et al., 1997) to both rat and human kinetic data for methylmercury. This model consists of four
submodels and incorporates the changes observed in both the mother and the fetus during the time course
of pregnancy. Both rat and human data have been simulated using the model following various routes of
exposure to methylmercury.
Stern (1997) identified data on the distribution of parameters in the one-compartment model from
the published literature. Available data specific to women between the ages of 18 and 40 were used; data
between men and women were also used to determine statistical differences, if any. Blood volume and
body weight were assumed to be correlated. A similar approach was used by Swartout and Rice (2000).
In that analysis, however, some of the parameters are described by different distributional shapes or by
distributions from different data sources than those used by Stern (1997).
Swartout and Rice (2000) performed an uncertainty analysis of the estimated ingestion rates used to
derive the methylmercury reference dose. The uncertainty arising from the calculation of ingestion dose
levels in mg/kg per day corresponding to measured concentrations of mercury in hair is estimated
through a Monte Carlo analysis of the EPA dose conversion model. The Monte Carlo model was
modified to include a methylmercury elimination concentration that was converted to an equivalent half-
Methylmercury Water Quality Criterion 1/3/01
2-9
-------
life, and a term was added to account for measurement error of hair-mercury concentrations. The authors
assumed correlations between several pairs of parameters: the hair-to-blood ratio and the elimination-rate
constant, body weight and blood volume, and the fraction of the absorbed dose in the blood and body
weight. Applying the results of this analysis and assuming the input correlations to the benchmark dose
of 11 ppm mercury in hair used in the derivation of the methylmercury RfD results in a lower 95%
confidence limit of 4.07 x 10"4 mg/kg-day. The dose conversion factor simulation is 8.0 x 10'5 with a
90% confidence interval of 3.7 x 10"5 to 1.6 x 10"4. The corresponding dose conversion value used in the
derivation of the methylmercury reference dose is 9.8 x 10'5. The 90% confidence interval spans a three
fold to five fold range of ingestion doses for any given concentration of mercury in hair. The hair-to-
blood mercury concentration ratio contributed to the variance of the output.
Gearhart et al. (1995) developed a multicompartment adult and fetal model to analyze
epidemiological data for a methylmercury risk assessment. This model was recently reparameterized by
Clewell et al. (1999) for use in a Monte Carlo variability and sensitivity analysis. The model structure, a
modification of the model developed by Farris et al. (1993), consists of a maternal model with a fetal
submodel. Changes in both maternal and fetal tissues during gestation are described. The model has the
capability to estimate maternal hair and blood concentrations following ingestion of methylmercury, as
well as the resulting fetal cord blood concentrations. This model was used to address the relationship
between mercury hi maternal hair and daily ingested dose, which has been identified as a major issue in
conducting a risk assessment for methylmercury. The results of Monte Carlo analysis using the model
provided an estimate of the variability in ingestion rates associated with a measured hair concentration.
The predicted variability (ratio of median to 5th percentile equals 1.5) is comparable to similar analyses
performed using a simple compartmental model (U.S. EPA, 1997e; Stern, 1997). The results of a
sensitivity analysis of the model suggest that the most important determinants of pharmacokinetic
variability for methylmercury are the hainblood partition, body weight, and hair growth rate.
2-10
Methylmercury Water Quality Criterion 1/3/01
-------
3.0 TOXICOLOGICAL BASIS FOR CRITERIA
This section of the Water Quality Criteria for the Protection of Human Health document for
methylmercury relies heavily on information provided in the Mercury Study Report to Congress (MSRC)
(U.S. EPA, 1997e) for summaries of studies published before 1997. Data published after 1997 are
summarized in this chapter. The Water Quality Criteria for the Protection of Human Health document
for methylmercury is not intended to be an exhaustive survey of the voluminous health effects literature
available; rather, it includes detailed information on studies that form the basis for EPA's hazard
identification and dose-response assessment. The database on neurodevelopmental effects of
methylmercury is quite extensive. Developmental neurotoxicity is currently considered the most
sensitive health endpoint. Data on cardiovascular and immunological effects are beginning to be
published and may provide a more sensitive endpoint for low-dose methylmercury effects. This chapter
will focus on developmental neurotoxic, cardiovascular, and immunological toxic effects of
methylmercury exposure. The reader is referred to the MSRC for information on other toxic effects of
methylmercury.
3.1 INTRODUCTION
Methylmercury is a highly toxic substance with a number of adverse health effects associated with
its exposure in humans and animals. Human exposure following high-dose poisonings in Japan and Iraq
resulted in effects that included mental retardation, cerebral palsy, deafness, blindness, and dysarthria in
individuals who were exposed in utero and sensory and motor impairment in exposed adults. Chronic,
low-dose prenatal methylmercury exposure from maternal consumption offish has been associated with
more subtle endpoints of neurotoxicity in children. Results from animal studies also show effects on
cognitive, motor, and sensory functions. The following section focuses on studies reporting
neurotoxicity as an endpoint for methylmercury exposure.
Methylmercury Water Quality Criterion 1/3/01
3-1
-------
3.2 NEUROTOXICITY
3.2.1 Human Studies
3.2.1.1 Minamata and Niigata, Japan
Minamata Bay, Japan
The first documented widespread human methylmercury poisoning occurred in Minamata, Japan,
between 1953 and 1960. Over time the source of the poisoning was traced to consumption of
contaminated fish and seafood from Minamata Bay. An industrial plant was found to have discharged
waste containing mercury directly into the waters of the bay. The initial cases of what was later called
Minamata disease were two young women with what appeared to be encephalitis. Public awareness of
the situation grew after the sudden deaths of cats in the surrounding area. Cats were brought into
Minamata in February 1957 to study the possible health impact of environmental exposure to
methylmercury. Within 32 to 65 days after arrival, all developed similar symptoms (e.g., excessive
salivation, violent rotational movements, inability to walk in a straight line, and collapsing death or
voluntarily jumping into the sea to drown) (Harada, 1995). This episode revealed the potential
neurotoxic effects on humans exposed to methylmercury.
Adult Minamata Disease
Officially, approximately 2,200 persons have Minamata disease. Many other cases of the disease
have either not been reported or were misdiagnosed. Many had eaten contaminated fish and shellfish for
quite some time before the symptoms appeared (Iwata et al., 1975). In human patients, the early stage of
Minamata disease brought gross disturbance of the central nervous system, which affected approximately
88 people living in the area around Minamata Bay. Of those 88 people, 12 died within 100 days, while
the others had permanent disability. Among those with permanent disability, symptoms included appallic
symptoms and idiotic disorders, with nervous symptoms resulting from widespread disturbance of brain
cortices. In those with advanced illnesses from moderate poisoning, symptoms included tremor,
disturbance of sensation, severe generalized ataxia, dysarthria, concentric constriction of the visual
fields, and difficultly hi hearing (Takeuchi et al., 1975).
3-2
Methylmercury Water Quality Criterion 1/3/01
-------
The most common clinical signs observed in adults were paresthesia, ataxia, sensory disturbances,
tremors, impairment of hearing, and difficulty in walking. Examination of the brains of severely affected
patients who died revealed marked atrophy of the brain (55% normal volume and weight), with lesions in
the cerebral cortex and cerebellar cortex, and changes in the nerve fibers, cystic cavities, and spongy foci
(Harada, 1995). Microscopically, entire regions of the brain were devoid of neurons, granular cells in
the cerebellum, Golgi cells, and Purkinje cells. In addition to effects on the brain, methylmercury is
known to have direct effects on the visual field. Korogi et al. (1997) presented results from a study on
the comparison of magnetic resonance imaging findings of the striate cortex with visual field deficits in
patients with Minamata disease. Results from this study indicated that the central 10° and 15° of vision
represent 20% and 30% of the surface area of the striate cortex, respectively. The central portion of the
visual fields occupied the posterior area as well as a greater proportion of the striate cortex. The visual
field deficits in patients with Minamata disease correlated well with the magnetic resonance findings of
the striate cortex. In severe cases of Minamata disease, the visual fields are identical with bilateral
homonymous hemianopsia, with sparing of central vision (Korogi et al., 1997).
Delayed Onset-Type Minamata Disease
Mercury content in the hair and blood samples of Minamata patients was not analyzed until 1959.
This was due in large part to the latency of the disease; the Minamata incident had apparently continued
for such a protracted period that symptoms were delayed in appearing. In some cases, symptoms
appeared more than 5 years after methylmercury intake ceased. Symptoms of delayed Minamata also
were complicated by other diseases or aging. In the case of maternal exposure, symptoms usually did not
appear until 5 to 8 years after the birth of the child. At this time, hair samples from mothers ranged from
1.82 to 191 ppm, while that of their offspring (congenital patients) ranged from 5.25 to 110 ppm (Harada,
1995).
Congenital Minamata Disease
Awareness of the developing fetus as a sensitive subpopulation came to light when a number of
children were bom with congenital cerebral palsy. These patients experienced symptoms such as mental
retardation, primitive reflex, cerebellar ataxia, disturbances in physical development and nutrition,
dysarthria, deformity of the limbs, hyperkinesia, hypersalivation, paroxysmal symptoms, strabismus, and
pyramidal symptoms. Pathological findings of congenital Minamata disease patients include general
atrophy and hypoplasia of the brain cortex and abnormality of the cytoarchitecture, remaining matrix
Methylmercury Water Quality Criterion 1/3/01
3-3
-------
cells, hypoplasia of the corpus callosum, intramedullary preservation of the nerve cells, and
dysmyelination of the pyramidal tract. In the cerebellum, hypoplasia of the granular cell layer and other
layers as well as degeneration of granular cells were observed (Harada, 1995).
In a small fishing village called Yudo, 7 cases of cerebral palsy and 10 cases of infantile Minamata
disease were found in a total of 50 households. Between 1955 and 1958, there were 188 births in the
small fishing villages of Yudo, Tsukinowa, and Modo, with a 9.0% incidence of Cerebral palsy, while the
overall national incidence ranged from 0.2% to 2.3% (Harada, 1995).
Extensive investigations of congenital Minamata disease were undertaken and 20 cases that
occurred over a 4-year period were documented. The exact number of congenital Minamata disease
patients is not known, as some undiagnosed patients were already deceased. At present, 64 cases have
been confirmed as congenital Minamata disease. In all instances congenital cases showed a higher
incidence of symptoms than did the cases where exposure occurred as an adult. The congenital patients
are unable to perform ordinary functions of living (Harada, 1995).
From 1950 to 1969, a total of 151 umbilical cords were collected from residents of the Minamata
area. Included in this pool were 25 patients with congenital Minamata disease. Levels of methylmercury
in the umbilical cords ranged from 0.35 ppm in 1952 to 0.96 ppm in 1955. The methylmercury levels in
the cords from patients with congenital Minamata disease showed higher values than the cords of patients
who had Minamata disease (0.72 ppm), mental retardation (0.74 ppm), other diseases (0.22 ppm), and no
symptoms (0.28 ppm) (Harada et al., 1999).
Kinjo etal. (2993)
*
A case-control study examined the relationship between health complaints of patients with
Minamata disease and exposure to methylmercury. A total of 1,144 Minamata disease patients older than
40 years of age were surveyed. A control group was also established; this group included nonexposed
people living in neighboring towns, matched by age and sex. A questionnaire was used to obtain
information on subjective complaints and activities of daily living (ADL). Results from analysis of the
data indicated that Minamata disease patients had significantly higher rates of all complaints than did
controls. Subjective complaints of Minamata disease patients, overall, were more prevalent than in
controls. The results remained unchanged with age when the subjective complaints were categorized into
two groups: those where frequency increased with age and those related to sensory disturbance. The
3-4 Methylmercury Water Quality Criterion 1/3/01
-------
authors noted that the reason for the high prevalence rate of sensory disturbance among current
Minamata disease patients is unclear. The data from the ADL questionnaire, when analyzed, were used
to estimate functional capability in the elderly. Results indicate that ADL was significantly lower for
Minamata disease patients aged 60 and over in comparison with controls. The authors conclude that
ADL disability in Minamata disease patients is accelerated by aging. Overall, the prevalence of deficits
was relatively greater in cases compared with controls as a function of increasing age.
Haradaetal (1998)
In 1995, Harada et al. (1998) measured mercury concentration in hair samples from 191 fishermen
and family members living in mercury-polluted areas in the Minamata region of Japan. The study
participants fished for a living and had previously consumed methylmercury-contaminated fish and
shellfish caught in this region. Estimates offish consumption were not provided. The study population
comprised 83 men and 108 women who ranged in age from 32 to 82 years. Data on subjective symptoms
and lifestyle factors were collected by questionnaire. In addition, each participant was administered
relevant neurological tests (test details not provided) by a group of neurologists. Mercury concentrations
in hair were less than 10 ppm in 185 out of 191 subjects. The mean concentrations were 5.0 ± 3.4 ppm
and 2.1 ± 1.1 ppm for men and women, respectively. All six subjects with hair concentrations greater
than 10 ppm were men. The mean concentration for men in the study was only slightly higher than the
mean value of 4.6 ppm for normal nonexposed Japanese men. There appeared to be an upward trend in
hair mercury concentration associated with increased frequency offish consumption. Although the hair
mercury concentrations approached what was considered normal (z 10 ppm in hair samples), the study
participants exhibited a high incidence of a variety of neurological conditions. More than 85% of
subjects reported subjective symptoms including numbness, forgetfulness, pain in the extremities, focal
cramps, headache, and motor disturbances. Clinical findings included sensory disturbance, ataxia,
speech impediment, hearing impairment, constriction of visual fields, and tremor. "Stocking and glove"
sensory disturbance (a hallmark of Minamata disease) occurred in 69% of the participants. A dose-
response relationship between clinical symptoms and hair concentration was not evident, indicating that
hair level data were of limited use for diagnosis of chronic Minamata disease.
Fukuda et al. (1999)
A study was completed in Kumamoto, Japan, near Minamata City, to evaluate the relationship
between the number of neurological complaints from symptoms and methylmercury exposure. A total of
Methylmercury Water Quality Criterion 1/3/01
3-5
-------
1,304 exposed adults living in a methylmercury-polluted area and 446 nonexposed age-matched .adults,
living in an area not known to be polluted with methylmercury, participated in an interview and
questionnaire survey. The data from 64 participants of the survey were analyzed by comparison of
prevalence, factor analysis, and cluster analysis. Results indicated that the exposed population had more
neurological complaints in comparison with those not exposed. The factor analysis proposed four
factors: arthritic, muscular, sensory, and nonspecific complaints. All four were higher in the exposed
population in comparison with the nonexposed. The authors suggest that the increased neurological and
nonspecific complaints may be due to past exposure to methylmercury.
Futatsuka et al. (2000)
A case-control study was conducted to estimate the role of various risk factors, including
methylmercury exposure, for diseases such as liver disease, renal disease, and diabetes mellitus. The
study population included 1,500 subjects over 40 years of age living in the town of Tsunagi since 1984.
The town of Tsunagi was methylmercury polluted, with 36.9 diagnosed Minamata disease patients for
every 1,000 population. Urine, blood, physical, and ultrasonographic examinations were administered to
determine evidence of liver disease, renal disease, and diabetes mellitus. Personal interviews were
conducted to collect information on risk factors and specific details on the complaints. Results from this
study indicated that prevalence of disease, liver disease, renal disease, and diabetes mellitus was not
higher in the methylmercury-polluted area compared with other areas in Japan. However, subjects in the
polluted area had more complaints than those in the nonpolluted area. The authors concluded that past
exposure to methylmercury may have influenced these results.
Niigata, Japan
From 1963 to 1965, patients with Minamata disease-like symptoms were reported in the basin of
the Agano River in Niigata. Methylmercury, a residual product from acetoaldehyde synthesis, was
released from a manure factory located 70 km up the river. Untreated wastewater from the factory
drained into the Agano River, contaminating the fish and shellfish population. By 1973, 325 patients
with Minamata disease were identified. This poisoning was later named "Niigata Minamata disease."
Similar to the incident in Minamata, the symptoms progressed even after cessation of exposure.
Numbness in the extremities and in the perioral area was the most frequently reported (Iwata et al.,
1975). In the Niigata incident, the maternal hair mercury concentration immediately after giving birth to
a congenital patient was 293 ppm. The maternal symptoms associated with this level of exposure were
3-6
Methylmercury Water Quality Criterion 1/3/01
-------
mild, with sensory disturbances and other Minamata disease-related symptoms. The level of mercury
exposure required to initiate the onset of Minamata disease was established at 50 ppm maternal mercury
hair level. Because of the previous experience in Minamata with methylmercury poisoning, women with
hair mercury levels above 50 ppm were advised not to become pregnant. As a consequence, there was
only one case of congenital Minamata disease in the Niigata incident (Harada, 1995).
3.2.7.2 Iraq Outbreak
In fall 1971, 90,000 metric tons of methylmercury-treated seed grain were imported through the
southern seaport of Basra, Iraq, and distributed freely throughout the countryside. Because the grain was
delivered at planting time, residents of the area baked the grain into bread. There are no records on the
size of the population who consumed grain treated with methylmercury fungicide. Nor are there reliable
estimates of the number of people who ate methylmercury-treated grain and developed signs and
symptoms but did not seek medical attention. It was not until late December 1971 that the first case of
methylmercury poisoning was recorded. Within 2 months, 6,530 hospital admissions and 459 hospital
deaths were recorded from methylmercury ingestion. Included in this exposed population were pregnant
women (Bakir et al., 1973). Children exposed in utero manifested severe sensory impairments such as
blindness and deafness, general paralysis, hyperactive reflexes, cerebral palsy, and impaired mental
development (Amin-Zaki et al., 1974).
A study was conducted by Marsh et al. (1987) to investigate the relationship between
methylmercury exposure, as measured by maternal hair concentrations during pregnancy, and associated
adverse effects in offspring. A total of 81 mother-infant pairs participated; maternal hair mercury levels
served as the index for prenatal exposure and were measured by x-ray fluorescent spectrometric analysis
to range from 1 to 674 ppm. Clinical evaluations were conducted along with interviews with the mother
about labor, delivery, any abnormalities at birth, size of the baby, early childhood development, and age
at which infants achieved developmental milestones. These milestones included sitting without support,
standing and walking unaided, and speaking two or three meaningful words. Developmental retardation
was indicated by the child's inability to walk a few steps unsupported by 18 months of age or to speak
two or three meaningful words by 24 months of age. Additional questions included any observations of
involuntary movements, seizures, impaired vision or hearing, lack of coordination, and the mother's
general impression of the child's physical and mental development. The interview was limited by the
mothers' recall of the age of their children; moreover, this culture did not use Western calendars to
record family events. The physical examination of the child included observation; head circumference
Methylmercury Water Quality Criterion 1/3/01
3-7
-------
and body length measurements; cranial nerve signs; speech; limb tone, strength; deep tendon reflexes;
plantar responses; coordination; dexterity; primitive reflexes; sensation; posture; and ability to sit, stand,
walk, and run. Neurological examinations scored 0 to indicate normal functions and 3 to indicate
definite abnormality. Unclear readings were denoted with points for borderline findings, whereas scores
of 0-3 reflect no definite abnormality. The highest score in the most severely affected child was 11.
The impact of methylmercury on neurological function of infants exposed in utero during the Iraqi
poisoning incident is described hi a series of reports by Amin-Zaki et al. (1974, 1976, 1979, 1981),
Marsh et al. (1980, 1981, 1987), and Seafood Safety (1991). The major symptoms observed in this
epidemic closely resembled those recorded in Minamata, Japan. The predominant symptom noted in
adults was paresthesia, and it usually occurred after a latent period of 16 to 38 days following initiation
of exposure. Additional dose-dependent symptoms observed in the more severely affected individuals
included ataxia, blurred vision, and constriction of the visual field leading to blindness in severe cases,
slurred speech and hearing difficulties. Fatalities from methylmercury exposure usually resulted from
failure of the central nervous system (Bakir et al., 1973). Of the 28 children with the highest exposures,
7 had seizures, whereas none of the 53 children with the lowest exposures experienced seizures.
Maternal hair mercury levels for those seven children ranged between 78 and 674 ppm.
Results indicate that boys appeared to be more severely affected than girls. Statistically significant
differences were apparent for regressions for boys and girls, where boys had the steeper slope to indicate
increased severity in late walking and talking than girls.
Cox et al. (1989) performed an analysis of the Iraqi data to identify the threshold for adverse
neurodevelopmental effects if one existed. A variety of statistical models such as logit, hockey-stick, and
nonparametric kernel-smoothing methods were used in the attempt. Analyses were limited by the lack of
data on the background prevalence of poor outcomes among Iraqi children. The authors estimated a
population threshold of approximately 10 ppm for the outcomes investigated. The uncertainty associated
with such an estimate, however, is highly dependent upon the assumed background prevalence of poor
outcomes (e.g., motor retardation, neurological abnormality) (Cox et al., 1989). In another attempt at
reanalyzing the data, Crump et al. (1995) reported that the estimate of the population threshold was
highly dependent on the choice of the model and highly sensitive to the definition of abnormality. For
example, delayed walking was heavily influenced by four cases of delayed walking among children with
corresponding maternal hair mercury levels below 150 ppm. Crump et al. (1995) concluded that the
statistical upper limit of the threshold could be as high as 255 ppm. Furthermore, their maximum
3-8
Methylmercury Water Quality Criterion 1/3/01
-------
likelihood estimate of the threshold using a different parametric model was said by the authors to be
virtually zero.
Cox et al. (1995) analyzed the Iraqi data on late walking in children exposed to methylmercury in
utero. The results indicated that dose-response analyses based on late walking endpoints were unreliable
because of four influential observations in the group of responders with hair mercury levels below 150
ppm. Based on visual interpretation of the plot of the data, the four observations are isolated from the
remainder of the responders and would be expected to have considerable influence on the threshold
estimate. No quantitative sensitivity analysis was performed to further investigate the effect of removing
one or more of these data points. The authors point out that if the four data points were to represent
background, the threshold for late walking would be greater than 100 ppm. This is, however, considered
unlikely given that no responses were observed in the 37 individuals with lower levels of exposure.
3.2.1.3 Peru
A prospective study (Marsh et al., 1995) was conducted in Mancora, Peru, between 1981 and 1984
but not published until 1995. Mancora was selected as the study site based on a number of criteria, but
mainly for its dependence on marine fish as a large source of dietary protein. A diet high in seafood was
presumed to be associated with methylmercury exposure. Study participants consisted of 369 pregnant
women and 194 of their children. Maternal hair samples were collected from the final group of 131
mother-infant pairs to analyze for methylmercury content. The geometric mean hair level was 7.05 ppm,
with a range of 0.9 to 28.5 ppm. The peak maternal hair methylmercury levels during pregnancy ranged
from 1.2 to 30 ppm, with a geometric mean of 8.3 ppm. Neurological examinations were administered to
children. Frequencies were reported for tone decreased; tone increased; limb weakness; reflexes
decreased; Babinski's sign, which is an indicator of a pyramidal-tract abnormality; primitive reflexes;
and ataxia. This study identified no significant relationship between maternal hair methylmercury levels
and measures of infant development or neurological signs. The authors suggested that marine fish may
contain elements, such as selenium, that reduce the toxicity of methylmercury, thereby masking any
neurological effects associated with methylmercury exposure.
3,2.1.4 Northern Quebec, Canada
A cross-sectional study of 234 Cree Indian children between the ages of 12 and 30 months on July
1, 1978, was conducted by McKeown-Eyssen et al. (1983). These children resided in four northern
Methylmercury Water Quality Criterion 1/3/01
3-9
-------
Quebec communities known to have the highest levels of methylmercury exposures within Quebec.
Maternal hair mercury level was the index to reflect prenatal exposure. Methylmercury levels of the hair
were measured in alternate 1-cm segments, beginning with the scalp-end segment. The average maternal
hair methylmercury concentration was 6 ppm, with only 6% of the samples exceeding 20 ppm. Physical
and neurologic examinations were administered to the children, with the additional measures of special
senses, cranial nerve function, sensory function, muscle tone, stretch reflexes, coordination, persistence
of Babinski's response, and a summary of signs for the absence or presence of neurologic abnormality.
At 4 years of age, four measures of the Denver Developmental Scale (gross and fine motor development,
language development, and personal and social skills) were administered to assess the child's
development. Associations between exposure and neurological outcome were analyzed by multiple
regression analyses adjusted for alcohol and caffeine intake, tobacco use, age of mother, and multiparity.
No significant association between methylmercury exposure and neurological deficits was
identified hi girls. Abnormality of tendon reflexes was evidenced in 11.4% of the boys and 12.2% of the
girls, but was only significantly associated with maternal hair mercury in boys. The prevalence of
abnormality of muscle tone or reflexes was found to increase seven times with each increase of 10 ppm
of the prenatal exposure index. However, the authors caution the interpretation of the results on boys
because the abnormality of muscle tone or reflexes tended to consist of isolated abnormalities of mild
severity that are of doubtful clinical importance. In addition, there was no dose-response relationship.
3.2.1.5 Seychelles Islands
The Seychelles Child Development Study (SCDS) was initiated in 1981 to examine the effects of
low-dose fetal exposure to methylmercury from maternal consumption of fish. The SCDS was planned
and conducted in two separate stages. The preliminary cross-sectional stage of the study sought to
provide additional detail and guidance on how to design the main study. The main study, started in 1989,
was a double-blind, prospective, longitudinally designed study that followed a cohort of infant-mother
pairs from 6 months to 66 months postgestation.
Demographics
The Seychelles Islands is a Westernized archipelago in the middle of the Indian Ocean, more than
1,500 kilometers from the eastern coast of mainland Africa. The Seychellois population is of African
and European origin with some minority groups from India and China. English, French, and Creole are
3-10 Methylmercury Water Quality Criterion 1/3/01
-------
the three official national languages, with Creole being the most popular language at home. A majority
(-85%) of the population consume a high amount of marine fish on a daily basis. In general, the
Seychellois population is considered quite healthy, with easy access to good health care and education
(Marsh et al., 1995).
Cross-Sectional Pilot Study (Myers et al, 1995b,c)
From 1987 to 1988, a cohort of 789 mother-infant pairs was selected after exclusion criteria were
exercised. The fetal exposure index used was maternal hair total mercury. The levels ranged from 0.59
to 36.4 ppm, while the median level in this study was 6.6 ppm total mercury. The Denver Developmental
Screening Test-Revised (DDST-R) was administered and a medical and neurological examination was
performed for each child between 5 and 109 weeks of age. Covariates were selected for statistical
analysis because of their potential to bias the assessment of the association between maternal mercury
and developmental outcomes. These covariates included gender, birth weight, Apgar score, age at
testing, and medical history. Mother's age, use of alcohol and tobacco, and medical history also were
used. When DDST-R scores of questionable and abnormal results were grouped, mercury effects were
seen and were more pronounced in boys and declined as age of testing increased. In general, males had
higher response rates on the DDST-R than females, independent of mercury level. No association,
however, was observed between mercury exposure and overall neurological examination results. The
authors cautioned the interpretation of the results because the developmental association with fetal
mercury exposure disappeared when DDST-R scores of "questionable" were treated in the standard
manner as passes.
A subset (217 children) of the children from the pilot study cohort (Myers et al., 1995a) was tested
at 66 months of age with the same battery of tests as planned for the main study at similar age. Maternal
hair mercury levels during pregnancy ranged from 1.0 to 36.4 ppm, while the median level was 7.1 ppm.
Nine endpoints were evaluated in this second evaluation: the McCarthy Scales of Children's Abilities'
that yield the general cognitive index (GCI), perceptual performance, memory, and motor ability; the
Preschool Language Scale that yields total language score and subscores for verbal ability and auditory
comprehension; and the letter-word identification and applied problems subscales of the Woodcock-
Johnson Tests of Achievement. The association between maternal hair mercury concentration and
outcome was assessed by multiple regression analysis. Prenatal mercury exposure correlated with
outcomes at 66 months on the McCarthy GCI and perceptual performance subscale and with total
language and auditory comprehension scores. After removing outliers and influential points, however,
Methylmercury Water Quality Criterion 1/3/01
3-11
-------
mercury effects were no longer significant except for the Preschool Language Scale auditory
comprehension subscale.
Prospective Longitudinal Main Study
A double-blinded, prospective longitudinal study was initiated with a new cohort of 740 mother-
infant pairs that were selected between 1989 and 1990. These participants resided on the island of Mahe,
which is one of the largest islands in the archipelago of the Seychelles where 90% of all Seychellois
citizens live. Maternal hair mercury level was used as the marker of fetal mercury exposure. The levels
ranged from 0.5 ppm to 26.7 ppm, with a median of 5.9 ppm. The cohort was followed from ages 6.5
months to 66 months, with evaluations occurring uniformly at four critical periods (6.5, 19, 29, and 66
months of age) (Myers et al., 1995). Tests of 7-year-old children have also been done, but results are not
yet published. Age-appropriate tests were administered at the time points indicated in Table 3-1.
6-Month Evaluation (Myers et al., 1995c)
At 6 months of age, all children were administered a standardized test of visual recognition memory
(Pagan Infantest); a standardized screening test to measure personal-social, fine motor adaptive,
language, and gross motor development (DDST-R); and a general medical and neurological examination.
Covariates of this main study included those evaluated in the pilot study, with the addition of birth order,
gestational age of the child, primary caregiver intelligence, maternal and paternal educational levels,
history of breastfeeding, language spoken at home, and family income. Medical conditions related to
poor neurodevelopmental outcomes were also included as covariates in the statistical analysis. The study
results indicate no association at 6 months of age with DDST-R, neurological examination, and Pagan
Infantest. However, males had lower scores on both tests than females.
19- and 29-Month Evaluations (Davidson et al., 1995)
At 19 months of age, children were evaluated with the Bayley Scales of Infant Development
(BSID), while the primary caregiver was administered the Raven Standard Progressive Matrices. The
cohort was evaluated again at 29 months. Infant intelligence was measured by BSID Mental and
Psychomotor Scales. To measure adaptive behaviors, a modified version of the BSID Infant Behavior
Record was completed at 29 months. Between the ages of 42 and 56 months, children were administered
3-12
Methylmercury Water Quality Criterion 1/3/01
-------
Table 3-1. Developmental domains evaluated and tests applied in the Seychelles Islands Child
Development Main Study
Developmental Domain
Age of Child (months)
6.5
Marsh etal. (1995)
Global-cognitive
Visual-perceptive
Speech-language
Memory
Visual attention
Neuromotor exam
Behavioral
Learning-achievement
Auditory response
DDST-R
—
DDST-R
Pagan Infantest
Pagan Infantest
Neurological
DDST-R
DDST-R
—
—
19
29
66
BSID MDI
Kohen-Raz
"
—
—
BSID PDI
—
—
—
BSID MDI
Kohen-Raz
^~~
—
—
BSID PDI
BSID IBR
—
—
MSCAGCI
Bender-Gestalt
MSCA Perceptual
MSCA Verbal
PLS Total Language
Aud. Comprehension
Verbal Ability
MSCA Memory
—
Bender-Gestalt
MSCA Motor
CBCL
Woodcock- Johnson
Audiometry
Tympanometry
Davidson etal (1998)
Global-cognitive
Visual-perceptive
Speech-language
Behavioral
Learning-achievement
—
—
—
—
—
• —
—
—
—
—
—
—
"
MSCA GCI
Bender-Gestalt
PLS Total Score
CBCL
Woodcock-Johnson
Letter and Word
Recognition,
Applied Problems
Symbols and Abbreviations: — = No test administered; BSID = Bailey Scales of Infant Development; IBR = Infant Behavior
Record; MDI = Mental Developmental Index; PDI = Psychomotor Developmental Index; CBCL = Child Behavior Checklist;
DDST-R = Denver Developmental Screening Test - Revised; GCI = General Cognitive Index; MSCA = McCarthy Scales of
Children's Abilities; PLS = Preschool Language Scale.
Source: Marsh et al. (1995); Davidson et al. (1998).
Methylmercury Water Quality Criterion 1/3/01
3-13
-------
the Pre-School Caldwell-Bradley Home Observation for Measurement of the Environment (HOME).
Hair samples were collected from all children at both 19 and 29 months of age for analysis of total
mercury concentration to determine postnatal exposure. The median maternal hair mercury
concentration during pregnancy for the 738 mother-infant pairs in the cohort at 19 months was 5.8 ppm.
Twenty-two percent of the children at 19 months had child hair mercury levels ;> 10 ppm (Myers et al.,
1997). The same covariates and modeling strategy were used as in the primary analysis. No effects of
mercury were detected on the BSID scores at either age. Results of this study indicate that one functional
behavior—the examiner's subjective rating of the child's test session activity level—was related to •
maternal hair mercury levels in the mothers of male children: activity level decreased as maternal hair
mercury level increased. Independent of mercury exposure; activity level was rated higher in males.
Authors of this study conclude that these two results suggest that prenatal exposure to mercury may
lower activity level in males. This result should be interpreted with caution as it is not yet clear whether
the lower activity in males is a direct result of increased mercury exposure.
19-Month Evaluation of Walking and Talking (Myers et al., 1997)
The 19-month cohort was selected for evaluations of two developmental milestones. Data for age
of first walking (n = 720) and talking (n = 680) were obtained from the primary caregiver of each child.
Age at walking was defined as the age when the child was able to walk without support, while age at
talking was defined as the age the child first said words other than "mama" and "dada." The mean age
for walking was 10.7 months for girls and 10.6 months for boys, while for talking it was 10.5 months for
girls and 11.0 months for boys. Multiple regression analysis was used to assess the relationships between
each developmental milestone, maternal hair mercury levels, and covariates. Covariates evaluated are
the same as those included in the study reported by Davidson et al. (1995) described in the previous
paragraph. In this study, there was a marginally significant relationship between prenatal mercury
exposure from eating fish and the age at which males started to walk, but this depended on four statistical
outliers. No association between prenatal mercury exposure and either the age at which females started
to walk or either gender started to talk was found.
Semiparametric Modeling of the 19-Month Data (Axtell et al., 1998)
In addition to the multiple regression analysis used in the prospective longitudinal main study of the
SCDS, a semiparametric generalized additive model was used to identify nonlinearities in the
relationship between prenatal methylmercury exposure and developmental milestone achievements. The
3-14 Methylmercury Water Quality Criterion 1/3/01
-------
specific milestones evaluated in the main SCDS cohort at 19 months of age (n = 738 children) were age
that children walked and said words. Walking was defined as the number of steps without support and
talking was any word except "mama" or "dada." Maternal hair total mercury was used as an index of
fetal exposure. No significant nonlinear relationships with mercury were identified in any of the models
for age at talking; this implies that the original linear regression models were appropriate for this
analysis. A General Additive Model analysis indicated that the relationship between maternal hair
mercury level and age at walking may not be linear. Walking appeared at a later age as exposure
increased in the range from 0 to 7 ppm. Walking appeared slightly earlier with increasing mercury levels
above 7 ppm. However, there was no evidence from any models that higher levels of mercury exposure
resulted in further delays in walking. There is no biological or developmental hypothesis to explain the
increase in age of walking at lower levels and not at higher levels.
66-Month Evaluation (Davidson et al., 1998)
An evaluation was conducted on 711 mother-child pairs at 66 months of age. At this age, six
neurobehavioral tests were administered: McCarthy Scales of Children's Abilities, the Preschool
Language Scale, the Woodcock-Johnson Applied Problems and Letter and Word Recognition Tests of
Achievement, the Bender Gestalt Test, and the Child Behavior Checklist (CBCL). Maternal hair
mercury and child hair mercury were measured. Mercury exposure was assessed by total mercury in
segments of maternal hair representing growth during pregnancy. The mean maternal hair total mercury
level was 6.8 ppm while the mean child hair total mercury level at age 66 months was 6.5 ppm. The
covariates evaluated include all those included in the previous study period, in addition to hearing status
of the child and Hollingshead socioeconomic status of the family. Two multiple linear regression
analyses were performed for each of the six primary measures. Secondary analyses tested the hypothesis
that associations between developmental outcomes and total mercury exposure might be nonlinear. Four
of the six measures (all except for Bender Gestalt and Woodcock-Johnson Applied Problems Tests of
Achievement) showed better scores in the highest methylmercury groups compared with lower groups for
both prenatal and postnatal exposure. For both prenatal and postnatal methylmercury exposure, no
adverse developmental effects were reported for toddlers. Postnatal exposure at 66 months, however,
was associated with a small but statistically significant increase on several developmental outcomes even
though there is no reason to suppose that such effects are associated with exposure to methylmercury.
There are studies, however, that indicate the methylmercury levels in the infant were surrogate for the
length of breastfeeding, which is reported to have a positive association with developmental outcomes
(Grandjean et al. 1992).
Methylmercury Water Quality Criterion 1/3/01
3-15
-------
New Analysis—CBCL Main Cohort 66 Months (Myers et al, 2000)
No effect of mercury was identified on the Child Behavior Check List (CBCL) at 66 months of age
in the main cohort of the Seychelles study as determined by the total T score (Davidson et al., 1998).
The CBCL is a report inventory scored by the caregiver that assesses eight domains: withdrawn, somatic
complaints, anxious/depressed, social problems, thought problems, attention problems, delinquent
behavior, and aggressive behavior. An analysis of these subscales was performed on the 711 children
assessed on this test (Myers et al., 2000). No effect of mercury was identified on individual subscales.
New Analysis—Main Cohort 66 Months (Axtell et al., 2000; Palumbo et al., 2000)
The investigators performed additional analyses of the 66-month data to evaluate the possibility of
nonlinear relationships associated with mercury exposure (Axtell et al., 2000). Endpoints included the
six primary variables analyzed previously: McCarthy GCI, Preschool Language Scale (PLS), Wodcock-
Johnson Applied Problems, Woodcock-Johnson Letter/Word Recognition, Bender copying errors, and
CBCL total T score. Generalized additive models, which make no assumptions about the relationship
between exposure and test score, were used. Maternal hair levels during pregnancy were used as a
measure of prenatal exposure and child's hair mercury at 66 months was used for postnatal exposure.
Nonlinearities were identified between prenatal exposure and PLS and CBCL, and between postnatal
exposure and McCarthy GCI. For the PLS the trend involved a decrement of 0.8 points (poorer
performance) from 0-10 ppm and an increase of 1.3 points above 10 ppm. For the CBCL there was an
increase (representing a poorer score) between 0 and 15 ppm and a decrease above 10 ppm. The GCI
increased (improved) by 1.8 points through 10 ppm mercury in the child's hair and declined by 3.1 above
10 ppm. Although these results are difficult to interpret, they provide limited evidence of an adverse
effect of mercury exposure below 10 ppm maternal hair on two measures, and a somewhat greater
association of adverse effects with child's hair mercury above 10 ppm on the GCI. As pointed out by the
authors, there are fewer data points above 10 ppm (this is especially true for child's hair mercury), and
therefore trends above this level are estimated less precisely.
The investigators hi the Seychelles study further examined by multiple linear regression the results
of the McCarthy GCI administered at 66 months (Palumbo et al., 2000). They analyzed the standard
MSCA subscales and also constructed subscales to approximate the domains of cognitive functioning
assessed in the Faroe Islands study: attention, executive function, expressive language, receptive
language, nonverbal memory, visuospatial ability, visuomotor ability, and gross motor ability. They
3-16 Methylmercury Water Quality Criterion 1/3/01
-------
found a positive association between child's hair mercury at 66 months and the standard memory
subscale, with no other associations identified. As with all previous analyses of these variables, the raw
scores were converted to "normative" scores. As pointed out by an OSTP panel (NIEHS 1998, Section
3.5 of the Confounders and Variables Section), the applicability of U.S. norms to this population is
unclear, and the use of standardized scores may decrease sensitivity by collapsing different raw scores to
one standard score.
Pilot Cohort Analysis at 108 Months (Davidson et al, 2000)
Further evaluation was performed on a portion of the Seychelles pilot cohort at 108 months of age
(Davidson et al., 2000). Eighty-seven children were tested on five subtests of the WISC-HI (Information,
Block Design, Vocabulary, Digit Span, and Coding), California Verbal Learning Test (CVLT), Boston
Naming Test (BNT), Beery-Buktenica Development Test of Visual Motor Integration (VMI) (copying
geometric figures), Finger Tapping, grooved pegboard, Trailmaking (tracing the correct route through a
form with a pencil), and the design memory subtest of the Wide Range Assessment of Memory and
Learning (WRAML) (drawing each of four geometric designs from memory). Performance on BNT,
VMI, and grooved pegboard showed a positive association (better performance) related to mercury
exposure in males, with no effects identified in females. There were trends toward poorer performance
related to mercury exposure for grooved pegboard in females (p = 0.07) as well as marginal/? values on
the full model that were not further analyzed (Finger Tapping, digit span). The investigators did not
report power calculations, but with such a small number of subjects the power was probably quite low, so
these largely negative results need to be interpreted with caution.
Benchmark Analysis (Crump etal., 2000)
A benchmark analysis (Crump et al., 2000) was conducted on data from the SCDS, with the goal of
providing an alternative basis for deriving an appropriate human exposure level for methylmercury. The
data modeled included responses from the neurological test batteries conducted at 6.5, 19, 29, and 66
months of age. In addition, data for developmental milestones (age first walked and age first talked)
were analyzed. Maternal hair mercury concentrations measured in this study ranged from 0.5 to 26.7
ppm and averaged 6.8 ppm.
Most of the measured endpoints in the SCDS were recorded as continuous responses, and the k-
power model, the Weibull model, and the logistics models for continuous data were applied. Test scores
Methylmercury Water Quality Criterion 1/3/01
3-17
-------
below a predetermined value, P0 = 0.05, were considered abnormal. For this analysis, the BMR was
defined as 10% (BMR = 0.1). (For a description of modeling terms see Section 4.3).
In cases where responses were recorded as quanta! responses (abnormal/normal), the data were
modeled using the Weibull dose-response model for quanta! data. Quantal responses reported in children
in the Seychelles study included deep tendon reflexes, limb tone, overall neurological responses, and
psychomotor index. In addition, each continuous response was converted to a quanta! response by
considering a response abnormal if it was more than 2 standard deviations away (in the adverse direction)
from the mean response of the entire cohort, and then analyzed using the Weibull model. In these
analyses, the BMD was defined in the same way as in the analyses of the continuous response.
The analyses of continuous response were conducted without covariates. Analyses with P0
specified were conducted using both an expanded set and a reduced set of covariates for the children:
sex, birth weight, birth order, whether or not the child was breastfed, medical history, maternal age,
maternal smoking and alcohol use during pregnancy, maternal medical history, language spoken in home,
score from home visit, Raven group (caregiver's intelligence quotient), maternal and paternal education
level, family income, gestational age, Hollingshead socioeconomic scale, auditory scores, and the child's
mercury level. Covariates were not included in the analyses of quantal responses or in the analyses of
continuous responses in which x0 was specified.
Parameter estimates were obtained using the maximum likelihood method, and statistical
confidence bounds were computed by the profile likelihood method. The BMDL was defined
conventionally as the 95% statistical lower confidence bound on the BMD. Results indicated that the
most reliable analyses were represented by 144 calculated lower statistical bounds on the BMD (BMDL,
or the lower statistical bound on maternal mercury hair level corresponding to an increase of 0.1 in the
probability of an adverse response) derived from the modeling of continuous responses.
The results of BMD modeling are shown in Table 3-2. The average value of the BMDL in these
144 analyses was 25 ppm mercury in maternal hair, with a range of 19 to 30 ppm. With the exception of
the linear model, which produced larger BMDLs, the dose-response models applied to continuous end
points all produced comparable BMDLs.
3-18
Methylmercury Water Quality Criterion 1/3/01
-------
Table 3-2. BMDL values (expressed as ppm mercury in maternal hair) for neurological responses and
developmental milestones from the Seychelles Child Development Study
Endpoint
Model
Weihull
Pf
None
Exp.c
vb
Xd
None
Quantal ,
None
JK-Power
p a
•*o
None
Exp.
Zo"
None
6.5 Months
Deep tendon reflexes
Limb tone
Overall neurological
Fagan visual recognition memory
Pagan attention
—
—
—
26.0
25.7
19 Months
Mental development index
Psychomotor index
23.7
—
—
—
—
26.0
25.9
—
—
—
27.4
27.0
22.8
20.9
15.8
19.7
23.7
—
—
—
26.0
25.5
23.4
—
26.0
—
22.6
22.3
24.3
—
—
—
—
26.0
25.6
—
—
—
26.9
26.4
24.1
—
25.6
—
29 Months
Mental development index
Psychomotor index
24.1
—
24.4
_
25.7
—
21.9
22.5
24.0
—
24.2
—
24.8
—
66 Months
Bender gestalt errors
Child behavior checklist total
McCarthy general cognitive index
Preschool language total score
Woodcock- Johnson
Applied problems
Letter-word recognition
Developmental milestones
Age first walked unassisted
Age first talked
26.9
27.2
24.4
25.2
26.7
27.2
24.2
25.1
28.5
29.0
26.5
26.8
22.7
19.4
22.7
22.7
26.7
20.0
24.7
24.7
26.7
26.9
24.6
24.7
27.5
27.8
25.9
25.5
23.1
23.7
23.5
23.7
25.3
25.3
22.7
22.7
23.9
23.8
24.3
23.9
25.5
24.7
24.9
24.6
24.0
23.5
25.9
25.9
22.7
20.3
24.4
25.0
23.2
24.1
26.8
25.9
" Abnormal defined as a response >2 standard deviations in adverse direction from mean response of entire cohort.
* Abnormal defined so that 5% of responses are abnormal (p0 = 0.05).
cExp. denotes use of an expanded range of covariates.
Source: Crump et al., 2000.
Methylmercury Water Quality Criterion 1/3/01
3-19
-------
3.2.1.6 New Zealand
A study was conducted in the northern New Zealand islands to study the effects of prenatal
methylmercury exposure on children exposed in utero from maternal fish consumption. Between 1982
and 1983,11,000 mother-infant pairs were requested to submit hair samples and fill out a detailed diet
questionnaire. Of those 11,000 pairs approximately 1,000 of these mothers had consumed fish more than
three times per week for the 9 months of pregnancy. Seventy-three had hair mercury levels above 6 ppm,
with the highest level being 86 mg/kg. This study was conducted in two stages.
Preliminary Tests at Age 4 (Kjellstrom et al, 1986)
From the 73 mothers with high mercury exposure (> 6 ppm) during pregnancy, a total of 31
matched pairs were selected to participate in a study on the effects of prenatal methylmercury exposure
on children exposed in utero from maternal consumption of fish. A reference child matched for mother's
ethnic group, age, and child's birthplace and birth date was located for each child selected from the high-
fish-consumption group. Mercury exposure during gestation was determined from maternal hair analysis.
The average hair concentrations for high-exposure mothers and the reference group were 8.8 ppm and 1.9
ppm, respectively. At 4 years of age, the children were tested using the DDST. Standardized vision tests
and sensory tests were also performed to measure development of these components of the nervous
system. The prevalence for developmental delay in children was 50% for progeny of high-mercury
mothers and 17% for progeny of mothers of the control group. These results were statistically
significant. Analysis of the DDST results by sector showed that developmental delays were most
commonly noted in the fine motor and language sectors, but the differences between the experimental
and control groups were not significant. The authors concluded that children born to mothers with mean
hair mercury levels above 6 ppm have twice the risk of delayed development, as tested by the DDST, in
comparison with the control group.
Psychological Tests at Age 6-7 (Kjellstrom et al., 1989)
In 1985 when the children were 6 to7 years of age, a follow-up study was conducted. In this study,
61 of the 74 high-exposure children were compared with three control groups with lower prenatal
mercury exposure. Average maternal hair mercury concentrations in the control groups were 3 to 6 ppm
and 0 to 3 ppm, respectively. The high-exposure group, with maternal hair mercury levels ranging from 6
to 86 ppm, was matched with controls for maternal ethnic group, age, smoking habits, residence, and sex
3-20 Methylmercury Water Quality Criterion 1/3/01
-------
of the child. Each child was tested with a battery of 26 scholastic, psychological, and behavioral tests,
which included Test of Language Development (TOLD), the Wechsler Intelligence Scale for Children
(WISC), and McCarthy Scale of Children's Abilities as described in Table 3-3. Confounding factors such
as language used at home, maternal and paternal occupation, maternal alcohol consumption, and number
of children in the household were controlled using linear multiple regression analysis.
Table 3-3. Developmental domains evaluated and tests applied in studies of New Zealand children with
prenatal exposure to mercury from fish
Developmental Domain
General cognitive
Visual-perceptual
Speech-language
Memory
Motor
Learning-achievement
Personal-social
Age of Child (years)
4
—
Sheridan-Gardiner Letter Matching
test
Miniature Toy Test
DDST
—
DDST
DDST
6
MSCA general
WISC-R Performance IQ, Total IQ
MSCA perceptual
TOLD Spoken Language Quotient
MSCA Verbal
WISC-R Verbal
Peabody Picture Vocabulary Test
(1981)
MSCA memory
MSCA motoric
Clay Diagnostic Survey Concepts,
Letter Test, and Word Test
MSCA quantitative
Burt Word Recognition Test
Key Math Diagnostic Arithmetic Test
Everts Behaviour Rating Scale
Symbols and Abbreviations: — = No test administered; DDST = Denver Developmental Screening Test; MSCA = McCarthy
Scales of Children's Abilities; TOLD = Test of Language Development; WISC-R = Wechsler Intelligence Scale for Children -
Revised.
Source: Kjellstrom et al., 1986; 1989.
An average hair mercury level of 13 to 15 ppm during pregnancy was consistently associated with
decreased test performance. Results of the psychological test variables were influenced by ethnic
background and social class. After controlling for confounding factors and eliminating outliers, the
association between prenatal methylmercury exposure and decreased performance in psychological tests
remained unchanged. The children who had the poorest performance in the WISC IQ test at age 6 also
had a high prevalence of abnormal or questionable DDST scores at age 4, indicating that the effects
evidenced in this follow-up study confirm those found in the preliminary study at age 4. The authors
Methylmercury Water Quality Criterion 1/3/01
3-21
-------
conclude that effects of methylmercury leading to developmental delays may later lead to deficits in
psychological tests.
Benchmark Modeling of the 1985 Data (Crump et al, 1998)
Crump et al. (1998) performed a reanalysis and BMD modeling of the Kjellstrom et al. study
results. Crump et al. used actual hair mercury levels as opposed to an indicator variable for mercury
level in hair; additional confounding factors, such as parent's education and age at which the child was
tested were also controlled for. They also and evaluated all 26 scholastic and psychological tests
(illustrated in Table 3-4) administered to the 237 6 to 7-year old children. No significant associations
between mercury exposure and children's test scores were identified. This finding, however, was highly
influenced by one child whose mother's hair mercury level was 86 ppm, fourfold higher than observed
for any other mother. When this outlier was omitted, scores on six tests were found to be significantly
associated with maternal hair mercury concentrations: Clay reading test-concepts, Clay reading test-letter
test, McCarthy-general cognitive test, McCarthy-perception, TOLD-grammar completion, and TOLD-
grammar understanding. BMDs calculated from five tests (TOLD-spoken language quotient, WISC-
performance IQ, WISC-full scale IQ, McCarthy perceptual, and McCarthy-motoric) ranged from 32 to 73
ppm and BMDL of 17 to 24 ppm, respectively. When the child with the highest maternal hair mercury
was excluded, the BMDs ranged from 13 to 21 ppm with BMDLs spanning 7.4 to 10 ppm (Table 3-4).
Table 3-4. BMD and BMDL values (expressed as maternal hair mercury concentration, ppm) for
neurobehavioral endpoints in New Zealand children evaluated at 6 to 7 years of age
Test
TOLD - spoken language
WISC - performance IQ
WISC-full-scale IQ
McCarthy-perception
McCarthy-motoric
All New Zealand children
BMD"
45
73
51
32
55
BMDLb
20
24
21
17
21
Child with highest maternal
mercury concentration omitted
BMD
15
15
15
13
21
BMDL
9.5
10
10
7.4
9.8
"A background prevalence CP0) of abnormal response of 5% and a benchmark response of 10% were used for these calculations.
h95% lower confidence bound on BMD.
Abbreviations: TOLD = Test of Language Development; WISC = Wechsler Intelligence Scale for Children.
Source: Crump et al. (1998).
3-22
Methylmercury Water Quality Criterion 1/3/01
-------
3.2.1.7 Faroe Islands
A large human prospective longitudinal study was conducted in the Faroe Islands to determine if
increased methylmercury exposure is related to decreased neurobehavioral function. Before the
prospective study, a pilot study was conducted to assess the magnitude of fetal mercury exposure in the
Faroes. At 12 months of age, a follow-up evaluation was conducted and then a prospective study was
initiated with children born at consecutive deliveries within a 22-month period at nearby hospitals.
Demographics
The Faroes is a group of 18 islands located hi the North Atlantic between Scotland and Iceland. The
Faroese population is homogenous with respect to cultural and socioeconomic factors. The culture is
mainly Scandinavian, with a traditional stable family unit that has easy access to good health care,
education, and social systems. Dietary deficiencies are virtually nonexistent, alcohol intake is low, rate
of preterm delivery of low-birth-weight infants is also low, and rate of breastfeeding is high for at least
12 months (Budtz-Jorgensen et al., 2000). Seafood constitutes a major part of the average diet in fishing
communities in the North Atlantic like the Faroe Islands (Grandjean et al., 1995). The major source of
methylmercury exposure is pilot whale, which according to ancient tradition was hunted and distributed
within the community (Grandjean et al., 1997). Other components of the Faroese diet include lamb,
potatoes, dairy products, and foods imported from other countries (Steurwald et al., 2000).
Pilot Study (Grandjean et al., 1992)
A pilot study was conducted by Grandjean et al. (1992) to assess the magnitude of fetal mercury
exposure in the small fishing village of Lorvik, Faroe Islands. Blood samples were collected from a
group of 53 women of fertile age, between 20 and 50, identified through a municipal register. Between
1986 and 1987, 1,023 umbilical cord blood samples were also collected at consecutive deliveries at three
local hospitals. Women had a median blood mercury level of 12.1 \ig/L, with values that ranged from 2.6
to 50.1 \ig/L. The median mercury concentration in cord blood for all 250 samples exceeded 40 u.g/L,
while 20 samples had levels higher than 100 |ig/L. Hair samples had mercury content that exceeded 10
ppm, and five samples exceeded 25 ppm. In 34 hair samples the measured mercury levels exceeded 15
ppm. Mercury concentrations tended to be 20% to 65% higher in cord blood than in the venous blood of
mothers. Highly increased mercury concentrations in maternal hair and umbilical cord blood were
related to maternal consumption of pilot whale.
Methylmercury Water Quality Criterion 1/3/01
3-23
-------
12-Month Evaluation (Grandjean et al, 1995b)
At 12 months of age, 583 children were selected for further evaluation. These children were
followed for 1 year after birth. Three age-appropriate developmental milestones were evaluated: sitting,
creeping, and standing. The age at which the child achieved a developmental milestone was not
associated with indices of prenatal mercury exposure, either from cord blood (average of 174 (xg/L) or
maternal hair (approximately 15% of mothers had concentrations above 50 nmol/g). Infants who reached
the milestone criteria early had significantly higher mercury concentrations in their hair at 12 months
than those who did not. The child's hair mercury concentration was found to be highly correlated to the
period of breastfeeding. Breast milk may transfer contaminants such as methylmercury, but it is also
known to confer certain advantages such as maternal antibodies. The authors concluded that if
methylmercury exposure from human milk had any adverse effect on milestone development in these 12
month-old infants, the effect was compensated for by advantages offered through breastfeeding.
Computer-Assisted Neurobehavioral Tests in 7-Year-Olds (Dahl et al., 1996)
In this study, 917 children were evaluated at 7 years of age. The study focused on computer-
assisted neurobehavioral tests and whether or not they could serve as meaningful parameters of
neurotoxicity; three Neurobehaviroal Evaluation System (NES) tests were administered with slight
modifications. The NES tests were selected to assess motor speed (Finger Tapping [FT]), sustained
attention (Continuous Performance Test [CPT]), and motor coordination (Hand-Eye Coordination [HEC]
Test). The CPT was modified to use animal silhouettes as a stimuli instead of letters to accommodate
those children who had not yet started school and were unfamiliar with the alphabet.
Finger Tapping was relatively easy for most children, but the HEC test was considered too difficult.
Of the 914 children who completed the full HEC, 755 had fewer than 25% nonresponses. Decreased
visual acuity, strabismus, use of eyeglasses, and contrast sensitivity were markedly associated with
decreased performance, especially on the CPT. Boys and older children performed better than girls and
younger children, but this was due to increased familiarity with computers and use of a joystick. The
authors concluded that maternal hair mercury and cord blood mercury were clearly associated with NES
results, especially in the FT and CPT tests.
3-24
Methylmercury Water Quality Criterion 1/3/01
-------
Main Prospective Longitudinal Study of 7-Ifear-Olds (Grandjean et al, 1997)
The cohort consisted of 917 children at 7 years of age who survived from the original cohort
established in the pilot study. Indices of prenatal exposure included cord blood and maternal hair, and
the index for postnatal exposure was children's hair mercury. The geometric mean cord blood mercury
concentration was 22.8 ng/L, and the concentration found in children's hair averaged 11.68 ppm.
Detailed neurobehavioral and physical examinations and neuropsychological and neurophysiological
testings were performed. The neuropsychological tests (Table 3-5) included NES FT Test, NES HEC
Test, Tactual Performance Test, NES CPT, Wechsler Intelligence Scale for Children - Revised (WISC-
R), WISC-R Similarities, WISC-R Block Designs, Bender Gestalt Test, California Verbal Learning Test-
Children [CVLT]), Boston Naming Test (BNT), and Nonverbal Analogue Profile of Mood States. These
tests were chosen for their sensitivities in detecting neuropathological abnormalities. The
neurophysiological tests were chosen to exclude those with electrical stimulation or long measurement
times. These tests include pattern reversal visual-evoked potentials with binocular full-field stimulation,
brain stem auditory-evoked potentials (BAEP), and postural sway.
Fewer than 60% of the children completed three of the most difficult tests. The WISC-R
Similarities Test, NES HEC Test, and Nonverbal Analoguous Profile of Mood States were found to be
too difficult for many of the children to reveal the subtle neurotoxic effects associated with
methylmercury. The geometric mean cord blood mercury concentration for the 85 children who failed or
refused to take the mood test was 29.5 [ig/L, compared with 22.3 ng/L in children who voluntarily
completed it. Reciprocal motor coordination and simultaneous finger movement showed no relation to
mercury exposure. In the finger opposition test, however, 465 children with geometric mean blood
concentrations of 21.8 ug/L mercury performed optimally, whereas those with blood concentrations of
23.9 (J-g/L had questionable or deficient performances.
Mercury-related abnormalities were not identified in either the neurophysiological or clinical
examination. However, in the neuropsychological testing, statistically significant mercury-related
dysfunction was observed. This was most pronounced in the areas of language, attention, and memory,
and to a lesser extent visuospatial and motor functions. After adjustment of covariates and exclusion of
children with maternal hair mercury above 10 ppm, the association remained. This indicates effects of
methylmercury at doses lower than that which result in 10 ppm maternal hair mercury. In the
neurophysiological test, girls showed significantly shorter latencies of evoked potentials than boys in the
electrophysiological tests. For the BAEP latencies, peak I at 40 Hz and 20 Hz was slightly delayed at
Methylmercury Water Quality Criterion 1/3/01
3-25
-------
Table 3-5. Developmental domains evaluated and tests applied in studies of Faroese children at age 7
years
Developmental Domain
Test
Grandjean et al (1997) - Main Prospective Study
General cognitive
Visuospatial
Attention
Speech-language
Memory
Motor
Personal-social
WISC-R Similarities
WISC-R Block Designs
Bender Motor Visual Gestalt Test
NES2 Continuous Performance
WISC-R Digit Spans Forward
Boston Naming Test
California Verbal Learning Test
NES2 Finger Tapping
NES2 Hand-Eye Coordination
NES2 Tactual Performance
Nonverbal Analogue Profile of Mood States
Grandjean etal (1998) • Nested Case Control Study
General cognitive
Visuospatial
Attention
Speech-language
Memory
Motor
Personal-social
WISC-R Similarities
WISC-R Block Designs
Bender Visual Motor Gestalt Test
NES2 Continuous Performance
WISC-R Digit Spans Forward
Boston Naming
California Verbal Learning Test
NES2 Finger Tapping
NES2 Hand-Eye Coordination
—
Symbols and Abbreviations: — = No test administered; NES2 = Neurobehavioral Evaluation System; WISC-R = Wechsler
Intelligence Scale for Children - Revised.
Source: Grandjean et al., 1997,1998.
increased prenatal mercury exposures and the delays for peaks HI and V were statistically significant, but
the interpeak latencies showed no associations with mercury. Body sway showed a slight negative
association with mercury exposure in all four conditions: eyes open, no foam; eyes closed, no foam; eyes
open with foam; and eyes closed with foam.
Four tests were selected for further analysis. Tests were chosen to reflect each of the following
brain functions: motor function (Finger Tapping with preferred hand), attention (CPT reaction time),
Visuospatial performance (error score on the Bender Visual Motor Gestalt Test), language (Boston
Naming Test after cues), and memory (long-delay recall on the California Verbal Learning Test). After
3-26 Methylmercury Water Quality Criterion 1/3/01
-------
adjustment for covariates using the Peters-Belson method, children with scores in the lowest quartile
were identified and distributed into quartile groups of mercury exposure (< 15, 15-30, 30-50, and > 50
lig/L). These results indicate that there is a statistically significant trend for the attention, language, and
memory test with increasing prenatal mercury exposure (Grandjean et al., 1997).
Pilot whale blubber is also consumed by the Faroese population, and this could result in increased
exposure to PCBs, a potential confounding factor. A subset (n = 436) of the cord tissue samples was
evaluated for PCBs; inclusion of PCB exposure as a covariate in the regression analysis affected only the
regression for the BNT. The authors conclude that results of the expanded data analysis do not suggest
that the mercury effect can be explained by concomitant PCB exposure, or that PCB exposure enhances
the mercury-associated effects.
Revaluation of the Evoked Potentials in the Prospective Study (Murata et al, 1999a)
Significant associations with delays in evoked potential latencies and mercury exposure (Murata et
al., 1999a) initiated the reanalyses of the data from the prospective longitudinal study. This analysis is
limited to only children born during the first half of the cohort generation in 1993. Data from the second
year were excluded because of shorter BAEP latencies and delayed latency on the visual-evoked
potentials. Three sets of mercury exposure data were utilized in regression analyses: (1) mercury in cord
blood (geometric mean of 23.0 \ig/L, range of 3.3-351 |J,g/L), (2) mercury in maternal hair at parturition
(geometric mean of 4.49 ppm, range of 0.9-39.1 ppm), and (3) mercury in the child's hair (geometric
mean of 3.42 ppm, range of 0.04-26.4 ppm). The mercury concentration in maternal hair was a
significant predictor for peak El latency and the I-ffl interval, where the child's own hair mercury
concentration at the time of examination was not associated with these response variables. The cord
blood concentration was, however, a significant predictor, supporting the notion that the latency delays
are related to increased prenatal methylmercury exposure.
Nested Case-Control Study (Grandjean et al., 1998)
Following the evaluation of 7-year-olds in the prospective longitudinal study, the data were
evaluated as a nested case-control study. From the original cohort of 1,022 established in the pilot study,
the cases and controls were selected based on maternal hair mercury concentration. The case group of
112 children whose mothers had hair mercury concentrations of 10 to 20 ppm was matched to children
with prenatal exposure below 3 ppm (control). Age, sex, time of examination, and maternal Raven score
Methylmercury Water Quality Criterion 1/3/01
3-27
-------
were matching criteria. The median maternal hair mercury concentrations in the two groups were
l.Sppm for the control group and 12.5 ppm for the cases, a sevenfold difference. The median cord blood
mercury concentrations for the control and cases were 11.9 and 59.0 |ig/L, respectively.
Neuropsychological tests evaluated were these: NES2 FT Test, NES2 HEC Test, NES2 CPT, WISC-R
Similarities, WISC Block Designs, Bender Visual Motor Gestalt Test, CVLT, and BNT. The case group
performed less satisfactorily than those in the control. On 6 of the 18 test outcomes, the inferior scores
achieved by the case group were statistically significant. In particular, the case group showed a deficit on
the Finger Tapping condition and the overall hand-eye coordination. Girls and boys scored differently on
the Bender Gestalt Test, California Verbal Learning Test, all three Finger Tapping conditions, CPT
reaction time, and the average hand-eye coordination score. No differences were reported between girls
in the cases versus controls, but boys in the case group scored poorer in the Finger Tapping reaction time
than the boys in the control group. The deficit in motor coordination, especially in Finger Tapping with
both hands, was highly significant for boys only. The author noted that the findings of this matched case-
control study are in accordance with regression analyses performed on all 900 children at the 7-year
evaluation; methylmercury effects appear in the several domains of the brain, focusing on motor
function, language, and memory.
Benchmark Modeling (Budtz-Jorgesen et al., 2000)
Benchmark modeling of the data from the Faroese children at 7 years of age was reported by Budtz-
Jorgesen et al. (2000). The exposure was modeled both as mercury concentration in cord blood and in
maternal hair. The number of children that completed neuropsychological tests varied between 837 and
901. One neuropsychological test was selected for evaluation of each of the five domains of brain
function:
1. Motor speed (NES FT Test)
2. Attention: NES2 CPT
3. Visuospatial performance: Bender Visual Motor Gestalt Test
4. Language: BNT
5. Short-term memory: CVLT
For tests of motor function, language, and memory, a logarithmic dose-response model tended to show a
better fit than a linear dose model using cord blood mercury concentration as the dose parameter. The
default po is 5%, which equates to the level (x0) of abnormal test performance as defined by a probability
3-28
Methylmercury Water Quality Criterion 1/3/01
-------
of 5% in the unexposed population. The Faroese cohort does not include an unexposed control group;
thus the performance level for an unexposed child is obtained by fitting a dose-response curve to all data
points, followed by extrapolating to zero exposure. Four different dose-response models were employed:
K power, linear, square root, and logarithmic.
The results from this analysis indicate that BMDs and BMDLs vary substantially. Of the four
models, the logarithmic dose-response model provided the best fit for some of the outcome variables that
showed the closest association with the cord blood mercury concentration. The lowest BMDLs averaged
approximately 5 fig/L cord blood, which is equivalent to approximately 1 ppm in maternal hair. Most
BMDLs for hair mercury concentrations were higher. However, the results for a BMR of 5% are the
same order of magnitude as the cord blood results at a BMR of 10%. The authors concluded that the
results of the benchmark calculation are highly dependent on the assumed dose-response model. Results
of this analysis are discussed further in the Risk Assessment chapter (Chapter 4). (For a description of
modeling terms see Section 4.3).
Second Cohort (Steurwald et al, 2000)
During a period from 1994 to 1995, a second cohort of 182 singleton term births was generated •
from consecutive births at the National Hospital in Thorshavn, Faroe Islands (Steurwald et al., 2000).
Maternal hair, serum, breast milk, and umbilical cord blood were analyzed for contaminants, while
selenium, thyroid hormones, and fatty acids were measured in cord blood. In addition to methylmercury,
PCBs were examined as a possible confounder in test outcome. At 2 weeks of age, infants were
administered a neurological examination. Assessment of functional abilities, reflexes and responses, and
stability of behavioral status during examination were completed with a score of optimal, questionable,
or suboptimal performance. The Neurologic Optimality Score (NOS) was the number of items rated as
optimal out of a total of 60. Results from this study indicate that prenatal exposure to methylmercury and
PCBs increased from maternal intake of seafood. After adjustment for confounders, a tenfold increase of
the cord blood mercury concentration was associated with a decreased NOS of 2.0. This effect
corresponds to a decrease in gestational age of about 3 weeks. The authors conclude that prenatal
exposure to methylmercury from contaminated seafood was associated with an increased risk of
neurodevelopmental deficit. No evidence for a protective or beneficial effect with respect to
neurological optimality score (the number of main items rated optimal out of 60) was observed for
essential fatty acids or selenium.
Methylmercury Water Quality Criterion 1/3/01
3-29
-------
3.2.1.8 Germany
Cross-Sectional Study (Altmann et al, 1998)
From a larger comparative environmental screening study, 384 children between the ages of 5 and 8
years were selected to participate hi a smaller field experiment to investigate the effects of low-level lead
and mercury exposure on the functions of the developing visual system. Blood lead levels and urinary
excretion of lead and mercury were used as exposure indices. Neurophysiological and psychophysical
measurements were administered to the children. Visual functions were assessed for neurophysiological
measurements, while psychophysical measurements were assessed by visual-evoked potentials and
contrast sensitivity. Linear regression analyses were used to analyze the possible relationship between
exposure to lead and mercury and outcome variables. Adjustments were made for potential confounding
factors such as parental education, birth weight, length of lactation, and premature birth.
After adjustment for potential confounding factors, contrast sensitivity values were significantly
reduced with increasing urinary mercury levels; four of the ten contrast sensitivity values tested showed a
statistically significant decrease with increasing urinary mercury. Very subtle changes in the visual
system function were noted at very low levels of urinary mercury. However, no significant associations
were found between urinary mercury output and any visually evoked potential outcome variables.
3.2.1.9 Nambija, Ecuador
Cross-Sectional Study on Neurosensory Dysfunction (Counter et al., 2000)
A cross-sectional study was conducted in the remote Andean settlement of Nambija, Ecuador, to
investigate whether blood mercury levels are associated with auditory neurosensory dysfunction.
Participants in this study included 36 children and 39 adults living in Nambija, an area known to have
extensive gold-mining operations where mercury is used in the extraction process. Mercury exposure
was measured in whole blood. The mean blood mercury level was 17.5 ug/L. A group of 34 subjects (15
children and 19 adults) from a non-gold-mining area were selected as the control group. Their mean
blood mercury level was 3.0 ug/L. A neuro-otological examination was administered; a neurological
examination of the cranial nerves was administered using standard procedures and an audiological test
was administered to 21 children and 19 adults.
3-30
Methylmercury Water Quality Criterion 1/3/01
-------
Of those examined, 45% of the group complained of headaches and/or memory loss, three cases
involved severe neurological impairment and four cases involved middle ear pathology. A statistically
significant relationship was identified between blood mercury level and hearing level in children at 3 kHz
in the right ear only. Adults were not affected. BAEP responses showed a significant correlation
between blood mercury and the I-DI interpeak latency on the left side. The authors conclude that the
findings of this study suggest that overall auditory sensory-neural function and neural conduction time at
the brain stem level were generally unaffected by elevated blood mercury levels in either children or
adults.
3.2.1.10 Amazonian Basin
The conditions in the Amazon—extremely high temperatures and humidity with seasonal
fluctuation of water during rainy and dry seasons—are conducive for mercury methylation because of
high quantities of suspended organic matter, high temperature, acidity, and redox potential. These
elements influence the availability of fish as a food resource. In 1996, Lebel and colleagues published
results from a small preliminary study on individuals from the Amazonian basin to determine the
relationship between mercury exposure and neurological outcomes and reported the decrease of visual
and motor functions with increasing hair mercury levels. In 1998, Lebel and colleagues published
another study to determine the neurofunctional and clinical manifestations of nervous system dysfunction
in relation to hair mercury levels below 50 ppm. In 1999, Grandjean et al. published results from a study
of populations living in four comparable Amazonian riverine communities located upstream of gold-
mining fields, while in 2000, Dolbec et al. published results from a cross-sectional study in a village on
the Tapajos River.
Lebel et al. (1996)
Lebel et al. (1996) published a study of 29 adult residents living in two villages located on the
Tapajos River, a tributary of the Amazon, located approximately 200 kilometers from several gold-
mining sites. Total hair mercury concentration ranged from 5.6 to 38.4 ppm; methylmercury constituted
between 72.2% and 93.3% of the total mercury measured in hair samples. A quantitative behavioral
neurophysiological battery was modified for administration to persons with minimal formal education
living in an area without electricity. Women exhibited a decrease in manual dexterity, as measured in the
Santa Ana Test (Helsinki version) that was correlated with increased mercury concentration in hair. For
both men and women, there was a statistically significant decrease in color discrimination capacity with
Methylmercury Water Quality Criterion 1/3/01
3-31
-------
increasing hair mercury concentrations. Near visual contrast sensitivity profiles and peripheral visual
field profiles were both reduced in the individuals with the highest hair mercury concentrations. The
authors note that constriction of the visual field has been observed in other instances of mercury
intoxication and that changes in contrast sensitivity have been noted in nonhuman primates exposed to
methylmercury (Rice and Gilbert 1982,1990).
Lebeletal. (1998)
A later study was conducted in a Tapajos River village that depends on fish as its main source of
protein. A total of 91 adults (45 men and 46 women between the ages of 15 and 81) of the 98 voluntary
participants were examined. Four measures of hair mercury concentrations were used: (1) mean total
hair mercury, (2) total hair mercury, (3) total hair mercury in the highest value obtained out of all
centimeters analyzed, and (4) total hair mercury in the first centimeter and methylmercury in the first
centimeter. Several tests were administered to score for neuropsychological dysfunction. Motor strength
was determined with a dynameter for grip test; manual dexterity was measured with the Santa Ana Test
(Helsinki version); and visual functions, color vision, and contrast sensitivity were assessed with a
battery of sensitive neurofunctional tests. Results were analyzed by multiple regression.
There was no difference between genders for all tests except the grip strength test. Women also
exhibited decreased grip strength with increasing peak mercury levels. Intermediate and higher
frequencies of near visual contrast sensitivity and manual dexterity (measured with the Santa Ana Test)
varied with the level of mercury in hair. Gender-nonspecific muscular fatigue was also noted with
increasing mercury levels. The authors suggest that there appears to be a dose-effect relationship for
certain motor and visual functions. Manual dexterity, alternating hand coordination, and muscular
fatigue were associated with hair mercury levels, while near visual contrast sensitivity and restricted
visual fields were dose-dependently altered.
Cross-Sectional Study (Grandjean et al, 1999)
A cross-sectional study was conducted in four comparable Amazonian riverine communities
located upstream toward gold-mining fields. Fish is consumed as a large part of the population's staple
diet. Of the 420 eligible children between the ages of 7 and 12, 351 were examined for neurobehavioral
dysfunction. Mercury exposure was measured through children's hair mercury levels because only 37%
of the participants had maternal hair mercury samples. Children's hair mercury concentrations had an
3-32 Methylmercury Water Quality Criterion 1/3/01
-------
overall geometric mean of 11.0 ppm and a median of 12.8 ppm, while mothers had geometric mean hair
mercury levels of 11.6 ppm and a median value of 14.0 ppm. Maternal hair mercury concentrations were
highly correlated with those of their children. Several neuropsychological tests of motor function,
attention, and visuospatial capability were administered. These included Finger Tapping, Santa Ana
form board, WISC-HI Digit Spans Test, and two subtests of the Stanford-Binet Intelligence Scale (the
copying test and memory condition). The relation between mercury exposure and neurobehavioral
function was analyzed by multiple regression analyses with adjustment for covariates including, age, sex,
health status, maternal education, and maternal marital status.
The Santa Ana form board and Stanford-Binet copying test showed the clearest associations with
the hair mercury concentration. The authors note that the effect of mercury was significantly greater in
younger children only for the nonpreferred hand condition of the Santa Ana Test. In interpreting these
results, the authors caution that there were no data for the level of prenatal exposure experienced in the
test children because of the lack of maternal hair samples. Additional sources of uncertainty in this study
include nutritional deficiencies that occurred in the past and possible infection of tropical diseases that
may have influenced the capabilities of these children at the time of neurological evaluation.
Cross-Sectional Study (Dolbec et al, 2000)
A cross-sectional study was conducted in May of 1996 in a village on the banks of the Tapajos river
in the Amazonian Basin, Brazil (Dolbec et al., 2000). This study was conducted on 84 fish-eating adults
between the ages of 15 and 79, to evaluate the effect of mercury exposure on motor performance. The
mean hair total mercury level was 9 ppm. Pychomotor performance was evaluated using the Santa Ana
Test for manual dexterity, the Grooved Pegboard Fine to test fine motor skills and NES Finger Tapping
Test for motor speed. Motor strength was measured by dynamometry for grip and pinch strength.
Multivariate analysis of the variance indicated that the hair mercury levels were inversely
associated with overall performance on the psychomotor tests, whereas an association was reported with
blood mercury. Semipartial regression analyses reported that hair total mercury accounted for 8%-16%
of the variance of psychomotor performance. The authors conclude that the findings of this study
demonstrated neurobehavioral manifestations of subtle neurotoxic effects on motor functions associated
with low-level methylmercury exposure.
Methylmercury Water Quality Criterion 1/3/01
3-33
-------
3.2.1.11 Madeira
Cross-Sectional Study (Murata et al, 1999b)
A cross-sectional study (Murata et al., 1999) was conducted in the Madeiran community to
determine possible—mercury exposure-related effects on evoked potentials in 149 children between the
ages of 6.4 and 7.4 years. Children's hair mercury concentrations were used to reflect current exposure
levels, while maternal hair levels from mothers who had followed consistent diets since pregnancy
represented prenatal mercury exposures 7 years ago. The use of maternal hair concentration as a
substitute for exposure during pregnancy is based on the assumption that mercury exposure has changed
very little over time. The authors acknowledge, however, that current maternal hair mercury levels
provide an imprecise indication of exposure during pregnancy and any recent dietary change would tend
to weaken the association with the outcome variables. The 149 children were administered physical and
functional neurological examinations, with an emphasis on motor coordination and perceptual motor
performance. Tests included these:
NES2FT
NES2HEC
NES2 CPT
• WISC-R subtests: Digit Spans forward condition and Block Designs
• Stanford-Binet Bead Memory Test
Evoked potentials were determined with a four-channel electromyograph, while pattern reversal visual-
evoked potentials with binocular full-field stimulation were conducted in a darkened room. Associations
between these outcomes and exposure to methylmercury were assessed by multiple regression analysis
and were adjusted for possible confounding variables: age, sex, maternal and paternal education and
employment, maternal alcohol use and smoking during pregnancy, numbers of older and younger
siblings, school, and the level of the child's computer acquaintance.
Increased exposure to methylmercury was associated with delays in evoked potential latencies;
peak III on the BAEP at 40 Hz, and N145 on the pattern reversal visual-evoked potentials at the 15-
minute condition. When the maternal hair mercury concentration exceeded 10 ppm, the increase of the
N145 visual-evoked potential latency at 15 minutes was 3.16 milliseconds (ms). The N75-N145 and
P100 and N145 interval latencies showed similar regression coefficients for mercury, although
3-34
Methylmercury Water Quality Criterion 1/3/01
-------
significance was evident only for the 15-minute condition. The authors suggest that this may indicate
that there is a mercury-associated delay occurring between P100 and N145. Weak associations were also
evidenced between maternal hair mercury levels and deficits on Digit Spans and Bead Memory tests.
3.2.1.12 French Guiana
Case-Control Study (Cordier and Garel, 1999)
High-exposure areas were selected in the Amerind villages in the Upper Maroni, with two other
Amerind villages with less mercury contamination to serve as reference groups (Cordier and Garel, 1999.
261 children participated in the study, 69 from the village of Camopi (control), 82 from Awala (control) a
total of and 110 in the Upper Maroni (cases). Hair samples were collected from both children and
mothers to represent exposure indices. Maternal hair mercury levels ranged from 2.5 to 6.7 ppm. This
was used as a surrogate for prenatal exposure. Children had slightly lower hair mercury levels than
adults, but this did not vary with age. Neurological examinations were administered to children from 9
months to 6 years of age with special emphasis on neuromotor examination of the upper and lower limbs,
axis of the body, deep reflexes, postural reactions, examination of the effects on neuromotor functions,
neurosensory examination, and cranial growth. The battery of tests was selected to measure the child's
abilities outside of educational or cultural influences; these include the NES FT Test to measure fine
motor function, coordination, and speed of execution; and the Stanford-Binet Intelligence Scale, with
subtests of immediate memory (bead memory) and ability to assess visuospatial and visuoconstructional
function (block-copying). In addition, the McCarthy memory test for digits (backward and forward) and
the McCarthy leg coordination test were utilized. Associations were analyzed by linear regression,
adjusting for potential confounding factors (alcohol consumption during pregnancy, parity, place of birth
of the child, and illnesses during childhood).
Within the case group, there is a significant decrease in the scores with exposure category for the
Leg Coordination test and close to significance for the Copying test. When boys and girls were
examined separately for the FT test, boys had higher scores than the girls, while a significant decrease is
observed in the score on the Block Design test correlated with exposure in girls. Boys also exhibited
greater incidence of increased reflexes correlated with maternal hair mercury concentrations. The
authors conclude that results of this study suggest a link between exposure to mercury and perturbations
of the child's neurological and intellectual development.
Methylmercury Water Quality Criterion 1/3/01
3-35
-------
3.2.2 Animal Studies
Substantial information on the neurotoxicity of methylmercury has been generated from animal
studies that support neurological effects reported in humans. Relatively brief, high-level exposures in
rats have been shown to cause characteristic signs of neurotoxicity (flailing and hindlimb crossing when
the animal is lifted by the tail), as well as neuronal degeneration in the cerebellum, cerebral cortex, and
dorsal root ganglia (Ihouye and Murakami, 1975; Leyshon and Morgan, 1991; Magos et al., 1985; Yip
and Chang, 1981). As observed in humans, there is a latency period before onset of neurological
symptoms. Toxic effects may not be observed or may not show maximal severity until several days after
the initiation of dosing. In short-term studies, toxicity may not become evident until after the cessation
of dosing. This section summarizes a few selected animal studies on neurotoxicity. For additional detail,
please refer to Volume V of the MSRC (U.S. EPA, 1997e) and the Toxicological Effects of
Methylmercury (NRC, 2000).
3.2.2.1 Acute Toxicity
In an acute study, exposure of rats to a single gavage dose of 19.9 mg mercury/kg as
methylmercuric chloride resulted in impaired open-field tests such as decreases in standing upright, area
traversed, and activity compared with the control group (Post et al., 1973). Animals were lethargic and
ataxic initially, but symptoms disappeared within 3 hours.
3.2.2.2 Chronic Toxicity
Longer term, low-level exposures revealed that evidence of neuronal degeneration may be observed
before the onset of overt signs of toxicity. Degeneration in the cerebellum was found in rats given 10 mg
mercury/kg as methylmercuric chloride once every 3 days for 15 days (Leyshon and Morgan, 1991).
Severe degenerative changes in the dorsal root fibers were observed in rats given 1.6 mg mercury/kg-day
as methylmercuric chloride for 8 weeks (Yip and Chang, 1981). Munro et al. (1980) observed
demyelination of dorsal nerve roots and damage in sciatic nerves with oral exposure to 0.25 mg
mercury/kg-day as methylmercuric chloride for up to 26 months. In mice given 1.9 mg mercury/kg-day
as methylmercury, cerebellar lesions were observed as early as 8 days after the start of dosing, but
changes in motor activity did not develop until after 24 weeks of exposure (MacDonald and Harbison,
1977). Similarly, cats receiving methylmercury in the diet for 11 months displayed degenerative changes
3-36
Methylmercury Water Quality Criterion 1/3/01
-------
in the cerebellum and cerebral cortex, but uncoordinated movements or weakness were observed only in
a small number of animals with histopathological changes (Chang et al., 1974).
A 2-year feeding study of methylmercuric chloride was conducted in B6C3F1 mice (60
mice/sex/group) at doses of 0, 0.4,2, and 10 ppm (0,0.03, 0.15, and 0.73 mg mercury/kg-day in males; 0,
0.02, 0.11, and 0.6 mg mercury/kg-day in females) to evaluate chronic toxicity and carcinogenic effects
(Mitsumori et al., 1990). Mice were examined clinically during the study, and neurotoxic signs
characterized by posterior paralysis were observed in 33 males after 59 weeks and in 3 females after 80
weeks in the 0.6 mg mercury/kg-day group. A marked increase in mortality and a significant decrease in
body weight gain were also observed in the high-dose males, beginning at 60 weeks. Postmortem
examination revealed toxic encephalopathy consisting of neuronal necrosis of the brain and toxic
peripheral sensory neuropathy in both sexes of the high-dose group. An increased incidence of chronic
nephropathy was observed in the 0.11- and 0.6-mg mercury/kg-day males.
Groups of Wistar rats (50/sex/group) were administered daily doses of 0.002, 0.02,0.05, and 0.25
mg mercury/kg-day as methylmercuric chloride for 26 months (Munro et al., 1980). Female rats that
received 0.25 mg/kg-day had reduced body weight gains and showed only minimal clinical signs of
neurotoxicity. Male rats that received this dose did show overt clinical signs of neurotoxicity, had
decreased hemoglobin and hematocrit values and reduced weight gains, and showed increased mortality.
Histopathologic examination of rats of both sexes receiving 0.25 mg/kg-day revealed demyelination of
dorsal nerve roots and peripheral nerves. Males showed severe kidney damage and females had minimal
renal damage. This study identified a NOAEL of 0.05 mg/kg-day and a LOAEL of 0.25 mg/kg-day,
based on the observed demyelination effect.
Bornhausen et al. (1980) reported a decrease in operant behavior performance in 4-month-old rats
whose dams had received methylmercuric chloride on gestation days 6 to 9. A statistically significant
effect was seen in offspring whose dams had received 0.01 and 0.05 mg/kg five times during gestation..
The authors postulated that more severe effects of in utero exposure would be seen in humans because
the biological half-life of mercury in the brain of humans is five times longer than in the rat. In addition,
much longer in utero exposure to mercury would occur in humans because gestation is much longer.
In a study of prenatal coexposure to metallic mercury vapor and methylmercury and their effects on
the developing central nervous system, Fredriksson et al. (1996) reported interactive behavioral effects
following exposure of pregnant female Sprague-Dawley rats to methylmercury and metallic mercury
Methylmercury Water Quality Criterion 1/3/01 3-37
-------
vapor. Between 4 and 5 months, testing of behavioral function, spontaneous motor activity, spatial
learning in a circular bath, and instrumental maze learning for food were performed. Exposure to
mercury vapor at 1.8 nig/m3 for 1.5 hours per day on gestation days 14 to 19 was related to hyperactivity
and decreased spatial learning. Although exposure to methylmercury at 2 mg/kg per day on gestation
days 6 to 9 was not related to adverse behavioral effects, coexposure to methylmercury and mercury
vapor potentiated the activity and spatial learning effects observed with mercury vapor alone. The results
of this study indicate that mercury vapor causes central nervous system functional disturbances in
offspring after both prenatal and postnatal exposure. The authors also suggest that coexposure to
methylmercury served to significantly aggravate the changes, whereas methylmercury alone did not cause
any significant functional alterations in this study.
Ramussen and Newland (1999) studied the acquisition of Multiple Differential Reinforcement of
High-Rate Extinction (MULT DRH-N:T EXT) schedules of reinforcement in female rats exposed to
methylmercury during development. Female rats were administered methylmercury (0, 0.5, or 6.4 ppm)
in drinking water from 4 weeks premating to postnatal day 16. Postnatal methylmercury concentrations
in the brain at birth were 0.49 and 9.8 ppm for two exposure groups. Li the MULT DRH-N:T EXT,
female offspring were trained to press levers under schedules of reinforcement. Whenever a response
occurred within a specific time measured in seconds, a food pellet was given. Two acquisition protocols
were examined; one imposed three successive sessions in a 3:1, 5:2, and 9:4 ratio. Values were chosen
so that the same rate of response was required by the schedules. The second acquisition protocol
required lever repressing as reestablished and the three schedules were continued until the behavior
became stable, which required more than 10 sessions. This study was not able to replicate the finding of
abnormal response patterns using the DRL paradigm used by Bornhausen (1980).
Cholinergic systems also play an important role in learning and memory. Coccini et al. (2000)
investigated the effect of low-level methylmercury exposure on muscarinic cholinergic receptor
(mAChR) binding characteristics in adult female Sprague-Dawley rats. The rats (4/dose) were
administered methylmercury in the drinking water at nominal concentrations of 0, 2.5, and 10 u-g/L for 16
days. Mean daily intake in the methylmercury-exposed groups was 0.45 and 1.8 mg/kg-day, respectively.
mAChR binding was assessed using the muscarinic antagonist [3H]quinuclidinyl benzilate (QNB) to label
receptors in excised brain tissues (cerebral cortex, hippocampus, and cerebellum). Exposure to
methylmercury selectively increased mAChR density in the hippocampus and cerebellum by 20% to
44%. This response was characterized by a 2-week latency period before onset. Receptor affinity was
3-38
Methylmercury Water Quality Criterion 1/3/01
-------
unaffected, as indicated by values for the dissociation constant. No significant effect on mAChR in
cerebral cortex was observed.
Nonhuman Primates—Macaco. Fascicularis Monkeys
Monkeys appear to be more sensitive to the neurotoxic effects of methylmercury than are rodents.
The primate model is particularly useful for studies of developmental exposures because monkeys, like
humans, have relatively prolonged periods of gestation, infancy, and adolescence (Burbacher and Grant,
2000). Long-term studies in primates have shown neurological impairment at doses as low as 0.05 mg
mercury/kg-day. Exposure of monkeys to 0.03 mg mercury/kg-day as methylmercury for approximately
4 months caused no detectable changes in motor activity or effects on vision or hearing, but degenerative
changes were observed in neurons of the calcarihe cortex and sural nerve when these were examined by
electron microscopy (Sato and Hcuta, 1975). At higher doses (0.08 mg mercury/kg-day), slight tremor,
lack of motor coordination, and blindness were observed in monkeys after 4 months of exposure
(Burbacher et al., 1988).
Gunderson et al. (1986) administered daily doses of 0.04-0.06 mg mercury/kg as methylmercuric
hydroxide to 11 crab-eating macaques (Macacafascicularis) throughout pregnancy. This dosing
protocol resulted in maternal blood levels of 1,080-1,330 ng/L in mothers and 1,410-1,840 (ig/L in the
offspring. Infants of treated mothers exhibited visual recognition deficits when tested 35 days after birth.
Rice (1989b) dosed five cynomolgus monkeys (Macacafascicularis) with 0.05 mg mercury/kg-day
as methylmercuric chloride from birth to 7 years of age. Clinical and neurological examinations were
performed during the dosing period and for an additional 6 years. Impairment of spatial visual function
was observed after 3 years. In the later stages of the observation period, monkeys dosed with
methylmercury were clumsier and slower to react when placed in the exercise cage than were unexposed
monkeys. Decreased fine motor performance, touch, and pinprick sensitivity, and impaired high-
frequency hearing were observed 6-7 years after cessation of dosing (Rice 1989a; Rice and Gilbert, 1982,
1990).
Rice (1998) did auditory testing of Macacafascicularis monkeys exposed to methylmercury
chloride at 10, 25, or 50 ug/kg per day in utero, throughout gestation, plus 4 years postnatally at 11 and
19 years of age. Results from this study indicated that at 19 months of age, all five Macacafascicularis
monkeys experienced deterioration in auditory function and elevated pure-tone thresholds throughout the
Methylmercury Water Quality Criterion 1/3/01 3-39
-------
full range of frequencies tested (0.125 to 31.5 kHz) when compared with age-matched controls. The
elevation of thresholds was in some cases 50 dB or higher. Because the auditory deficits are experienced
approximately 7 to 15 years after cessation of methylmercury exposure, they are considered irreversible
and permanent. The author concluded from this study that the high-dose monkeys experience an earlier
onset of effect on the auditory function than do low-dose monkeys. The group of monkeys that showed
delayed neurotoxicity at 15 years also had visual deficits identified at 3 years, as well as auditory and
somatosensory impairment. The high-dose monkeys were also impaired at 11 years, and relatively more
impaired than controls at 19 years, thus providing evidence for accelerated aging. These results provide
evidence for the accelerated impairment of auditory function during aging as a consequence of
developmental methylmercury exposure.
Li another study by Rice (1998), monkeys with robust methylmercury-induced deficits in visual,
auditory, and somatosensory function were tested on a series of tasks assessing central processing speed.
This task is thought to be similar to tests measuring human intelligence. Five Macaca fascicularis
monkeys were dosed with 50 u.g/kg per day methylmercuric chloride from birth until 7 years of age.
Blood mercury levels ranged from 0.8 to 1.1 yg/g until cessation of dosing. At 20 years of age, the
monkeys and four age-matched and rearing-matched controls were tested on a series of simple and
complex reaction-time tasks. In the simple reaction-time test, the monkeys were required to press a
button when it changed from off to on (bright red light). The monkeys then performed a sequence of
complex reaction-time tasks: two-button pressing, four-button pressing, and several tasks of increasing
complexity using four buttons and multiple colors. The results indicated no differences between groups
on any aspect of the experiment. The author concluded that the data provide further evidence for the
absence of cognitive impairment in monkeys exposed developmentally to methylmercury.
In 1999, Burbacher et al. published a study that assessed visual and auditory functions in adult
Macaca fascicularis monkeys exposed to methylmercury in utero. Maternal doses were 0, 50, 70, or 90
Hg/kg per day; this resulted in infant blood mercury levels that ranged from 1.04 to 2.45 ppm. When the
monkeys reached 15 years of age, they were tested on spatial visual contrast sensitivity tasks at spatial
frequencies of 1,4, 10, and 20 cycles per degree of visual angle and auditory pure tone detection tasks at
frequencies of 125,500, 1,000,4,000,10,000, 25,000, and 31,500 Hz. The results of these tests indicated
that in utero exposure to methylmercury has long-term effects on visual contrast sensitivity thresholds.
Preliminary results from the auditory task suggest that auditory thresholds are not affected by
methylmercury exposure. The authors suggest that results from this study point to the postnatal period as
a possible critical window for methylmercury induced auditory neurotoxicity.
3-40
Methylmercury Water Quality Criterion 1/3/01
-------
3.3 CARDIOVASCULAR TOXICITY
3.3.1 Human Studies
3.3.1.1 Cardiovascular Effects From the Faroe Islands (Sorensen et al, 1999)
S0rensen et al. (1999) evaluated the relationship between prenatal exposure to methylmercury and
occurrence of cardiovascular effects at 7 years of age in a birth cohort (n = 1,000) of children from the
Faroe Islands. Prenatal exposure was assessed by analysis of cord blood and maternal hah- collected at
parturition. More than 80% of the hair samples exceeded a methylmercury concentration of 2 ppm, which
corresponded to a cord blood concentration of approximately 10 p.g/L. The cardiovascular endpoints
evaluated at 7 years included systolic and diastolic blood pressure, heart rate, and heart rate variability.
Weight, height, body mass index, sex, and maternal hypertension were examined as predictors of blood
pressure and heart rate in approximately 900 children. Birth weight and placental weight were also
examined as predictors of blood pressure. Following adjustment for body weight, diastolic and systolic
blood pressure increased by 13.9 mm mercury (95% confidence limits [CL] = 7.4, 20.4) and 14.6 mm
mercury (95% CL = 8.3, 20.8), respectively, as cord blood mercury concentrations increased from 1 to 10
\ig/L. No further increase was noted at higher concentrations of mercury. Low-birth-weight children
were more likely to experience methylmercury-related increase in blood pressure. A gender-specific
decrease in heart rate variability was also noted with increasing mercury exposure. This effect was most
pronounced in boys, where a 47% reduction in heart rate variability was observed when cord blood
mercury concentrations increased from 1 to 10 |o,g. The authors concluded that the findings suggest that
prenatal exposure to methylmercury may influence the development of cardiovascular regulatory
mechanisms.
3.3.1.2 Cross-Sectional Study (Salonen etal, 1995)
Salonen et al. (1995) examined the relationship between dietary intake of fish and mercury and risk
of acute myocardial infarction (AMI), death from coronary heart disease (CHD), and other
cardiovascular diseases (CVD). Participants of this study included 1,833 men in eastern Finland between
the ages of 42 and 60 with no clinically diagnosed CHD, claudication, stroke, or cancer. Baseline
examinations were administered between March 1984 and December 1989. Fish consumption was
assessed at time of blood sampling with an interview-verified 4-day food record. The food recording was
repeated approximately 12 months after the baseline examination in a random sample of 50 men in the
Methylmercury Water Quality Criterion 1/3/01
3-41
-------
cohort. Daily fish intake ranged from 0 to 619.2 g (mean of 46.5 g/day). Mercury in hair and urine was
determined by flow injection analysis-cold vapor atomic absorption spectrometry and amalgamation.
Hair mercury concentrations ranged from 0 to 15.67 ppm (mean of 1.92 ppm) while dietary mercury
intake ranged from 1.1 to 95.3 ug /day (mean of 7.6 ug per day). In 2 to 7 years, 73 of the 1,833 men
experienced an AMI; 18 of the 73 patients with AMI died of CHD, while 24 of the 73 died of CVD.
Covariates included these: age; examination year; family history of CHD; place of residence (rural vs.
urban); diabetes; socioeconomic status; iron intake; number of cigarettes, cigars, and pipefuls of tobacco
currently smoked daily; duration of regular smoking in years; alcohol consumption; history of myocardial
infarction; angina pectoris and other ischemic heart disease; presence of hypertension; and current
antihypertensive medication. The Cox models reported dietary intakes of fish and mercury associated
with increased risk of AMI and death from CHD, CVD, and any death. Results from this study indicated
that eastern Finnish men with hair mercury levels exceeding 2 ppm had a twofold age- and CHD-adjusted
risk of AMI and a 2.9-fold adjusted risk of cardiovascular death compared with those having lower hair
mercury content.
3.3.1.3 Nested Case-Control Study (Salonen etal, 1995)
A nested case-control study was also conducted using a subsample of the original study
participants. Serum immune complexes containing oxidized LDL were measured in a subsample of 187
control subjects using an ELISA assay with copper-oxidized LDL as the antigen. Pearson correlation
coefficients adjusted for age and year of baseline examination were used to determine the association
between hair mercury content and dietary intakes offish and mercury. Partial associations of hair and
urinary mercury with liters of immune complexes against oxidized LDL were estimated by SPSS step-up
least-squares regression analysis. A multivariate logistic model included the following covariates:
cigarette-years, serum ferritin concentration, ischemic exercise ECG, serum apolipoprotein, family
history of CHD, maximal oxygen uptake, and serum HDL2 cholesterol. There was a statistically
significant association between urinary mercury excretion and the risk of AMI was reported. For each
microgram of mercury excreted daily, the risk of AMI increased by 36%. From the immunotoxicity test,
both the hair and urinary excretion mercury levels were associated with immune complex titers measured
with a rabbit antiserum against oxidized LDL and the y-globulin fraction of a rabbit antiserum against
oxidized LDL. Overall, hair mercury was the strongest predictor of both immune complex titers.
3-42
Methylmercury Water Quality Criterion 1/3/01
-------
On the basis of these data, the authors concluded that a high intake of mercury from nonfatty
freshwater fish, and the consequent excess risk of AMI as well as death from CHD and CVD in eastern
Finnish men, may be due to the promotion of lipid peroxidation by mercury.
3.3.2 Animal Studies
Data on cardiovascular effects following oral methylmercury exposure were obtained from two
studies in rats. Rats given two daily doses of methylmercuric chloride exhibited decreases in heart rates
following two daily doses of methylmercury at 12 mg/kg per day (Arito and Takahashi, 1991). Wistar
rats (n = 80) treated by subcutaneous injection with 0.5 mg/kg-day methylmercuric chloride for 1 month
had increased systolic blood pressures beginning 42 days after cessation of dosing (WaMta, 1987). This
effect persisted for more than a year.
Mitsumori et al. (1983, 1984) fed Sprague-Dawley rats diets containing methylmercuric chloride
(males 0, 0.011, 0.05, or 0.28 mg/kg/day; females 0.014, 0.064, or 0.34 mg/kg/day) for up to 130 weeks.
Polyarteritis nodosa and calcification of the arterial wall were seen at the highest dose. Histological
examination revealed evidence of hemosiderosis and extramedullary hemotopoiesis of the spleen.
In a study on 7-week-old, hypertensive SHR/NCrj rats, Tamashiro et al. (1986) reported an increase
in blood pressure resulting from exposure to methylmercury chloride once a day at 2 mg/kg/day for 26
consecutive days. Body weight loss, an early sign of methylmercury intoxication, was more marked in
males than females. All male rats died by the 29th day posttreatment. Neurological signs, hindleg
crossing, disturbed righting movement and abnormal gait always preceded death. No mortality was
reported for the female rats. However, increase in blood pressure was sex-specific, being observed only
in females. The authors noted that considerable variation was observed in blood pressure for both the
methylmercury-exposed and the control rats; and that these findings suggest strain differences in male-
female toxicity of methylmercury chloride.
3.4 IMMUNOTOXICITY
3.4.1 Human Studies
At this time, there are no studies published on the effect of methylmercury on the human immune
system. In occupational exposure studies, elemental mercury has been found to affect particular immune
Methylmercury Water Quality Criterion 1/3/01
3-43
-------
parameters. A study by Queiroz and Dantas (1997) evaluated B-lymphocyte, T-helper, T-suppressor, and
T-cell proliferative response to phytohemagglutinin in 33 male workers in a Brazilian mercury
production facility. These workers had a mean age of 29 and a mean mercury exposure period of 19
months. All of the workers had urinary mercury concentrations below 50 ug/g of creatinine. Analysis of
the T-cell populations found a reverse CD4+ to CD8+ ratio that was characterized by a reduction in the
number of CD4 lymphocytes. B-lymphocytes were also significantly reduced. Analysis of serum
antibody levels found increased immunoglobulin E levels but did not detect anti-DNA or anti-nucleolar
antibodies. No changes were observed in the proliferative response to phytohemagglutinin of
lymphocytes from exposed individuals. The authors reported a negative correlation between the length
of exposure to mercury and IgE levels, and no correlations between lymphocyte changes and urinary
mercury concentrations, time of exposure, or the age of the workers. (Queriroz and Dantas, 1997)
Another occupational exposure study by Moszczynski et al. (1995) examined the lymphocyte
subpopulation of T-cells, T-helper cells, T-suppressor cells, and natural killer cells in the peripheral
blood of 81 men exposed to metallic mercury vapors and 36 unexposed men. The average workplace
exposure to mercury in air was 0.0028 mg/m3. Urinary mercury concentrations ranged from 0 to 240
jig/L and concentrations in the blood varied from 0 to 30 ug/L. Stimulation of the T-lymphocytes
manifested by an increased number of T-cells, T-helper cells, and T-suppressor cells was observed.
3.4.2 Animal Studies
Data on the potential immunotoxic effects of methylmercury are available from several animal
studies. Suppression of humoral and cellular immune responses has been observed in animals after oral
exposure to methylmercury or methylmercuric chloride. Decreases in the production of antibody-
producing cells and/or decreased antibody liter following inoculation with immune-stimulating agents
(such as sheep red blood cells) have been observed in mice and rabbits (Blakley et al., 1980; Koller et al.,
1977; Ohi et al., 1976). Decreases in natural killer T-cell activity and reduced thymus weight have been
observed in female mice after 14 weeks of exposure to methylmercury (flback, 1991). Bernaudin et al.
(1981) observed IgG deposits along the glomerular capillary wall of Brown Norway rats treated with
methylmercury for 2 months and noted that these deposits were suggestive of autoimmune disease. The
following sections include summaries of selected studies.
Wild et al. (1997) evaluated immune function in the offspring of Sprague-Dawley rats exposed to
methylmercuric chloride (5 or 500 (J-g/L) or methylmercury sulfide (5 ng/L) via drinking water. There
3-44
Methylmercury Water Quality Criterion 1/3/01
-------
were three exposed groups and one control group. The control group was fed plain tap water. Rats of
both sexes were treated for 8 weeks prior to mating and treatment of female rats continued throughout
pregnancy and nursing. The total duration of indirect exposure of the offspring to methylmercury was 42
days. Immunological function was assessed in six offspring per treatment group at 6 and 12 weeks of age
(3 and 9 weeks after termination of methylmercury exposure at weaning, respectively). At 6 weeks, total
body weights, splenic weights, and thymic weights were increased in the methylmercury chloride-
exposed rats, whereas the rats exposed to methylmercury sulfide experienced only an increase in thymic
weight at 6 weeks. At 12 weeks, natural killer cell activity was markedly depressed (56%) for rats
exposed to methylmercury chloride in comparison with controls. Methylmercury sulfide appeared to
have different effects on the immune system than did methylmercury chloride. For example, the sulfide
form affected only thymic weight and had no significant effect on NK or splenocyte cell activity or
splenocyte LPR. Whether this result reflects differential distribution of the sulfide form or affinity for
different targets in the immune system is unknown. The authors concluded that methylmercury chloride
seems to have an effect on splenocytes and natural killer cell activity.
Inorganic mercury has been observed to induce a variety of immune effects in mice. However,
until recently there has been limited investigation of the ability of methylmercury to induce similar
immune responses. Hultman and Hansson-Georgiadis (1999) investigated the ability of subcutaneously
injected methylmercury to induce systemic autoimmunity in five genetically susceptible and resistant
strains of mice. Female SJN/L, A.SW, B10.S (H-25), BALB/C, DBA/2 (H-2*), A.TL, and B10.TL (H-2")
mice were administered subcutaneous injections of 1 mg/kg methylmercury every third day for 4 weeks.
This treatment protocol resulted in an average daily dose of approximately 350 p.g mercury/kg-day. The
immune response to methylmercury differed qualitatively and quantitatively from the response to
inorganic mercury. Treatment with methylmercury induced at most a small increase in serum Ig
concentrations after 4 weeks of treatment. The observed increases during the treatment period were
generally marginal when compared with increases induced by mercuric chloride. Treatment with
methylmercury induced development of antinucleolar antibodies (ANoA) targeting the nucleolar protein
fibrillarin in the susceptible SJL, A.SW, and B10.S strains. Susceptibility to development of ANoA was
linked to the mouse major histocompatibility complex H-2. However, background genes determined the
strength of the response in susceptible strains. Serum IgE concentration and ANoA liter increased 2 to 3
weeks after cessation of treatment with methylmercury. In H-2* mice, methylmercury induced a weaker
general (polyclonal) and specific (ANoA) response when compared to mercuric chloride. Unlike
mercuric chloride-treated mice, animals administered methylmercury did not develop systemic or renal
immune system deposits.
Methylmercury Water Quality Criterion 1/3/01
3-45
-------
3.5 REPRODUCTIVE TOXICITY
3.5.1 Human Studies
There are no studies of reproductive deficits in humans exposed to low-dose methylmercury.
3.5.2 Animal Studies
There are no two-generation reproductive assays for methylmercury.
3.6 GENOTOXICITY
3.6.1 Human Studies
Data from several studies in humans suggest that ingesting methylmercury may cause chromosomal
aberrations and sister chromatid exchanges (SCE) (Skerfving et al., 1970; Wulf et al., 1986; Franchi et
al., 1994).
A study of nine Swedish subjects who consumed mercury-contaminated fish and four controls
showed a statistically significant rank correlation between blood mercury and percentage of lymphocytes
with chromosome breaks (Skerfving et al., 1970). An extension of this study (Skerfving et al. 1974)
included 23 exposed (5 females and 18 males) and 16 controls (3 females and 13 males). The authors
reported significant correlations between blood mercury level and frequency of chromatid changes and
"unstable" chromosome aberrations; there was no correlation with "stable" chromosome aberrations.
The Wulf et al. (1986) study was of 92 Greenlander Eskimos. Subjects were divided into three
groups based on intake of seal meat (six times per week; two to five times per week, once a week, or no
consumption of seal meat). Higher frequency of SCE in lymphocytes was correlated with blood mercury
concentration; an increase of 10 ug mercury per liter of blood was associated with an increase of 0.3
SCE/cell. Positive correlations were also found for smoking, diet, living district, and cadmium exposure.
Franchi et al. (1994) evaluated formation of micronuclei in peripheral blood lymphocytes of
Mediterranean fishers, a group with presumed high exposure to methylmercury. Fifty-one subjects were
3-46 ' Methylmercury Water Quality Criterion 1/3/01
-------
interviewed on age, number of seafood-based meals/week, and habits such as smoking and alcohol
consumption. Total blood mercury was measured; the range was 10.08-304.11 ng/g with a mean of 88.97
± 54.09 ng/g. There was a statistically significant correlation between blood mercury concentration and
micronucleus frequency and between age and micronucleus frequency (U.S. EPA, 1997e)
3.6.2 Animal Studies
In a study with cats (Charbonneau et al. 1976), methylmercury did not induce dose-related
unscheduled DNA synthesis in lymphocytes or chromosomal aberrations in bone marrow cells after oral
exposure for up to 39 months (Miller et al., 1979). Statistically significant decreases in unscheduled
DNA synthesis and increases in chromosomal aberrations were observed, but there was no dose-
response.
Strain-specific differences exist with respect to the ability of methylmercury to produce dominant
lethal effects in mice (Suter, 1975). When (SEC x CSlE^Fi males were injected with 10 mg/kg
methylmercury hydroxide, there was a slight reduction in the total number of implantations and a
decrease in the number of viable embryos. This was not observed when (101 x C3H)Fj males were
exposed in a similar fashion. When female (10 x C3H)Fj mice were treated with methylmercuric
hydroxide, no increase in the incidence of dead implants was observed (unlike the case for mercuric
chloride). Changes in chromosome number, but no increase in chromosome aberrations, were observed
in oocytes of Syrian hamsters treated with one interperitoneal injection of 10 mg/kg methylmercuric
chloride (Mailhes, 1983). Methylmercury was administered subcutaneously to golden hamsters at doses
of 6.4 mg or 12.8 mg mercury/kg/body weight. Polyploidy and chromosomal aberrations were increased
in bone marrow cells, but there was no effect on metaphase n oocytes. There was an inhibitory effect on
ovulation, which the authors noted was not as severe as that induced by mercuric chloride in the same
study (Watanabe et al., 1982). Nondysjunction and sex-linked recessive lethal mutations were seen in
Drosophila melanogaster treated with methylmercury in the diet (Ramel, 1972).
As reviewed in WHO (1990), methylmercury is not a point mutagen but is capable of causing
chromosome damage in a variety of systems. In vitro studies have generally shown clastogenic activity
but only weak mutagenic activity. Methylmercuric chloride and dimethylmercury were both shown to
induce chromosome aberrations and aneuploidy in primary cultures in human lymphocytes;
methylmercuric chloride was the more potent clastogen at equally toxic doses (Betti et al., 1992). Both
methylmercury and mercuric chloride induce a dose-dependent increase in SCE in primary human
Methylmercury Water Quality Criterion 1/3/01 3-47
-------
lymphocytes and muntjac fibroblasts; methylmercury was about five times more effective in this regard
(Verschaeve et al., 1984; Morimoto et al., 1982).
Methylmercury has been shown to inhibit nucleolus organizing activity in human lymphocytes
(Verschaeve et al., 1983). Methylmercury can induce histone perturbation and has been reported to
interfere with gene expression in cultures of glioma cells (WHO, 1990). Impaired growth and
development was noted in cultured mouse embryonic tissue treated in vitro with methylmercuric
chloride, but there was no increase in SCE (Matsumoto and Spindle, 1982). Costa et al. (1991) showed
that methylmercuric chloride caused DNA strand breaks in both V79 and rat glioblastoma cells treated in
vitro. Methylmercuric chloride produced more strand breaks than did mercuric chloride.
Evidence of DNA damage has been observed in the Bacillus subtilis rec-assay (Kanematsu et al.,
1980). These authors reported negative results for methylmercury in spot tests for mutagenicity in the
following bacterial strains: E. coli B/r WP2 and WP2; and Salmonella typhimurium strains TA1535,
TA1537, TA1538, TA98, and TA100. Jenssen and Ramel (1980) indicated in a review article that
methylmercury acetate was negative in both micronucleus assays and mutagenicity tests in Salmonella;
the article referred to Heddle and Bruce (1977) and provided no experimental details. Weak mutagenic
responses for methylmercuric chloride and methoxyethyl mercury chloride were observed in Chinese
hamster V79 cells at doses near the cytotoxic threshold (Fiskesjo, 1979), and methylmercury produced a
slight increase in the frequency of chromosomal nondysjunction in Saccharomyces cerevisiae (Nakai and
Machida, 1973). Methylmercury, however, caused neither gene mutations nor recombination in S.
cerevisiae (Nakai and Machida, 1973). Methylmercury retarded DNA synthesis and produced single-
strand breaks in DNA in L5178Y cells (Nakazawa et al., 1975).
3.7 CARCINOGENICITY
3.7.1 Human Studies
At this time, no human studies have reported an association between methylmercury exposure and
overall cancer rates. Three studies were identified that examined the relationship between
methylmercury exposure and cancer. No persuasive evidence of increased carcinogenicity attributable to
methylmercury exposure was observed in any of the studies. Interpretation of these studies, however,
was limited by poor study design and incomplete descriptions of methodology and/or results.
3-48
Methylmercury Water Quality Criterion 1/3/01
-------
3.7.2 Animal Studies
The results from three dietary studies in two strains of mice indicate that methylmercury is
carcinogenic. Interpretation of two of the positive studies was complicated by observation of tumors
only at doses that exceeded the Maximum Tolerated Dose (MTD). Therefore, only one positive animal
study is appropriate for consideration. A fourth dietary study in mice, three dietary studies in rats, and a
dietary study in cats failed to show carcinogenicity of methylmercury. Interpretation of four nonpositive
studies was limited because of deficiencies in study design or failure to achieve an MTD.
Methylmercuric chloride was administered in the diet at levels of 0, 0.4, 2, or 10 ppm (0, 0.03, 0.14,
and 0.69 mg Hg/kg-day in males and 0, 0.03, 0.13, and 0.60 mg Hg/kg-day in females) to B6C3F1 mice
(60/sex/group) for 104 weeks (Mitsumori et al., 1990). In high-dose males, a marked increase in
mortality was observed after 60 weeks (data were presented graphically; statistical analyses not
performed). Survival at study termination was approximately 50%, 60%, 60%, and 20% in control, low-,
mid-, and high-dose males, respectively, and 58%, 68%, 60%, and 60% in control, low-, mid-, and high-
dose females, respectively. The cause of the high mortality was not reported. At study termination, the
mean body weight in high-dose males was approximately 67% of controls and in high-dose females was
approximately 90% of controls (data presented graphically; statistical analyses not performed). Focal
hyperplasia of the renal tubules was significantly (p<0.01) increased in high-dose males (14/60; the
incidence was 0/60 in all other groups). The incidence of renal epithelial carcinomas (classified as solid
or cystic papillary type) was significantly (p<0.01) increased in high-dose males (13/60; the incidence
was 0/60 in all other groups). The incidence of renal adenomas (classified as solid or tubular type) was
also significantly (p<0.05) increased hi high-dose males; the incidence was 0/60, 0/60, 1/60, and 5/60 in
control, low-, mid-, and high-dose males, respectively, and 0/60, 0/60, 0/60, and 1/60 in control, low-,
mid-, and high-dose females, respectively. No metastases were seen in the animals. The incidences of a
variety of nonneoplastic lesions were increased in the high-dose rats including these: sensory
neuropathy, neuronal necrosis in the cerebrum, neuronal degeneration in the cerebellum, and chronic
nephropathy of the kidney. Males exhibited tubular atrophy of the testis (1/60, 5/60, 2/60, and 54/60 in
control, low-, mid-, and high-dose, respectively) and ulceration of the glandular stomach (1/60, 1/60,
0/60, and 7/60 in control, low-, mid-, and high-dose males, respectively). An MTD was achieved in
middose males and high-dose females. High mortality in high-dose males indicated that the MTD was
exceeded in this group.
Methylmercury Water Quality Criterion 1/3/01
3-49
-------
Mitsumori et al. (1981) administered 0, 15, or 30 ppm of methylmercuric chloride (99.3% pure) in
the diet (0,1.6 and 3.1 mg Hg/kg-day) to ICR mice (60/sex/group) for 78 weeks. Interim sacrifices of up
to 6/sex/group were conducted at weeks 26 and 52. Kidneys were microscopically examined from all
animals that died or became moribund after week 53 or were killed at study termination. Lungs from
mice with renal masses and renal lymph nodes showing gross abnormalities were also examined.
Survival was decreased in a dose-related manner; at week 78 survival was 24/60, 6/60, and 0/60 in
control, low-, and high-dose males, respectively, and 33/60, 18/60, and 0/60, in control, low-, and high-
dose females, respectively (statistical analyses not performed). The majority of high-dose mice (51/60
males and 59/60 females) died by week 26 of the study. Examination of the kidneys of mice that died or
were sacrificed after 53 weeks showed a significant (p<0.001) increase in renal tumors in low-dose males
(13/16 versus 1/37 in controls). The incidence of renal epithelial adenocarcinomas in control and low-
dose males was 0/37 and 11/16, respectively (p<0.001). The incidence of renal epithelial adenomas in
control and low-dose males was 1/37 and 5/16, respectively (p<0.01). No renal tumors were observed in
females in any group. No metastases to the lung or renal lymph nodes were observed. Evidence of
neurotoxicity and renal pathology was observed in the treated mice at both dose levels. The high
mortality in both groups of treated males and in high-dose females indicated that the MTD was exceeded
in these groups.
A followup study to the Mitsumori et al. (1981) study was reported by Hirano et al. (1986).
Methylmercuric chloride was administered in the diet to ICR mice (60/sex/group) at levels of 0, 0.4, 2, or
10 ppm (0, 0.03,0.15, and 0.73 mg Hg/kg-day in males and 0, 0.02, 0.11, and 0.6 mg Hg/kg-day in
females) for 104 weeks. Interim sacrifices (6/sex/group) were conducted at 26, 52, and 78 weeks.
Complete histopathological examinations were performed on all animals found dead, killed in extremis,
or killed by design. Mortality, group mean body weights and food consumption were comparable to
controls. The first renal tumor was observed at 58 weeks in a high-dose male, and the incidence of renal
epithelial tumors (adenomas or adenocarcinomas) was significantly increased in high-dose males (1/32,
0/25, 0/29, and 13/26 in the control, low-, mid-, and high-dose groups, respectively). Ten of the 13
tumors in high-dose males were adenocarcinomas. These tumors were described as solid type or cystic
papillary types of adenocarcinomas. No invading proliferation into the surrounding tissues was seen.
The incidence of renal epithelial adenomas was not significantly increased in males, and no renal
adenomas or adenocarcinomas were observed in any females. Focal hyperplasia of the tubular
epithelium was reported to be increased in high-dose males (13/59; other incidences not reported).
Increases in nonneoplastic lesions in high-dose animals provided evidence that an MTD was exceeded.
Nonneoplastic lesions reported as increased in treated males included the following: epithelial
3-50 Methylmercury Water Quality Criterion 1/3/01
-------
degeneration of the renal proximal tubules; cystic kidney; urinary cast and pelvic dilatation; and
decreased spermatogenesis. Epithelial degeneration of the renal proximal tubules and degeneration or
fibrosis of the sciatic nerve were reported in high-dose females.
No increase in tumor incidence was observed in a study using white Swiss mice (Schroeder and
Mitchener 1975). Groups of mice (54/sex/group) were exposed from weaning until death to
methylmercuric acetate in the drinking water at two doses. The low-dose group received 1 ppm
methylmercuric acetate (0.19 mg Hg/kg-day). The high-dose group received 5 ppm methylmercuric
acetate (0.95 mg Hg/kg-day) for the first 70 days and then 1 ppm, thereafter, due to high mortality (21/54
males and 23/54 females died prior to the dose reduction). Survival among the remaining mice was not
significantly different from controls. Significant (p<0.001) reductions in body weight were reported in
high-dose males (9-15% lower than controls) and high-dose females (15-22% lower than controls)
between 2 and 6 months of age. Mice were weighed, dissected, gross tumors were detected, and some
sections were made of heart, lung, liver, kidney, and spleen for microscopic examination. No increase in
tumor incidence was observed. This study is limited because complete histological examinations were
not performed, and pathology data other than tumor incidence were not reported.
Mitsumori et al. (1983, 1984) conducted a study in Sprague-Dawley rats. They administered diets
containing 0, 0.4, 2, or 10 ppm of methylmercuric chloride (0, 0.011, 0.05, and 0.28 mg Hg/kg-day in
males; 0, 0.014, 0.064, and 0.34 mg Hg/kg-day in females) to Sprague-Dawley rats (56
animals/sex/group) for up to 130 weeks. Interim sacrifices of 10/group (either sex) were conducted at
weeks 13 and 26 and of 6/group (either sex) at weeks 52 and 78. Mortality was increased in high-dose
males and females. At week 104, survival was approximately 55%, 45%, 75%, and 10% in control, low-,
mid-, and high-dose males, respectively, and 70%, 75%, 75%, and 30% in control, low-, mid-, and high-
dose females, respectively (data presented graphically). Body weight gain was decreased in high-dose
animals (approximately 20-30%; data presented graphically). No increase in tumor incidence was
observed in either males or females. Noncarcinogenic lesions that were significantly increased (p< 0.05)
in high-dose rats included the following: degeneration in peripheral nerves and the spinal cord (both
sexes); degeneration of the proximal tubular epithelium of the kidney (both sexes); severe chronic
nephropathy (females); parathyroid hyperplasia (both sexes); polyarteritis nodosa and calcification of the
abdominal arterial wall (females); bone fibrosis (females); bile duct hyperplasia (males); and
hemosiderosis and extramedullary hematopoiesis in the spleen (males). In addition, mid-dose males
exhibited significantly increased degeneration of the kidney proximal tubular epithelium and hyperplasia
Methylmercury Water Quality Criterion 1/3/01
3-51
-------
of the parathyroid. An MTD was achieved in mid-dose males and in high-dose females; the MTD was
exceeded in high-dose males.
No increase in tumor incidence or decrease in tumor latency was observed in another study using
rats (strain not specified) (Verschuuren et al., 1976). Groups of 25 female and 25 male rats were
administered methylmercuric chloride at dietary levels of 0,0.1, 0.5, and 2.5 ppm (0, 0.004, 0.02, and 0.1
mg Hg/kg-day) for 2 years. No significant effects were observed on growth or food intake except for a
6% decrease (statistically significant) in body weight gain at 60 weeks in high-dose females. Survival
was 72%, 68%, 48%, and 48% in control, low-, mid- and high-dose males, respectively; and 76%, 60%,
64%, and 56% in control, low-, mid- and high-dose females, respectively (statistical significance not
reported). Increases in relative kidney weights were observed in both males and females at the highest
dose. No effects on the nature or incidence of pathological lesions were observed, and tumors were
reported to have been observed with comparable incidence and latency among all of the groups. This
study was limited by the small sample size and failure to achieve an MTD.
No tumor data were reported in a study using Wistar rats (Munro, 1980). Groups of 50 Wistar
rats/sex/dose were fed diets containing methylmercury; doses of 2, 10, 50, and 250 micrograms Hg/kg-
day were fed for 26 months. High-dose female rats exhibited reduced body weight gains and showed
minimal clinical signs of neurotoxicity; however, high-dose male rats showed overt clinical signs of
neurotoxicity, decreased hemoglobin and hematocrit values, reduced weight gains and significantly
increased mortality. Histopathologic examination of the high-dose rats of both sexes revealed
demyelination of dorsal nerve roots and peripheral nerves. Males showed severe dose-related kidney
damage, and females had minimal renal damage.
No increase in tumor incidence was observed in a multiple generation reproduction study using
Sprague-Dawley rats (Newberne et al., 1972). Groups of rats (30/sex) were given semisynthetic diets
supplemented with either casein or a fish protein concentrate to yield dietary levels of 0.2 ppm
methylmercury (0.008 mg Hg/kg-day). Another group of controls received untreated rat chow. Rats that
received diets containing methylmercury during the 2-year study had body weights and hematology
comparable to controls. Detailed histopathologic analyses revealed no lesions of the brain, liver, or
kidney that were attributable to the methylmercury exposure. Mortality data were not presented.
Interpretation of these data is limited by the somewhat small group sizes and failure to achieve an MTD.
3-52
Methylmercury Water Quality Criterion 1/3/01
-------
No increase in tumor incidence was observed in a study using random-bred domestic cats
(Charbonneau et al., 1976). Groups of cats (4-5/sex/group) were given doses of 0.0084, 0.020, 0.046,
0.074 or 0.176 mg Hg/kg-day either as methylmercury-contaminated seafood or as methylmercuric
chloride in the diet for up to 2 years. Controls were estimated to have received 0.003 mg Hg/kg-day.
Food consumption and body weight were not affected by treatment with methylmercury. Due to
advanced signs of neurotoxicity (loss of balance, ataxia, impaired gait, impaired reflexes, weakness,
impaired sensory function, mood change and tremor), cats at the highest dose tested were sacrificed after
approximately 16 weeks, and cats at the next highest dose were sacrificed after approximately 54—57
weeks. Cats at the next highest dose generally exhibited mild neurological impairment (altered hopping
reaction and hypalgesia). One cat at this dose was sacrificed after 38 weeks because of neurotoxicity,
and one cat died of acute renal failure after 68 weeks. Cats at the two highest doses had pathological
changes in the brain and spinal cord, but no histopathological changes were noted in other tissues
examined. Interpretation of the results of this study is limited because of the small group sizes, early
sacrifice of cats at the two highest dose levels and no available data regarding pathological changes in
cats at the three lowest dose levels. This study was also limited by its short duration when compared to
the lifespan of a cat.
Blakley (1984) administered methylmercuric chloride to female Swiss mice (number/group not
specified) in drinking water at concentrations of 0,0.2,0.5 or 2.0 mg/L for 15 weeks. This corresponded
to approximately 0, 0.03, 0.07 and 0.27 mg Hg/kg-day. At the end of week 3, a single dose of 1.5 rag/kg
of urethane was administered intraperitoneally to 16-20 mice/group. No effects on weight gain or food
consumption were observed. Lung tumor incidence in mice not administered urethane (number/group
not specified) was less than 1 tumor/mouse in all groups. Statistically significant trends for increases in
the number and size of lung adenomas/mouse with increasing methylmercury dose were observed; the
tumor number/mouse was 21.5, 19.4,19.4, and 33.1 in control, low-, mid- and high-dose mice,
respectively, and the tumor size/mouse was 0.70, 0.73, 0.76 and 0.76 mm in control, low-, mid- and high-
dose mice, respectively. The study authors suggest that the increase in tumor number and size may have
been related to immunosuppressive activity of methylmercury. It should be noted that this is considered
a short-term assay and that only pulmonary adenomas were evaluated.
Methylmercury Water Quality Criterion 1/3/01
3-53
-------
-------
4.0 RISK ASSESSMENT FOR METHYLMERCURY
4.1 BACKGROUND
Methylmercury is highly toxic to mammalian species and causes a variety of adverse effects. It is a
developmental toxicant in humans and animals. It causes chromosomal effects but does not induce point
mutations. The Mercury Study Report to Congress (MSRC) (U.S. EPA, 1997) concluded that because
there are data for mammalian germ-cell chromosome aberration and limited data from a heritable
mutation study, methylmercury is placed in a group of high concern for potential human germ-cell
mutagenicity. There is no two-generation study of reproductive effects, but shorter term studies in
rodents, guinea pigs, and monkeys have reported observations consistent with reproductive deficits.
There are no data to indicate that methylmercury is carcinogenic in humans, and it induces tumors in
animals only at highly toxic doses. Application of the revised Guidelines for Cancer Risk Assessment
leads to a judgment that methylmercury is not likely to be carcinogenic for humans under conditions of
exposure generally encountered in the environment.
The quantitative health risk assessment for a noncarcinogen is the reference dose (RfD). This is an
estimate (with uncertainty spanning perhaps an order of magnitude) of a daily exposure to the human
population (including sensitive subgroups) that is likely to be without an appreciable risk of deleterious
health effects during a lifetime.
EPA has published two RfDs for methylmercury that represented the Agency consensus at that
time. The original RfD of 0.3 ng/kg/day was determined in 1985. The current RfD of 0.1 |ig/kg/day was
established as the Agency consensus estimate in 1995. While EPA was developing the MSRC (U.S. EPA,
1997), it became apparent that considerable new data on the health effects of methylmercury in humans
were emerging. Among these data sources were large studies of seafood-consuming populations in the
Seychelles and Faroe Islands. Smaller scale studies were being reported on effects in populations around
the U.S. Great Lakes and in the Amazon basin. Publications also included novel statistical approaches
and applications of physiologically based pharmacokinetic (PBPK) models.
In 1997 the MSRC was undergoing final review; at that time many of the new data had either not
been published in the peer-reviewed press or not been subjected to rigorous review. EPA decided that it
was premature to make a change in the 1995 methylmercury RfD for the MSRC. This decision was in
accordance with the advice of the Science Advisory Board (SAB). Since 1997 the field of
Methylmercury Water Quality Criterion 1/3/01
4-1
-------
methylmercury toxicology and assessment has expanded dramatically. This criteria document presents a
revised RfD that considers data from the human studies published in the 1990s, recent evaluations of
health and pharmacokinetic data, and recent statistical and modeling approaches to assessing those data.
The following sections include brief descriptions of the previously published EPA RfDs as well as
descriptions of some of the evaluation processes that took place at the end of the 1990s.
For this document the following definitions apply. These reflect usage in the National Research
Council publication Toxicological Effects of Methylmercury (NRC, 2000) (see Section 1.5).
NOAEL No-observed-adverse-effect level. An exposure level at which there are no statistically
or biologically significant increases in the frequency or severity of adverse effects in a
comparison between an exposed population and a control group. Effects may be seen at
this level of exposure, but they are not considered to be adverse. For risk assessment the
NOAEL is generally the highest level at which no adverse effects are seen.
LOAEL Lowest-observed-adverse-effect level. The lowest exposure level at which there are
statistically or biologically significant increases in frequency or severity of adverse
effects in a comparison between an exposed population and a control group.
BMD Benchmark dose. In common parlance this term refers to a quantitative assessment for
noncancer health effects that uses a curve-fitting procedure to determine a level
functionally equivalent to a NOAEL. In this chapter, BMD will be used to mean an
estimated dose that corresponds to a specified risk above the background risk.
BMDL Benchmark dose lower limit, a statistical lower limit on a calculated BMD. In this
document that will be the 95% lower confidence limit. The BMDL will be used as the
starting point for the calculation of the methylmercury RfD.
4.1.1 Other RfDs Published by EPA
Two RfDs based on human studies have been published as consensus values for EPA. In addition,
the MSRC (EPA, 1997) describes an RfD that could be estimated from animal data.
4-2
Methylmercury Water Quality Criterion 1/3/01
-------
4.1.1.1 1985 RfD
A hazard identification and dose-response assessment was proposed for methylmercury in 1980
(U.S. EPA, 1980). This assessment was reviewed and consensus was achieved by the EPA RfD/RfC
(reference concentration) Work Group on December 2, 1985. This RfD was published on EPA's
Integrated Risk Information System (IRIS) in 1986. The critical effects were multiple central nervous
system (CNS) effects, including ataxia and paresthesia in populations of humans exposed to
methylmercury through consumption of contaminated grain (summarized by Clarkson et al., 1976;
Nordberg and Strangert, 1976; and WHO, 1976).
The RfD for methylmercury was determined to be 3 x 10"4 mg/kg-day (0.3 |j,g/kg/day), based on a
LOAEL of 0.003 mg/kg-day (corresponding to 200 |ig/L blood concentration) and an uncertainty factor
of 10 to adjust the LOAEL to what is expected to be a NOAEL. An additional uncertainty factor (UF) of
10 for sensitive individuals for chronic exposure was not deemed necessary, as the adverse effects were
seen in what was regarded as a sensitive group of individuals: adults who consumed methylmercury-
contaminated grain.
The RfD/RfC Work Group ascribed medium confidence to the choice of study, the database, and
the RfD. The blood levels associated with the LOAEL were well supported by more recent data, but
neither the chosen studies nor supporting database described a NOAEL. Medium confidence generally
indicates that new data may change the assessment of the RfD.
4.1.13 1995 RfD
After publication of the RfD of 0.3 |ig/kg/day, questions were raised as to its validity; some of these
questions were in formal submissions requesting a change on the IRIS entry. In particular it was asked
whether the RfD based on effects in exposed adults was protective against developmental effects.
Subsequent to the RfD publication, the effects in Iraqi children of in utero exposure to methylmercury
were reported by Marsh et al. (1987). The RfD/RfC Work Group discussed the methylmercury RfD in
1992 and again in 1994. Consensus on a revised RfD was reached in January 1995. Detailed description
of the RfD derivation can be found in Volume V of the MSRC (U.S. EPA, 1997e).
Marsh et al. (1987) was chosen as the most appropriate study for determination of an RfD
protective of a putative sensitive subpopulation, namely infants born to mothers exposed to
Methylmercury Water Quality Criterion 1/3/01
4-3
-------
methylmercury during gestation. The data collected by Marsh et al. (1987) summarize clinical
neurologic signs of 81 mother-and-child pairs. Maternal hair mercury concentrations were collected as
the exposure metric. Concentrations ranging from 1 to 674 ppm mercury were determined from X-ray
fluorescent spectrometric analysis of selected regions of maternal scalp. These were correlated with
clinical signs observed in the affected members of the mother-child pairs. The hair concentration at a
hypothetical NOAEL for developmental effects was determined by application of a BMD approach (see
subsequent section for discussion of methods and data used). The analysis used the combined incidence
of all neurological effects in children exposed in utero as reported in the Marsh et al. (1987) study. A
Weibull model for extra risk was used to determine the BMD; in current terminology, this was a BMDL
(95% lower confidence limit) on the dose corresponding to a 10% risk level. This level was calculated to
be 11 ppm mercury in maternal hair (11 mg/kg hair). A description of BMD determination, choice of
model, and issues on grouping of data is on pages 6-25 to 6-31 of Volume V of the MSRC.
The BMD of 11 ppm maternal hair mercury was converted to an exposure level of 44 ^g mercury/L
blood using a 250:1 ratio as described in the MSRC (U.S. EPA, 1997e, pp. 6-22 to 6-23):
11 mg/kg hair / 250 =44 (ig/L blood
To obtain a daily dietary intake value of methylmercury corresponding to a specific blood
concentration, factors of absorption rate, elimination rate constant, total blood volume, and percentage of
total mercury present in circulating blood were taken into account. Calculation was by the following
equation, based on the assumptions that steady-state conditions exist and that first-order kinetics for
mercury are being followed:
d \ig/day =
C x b
where:
d = daily dietary intake (expressed as jig of methylmercury)
c = concentration in blood (expressed as 44 ng/L)
b = elimination constant (expressed as 0.014 days"1)
V = volume of blood in the body (expressed as 5 L)
4-4
Methylmercury Water Quality Criterion 1/3/01
-------
A = absorption factor (expressed as a unitless decimal fraction of 0.95)
f = fraction of daily intake taken up by blood (unitless, 0.05)
Solving for d gives the daily dietary intake of mercury that results in a blood mercury concentration of 44
ug/L. To convert this to daily ingested dose (ng/kg-day), a body weight of 60 kg was assumed and
included in the equation denominator:
d =
d =
d =
c x
A xfx- bw
_i
x 0-Q14 days' x 5 L
0.95 x 0.05 x 60 kg
1.1 iig/kg-day
The dose d (1.1 ug/kg-day) is the total daily quantity of methylmercury that is ingested by a 60-kg
individual to maintain a blood concentration of 44 |ig/L or a hair concentration of 11 ppm. The
rationales for use of the hainblood ratio and specific values for equation parameters can be found on
pages 6-21 to 6-25 of Volume V of the MSRC.
A composite uncertainty factor of 10 was used. This uncertainty factor was applied for variability
in the human population, in particular the wide variation in biological half-life of methylmercury and the
variation that occurs hi the hair-to-blood ratio for mercury. In addition, the factor accounts for lack of a
two-generation reproductive study and lack of data for possible chronic manifestations of adult effects
(e.g., paresthesia observed during gestation). The default value of 1 was used for the modifying factor.
The RfD was calculated using the following equation:
BMD
RfD =
UF * MF
= 1.1 iig/kg-day
10
= 1 x 10'4 mg/kg-day
or 0.1 p.g/kg/day.
Methylmercury Water Quality Criterion 1/3/01
4-5
-------
Confidence in the supporting database and in the RfD were considered medium by the RfD/RfC
Work Group. The MSRC (U.S. EPA, 1997e) says the following:
The principal study (Marsh et al. 1987) is a detailed report of human exposures with quantitation of
methylmercury by analysis of specimens from affected mother-child pairs. A strength of this study is that the
quantitative data are from the affected population and quantitation is based upon biological specimens
obtained from affected individuals. A threshold was not easily defined; extended application of modeling
techniques was needed to define the lower end of the dose-response curve. This may indicate high variability
of response to methylmercury in the human mother-child pairs or misclassification in assigning pairs to the
cohort.
Further discussion of areas of uncertainty and variability are on pages 6-31 to 6-51 of Volume V of
the MSRC (U.S. EPA, 1997e). A quantitative analysis of uncertainty in an RfD based on the Iraqi data is
found in Appendix D of Volume V, and additional discussions of areas of uncertainty are in Volume VH,
Risk Characterization, of the MSRC (U.S. EPA, 1997g).
4.1.1.3 Reference Values Derived From Animal Data
There are issues inherent to epidemiological studies, including the possibility of coexposure to
other potential toxicants, that are not of concern in controlled experimental animal studies. It is therefore
informative to compare RfDs that may be derived from animal studies to those derived from the
epidemiological literature. RfDs derived from monkey studies are particularly relevant* as the neurotoxic
effects produced by developmental methylmercury exposure in monkeys are similar to those identified in
humans (Burbacher et al., 1990a; Gilbert and Grant-Webster, 1995). The studies at the University of
Washington were of a relatively large cohort of macaque monkeys whose mothers were exposed
throughout pregnancy to 50 ng/kg/day of methylmercury. The studies revealed deficits on cognitive tests
during infancy, which may represent retarded development (Burbacher et al., 1986; Gunderson et al.,
1986, 1988). These methylmercury-exposed monkeys also displayed aberrant play and social behavior
(Burbacher et al., 1990b). Studies at the Canadian Health Protection Branch in the same species of
monkey, dosed with 50 ug/kg/day from birth to 7 years of age, revealed visual, auditory, and
somatosensory deficits, including evidence of delayed neurotoxicity identified in middle age (Rice and
Gilbert, 1995, 1992, 1982; Rice, 1989a). Research in a cohort of monkeys dosed beginning in utero and
continuing until 4 years of age revealed similar sensory system impairment (Rice, 1998; Rice and Gilbert,
1995,1990). Three individuals dosed at 10 or 25 ug/kg/day all exhibited impaired function in at least
4-6
Methylmercury Water Quality Criterion 1/3/01
-------
one sensory system in addition to evidence of delayed neurotoxicity (Rice, 1998). In none of these
studies was a NOAEL identified.
Calculation of an RfD from these data according to the method typically used by the EPA would
include application of a number of UFs, including dividing the LOAEL by a factor of 10 (because no
NOAEL was identified), division by 10 again for extrapolation from animal to human data, and division
by another factor of 10 in consideration of individual variation in sensitivity. Monkeys and humans have
approximately the same brain:blood mercury ratio following chronic exposure (Burbacher et al., 1990a),
although the ratio in humans may be slightly higher than in monkeys (Rice, 1989b). However, the half-
life of mercury in the blood of monkeys is about 15 days (Rice, 1989c), whereas clearance times for
humans averaged 45-70 days in several studies, with some individuals having even longer clearance
times (see Section 4.2.3). The shorter clearance time in monkeys would result in an UF of at least 5
based on pharmacokinetic considerations alone; therefore an overall factor of 10 appears appropriate for
interspecies extrapolation. This calculation would yield an RfD of 0.05 ug/kg/day from the in utero and
postnatal exposure studies, and an RfD as low as 0.01 ug/kg/day based on combined in utero and
postnatal exposure (Rice, 1996). Gilbert and Grant-Webster (1995) suggested an RfD of 0.025
Hg/kg/day based on the same data.
4.1.2 Risk Assessments Done by Other Groups
Quantitative estimates of hazards of oral exposure to methylmercury have been considered by the
Food and Drug Administration (FDA), Agency for Toxic Substances and Disease Registry (ATSDR),
and other countries (WHO/IPCS), among others.
4.1.2.1 Food and Drug Administration
In 1969, in response to the poisonings in Minamata Bay and Niigata, Japan, the U.S. FDA proposed
an administrative guideline of 0.5 ppm for mercury in fish and shellfish moving in interstate commerce.
This limit was converted to an action level in 1974 (Federal Register 39, 42738, December 6, 1974) and
increased to 1.0 ppm in 1979 (Federal Register 44, 3990, January 19, 1979) in recognition that exposure
to mercury was less than originally considered. In 1984, the 1.0 ppm action level was converted from a
mercury standard to one based on methylmercury (Federal Register 49; November 19, 1984).
Methylmercury Water Quality Criterion 1/3/01
4-7
-------
The action level takes into consideration the tolerable daily intake (TDI) for methylmercury as well
as information on seafood consumption and associated exposure to methylmercury. The TDI is the
amount of methylmercury that can be consumed daily over a long period of time with a reasonable
certainty of no harm. FDA established a TDI based on a weekly tolerance of 0.3 mg of total mercury per
person, of which no more than 0.2 mg should be present as methylmercury. These amounts are
equivalent to 5 and 3.3 |ig, respectively, per kilogram of body weight. Using the values of
methylmercury, this tolerable level would correspond to approximately 230 [ig/week for a 70-kg person,
or 33 fig/person/day (0.47 H-g/kg bw/day). The TDI was calculated from data developed in part by
Swedish studies of Japanese individuals poisoned in the Niigata episode, which resulted from the
consumption of contaminated fish and shellfish and the consideration of other studies of fish-eating
populations.
Based on observations from the later poisoning event in Iraq, FDA has acknowledged that the fetus
may be more sensitive than adults to the effects of mercury (Federal Register 44, 3990, January 19, 1979;
U.S. FDA Consumer, September 1994). In recognition of these concerns, FDA has provided advice to
pregnant women and women of childbearing age to limit their consumption of fish known to have high
levels of mercury (U.S. FDA Consumer, 1994). FDA believes, however, that given existing patterns of
fish consumption, few women (less than 1%) eating such high-mercury fish will experience slight
reductions in the margin of safety. However, because of the uncertainties associated with the Iraqi study,
FDA has chosen not to use the Iraqi study as a basis for revising its action level. Instead, FDA has
chosen to wait for findings of prospective studies offish-eating populations in the Seychelles Islands.
4.1.2.2 World Health Organization
The International Programme on Chemical Safety (IPCS) of the World Health Organization
published a criteria document on mercury (WHO, 1990). In that document, it was stated that "a daily
intake of 3 to 7 u£ Hg/kg body weight would cause adverse effects of the nervous system, manifested as
an approximately 5% increase in the incidence of paraesthesias." The IPCS expert group also concluded
that developmental effects in offspring (motor retardation or signs of CNS toxicity) could be detected as
increases over background incidence at maternal hair levels of 10-20 ppm mercury. These levels of
concern were based on evaluation of data including the human poisoning incident in Iraq.
4-8
Methylmercury Water Quality Criterion 1/3/01
-------
4.1.2.3 ATSDR
In 1993, ATSDR first published a Minimal Risk Level (MRL) for methylmercury. An MRL is
derived in a manner similar to the RfD; it is defined as an estimate of the daily human exposure to a
hazardous substance that is likely to be without appreciable risk of adverse noncancer health effects over
a specified duration of exposure. In 1999 ATSDR published a revised methylmercury MRL using the
Seychelles Islands study (SCDS) (Davidson et al., 1998) as the starting point (ATSDR, 1999). In this
study (described in detail in Section 3.2.2.5 and summarized in Section 4.2.13), the investigators
examined the correlation between subtle neurological effects and low-dose chronic exposure to
methylmercury. No correlation between maternal hair mercury concentrations and neurological effects
was seen in the SCDS 66-month-old children. ATSDR determined a minimal risk level of 0.3 [igfkg per
day, based on a dose of 1.3 jig/kg per day, which reflects the average concentration of the upper quintile
of the exposed population but does not necessarily correspond to a NOAEL. ATSDR used a UF of 1.5 to
account for pharmacokinetic variability within the human population; they made their choice based on
the analyses of Clewell et al. (1998). An additional factor of 1.5 was applied to account for any other
individual variability (e.g., pharmacodynamics) as well as a modifying factor of 1.5 to account for the
possibility that domain-specific tests used in the Faroe Islands study might have allowed detection of
subtle neurological effects that were not evaluated in the Seychelles cohort. Although the conventional
risk assessment approach is to multiply UFs, ATSDR summed these factors to develop an overall safety
factor of 4.5.
4.1.3 SAB Review of the Mercury Study Report to Congress
The Science Advisory Board (SAB) is a public advisory group providing extramural scientific
information and advice to the Administrator and other officials of the EPA. The SAB is structured to
provide balanced, expert assessment of scientific matters relating to problems facing the Agency. The
SAB reviewed a draft of the eight-volume MSRC (U.S. EPA, 1997a-h) in the context of a public meeting
held February 13 and 14,1997. A panel of 33 scientists reviewed the entire MSRC. A subgroup focused
on the health effects data, and in particular EPA's use of those data to derive the methylmercury RfD of
0.1 ng/kg/day, based on effects observed in Iraqi children exposed in utero.
The SAB report was published in October 1997 (EPA-SAB-EC-98-001). It made the following
statement:
Methylmercury Water Quality Criterion 1/3/01
4-9
-------
In general, from the standpoint of looking at human health effects and the uncertainties, the draft report
[MSRC] 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 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.
The SAB report continued:
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 paragraph in that they have much larger sample 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 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. [Emphasis theirs]
4.1.4 Interagency Consensus Process
Among the many reviews of the MSRC was one by scientists and policy-makers from interested
Federal agencies, sponsored by the Committee on Environment and Natural Resources (CENR), Office
of Science and Technology (OSTP). This review highlighted many divergent points of view as to the
appropriate basis for quantitative assessment of the low-dose effects of methylmercury exposure. It was
decided that an interagency process with external involvement would be undertaken to review new
methylmercury data and evaluate new and existing data. EPA committed to participate in this process
and, at its conclusion, to assess its 1995 RfD for methylmercury to determine if a change was warranted.
Subsequently a workshop was organized by an interagency committee at the request of OSTP. The
organizing committee was chaired by the National Institute of Environmental Health Sciences (NIEHS)
and included representatives from several agencies:
Department of Health and Human Services (DHHS)
Office of the Assistant Secretary for Planning and Evaluation
Centers for Disease Control and Prevention (CDC)
Agency for Toxic Substances Disease Registry (ATSDR)
Food and Drug Administration (FDA)
Environmental Protection Agency (EPA)
4-10 Methylmercury Water Quality Criterion 1/3/01
-------
National Oceanic and Atmospheric Administration (NOAA)
Office of Science and Technology Policy (OSTP)
Office of Management and Budget (OMB)
The Methylmercury Workshop was a response to the suggestion that the emerging Seychellois and
Faroese data undergo a level of scrutiny beyond journal peer review if they were to be used in policy
setting.
The Workshop on the Scientific Issues Relevant to Assessment of Health Effects from Exposure to
Methylmercury was held in Raleigh, North Carolina, November 18-20, 1998. The purpose of the
workshop was to discuss and evaluate the major epidemiologic studies associating methylmercury
exposure with an array of developmental measures in children. The workshop did not attempt to derive a
risk assessment, but it was assumed by participants that the workshop evaluation would facilitate
agreement on risk assessment issues. The major studies considered were those that have examined
populations in Iraq, the Seychelles, the Faroe Islands, and the Amazon, along with the most relevant
animal studies. Study authors made detailed presentations to respond to a series of questions on study
exposures, potential confounders, measurements of effect, and other related topics. Five expert panels
discussed the presentations and published data; panels covered the following areas: exposure,
neurobehavioral endpoints, confounders and variables, design and statistics, and experimental (animal
and in vitro) data. The results of their deliberations were published in the Spring of 1999 (NIEHS,
1999). Conclusions of the report were reviewed by workshop panelists and by Federal scientists who had
attended the workshop. The conclusions are quoted below.
1. Methylmercury is a developmental neurotoxin, but effects at low doses encountered by eating fish are
difficult to evaluate.
2. All the studies reviewed were considered of high scientific quality, and the panel recognized that each of
the investigations had overcome significant obstacles to produce important scientific information. The panel
also stated that continued funding of the studies in the Seychelles, Faroes, and Amazon is necessary for the full
potential of those studies to be realized. This is particularly the case for the Faroes and Seychelles studies,
which have assessed and are currently assessing the potential developmental neurotoxic effects of
methylmercury in fish-eating populations. The developmental studies would benefit by evaluation of common
endpoints using similar analytical methods. It is important to note that the Amazon study did not assess
developmental endpoints but assessed effects in adults.
3. Results from the Faroes and Seychelles studies are credible and provide valuable insights into the
potential health effects of methylmercury.
4. Some differences are clearly present in results from the Faroes, Seychelles, and Amazon, but the panel
was not able to clearly identify the sources of these differences. Among possible sources are the different effects
of episodic versus continuous exposure, ethnic differences in methylmercury responses, lack of common
Methylmercury Water Quality Criterion 1/3/01
4-11
-------
endpoints in the Faroes and Seychelles studies, and several other confounders or modifying factors such as those
found in diet and lifestyle, as well as in chemicals present in seafood, which is the source of methylmercury to
these populations. The other chemical constituents of seafood that may be explanatory include those that may
be beneficial to fetal neurodevelopment (i.e., omega-3 fatty acids) and those that may be harmful to fetal
neurodevelopment (e.g., PCBs).
5. These studies have provided valuable new information on the potential health effects of methylmercury,
but significant uncertainties remain because of issues related to exposure, neurobehavioral endpoints,
confounders and statistics, and design.
The interagency organizing committee agreed unanimously that the deliberations of the panels and the
workshop report will be a key factor in subsequent public health policy actions taken by each of the
participating agencies.
4.1.5 National Academy of Sciences Review
Congress directed EPA, through the House Appropriations Report for FY99, to contract with the
National Research Council (NRC, a body of the National Academy of Sciences) to evaluate the body of
data on the health effects of methylmercury, with particular emphasis on new data since the publication
of the MSRC. NRC was asked to provide recommendations regarding issues relevant to the derivation of
an appropriate RfD for methylmercury.
The NRC empaneled a group of scientific experts who held public meetings at which there were
presentations from methylmercury researchers, government agencies, trade organizations, public interest
groups, and concerned citizens. The panel evaluated the scientific basis for risk assessments done by
EPA and other groups as well as new data and findings available since publication of the MSRC. The
committee was not charged with developing an RfD as an alternative to the EPA assessment, but rather
provided scientific guidance that would inform such an assessment. The NRC report, lexicological
Effects of Methylmercury, was released to the public on July 11, 2000 (NRC, 2000). Conclusions of that
report are summarized below.
The report concludes that methylmercury is a highly toxic substance; a number of adverse health
effects associated with methylmercury exposure have been identified in humans and in animal studies.
Most extensive are the data for neurotoxicity, particularly in developing organisms. The nervous system
is considered by the NRC committee to be the most sensitive target organ for which there are data
suitable for derivation of an RfD. The committee also concludes on the basis of data from humans and
from animal studies that exposure to methylmercury can have adverse effects on the developing and adult
cardiovascular system. They note that some research demonstrated adverse cardiovascular effects at or
4-12
Methylmercury Water Quality Criterion 1/3/01
-------
below levels associated with effects on the developing nervous system. The NRC also cites evidence of
low-dose methylmercury effects on the immune and reproductive systems.
The NRC report presents some conclusions on the public health implications of methylmercury
exposure; one conclusion is quoted below:
The committee's margin-of-exposure analysis based on estimates of MeHg exposure in the U.S. population
indicates that the risk of adverse effects from current MeHg exposure in the majority of the population is low.
However, individuals with high MeHg exposure from frequent fish consumption might have little or no margin
of safety (i.e., exposures of high-end consumers are close to those with observable adverse effects). The
population at highest risk is the children of women who consumed large amounts offish and seafood during
pregnancy. The committee concludes that the risk to that population is likely to be sufficient to result in an
increase in the number of children who have to struggle to keep up in school and who might require remedial
classes or special education. (NRC, 2000 p. 9)
The NRC report gives an evaluation of the 1995 EPA RfD. Their conclusion is as follows:
On the basis of its evaluation, the committee's consensus is that the value of EPA's current RfD for MeHg, 0.1
(ig/kg/day, is a scientifically justifiable level for the protection of public health. However, the committee
recommends that the Iraqi study no longer be used as the scientific basis of the RfD (NRC, 2000 p. 11).
The NRC report made several recommendations on the appropriate basis for a revised RfD. The
Committee thoroughly reviewed three epidemiological longitudinal developmental studies: the
Seychelles Islands, the Faroe Islands, and New Zealand. The Seychelles study yielded scant evidence of
impairment related to in utero methylmercury exposure through 5.5 years of age, whereas the other two
studies found dose-related effects on a number of neuropsychological endpoints. The Faroe Islands study
is the larger of the latter two studies and has been extensively peer-reviewed. NRC recommended use of
data from the Faroe Islands study for derivation of the RfD (NRC, 2000 p. 11).
NRC recommended BMD analysis as the most appropriate method of quantifying the dose-effect
relationship. They recommend the lower limit on a 5% effect level obtained by applying a K-power
model (K ;> 1) to dose-response data based on Hg in cord blood. NRC noted that for the Faroe Islands
data the results of the K-power model under this constraint are equivalent to a linear model (NRC, 2000,
pp. 11-12).
NRC recommended use of the Boston Naming Test (BNT) as the critical endpoint. This endpoint
yields the second-lowest BMDL but was judged by the Committee to be more reliable than the endpoint
Methylmercury Water Quality Criterion 1/3/01
4-13
-------
that yields the lowest BMDL. The BMDL for the BNT from the Faroe Islands study is 58 ppb Hg in cord
blood.
NRC described alternative dose conversion processes using a one-compartment model similar to
that used in the MSRC.
In their discussion of uncertainty factors, NRC reviewed several sources of variability and
uncertainty and recommended that an uncertainty factor of at least 10 be used. NRC recommended a
factor of 2 to 3 for biological variability in dose estimation. They also recommended an additional factor
to account for data gaps relating to possible long-term neurological effects not evident in childhood, as
well as possible effects on the immune and cardiovascular systems (NRC, 2000, p. 327).
4.1.6 External Peer Review of Draft RfD
A draft EPA RfD document was submitted for external scientific peer review in late October 2000;
the reviewers are listed at the front of this document. At the same time the draft RfD document was
ckculated for comment to other Federal Agencies through CENR and OSTP. A public scientific review
meeting was held November 15, 2000; the final peer review report was delivered to EPA on December 7,
2000, and is available in the docket. The external peer reviewers supported the use of the Faroes data,
derivation of a BMD as described by NRC, and application of a tenfold uncertainty factor to the BMDL.
They agreed with EPA's use of a one-compartment model for dose conversion as well as with most of the
parameter estimates; they commented correlation among some of the parameters. The peer reviewers
disagreed with NRC's recommendation to set the RfD on the BNT results from the full Faroese cohort.
They felt that the BNT scores showed an effect of concomitant PCB exposure in some analyses. They
preferred a PCB-adjusted BMDL of 71 ppb mercury in cord blood for the BNT. They also offered
suggested alternatives to use of the BNT test resits. The peer reviewers validated a final RfD of 0.1
jj.g/kg bw /day.
4.1.7 Revised RfD
The development of this RfD considered the NRC recommendations and followed them for the
most part. Most recommendations of the peer-review panel were incorporated as well. The following
sections provide rationales for choices made by EPA in determining the basis for the RfD.
4-14
Methylmercury Water Quality Criterion 1/3/01
-------
4.2 CHOICE OF CRITICAL STUDY AND ENDPOINT
NRC concluded, and EPA agrees, that the data from human studies showing developmental
neurotoxicity are the most appropriate basis for the RfD. NRC concluded that human studies on
methylmercury carcinogenicity are inconclusive and that the renal tumors observed in mice were found
only when animals were exposed at or above the maximally tolerated dose (MTD). In the MSRC, EPA
noted that if one applied the principles of the revisions to the Risk Assessment Guidelines for
Carcinogenicity, the following conclusions would be reached:
Methylmercury is not likely to be a human carcinogen under conditions of exposure generally encountered in
the environment. Data in humans were inadequate; interpretation is limited by inappropriate study design and
incomplete descriptions of methodology. Dietary exposure in two strains of mice resulted in increased renal
adenomas and adenocarcinomas. Tumors were observed only in dose groups experiencing profound
nephrotoxicity. Studies in rats exposed to an MTD showed no increased tumor incidence. Several studies
show that methylmercury can cause chromosomal damage in somatic cells. While evidence is good for
chromosomal effects, it does not appear that methylmercury is a point mutagen. The mode of action in renal
tumor induction is likely to be related to reparative changes in the tissues. Human exposure is likely to be
from consumption of contaminated foods, especially fish. It is expected that exposure, even in groups
consuming large amounts of fish from contaminated sources, will be to levels far below those likely to cause
the tissue damage associated with tumor formation in animals (U.S. EPA, 1997).
NRC concluded that human data, as well as results of animal tests, indicate the cardiovascular
system is a sensitive target for methylmercury effects. This is particularly true for developing organisms.
Their report also cites animal and in vitro data, linking methylmercury exposure to immunotoxic and
reproductive effects (summarized in NRC, 2000, pp. 190-191). It is clear, however, that at the current
time the human data set on developmental neurotoxicity is the most extensive, best reviewed, and most
thoroughly evaluated. The RfD will thus rely on those data. It is expected that an RfD based on
developmental neurotoxicity will be protective against adverse effects likely to occur at higher levels of
mercury exposure. Following NRC's recommendation, EPA's choice of critical study was limited to
those developmental studies of populations experiencing long-term, low-dose exposure. Only those
studies are summarized in subsequent sections of this document.
4.2.1 Summary of Available Data
This section gives brief summaries of studies on the developing central nervous system that were
described by NRC. This section follows the format used by the NRC report; studies are grouped into
subsections by endpoint and chronologically within subsection. Section 4.2.1.1 describes the evidence
for effects of methylmercury on neurological status; Section 4.2.1.2 describes the effects on attainment of
Methylmercury Water Quality Criterion 1/3/01
4-15
-------
developmental milestones during infancy; Section 4.2.1.3 describes other effects during infancy and early
childhood; Section 4.2.1.4 presents evidence for cognitive deficits during childhood (school age); and
Section 4.2.1.5 describes sensory and other effects of methylmercury.
For more detailed study descriptions refer to Section 3 of this document or to the MSRC.
4.2.1.1 Status on Neurological Examination
Cree Population—McKeown-Eyssen et al. (1983)
McKeown-Eyssen et al. (1983) studied a population of 234 12- to 30-month-old Cree Indian
children for whom prenatal methylmercury exposure was estimated on the basis of maternal hair samples.
The subjects lived in four communities in northern Quebec. Hair samples were collected on 28% of the
mothers during pregnancy; prenatal exposure for the rest of the cohort was estimated from hair segments
assumed to date from the time the study child was in utero. No child was judged to have any abnormal
physical findings. Overall, 3.5% (4) of the boys and 4.1% (5) of the girls were considered to have
abnormal neurological findings. The most frequent abnormality (observed in 11.4% [13] of the boys and
12.2% [14] of the girls) involved tendon reflexes. Abnormalities of muscle tone or reflexes in boys were
the only neurological finding for which there was a statistically significant association with prenatal
methylmercury exposure, either before or after adjustment for confounding. The risk of an abnormality
of tone or reflexes increased seven times with each 10 ppm increase in maternal hair mercury. When
exposure was categorized, the prevalence of tone or reflex abnormality did not increase in a clear dose-
response manner across categories. In girls, incoordination was negatively associated with prenatal
methylmercury exposure. The authors noted that these mild, isolated neurological findings were
different from those described in previous reports of neurological abnormalities after prenatal exposure
to higher levels of methylmercury.
Mancora, Peru—Marsh et al. (1995)
Neurological examination was done on 194 children in Mancora, Peru. Although the study was
conducted in the early 1980s, it was not published until 1995 (Marsh et al., 1995). Fish consumption was
the primary route of methylmercury exposure and maternal hair was used as the index of exposure
(geometric mean 7.05 ppm; range 0.9 to 28.5 ppm). Comparison of peak and mean hair-mercury
concentration suggested that the women's exposure was at steady state because of stability in their fish-
4-16 Methylmercury Water Quality Criterion 1/3/01
-------
consumption patterns. Maternal hair samples and data on child neurological status were available for
131 children. Several elements of the study design are not described: the size of the eligible population
from which the 131 children were sampled, the specific elements of the neurological assessment
conducted, and the ages at which the children were examined. Frequencies were reported for the
following endpoints: tone decreased, tone increased, limb weakness, reflexes decreased, reflexes
increased, Babinski's sign, primitive reflexes, and ataxia. No endpoint was significantly associated with
either mean or peak maternal hair mercury.
SCDS Pilot Study—Myers et al. (1995b)
In the cross-sectional or pilot study of the SCDS (Myers et al., 1995), 789 infants and children
between the ages of 5 and 109 weeks were evaluated by a pediatric neurologist. Mean maternal hair
mercury in the cohort was 6.1 ppm (range 0.6 to 36.4 ppm). The endpoints assessed were mental status,
attention, social interactions, vocalizations, behavior, coordination, postures and movements, cranial
nerves, muscle strength and tone, primitive and deep tendon reflexes, plantar responses, and age-
appropriate abilities such as rolling, sitting, pulling to stand, walking, and running. The statistical
analyses focused on three endpoints chosen on the basis of their apparent sensitivity to prenatal
methylmercury exposure in the Iraq and Cree studies: overall neurological examination, increased muscle
tone, and deep tendon reflexes in the extremities. There was no association between maternal hair
mercury and questionable and abnormal results. The frequency of those results ranged from 16.5% in the
group with hair mercury at 0 to 3 ppm to 11.7% in the group with Hair mercury at more than 12 ppm.
The frequencies of abnormalities of limb tone or deep tendon reflexes were about 8%; there was no dose-
dependent variation in frequency of either endpoint.
SCDS Main Study—Myers et al. (1995c)
The main cohort of the SCDS consisted of 779 mother-infant pairs, representing approximately
50% of all live births during the period of recruitment. The final sample size was 740. When the infants
were 6.5 months old, a pediatric neurologist administered essentially the same neurological examination
that had been used in the pilot phase; testing was blinded as to child's exposure. A total of 3.4% (25) of
the children had overall neurological scores considered abnormal or questionable; this frequency was too
low to permit statistical analysis of the overall neurological examination. The frequency of abnormalities
was 2% for both limb tone and abnormal deep tendon reflexes. Questionable limb tone was identified in
approximately 20% of the children, and questionable deep tendon reflexes in approximately 15%.
Methylmercury Water Quality Criterion 1/3/01
4-17
-------
Although such findings were not considered pathological, they were combined with abnormal findings
for statistical analyses. The frequency of abnormal and questionable findings for limb tone or deep
tendon reflexes was not significantly associated with maternal hair mercury concentrations.
Faroes Population—Dahl et al. (1996)
A functional neurological exam was part of a general physical examination administered to a
cohort of 7-year-old children from the Faroe Islands. Of 1,386 infants eligible at recruitment, cord-blood
and maternal hair samples were obtained from 1,022 singleton births (75%), and 917 children were
examined (66%) (Grandjean et al., 1992). The mean cord-blood concentration was 22.9 |ig/L; the mean
maternal hair mercury concentration was 4.3 ppm. The examination focused on motor coordination and
perceptual-motor performance (Dahl et al., 1996). Results were scored as automatic, questionable, or
poor. There was no association between cord-blood mercury and the number of tests on which a child's
performance was considered automatic or performed optimally. On the tests of reciprocal motor
coordination, simultaneous finger movement, and finger opposition, fewer than 60% of the children
achieved a score of automatic for optimal performance. On the finger opposition test, children with
questionable and poor performance (425 children) had a significantly higher mean cord-blood mercury
concentration than children with automatic performance (465 children) (23.9 versus 21.8 ng/L, p = 0.04)
(Grandjean et al., 1997).
Faroes Population—Steurwald et al. (2000)
A cohort of 182 singleton, full-term infants born in the Faroe Islands between 1994 and 1995 was
recruited. The cohort represented 64% of all births in the study area. Data were collected on maternal
hair mercury, cord whole-blood mercury, and cord serum mercury. A total of 15 maternal hair
measurements exceeded 10 ppm. Measurements were also taken of 18 pesticides or metabolites and 28
polychlorinated biphenyl (PCB) congeners in maternal serum. At 2 weeks of age infants were given a
neurological examination designed to assess functional abilities, reflexes and responses, and stability of
behavioral status during examination. Responses were categorized as optimal, questionable, or
suboptimal. The neurological optimality score (NOS) was the number of items rated as optimal out of a
total of 60. Two subscores were generated (muscle tone and reflexes) and a variety of thyroid-function
indices were also assessed. Maternal hair mercury concentrations were not significantly associated with
NOS score, but there was a significant inverse relationship between NOS scores and cord whole-blood
mercury. The mean mercury concentration was 20.4 |j,g/L (range 1.9 to 102 fxg/L). Based on NOS score,
4-18
Methylmercury Water Quality Criterion 1/3/01
-------
a tenfold increase in cord-blood mercury was associated with the equivalent of a 3-week reduction in
gestational age. Adjustments for total PCBs and fatty acid concentrations had no effect on results, and
selenium was not an effect modifier. Muscle-tone and reflexes subscores were not significantly
associated with any exposure biomarker.
Cordier and Garel (1999)
Cordier and Garel (1999) studied a cohort of Amerind children from a gold-mining area in French
Guiana. Median maternal hah- concentration was 6.6 ppm with a range of 2.9 to 17.8 ppm; 35% of
maternal hair mercury levels were greater than 10 ppm. Neurological examination included the
following: neuromotor examination of the upper an lower limbs, body axis, deep reflexes, and postural
reactions; neuromotor functions; neurosensory examination; and cranial growth. The authors report that
for children greater than 2 years of age, increased reflexes were found with greater incidence as a
function of maternal hair mercury; the effect was greater in boys than in girls. When 10 children were
retested 9 months later by a different examiner, only 3 were found to have the increased reflex response.
The authors commented that this poor reproducibility makes the reflex response difficult to interpret.
Conclusions
There is some evidence that neurological status in children is associated with low-dose in utero
exposure: (1) an increased incidence (not dose dependent) of tone or reflex anomalies in boys associated
with increased maternal hair mercury (McKeown-Eyssen et al., 1983); (2) an inverse association between
newborn neurological optimality score and cord-blood mercury in Faroese children (Steurwald et al.,
2000); (3) a statistically significant increase in the mean cord-blood mercury of 7-year-old Faroese
children who performed less than optimally on a finger opposition test, compared with Faroese children
with normal performance (Grandjean et al., 1997); (4) the association of increased reflexes with
increasing maternal hair mercury in a group of children aged 9 months to 6 years in French Guiana
(Cordier and Garel, 1999). NRC notes that a particular limitation of the use of neurological status is the
categorical nature of the response; in other words, the subject has either an abnormal response or a
normal response. This may have been a factor in the evaluation of results from the SCDS. The number
of abnormal responses in this population was very low; thus there was reduced statistical power for
hypothesis testing.
Methylmercury Water Quality Criterion 1/3/01
4-19
-------
4.2.1.2 Age at Achievement of Developmental Milestones
SCDS—Myers et al. (1997) andAxtell et al. (1998)
The association between achievement of developmental milestones and prenatal methylmercury
exposure was evaluated in the main cohort of the SCDS (Myers et al., 1997). Data were available for
738 of the 779 children enrolled. The mean average age for walking was 10.7 months for girls and 10.6
months for boys; for talking it was 10.5 months for girls and 11.0 months for boys. The mean age at
which a child was considered to talk was not significantly associated with maternal hair mercury in any
of the regression models used. In regressions stratified by child sex, a positive association was found
between age at walking and maternal hair mercury in boys only. The interaction between mercury and
sex was not statistically significant in the analyses of the complete cohort. The authors considered the
magnitude of the delay in boys' walking to be clinically insignificant; a 10-ppm increase in maternal hair
mercury was associated with approximately a 2-week delay. This association in boys was not significant
when four statistical outliers were excluded from the analysis. Authors concluded that hockey-stick
models provided no evidence of a threshold for developmental delay, as the fitted curves were essentially
flat.
Axtell et al. (1998) reanalyzed the milestone data, applying semiparametric generalized additive
models that are less restrictive than the approaches used by Myers et al. (1997). Their major finding was
that the association between age at walking and maternal hair mercury in boys was nonlinear. In their
modeled estimates, walking was delayed as maternal hair concentrations increased from 0 to 7 ppm but
was observed at a slightly earlier age as mercury concentration increased beyond 7 ppm. The size of the
effect associated with the increase from 0 to 7 ppm was very small, corresponding to a delay of less than
1 day in the achievement of walking. Because of the contradictory nature of the dose-response
relationships above and below 7 ppm, the authors expressed a doubt that the association found below 7
ppm reflected a causal effect of mercury exposure on age at walking.
Mancora, Peru—Marsh et al. (1995)
Data on developmental milestones were collected in the Peruvian study conducted by Marsh et al.
(1995). The study was conducted prospectively, and data were apparently collected in an ongoing
manner over the course of a mother's visits to a postnatal clinic. Regression analyses, including analyses
stratified by child sex, did not reveal any significant associations between maternal hair mercury
4-20 Methylmercury Water Quality Criterion 1/3/01
-------
concentrations and the ages at which children sat, stood, walked, or talked. The rates of developmental
retardation, especially in speech (13 of 131), were substantial. Children's birthweight, height, and head
circumference were unrelated to maternal hair mercury concentrations.
Faroes Population—Grandjean et al. (1995)
Ages at achievement of motor development milestones were investigated in a 21-month birth cohort
(1,022 infants born in 1986-1987) of children in the Faroe Islands. Complete data were available for 583
children. Three motor-development milestones commonly achieved between 5 and 12 months of age
were selected for analysis: "sits without support," "creeps," and "gets up into standing position with
support." There was no significant association between age at achievement and either cord-blood or
maternal hair mercury for any of the three milestones. For all three, however, the authors reported a
significant inverse association between age at achievement and the child's hair mercury concentration at
12 months. Children's hair mercury was interpreted as an index of postnatal exposure to methylmercury.
Breastfeeding was associated with both increased hair mercury concentrations and more rapid
achievement of milestones. Therefore, the authors concluded that the inverse association reflected
residual confounding by duration of breastfeeding.
Conclusions
The recent human studies provide little evidence of an association between maternal hair mercury
below 30 ppm and delayed developmental milestones. The NRC report noted that in the SCDS, mean
age of walking was higher in the part of the population born to mothers with higher hair mercury. The
association was for male children only and it was not dose related. In the Faroese population, there was a
negative association for maternal hair mercury and three developmental milestones. The study authors
attributed this to higher mercury exposure in the breastfed population and the salutary effect of breast
milk on development. The NRC report commented on the reported developmental delays in the Iraqi
population, which has been the subject of much discussion as to the degree of uncertainty in the estimates
(see also MSRC Volumes V and VII). NRC cites analyses by Cox et al. (1995) and Crump et al. (1995),
which indicate that the earlier estimates of the Iraqi threshold for late walking were too low. The
threshold for late walking appears highly dependent on assumptions on background incidence, the
definition of delayed walking, and the effect of a small number of influential data points.
Methylmercury Water Quality Criterion 1/3/01
4-21
-------
4.2.1.3 Infant and Preschool Development
Cree population—McKeown-Eyssen et al. (1983)
In the study of a Cree population, the Denver Developmental Screening Test (DDST) was
administered to the 12- to 30-month-old children in the cohort (n = 234). Scores were reported as the
percentage of items passed on each subscale as well as on the entire test. The authors did not provide
estimates of significance of association between test scores and maternal hair mercury concentrations;
they concluded that there was no significant association indicative of an adverse effect of methylmercury
before or after adjustment for confounding variables.
Ne\v Zealand population—Kjellstrom et al. (1986)
Kjellstrom et al. (1986) studied a cohort of New Zealand children for whom prenatal
methylmercury exposure was estimated on the basis of maternal hair samples as well as dietary
questionnaires collected during the period when the study child was in utero. Exposure information was
collected on nearly 11,000 women; the study focused on 935 women who reported eating fish more than
three times per week during pregnancy. Seventy-three women had hair mercury concentrations greater
than 6 ppm. The 74 children of those women were designated as the high-mercury group. Efforts were
made to match each child in the high-mercury group with a reference child on the basis of maternal
ethnicity, hospital of birth, maternal age, and child age. In the followup evaluations at 4 years of age, a
total of 38 exposed and 36 reference children were tested; this data set included 30 completely matched
pairs. Fifty-two percent of the children in the high-mercury group had an abnormal or questionable
DDST score compared with 17% of the children in the control group (p < 0.05). That result corresponds
to an odds ratio of 5.3. Results were similar when pairs that were poorly matched on ethnicity were
excluded.
SCDS pilot study—Myers et al. (1995b)
In the SCDS cross-sectional study, a revised version (DDST-R) of the DDST was administered to
789 children between the ages of 1 and 25 months. No association was found between maternal hair
mercury concentration during pregnancy (mean 6.6 ppm) and DDST-R results when normal and
questionable examinations were combined. The prevalence of abnormal findings was so low (three
children <1%) that the statistical analysis was not meaningful. When abnormal and questionable results
4-22
Methylmercury Water Quality Criterion 1/3/01
-------
were grouped (in 65 children, 8%), high maternal hair mercury concentrations were significantly
associated with poor outcomes (p = 0.04, one-tailed test). That result was largely attributable to the
higher frequency of abnormal and questionable results among children in the highest maternal hair
mercury category (greater than 12 ppm), by contrast to the frequency of approximately 7% among
children in each of the other four groups (0-3, 3-6, 6-9, and 9-12 ppm).
SCDS main study—Myers et al. (1995c)
In the main SCDS study, the DDST-R was administered to a cohort of 740 children at age 6.5
months. The frequency of examinations considered to be abnormal or questionable was very low,
precluding meaningful statistical analysis of the DDST-R data. The researchers also administered the
Fagan Test of Infant Intelligence, an assessment of visual-recognition memory or novelty preference.
Results were not related to maternal hair mercury concentrations.
SCDS main cohort at 19 and 29 months—Davidson et al. (1995)
The Bayley Scales of Infant Development (BSID) were administered to children in the SCDS
cohort at ages 19 and 29 months. In addition, at 29 months, six items of the Infant Behavior Record, a
rating scale, were completed by the examiner. There are two primary scores on the BSID: the mental
development index (MDI) and psychomotor development index (PDI). At both ages, MDI scores were
similar to the expected mean for U.S. children. At both ages, however, the Seychellois children
performed markedly better on PDI than the expected mean for U.S. children. There was no association
between MDI scores at 19 or 29 months with maternal hair mercury concentration during pregnancy.
Similar results were obtained in a secondary analysis that included only children with the lowest or
highest maternal hair mercury concentrations. Assessments of perceptual skills at 19 months were not
associated with mercury exposure. Scores on that test at 29 months could not be evaluated because of a
pronounced ceiling effect; that is, there were so many high scores on the test that no difference would be
detectable. Likelihood of a PDI score below the median was not significantly associated with maternal
hair mercury concentration in the full logistic regression model, but was associated with this exposure
index in a model that included limited covariates.
Methylmercury Water Quality Criterion 1/3/01
4-23
-------
Conclusions
There is some indication of low-dose mercury effects in very young children, but there are
difficulties in the measurement of such effects. The DDST was administered to four study populations.
When abnormal and questionable results were combined, there was a significant association with
increasing maternal hair mercury in the New Zealand cohort and in the SCDS cross-sectional study (but
not the main study). The NRC report comments on the bases for the different findings: age at
examination, different rates of abnormal and questionable scores, and the possibility that test items or
criteria for judging scores differed among studies. NRC offered the general conclusion that screening
tests such as the DDST are not useful in neurobehavioral toxicology studies; such tests are insufficiently
sensitive to variations hi the range of normal performance (NRC 2000, p. 200).
The NRC panel noted that the BSID is currently considered to be the best available instrument for
infant assessment and is useful for measurement of prenatal exposures to neurotoxicants (NRC 2000, p.
200). In the SCDS main study there was no significant association between young children's scores on
the BSID and maternal hair mercury. At 19 and 29 months, the Seychellois children scored higher than
the means for U.S. children on the PDI portion of the scales.
4.2.1.4 Childhood Development
New Zealand population—Kjellstrom et al. (1989)
Children in the New Zealand cohort were followed up at 6 years of age. Children were given a
battery of 26 psychological tests, tests of scholastic aptitude, and behavioral tests. The following
domains were assessed: general intelligence, language development, fine and gross motor coordination,
academic attainment, and social adjustment. Maternal hair mercury concentration was associated with
poorer scores on full-scale IQ tests (Wechsler Intelligence Scale for Children, Revised [WISC-R]),
language development (Test of Language Development, spoken language quotient), and visual-spatial
and gross-motor skills (McCarthy Scales of Children's Abilities). Multiple regression analyses were
done on these endpoints: Test of Language Development, spoken language quotient (TOLD-SL); WISC-
R, performance IQ; WISC-R full-scale IQ; McCarthy Scales, perceptual performance; and McCarthy
Scales, motor scales. Covariates in the regressions were these: maternal ethnic group, maternal age,
maternal smoking and alcohol use during pregnancy, length of maternal residence in New Zealand, social
class, primary language, siblings, sex, birthweight, fetal maturity, Apgar score, and duration of
4-24
Methylmercury Water Quality Criterion 1/3/01
-------
breastfeeding. Observations were weighted in the regression to deal with outliers. In the analyses there
were statistically significant associations between maternal hair mercury and poorer scores on the
following measures: full-scale IQ; language development (spoken language quotient), visual-spatial
skills (perceptual-performance scale), and gross motor skills (motor scale). The poorer mean scores of
the children in the high-mercury group were largely attributable to children of mothers with mercury
concentrations above 10 ppm. In this group, mean average hair mercury was 13 to 15 ppm and mean
peak was 25 ppm. Maternal hair mercury concentrations accounted for relatively small amounts of
variance in the outcome measures and generally accounted for less than covariates such as social class
and ethnic group.
In the original analyses of five test scores (Kjellstrom et al., 1986), hair mercury was used in
regression analyses as a binary variable; that is, either >6 ppm or between 3 and 6 ppm. Analyses found
an association between high prenatal mercury exposure and decreased test performance. Later regression
analyses by Crump et al. (1998), which used maternal hair mercury level as a continuous variable, did not
find significant associations between mercury and children's test scores. However, this finding was
highly influenced by a single child whose mother's mercury hair level (86 ppm) was more than four
times that of any other. When this child was excluded, there were significant associations between hair
mercury and TOLD-SL and MC-PP scores. When regression analyses were done on scores from all 26
scholastic and psychological tests, and the data on the influential point were omitted, scores on six tests
were significantly associated with mothers' hair mercury: Clay Reading Test-concepts, Clay Reading
Test-letter test, McCarthy Scales-general cognitive index, McCarthy Scales-perceptual-performance
scale, Test of Language Development-grammar completion, and Test of Language Development-
grammar understanding.
SCDS pilot study— Myers et al. (1995a), Davidson et al. (2000), Davidson et al. (1998), Myers et al.
(2000).
A portion of the pilot cohort of 789 children were given developmental assessments; these were
children who were 66 months old within a 1-year testing window (Myers et al., 1995a). Of the 247
eligible children, 217 were administered a test battery consisting of the McCarthy Scales of Children's
Abilities, the Preschool Language Scale, and two subtests of the Woodcock-Johnson Tests of
Achievement (letter-word identification and applied problems). The median maternal hair mercury
concentration in that subsample of the pilot cohort was 7.1 ppm. Maternal hair mercury was associated
with significantly lower general cognitive index (GCI) scores on the McCarthy scales. Scores declined
Methylmercury Water Quality Criterion 1/3/01
4-25
-------
approximately five points between the lowest and highest exposure categories. Similar associations were
found on the perceptual-performance scale of the McCarthy scales and on the auditory comprehension
scale of the Preschool Language Scale. Scores declined approximately 2.5 points across the range of
maternal hair mercury concentrations. When outliers and influential points were removed from the
regressions the statistical significance of the associations was lost for all except auditory comprehension
(Preschool Language Scale Auditory Comprehension subscale). In the pilot phase of the SCDS,
information was not collected on several key variables that frequently confound the association between
neurotoxicant exposures and child development. Those variables are socioeconomic status, caregiver
intelligence, and quality of the home environment.
Further evaluation was performed on a portion of the Seychelles pilot cohort at 108 months of age
(Davidson et al., 2000). Eighty-seven children were tested on five subtests of the WlSC-in (Information,
Block Design, Vocabulary, Digit Span, and Coding), CVLT, BNT, Beery-Buktenica Development Test
of Visual Motor Integration (VMT) (copying geometric figures), finger tapping, grooved pegboard,
Trailmaking (tracing the correct route through a form with a pencil), and the design memory subtest of
the Wide Range Assessment of Memory and Learning (WRAML) (drawing each of four geometric
designs from memory). Performance on BNT, VMI, and grooved pegboard showed a positive
association related to mercury exposure in males, whereas there were trends toward poorer performance
related to mercury exposure for grooved pegboard in females (p = 0.07). Given the small number of
subjects, the power of the study was probably quite low; these largely negative results should be
interpreted with caution.
No effect of mercury was identified on the Child Behavior Check List (CBCL) at 66 months of age
in the main cohort of Seychelles study as determined by the total T score (Davidson et al., 1998). The
CBCL is a report inventory scored by the caregiver that assesses eight domains: withdrawn, somatic
complaints, anxious/depressed, social problems, thought problems, attention problems, delinquent
behavior, and aggressive behavior. An analysis of these subscales was performed on the 711 children
assessed on this test (Myers et al., 2000). No effect of mercury was identified on individual subscales.
SCDS Main Study—Davidson et al. (1998), Axtell et al. (2000), Palumbo et al. (2000)
As part of the main SCDS, 711 children 66 months of age (from the original cohort of 779) were
evaluated with a battery of standardized neurodevelopmental tests. At this evaluation, mercury was
measured in a 1-cm segment of the child's hair as an indicator of postnatal exposure. The following were
assessed: general cognitive ability (McCarthy Scales of Children's Abilities), expressive and receptive
4-26 Methylmercury Water Quality Criterion 1/3/01
-------
language (Preschool Language Scale, PLS), reading achievement (letter-word recognition subtest of the
Woodcock-Johnson Tests of Achievement), arithmetic (applied problems subtest of the Woodcock-
Johnson Tests of Achievement), visual-spatial ability (Bender Gestalt Test), and social and adaptive
behavior (CBCL). The scores of the six primary endpoints indicated no adverse effect of either prenatal
or postnatal mercury exposure. The only significant associations were consistent with enhanced
performance among children with increased exposure to methylmercury. Increased pre- and postnatal
mercury concentrations were significantly associated with better scores on the total score of the
Preschool Language Scale. For the applied problem test, increased postnatal mercury concentrations
were associated with better scores. Among boys, increased postnatal mercury concentrations were
associated with fewer errors on the Bender Gestalt Test.
The investigators published additional analyses of the 66-month data evaluating the possibility of
non-linear relationships associated with mercury exposure (Axtell et al., 2000). Endpoints included the
six primary variables analyzed previously: McCarthy GCI, PLS, Woodcock-Johnson (WJ) applied
problems, WJ letter/word recognition, Bender copying errors, and CBCL total T score. Generalized
additive models, which make no assumptions about the relationship between exposure and test score,
were used. Nonlinearities were identified between prenatal exposure and PLS and CBCL, and between
postnatal exposure and McCarthy GCI. For the PLS the trend involved a decrement of 0.8 points (poorer
performance) from 0 to 10 ppm and an increase of 1.3 points above 10 ppm. For the CBCL there was an
increase (representing a poorer score) between 0 and 15 ppm and a decrease above 10 ppm. The GCI
increased (improved) by 1.8 points through 10 ppm in the child's hair and declined by 3.1 above 10 ppm.
Although these results are difficult to interpret, they provide limited evidence of an adverse effect of
mercury exposure below 10 ppm maternal hair on two measures, and are associated with child's hair
mercury concentration above 10 ppm on the GCI. As pointed out by the authors, there are fewer data
points above 10 ppm (this is especially true for child's hair mercury), and therefore trends above this
level are estimated less precisely.
The SCDS investigators used multiple linear regression to assess the results of the McCarthy GCI
administered at 66 months (Palumbo et al., 2000). They analyzed the standard MSCA subscales and also
constructed specific subscales to approximate the domains of cognitive functioning assessed in the Faroe
Islands study: attention, executive function, expressive language, receptive language, nonverbal memory,
visuospatial, and gross motor visuomotor development. They found a positive association between the
child's hair mercury at 66 months and the standard memory subscale, with no other associations
identified. As with all the previous analyses of these variables, the raw scores were converted to
"normative" scores. As pointed out by the OSTP panel (NIEHS, 1999, Section 3.5 of the Confounders
Methylmercury Water Quality Criterion 1/3/01
4-27
-------
and Variables Section), the applicability of U.S. norms to this population is unclear, and the use of
standardized scores may decrease sensitivity by collapsing different raw scores to one standard score.
Faroes Population—Grandjean et al. (1997)
Testing was done at approximately 7 years of age on 917 of the surviving members of a 1986-1987
birth cohort of 1,022 singleton births. Maternal hair was sampled at parturition (geometric mean 4.3
ppm); children's hair mercury was measured at 12 months (geometric mean =1.1 ppm) and 7 years of
age (geometric mean = 3.0 ppm). Mercury was also measured in cord blood. The neuropsychological
tests were these: computer-administered tests from the Neurobehavioral Evaluation System (NES) (finger
tapping, hand-eye coordination, and continuous performance test); Tactual Performance Test; three
subtests of the WISC-R (digit span, similarities, and block design); Bender Gestalt Test; CVLT; the
BNT; and Nonverbal Analogue Profile of Mood States. Not all children could complete the entire
battery; this was associated with increased mercury exposure for some tests such as the finger opposition
test and mood test.
In multiple-regression analyses, increased cord-blood mercury concentration was significantly
associated with worse scores on finger tapping, continuous performance test (CPT) (in the first year of
data collection), WISC-R digit span, BNT, and CVLT. The investigators estimated that a tenfold
increase in cord mercury concentration was associated with delays of 4 to 7 months in those
neuropsychological domains. The maternal hair mercury concentration showed regression coefficients
that were generally lower than those obtained with cord-blood mercury as the exposure indicator. For the
finger tapping test, maternal hair mercury was a better predictor of effect, especially for the both-hands
condition. The child's hair mercury measured at 12 months was a significant predictor for finger tapping
with both hands and CPT reaction time; by contrast, hair mercury at the time of examination was
significantly associated with continuous performance test reaction time, block designs, and Bender
Visual Motor Gestalt errors.
When the Peters-Belson method for covariate adjustment was used, two additional endpoints
(WISC-R block design, Bender Gestalt Test errors) were found to be associated with mercury exposure.
Associations remained significant when the part of the cohort with maternal hair mercury concentrations
greater than 10 ppm was excluded from the analyses. A term for the interaction between mercury and
sex was not statistically significant, indicating that the effects were similar among boys and girls. In
general, children's test scores were more strongly associated with cord-blood mercury concentration than
4-28
Methylmercury Water Quality Criterion 1/3/01
-------
with either maternal hair mercury concentration or mercury concentrations in samples of children's hair
collected at 1 and 7 years of age.
Grandjean et al. (1998) also analyzed the Faroese data in a case-control fashion. Two groups were
assembled: a case group of 112 children with maternal hair concentrations of 10 to 20 ppm at parturition,
and a control group of 272 children with maternal hair mercury concentrations less than 3 ppm. Controls
were matched to cases on age, sex, year of examination, and caregiver intelligence. The median maternal
hair mercury concentrations in the two groups were 1.8 and 12.5 |ig/g, constituting a sevenfold
difference. Median cord-blood mercury concentrations also differed substantially (59.0 \igfL in the case
group versus 11.9 [ig/L in the control group). On 6 of the 18 endpoints, the case group scored
significantly lower than did the control group. The results of those analyses differ hi certain respects
from those of the main analyses. First, the set of endpoints on which the cases and controls differed is
similar but not identical to the set of endpoints that was significantly associated with cord blood mercury
concentration found in the main analyses. In the case-control analyses, a term for the interaction between
mercury and sex was statistically significant for several scores: the Bender Gestalt Test error score, short-
term reproduction on the CVLT, all three finger tapping conditions, CjPT reaction time, and average
hand-eye coordination score. For all scores, adverse mercury effects were noted for boys but not girls.
Amazon Valley—Grandjean et al. (1999)
A study cohort was assembled numbering 351 children ages 7 to 12. The population, which was
drawn from four riverine communities in Amazonian Brazil, had increased exposures to methylmercury
because of their consumption offish contaminated by upstream gold-mining activities. When data on all
four villages were combined, children's hair mercury concentrations were significantly associated with
their scores on finger tapping, Santa Ana dexterity test, WISC-M digit span, Stanford-Binet copying and
recall, and Stanford-Binet bead memory. Adjustment for community generally reduced the magnitude of
the associations, sometimes dramatically. It was noted that hair mercury concentrations and village
residence were so highly confounded, however, that adjustment for village might be inappropriate.
French Guiana population—Cordier and Garel (1999)
Cordier and Garel (1999) studied a cohort from a gold-mining area in French Guiana. Median
maternal hair concentration was 6.6 ppm with a range of 2.9 to 17.8 ppm. Children ages 5 to 12 years old
(n = 206) were administered a battery of neuropsychological tests: finger tapping, three subtests from the
Stanford-Binet (block design, copying designs, bead memory), and two subtests from the McCarthy
Methylmercury Water Quality Criterion 1/3/01 4-29
-------
scales (numerical memory, leg coordination). After adjustment for potential cofounders, increased
maternal hair mercury concentrations were significantly associated with copying-design score; the effect
was greater in boys. The data were reanalyzed to include only those observations from the region with
highest mercury exposures (Upper Maroni). When observations were separated by gender, there was an
association in boys between mercury exposure and poorer leg coordination, and with poorer block-design
scores in girls.
Conclusions
There is ample evidence of low-dose in utero mercury effects on neuropsychological indices in
school-age children. In the New Zealand population, maternal hair mercury was associated with poorer
scores on several measures: full-scale IQ, language development (spoken language quotient), visual-
spatial skills (perceptual-performance scale), and gross motor skills (motor scale). The poorer mean
scores in the high-mercury group were largely attributable to the children of mothers with hair mercury
above 10 ppm. One analysis by Crump et al. (1998) used maternal hair mercury as a continuous, rather
than binary, variable; in this analysis there was no significant association with hair mercury. These
analyses were heavily influenced by a single data point (a child with purported high developmental
exposure who showed no abnormal scores). If data for this child are excluded, and parental education
and age at testing are included as covariates, there are significant associates between mercury exposure
and six scores.
In the SCDS pilot (cross-sectional) study, increasing maternal hair mercury was associated with the
GCI and the perceptual performance scale of the McCarthy scales. Exclusion from analyses of several
influential points reduced the significance of the mercury effect. As it was intended as a feasibility
study, the pilot SCDS did not collect information on socioeconomic status, caregiver intelligence, or
quality of home environment. In the SCDS main study there was no observation of any adverse effect of
prenatal or postnatal mercury exposure. The NRC report commented on the regression model for the
GCI score:
The R2 (square of the multiple correlation coefficient) value (0.10) of the reduced regression model for the
GCI score in the main SCDS study was identical to that in the pilot study. That also appeared to be true for
scores on the Preschool Language Scale.... That finding is puzzling because the pilot-study models—did not
include several key covariates...and because the regression coefficients for socioeconomic status and caregiver
intelligence were statistically significant for total scores of the GCI and Preschool Language Scale in the main
study cohort. Those differences suggest that maternal hair Hg concentration is very highly confounded with
4-30
Methylmercury Water Quality Criterion 1/3/01
-------
those key covariates in the Seychelles population, or they suggest that the associations between child
neurodevelopment and the covariates differ substantially in the pilot and main study cohorts, or both (NRC
2000, pp. 203, 205).
In the Faroes population, mercury exposure measured in cord blood was associated with deficits on
several measures: finger tapping, preferred hand; CPT (first year of data collection, two scores); mean
reaction time, WISC-R digit span; BNT (with and without cues); and CVLT (short-term and long-term
reproduction). The mercury effect was similar in males and females. Most test scores were more
strongly associated with cord-blood mercury than with maternal hair mercury. In the case-control
analysis, the case group scored significantly lower than the control group on 6 of 18 endpoints.
In two smaller populations there were observed effects of mercury exposure. Combining results
from four communities in the Amazon basin showed a significant association of children's hair mercury
with deficits on four measures. In a French Guiana cohort (n = 206), it was shown that maternal hair
mercury was associated with one measure (a Stanford-Binet subtest), particularly in boys.
4.2.1.5 Sensory, Neurophysiological, and Other Endpoints in Children
Faroes population—Grandjean et al. (1997)
In the Faroe Islands cohort, the evaluation of 7-year-old children included assessments of visual
acuity, near-contrast sensitivity, otoscopy and tympanometry, and some neurophysiological tests. Visual
acuity, contrast sensitivity, auditory thresholds, and visual-evoked potentials were not significantly
associated with prenatal methylmercury exposures. For brainstem auditory-evoked potential, peaks I, IE,
and V were slightly delayed at increased cord-blood mercury concentrations at both 20 and 40 Hz;
interpeak latencies were not associated with mercury at either frequency.
Madeira population—Murata et al. (1999B)
Many of the same neurophysiological tests that had been done in the Faroe Islands study were
administered to 6- to 7-year-old children living in Madeira. This was a cross-sectional study of 149
subjects. For brainstem auditory-evoked potential, maternal hair mercury was significantly associated
with I-m and I-V interpeak latencies at both 20 and 40 Hz, as well as with total latencies for peaks HI and
V at both frequencies. Those results are similar to the findings in the children tested in the first year of
the Faroes cohort. For visual-evoked potentials on a pattern-reversal task, maternal hair mercury
Methylmercury Water Quality Criterion 1/3/01
4-31
-------
concentration was significantly associated with one of the three latencies, as well as with the N75-N145
and P100-N145 latencies.
Ecuador—Counter et al. (1998)
Auditory function in children and adults was investigated by Counter et al. (1998). The study
sample consisted of 75 individuals (36 children and 39 adults) from a gold-mining region in Ecuador and
34 individuals (15 children and 19 adults) from nonmining areas as a control. Blood mercury
concentrations were significantly higher in individuals (both adults and children) from the gold-mining
area than in individuals from the control region (mean level of 17.5 \igfL versus 3.0 |ig/L). Neurological
examinations were carried out on all individuals. In children, blood mercury was significantly associated
with hearing threshold at 3 kHz in the right ear only. No association was found for adults. A borderline
association was found between blood mercury concentration and I-IH interpeak transmission time on the
left side in both children and adults. The authors concluded that overall auditory sensory-neural function
and neural conduction time at the brainstem level were generally unaffected by elevated blood mercury
levels in either children or adults.
Conclusions
There is increasing evidence of adverse endpoints other than cognitive development in mercury-
exposed children. In the Faroes cohort, there were delays in some auditory-evoked potential peaks as a
function of cord-blood mercury. Similar findings were reported for a smaller population from a fishing
village in Madeira. A population of children in a gold-mining region of Ecuador showed an association
between blood mercury and hearing threshold in the right ear at 3 kHz.
4.2.2 Choice of Study
Of the three large human developmental studies, two reported associations between low-dose in
utero exposure to methylmercury and performance on standardized neurobehavioral tests. The Faroes
investigators reported effects in the domains of attention, fine-motor function, confrontational naming,
visual-spatial abilities, and verbal memory. Although similar results were reported for the New Zealand
population (and in the Seychelles pilot study), there were no observations of adverse effects attributable
to methylmercury in the main SCDS.
4-32
Methylmercury Water Quality Criterion 1/3/01
-------
This section discusses issues relevant to the choice of critical study for calculation of a reference
dose from among these three studies.
4.2.2.1 Critique of New Zealand Study
The study by Kjellstrom et al. (1986) included 57 fully matched groups of four 6-year-old children
each as well as four incomplete sets, for a total of 237. As was the case for the Faroes study, these
authors reported deficits in measures associated with methylmercury exposure. NRC noted (NRC, 2000
p. 251) that the New Zealand population's sources of methylmercury exposure and the study endpoints
were similar to those examined in the Seychelles. While EPA was developing its RfD for the MSRC, the
New Zealand data were available as a report that had not been subjected to standard peer-review
procedures. In 1998, Crump and associates published a reanalysis of the New Zealand data that was peer
reviewed. This paper reported associations of prenatal methylmercury exposure with several endpoints
(when one extreme outlier was excluded), including four endpoints that were not found to be related to
methylmercury in the Seychelles study. The New Zealand study has been criticized for errors in
matching exposed children to controls and for testing exposed children and controls at different ages
(Myers et al., 1998). Those errors occurred in the 4-year followup but were corrected in the 6-year
followup. NRC notes (NRC, 2000, p. 209) that there is no reason to expect differential measurement
error across the studies. An error of that type is likely to be nondifferential (i.e., unbiased), and it would
reduce the likelihood of detecting associations between methylmercury exposure and neurobehavioral
test scores.
The Kjellstrom et al. (1986) study collected data on several potential confounding factors and used
a broad battery of standardized measures that were administered by trained examiners. It is likely that
the exposure was relatively low-dose and not episodic, reflecting well-established food consumption
patterns. The section below discussed controls for possible confounders in the SCDS and Faroes studies.
An important variable is the concomitant exposure to organochlorine compounds such as PCBs and
pesticides that could have neurotoxic effects. There is essentially no information on the extent of such
exposures in the New Zealand study population, either in the original report or in follow-up analyses (e.g.
Crump, 1998).
4.2.2.2 Control for Possible Confounding
Both the Faroes study and the SCDS evaluated most of the variables that have been linked to
childhood cognitive development. Table 6-2 of the NRC report lists these and notes which study
Methylmercury Water Quality Criterion 1/3/01 4-33
-------
controlled for the particular variable. Although neither study controlled for all potential confounders, it
was felt by the authors of the NRC report that the influences of those variables on cognitive outcome are
probably too weak to account for any major inconsistencies between the two studies. The Confounders
and Variables Panel of expert workshops sponsored by OSTP had earlier concluded that neither the
SCDS nor the Faroese study was critically flawed and that these studies were suitable for determination
of the upper limit of a methylmercury NOAEL (NIEHS, 1999).
Place of Faroese residence—town versus country
At the 1998 OSTP workshop, the Faroes investigators noted that the maternal Ravens scores and
the child verbal-test scores were generally higher among families residing in one of the three towns in the
Faroes compared with those living in the countryside (NIEHS, 1999). This was thought to be due to
social-class differences. It was suggested that because more fish and, in particular, whale meat was
consumed by rural residents, the associations of mercury exposure with child verbal-test scores could in
fact reflect those social-class differences. However, analyses presented at the workshop showed that
these associations remained significant even after controlling for a dichotomous town-country control
variable (Table 6-3 in the NRC report). NRC felt it would not be appropriate to control for town
residence in all analyses. They made the following statement:
Because fish and whale consumption constitute a large proportion of the rural diet, the disappearance of
associations after controlling for residence could be due to the fact that residing in a rural area leads to
increased Hg exposure which, in turn, causes an adverse outcome. It would not necessarily indicate that the
lower social class associated with rural residence is the true cause of the Hg-associated deficit. The
disappearance of an association between Hg and neurobehavioral effects under those circumstances would be
very difficult to interpret, because the interpretation would depend upon what condition is considered the
reason for the association between living in a rural area and poor outcome (i.e., lower social class or greater
Hg exposure) (NRC, 2000, p. 261).
Another source of town versus country difference could be the distance traveled to the testing site,
with resulting fatigue in the children from the countryside. However, analyses showed that the
regression coefficients for prenatal mercury exposure remain significant even after controlling for child's
residence.
4-34
Methylmercury Water Quality Criterion 1/3/01
-------
Test administration
The neuropsychological test examiner was routinely controlled for in the Faroe Islands study (see
NIEHS, 1999, Section 3.5), but not the SCDS. It was suggested at the OSTP workshop that if an
examiner who is less adept at eliciting optimal performance from the subjects tested a large proportion of
less-exposed children, the results could be affected (NIEHS, 1999). NRC noted:
If those children performed more poorly than they otherwise would have on the test, an association between
Hg concentration and test scores might be obscured by failure to control for the examiner. That result could
also occur if an adept tester tested a large proportion of the more heavily exposed children, leading them to
achieve higher scores than they would have if tested by other examiners (NRC, 2000 p. 263 ).
Age at testing
The SCDS controlled for age at testing by converting the raw test scores to age-corrected standard
scores with conversion tables based on U.S. norms (NIEHS, 1999). The Faroes investigators analyzed
the raw scores by adjusting statistically for the child's age (measured hi days since birth). NRC found
the latter approach to be preferable (NRC, 2000, p. 263). They noted, first, that the applicability of U.S.
norms to these study populations is uncertain. In this context it should be noted that the Seychellois
scores on the BSID were higher than U.S. averages at both 19 and 29 months. Second, NRC felt that the
use of age-corrected standard scores could reduce the sensitivity of the test, because several adjacent raw
scores are treated as equivalent in converting to standard scores. Last, they noted that age-corrected
standard scores use 3-month intervals, which introduces a degree of arbitrariness in assigning a child to a
particular group. The NRC report found the approach of controlling statistically for age by multiple
regression to be appropriate, because the effect of age is likely to be linear across the relatively short age
period (3 months in both studies); that is, over short time periods, development is most likely to take
place at a constant rate.
Some members of the scientific community have noted the possibility that the most important
difference in the design of the two studies is the age of the child at assessment; 7-year-olds were tested in
the Faroes as opposed to children 5.5 years of age in the SCDS. Developmental assessments are likely to
be less sensitive in detecting subtle neurotoxic effects when they are administered during a period of
rapid developmental change. Individual differences in the rate of neurocognitive maturation may mask
subtle differences in function attributable to toxic exposures. NRC (2000, pp. 257-258) also noted that
Methylmercury Water Quality Criterion 1/3/01
4-35
-------
infant assessments in the SCDS (namely the 19 and 29 month Bayley Scale examinations) were not given
at optimal age points for detecting effects, particularly in this developmentally robust population.
Selection bias from exclusion of individuals -with severe impairments
The OSTP workshop Confounders and Variables Panel (NIEHS, 1999) identified what they
considered a serious potential issue with the SCDS. They noted that recruitment was limited to children
with no severe debilitating conditions. This panel felt that such a restriction could lead to
underestimation of effect when the shape of the dose-response curve is not known.
PCB exposure in the Faroese population
PCB exposure through maternal consumption of whale blubber was discussed at length at the OSTP
workshop and in the report of the Confounders and Variables Panel (NIEHS, 1999). Using the data from
the part of the cohort for which cord PCB was measured, Grandjean et al. (1997) performed a series of
analyses to ascertain if the PCB and mercury effects could be separated. Of the eight outcomes for which
there was a significant association with cord-blood mercury, four were also associated (p<0.1) with log
transformed PCB levels in cord tissue before adjustment for mercury. These four endpoints were also
significantly related to mercury cord-blood concentrations. These were CPT reaction time, BNT with
and without cues, and CVLT long-term reproduction (Table 4-1). When PCBs were included in the
regression analysis, only the CPT reaction time remained significantly associated with mercury. CVLT
and BNT with no cues were not significantly associated with either agent, whereas BNT with cues was
about equally associated with both (p <. 0.10). It is important to recognize that such an analysis removes
the shared variance related to both mercury and PCBs, thereby reducing the/? value associated with
either agent.
The Faroes investigators considered CPT reaction time to be a test of attention, BNT to assess
language, and CVLT to assess memory (Grandjean et al., 1997). Deficits in overall cognitive functioning
and verbal comprehension have been found to be associated with in utero PCB exposure in a study of
4.5-year-old children in the Netherlands (Patandin et al., 1999a), whereas deficits on a vigilance task
similar to the CPT were associated with cord PCB levels (commission errors) as well as the child's
concurrent PCB exposure (reaction time) (Patandin et al., 1999b). In the Patandin et al. study, PCB and
dioxin exposure was through diet unrelated to fish consumption. Another study reported effects of
exposure to children through their mothers' consumption of contaminated Lake Michigan fish. Deficits
in attention, language processing (reading comprehension), and memory related to prenatal PCB
4-36 Methylmercury Water Quality Criterion 1/3/01
-------
Table 4-1. Regression coefficients (betas) for effects of logarithmic transformations of mercury before
and after adjustment for PCB concentrations on Faroese neuropsychological tests: results from 7-year-
old children from the first year of testing.
After Adjustment for PCB
Before Adjustment
Neuropsychological Test
Continuous Performance Test
Average reaction time (ms)
Boston Naming Text
No cues
With cues
California Verbal Learning
Test (Children)
Long-term reproduction
Beta
39.3
-1.58
-2.03
-0.99
p-Value
<0.001
0.04
0.007
0.03
Beta
37.8
-1.04
-1.36
-0.78
Mercury
0.002
0.21
0.10
0.11
p-Values
PCB
0.64
0.16
0.08
0.26
Both
0.001
0.05
0.008
0.05
From Grandjean et al., 1997.
exposure were identified in 11-year-old children (Jacobson and Jacobson, 1996). Other contaminants
undoubtedly present in the fish, including methylmercury, were not assessed in this study; the potential
contribution of methylmercury exposure to the observed effects could not be evaluated.
It is informative to compare PCB levels hi other studies reporting adverse effects associated with
PCBs with PCB levels in the Faroese women. No breast milk or blood PCB levels from the mothers or
infants in the Faroe Islands cohort have been published. However, a recent study compared levels of
PCB congener 153 in human blood in pregnant women from the Faroe Islands consuming 0-1 blubber
meals/month ("low") or 2-3 blubber meals/month ("high") with other populations (Fangstrom et al.,
2000). "Low" Faroese exposure was comparable to blood PCB levels in an unspecified number of
pregnant women in the Netherlands, whereas "high" Faroese blood PCB levels were comparable to those
in an identified highly exposed population in the Quebec Arctic. The Faroese samples in the Fangstrom
et al. (2000) analysis were collected in 1994-1995, and the cohort for the Faroe study of developmental
neurotoxicity was recruited in 1986-1987. It is unclear when the Dutch samples in the Fangstrom et al.
(2000) study were collected; the cohort in the Dutch developmental study was recruited in 1990-1992.
Blood levels cannot be directly compared between the Dutch study and the Fangstrom et al. (2000) data
because one was on lipid-adjusted serum and the other on non-lipid-adjusted plasma. Similarly, breast
Methylmercury Water Quality Criterion 1/3/01
4-37
-------
milk levels cannot be directly compared (Grandjean et al., 1995; Steurwald et al., 2000; Lanting et al.,
1998). lii general, human body burdens of PCBs have decreased by about 50% over the past decade, so it
is possible that blood levels in the Dutch study were higher than those reported in the Fangstrom et al.
(2000) paper. It is also quite probable that PCB levels in the Faroe Islands were higher in the mid-1980s
than the mid-1990s, suggesting that the "low" Faroe exposure is comparable to levels in the Dutch study.
It is important to reiterate that whereas there may have been effects of PCBs in addition to those of
methylmercury, statistical analyses indicated that the effects were independent in this population (Budtz-
J0rgensen et al., 1999).
The Confounders and Variables Panel at the OSTP meeting (NffiHS, 1998) concluded that both
PCB and mercury had adverse effects on the CVLT score and on the BNT scores with and without cues.
They felt that it was not possible to determine the relative contribution of each. NRC concluded that
there was no empirical evidence or theoretical mechanism to support the opinion that in utero Faroese
exposure to PCBs exacerbated the reported methylmercury effect. They note that statistical tests for
interaction between PCB and mercury show no interaction. NRC reached a similar conclusion to the
Confounders and Variables Panel; a likely explanation is that both PCB and mercury adversely affect
some test outcomes, but their relative contributions cannot be determined given their co-occurrence in the
Faroes population. NRC states it is unlikely that a difference in PCB exposure between the two
populations explains the lack of developmental neurotoxic effects in the Seychelles (NRC, 2000, pp. 220
and 223).
In a second set of analyses, Budtz-J0rgensen et al. (1999) found that the effect of prenatal PCB
exposure was reduced when the data were sorted into tertiles by cord PCB concentrations. Regressions
assessing mercury exposure and the five principal test outcomes were then run separately for each of the
three groups. The regression coefficients for a mercury effect in the lowest PCB tertile were no weaker
than those for the higher two PCB groups. This lends additional credence to a conclusion that the
associations between mercury and test outcomes are not attributable to confounding by prenatal PCB
exposure. Calculations of benchmark doses and lower limits (BMDLs) were done using the whole
cohort, after a PCB correction and for the portion of the cohort with the lowest PCBs (NRC 2000 , Table
7-4, reproduced here as Table 4-2). In this table results are reported separately for methylmercury
measured in hair and cord blood and are calculated using the K-power model described in Section 4.3.4.
NRC commented on the results for the low-PCB-exposed subset for the two endpoints that were related
to PCB exposure, the BNT and the CVLT. They noted that the BMDs for these outcomes did not differ
from the BMDs for the total sample by any more than the BMDs for the two endpoints that were not
related to PCB exposure. NRC opined that the variability seen in Table 4-2 is no more than that expected
4-38 Methylmercury Water Quality Criterion 1/3/01
-------
by chance; the BMDs and BMDLs for both the PCB-adjusted and the low-PCB subset analyses are
within the intervals defined by the BMDs and corresponding BMDLs derived for the full cohort. The
difference between the BMDs based on the full cohort and the low PCB subset is less than one standard
error of the low PCB subset (NRC, 2000, p. 288). These analyses support a conclusion that there are
measurable effects of methylmercury exposure in the Faroese children that are not attributable to PCB
toxicity.
PCB body burdens in the Seychellois are very low by comparison to North American and European
populations. In 28 serum samples obtained from Seychelles study children, there were no detectable
concentrations of any PCB congeners. In the Faroes study, prenatal PCB exposure was measured in 436
stored umbilical cord tissue samples. It was noted at the OSTP workshop that cord tissue PCB
concentration has never been validated in relation to blood or milk concentration; because cord tissue is
lean and PCBs are lipophilic, the panel felt that it may not be the most reliable indication of total PCB
body burden (NEEHS, 1999). The cord samples were analyzed for a small subset of PCB congeners that
were used to represent the biologically significant PCB exposure. In an earlier publication (Grandjean, et
al., 1995) it was shown that these congeners predominate in samples from the Faroes cohort; comprise
these three congeners comprise approximately 50% of the PCBs in breast milk lipid. These same three
congeners, along with one other, were used to quantify PCB body burdens in milk and plasma in a study
of children in the Netherlands (Lanting et al., 1998). The approach taken in the Faroes for quantifying
PCB exposure (adding three key congeners together and multiplying by 2) appears to be a reasonable
approach for estimating total PCB exposure and is not expected to introduce a bias into the analysis.
4.2.2.3 Population Differences in Susceptibility
Populations may be more or less susceptible to effects of a toxicant as a consequence of
predisposing factors, such as nutritional status, exposure to other agents (see Section 4.2.2.1), or genetic
susceptibility.
The SCDS cohort is predominantly African in descent; the Faroes cohort is Caucasian. The latter
population has been somewhat isolated and thought to be descended from a small number of "founders."
This homogeneity in the Faroes could increase or decrease genetic susceptibility to effects of toxic insult.
NRC noted that methylmercury neurodevelopmental effects were observed in a genetically
heterogeneous and racially diverse sample studied in New Zealand, a population that was predominantly
non-Caucasian.
Methylmercury Water Quality Criterion 1/3/01
4-39
-------
Table 4-2. BMD (BMDL) Estimates from the Faroe Islands Study with and without adjustment for
PCBs and in the subset of Low PCB-exposed children (reproduced from NRC 2000)
Exposure
Hair
Cord Blood
Endpoint
Finger tapping
CPT Reaction Time
Boston Naming Test
CVLT: Delayed Recall
Finger tapping
CPT Reaction Time
Boston Naming Test
CVLT: Delayed Recall
Full Cohort
BMD (BMDL)a
20 (12)
18 (10)
15 (10)
27 (14)
140 (79)
72 (46)
85 (58)
246 (103)
Adjusted for
PCBs
BMD (BMDL)
17 (9)
27 (11)
24 (10)
39 (12)
149 (66)
83 (49)
184 (71)
224 (78)
Low PCB subset
BMD (BMDL)
7 (4)
13 (5)
21 (6)
32 (7)
41 (24)
53 (28)
127 (40)
393 (52)
, aBMDs are calculated under the assumption that 5% of the responses will be abnormal in unexposed subjects
(P0= 0.05), assuming a 5% excess risk (BMR = 0.05).
Source: E. Budtz-J0rgensen, Copenhagen University, N. Keiding, Copenhagen University, and P. Grandjean,
University of Southern Denmark, unpublished material, April 28,2000.
Data on birthweight and gestation length in the Faroes and Seychelles show no indication of energy
or macronutrient (protein and carbohydrate) deficiency. It is possible that members of either population
could be deficient in micronutrients. It has been suggested that certain nutrients found in fish eaten by
the Seychelles residents (e.g., omega-3 fatty acids and selenium) could attenuate adverse effects of
methylmercury exposure. It should be noted that both the Faroese and New Zealand populations would
be considered "high fish consumers" by comparison to U.S. norms, and both populations were observed
to have measurable effects of mercury exposure. It is unlikely that general health status of the Faroese
and Seychellois was a factor in enhancement or attenuation of mercury effects. Both populations receive
excellent health care.
The point was made in Section 4.2.2.2 that recruitment in the SCDS was limited to children with no
severe debilitating conditions. In the opinion of some scientists this may contribute to making the Faroes
sample more representative of the population at risk in the United States in that it includes infants with
some degree of initial perinatal risk.
It has been noted in several scientific forums that the cohort in the main Seychelles study appears to
have been robust for psychomotor development at early ages. The SCDS authors report a number of
abnormal scores on the Denver Developmental Screening Test that are considered to be exceptionally
low by U.S. norms. The population also was observed to have an unusually high mean PDI score and a
very low rate of referral for mental retardation. The means and standard deviations of the cognitive
measures administered at later ages were similar to U.S. norms. It is not clear what, if any, effect this
4-40
Methylmercury Water Quality Criterion 1/3/01
-------
developmental robustness has on susceptibility to adverse effects of prenatal Hg exposure. Statistical
power to find an adverse effect is discussed in Section 4.2.2.8.
4.2.2.4 Assessment of Prenatal Mercury Exposure
In the Faroes study, mercury in cord blood and maternal hair was measured; in the Seychelles,
maternal hair mercury was the biomarker of exposure. »The maternal hair samples obtained in the Faroes
and Seychelles studies did not necessarily reflect the same period of pregnancy. The Seychelles samples
were 9-cm lengths of hair reflecting average mercury exposure during pregnancy. The Faroes study
analyzed mercury from hair samples of variable length, some 3 cm (reflecting late second and third
trimester) and some 9 cm (presumably reflecting the entire pregnancy).
In the analyses of the Faroese data, cord-blood mercury concentration was significantly associated
with a slightly larger number of endpoints than was maternal hair mercury. Given the estimated half-life
of methylmercury and what is known of PBPK, it could be assumed that cord-blood mercury reflects the
latter part of gestation. Hair mercury could reflect the entire pregnancy or could be segmentally analyzed
to provide snapshots of various times in gestation. Some of the effects reported in the Faroese cohort
could be related to toxic responses in the latter stages of prenatal development. However, hair mercury
concentrations in the Faroe Islands study were only a slightly weaker predictor of methylmercury effects
than was cord blood. NRC concluded that it would be reasonable to expect that, if children were affected
in the main Seychelles study, some indication of an association between child development and maternal
hair mercury concentration would have been observed (NRC, 2000, p. 252). It noted that the findings of
developmental effects reported in New Zealand were based solely on maternal hair sample data averaged
across the entire period of pregnancy. The difference in the observation of effects between the Faroes
study and the SCDS is thus not an artifact of biomarkers of exposure.
4.2.2.5 Level of Exposure
In their analyses the SCDS authors used maternal hair mercury as the biomarker of exposure; the
Faroes investigators used both cord blood and maternal hair mercury. A comparison of maternal hair
mercury levels indicates that exposure in the two studies was in the same range. For the main SCDS, the
median hair mercury was 5.9 ppm with a range of 0.5 ppm to 26.7 for the whole cohort. In the Faroes
birth cohort (n = 1,020), the median hair mercury was 4.5 ppm with a range of 2.7 to 42.6 ppm
(Grandjean et al., 1992) . That the Seychelles Islands study may entail a lower exposure level than the
Faroe Islands study could be concluded from two lines of evidence: the hainblood ratio from the
Methylmercury Water Quality Criterion 1/3/01
4-41
-------
Seychelles Islands and laboratory studies suggesting that dietary factors can influence tissue levels of
methylmercury.
The ratio of hair mercury to blood mercury in the Seychelles study was estimated to be 416, a value
that is higher than ratios reported elsewhere, which span 190 to 367 (Stern, 1997). The hair: cord blood
ratio for the Faroes cohort was 191 (Grandjean et al., 1992). The value commonly used in dose
conversion models is 250 (Stern, 1997; U.S. EPA, 1997e). If the value of 416 is used in estimating
maternal or fetal blood mercury then estimates of the dose experienced by the Seychellois fetuses would
be lower, by almost twofold, than assumed.
The hair: blood ratio of 416 is plausible for the Seychellois population considering their high fish
diet and suggestions in the literature that diet can influence tissue levels of mercury. Average fish
consumption in that population is 12 fish meals/week, which is likely to result in comparatively high
levels of n-3 fatty acids and selenium. Such a diet may alter the kinetics of mercury by lowering blood or
organ levels of mercury associated with a certain level of intake.
4.2.2.6 Episodic Versus Continuous Exposure
Exposure to methylmercury in the Seychelles is through daily consumption of fish. Although the
Faroese eat fish more frequently than does the average consumer in the United States (about three meals
a week), a significant source of methylmercury exposure in this population is from eating pilot whale
meat. Pilot whale meals are relatively infrequent (less than once per month on the average) (Grandjean
et al., 1992) with additional intermittent snacks of dried whale (Grandjean et al., 1998). The whale meat
mercury concentration varies with the pod. An analysis of 466 whales showed an average concentration
of 1.9 ppm, with a range of 0.59 to 3.30 ppm (Faroese Food Agency data quoted in NIEHS, 1999). There
is no evidence to indicate that methylmercury bioavailability from the muscle of pilot whale is any
different from that of fish tissue.
In the New Zealand study, there was the assumption of regular consumption of a relatively high-
mercury fish (shark) in fish and chips, the major fast food of the area; the actual frequency and pattern of
exposure are unavailable.
The degree to which differences in exposure pattern among studies account for differences in
outcome is uncertain. It has been suggested that the mercury body burden in the Faroe Islands study was
the consequence of a "spike" exposure pattern, in contrast to a more continuous exposure pattern in the
4-42 Methylmercury Water Quality Criterion 1/3/01
-------
Seychelles study, which nonetheless resulted in a similar body burden. The Faroese investigators did
segmental analyses of a small number of long hair strands from cohort mothers. Their results indicated a
few instances of hair mercury peaks that implied temporal variation or spiking. They noted, however,
that the peak level was only about twice the lowest hair mercury concentration (Budtz-J0rgensen et al.,
1999).
The pattern of exposure can be a critical determinant of in utero toxicity. For example, the NRC
report cites data in animals that showed that maternal ingestion of a given dose of alcohol over a short
time caused greater neuronal impairment (Bonthius and West, 1990) and behavioral impairment
(Goodlett et al., 1987) than that caused by gradual ingestion of the same total dose over several days.
The frequency of exposure has a significant influence on the variation in blood levels, even under steady-
state conditions, and is dependent on blood half-life (Rice et al., 1989).
It is probable that both episodic and continuous patterns of exposure are present in the population
of the United States. Individuals in some ethnic groups engage in a subsistence-type fishing pattern,
consuming fish as their major protein source. Most sport fishers, however, consume fish on an
intermittent basis. It is not uncommon for piscivorous fish in inland waters to have mercury levels
exceeding 1 to 2 ppm (U.S. EPA, 1997), so that the body burden of mercury in this group of fish
consumers would presumably be the result of episodic exposure to food sources with levels of mercury
similar to those in the Faroe Islands (see also Section 5.4.4 of this document). It may be that the
consumption pattern of the Faroe Islands population better represents the pattern of exposure in the
majority of the U.S. population exposed to elevated levels of methylmercury than does the consumption
pattern of the population of the Seychelles Islands.
4.2.2.7 Endpoints Assessed
As described in Section 4.2.1, there have been inconsistent indications of adverse effect in
newboms or preschool children of mothers experiencing low-dose, long-term exposure to
methylmercury. The lack of consistent positive findings using standard newborn neurological tests has
been considered unsurprising. Neurological examination of the newborn and young infant presents
testing challenges that are difficult to meet in large-scale studies. The state of the newborn determines to
a significant degree the quality and intensity of response to stimulation during an examination. "The state
of an infant is usually dependent upon factors that are often outside the examiner's control, such as
hunger, hydration, illness, and the temporal location of an infant in its sleep-wake cycle. The recognition
that state is a key variable in newborn behavior can be found in the fact that neonatal behavioral and
Methylmercury Water Quality Criterion 1/3/01
4-43
-------
neurologic assessments usually indicate what state the newborn should be in before a given item series is
administered..." (K. Deitrich, in U.S. EPA, 20000-
It has been observed that most of the deficits associated with low-level prenatal exposure to
developmental toxicants would not be revealed in a pediatric neurological examination and that gross
neurological findings are unlikely in such studies. It has also been shown in studies not related to
methylmercury that minor neonatal neurological deviations from the norm are not predictive of later
neurobehavioral morbidity (U.S. EPA, 2000f).
Screening tests such as the Denver Developmental Screening Test have been used with highly
variable results in methylmercury studies. Section 4.2.1 reports the differences in results among the New
Zealand, SCDS pilot, and SCDS main cohorts. Recent research suggests that screening tests are not as
sensitive as once believed and are no longer recommended for use in studies of low-level environmental
chemical exposures to the fetus or infant (U.S. EPA, 2000f).
In the opinion of most developmental scientists, the Faroes and Seychelles studies used very
different neurobehavioral test batteries. The tests selected for use in the SCDS are considered apical or
omnibus tests (e.g., the McCarthy Scales of Children's Abilities); these provide global scores that
integrate performance over many separate neuropsychological domains. The investigators studying the
Faroes population were working from a hypothesis that mercury would have multifocal domain-specific
neuropsychological effects.. The OSTP Neurobehavioral Endpoints Panel was similarly disposed. They
noted that it is plausible that prenatal exposure to methylmercury may not affect IQ, but rather domain-
specific areas such as memory deficits, motor delays, or effects on so called "executive functions" - the
complex domains that involve planning and cognitive flexibility (NIEHS, 1999). The Faroese test
battery consisted of highly focused tests selected from those commonly used in clinical neuropsychology
(e.g., CVLT and BNT) and did not include an apical test of global function. They observed effects in
areas of language, memory, motor skills, visual-spatial abilities, and attention.
Many of the subscales of the McCarthy Scales might be expected to provide measures comparable
to some tests administered to the Faroese children. However, there was no evidence from the McCarthy
subtests of domain-specific effects in the Seychelles. These included verbal, perceptual-performance,
quantitative memory, and motor scores. One conclusion is that if there were actually domain-specific
effects occurring in the 5-year-old Seychellois, they should have been observed in the analyses of the
McCarthy Scales results. The NRC panel came to a different conclusion: "Although the Faroe Islands
and SCDS test batteries include tests of language and memory, it is not appropriate to view the endpoints
4-44 Methylmercury Water Quality Criterion 1/3/01
-------
used in the studies to assess each domain to be equivalent either in terms of the specific skills assessed or
the test sensitivity." (NRC, 2000, pp. 256-257).
One test was administered to both populations: the Bender-Gestalt Test. The investigators used
different scoring systems; the SCDS used the Koppitz system whereas the Faroes used the Gottingen
system. The NRC report noted that in a paper by Trillingsgaard et al. (1985) scores derived using the
more detailed Gottingen system were significantly associated with low-dose lead exposure, whereas
scores on the Koppitz system were not. Thus the Gottingen system used in the Faroe Islands might be
more sensitive.
A second important difference in the assessment batteries used in the Faroes study and SCDS is the
age of the child at assessment; 7-year-olds were tested in the Faroe Islands in contrast to children 5.5
years of age in the SCDS. Assessments in the New Zealand cohort were done at 4 and 6 years of age. It
is generally thought that developmental assessments are likely to be less able to detect subtle neurotoxic
effects when they are administered during a period of rapid developmental change. The period covering
ages 60 to 72 months (when the SCDS and New Zealand cohorts were evaluated) is such a time;
individual differences in the rate of cognitive maturation are likely to eclipse subtle differences in
function attributable to a teratogenic exposure (Jacobson and Jacobson, 1991). The NRC panel also felt
that in the SCDS, assessments of infants (particularly the 19- and 29-month BSID) were not given at
optimal age points. Their report makes the following statement:
Studies of prenatal exposure to alcohol and other substances that have administered the Bayley scales at
multiple ages have repeatedly failed to detect effects at 18 months, probably because it too is a period of rapid
cognitive maturation, involving the emergence of spoken language. Twenty-nine months is likely to be an
insensitive testing point for the Bayley scales because it is at the end of the age range for which the version of
this test used in the Seychelles was standardized, leading to a substantial risk of a "ceiling effect" (i.e., too
many children receiving the highest possible scores on numerous items) (NRC, 2000 pp. 257-258).
The overall conclusion of NRC, however, was that discrepancies between the Faroe Islands and the
main Seychelles studies are probably not due to differences in the assessments. They point out that the
New Zealand study observed associations between methylmercury exposure and scores on the McCarthy
Scales of Children's Abilities (the primary outcome measure used in the SCDS) at about the same age of
assessment as in the Seychelles study (NRC, 2000, p. 258).
Methylmercury Water Quality Criterion 1/3/01
4-45
-------
4.2.2.5 Power of Studies
NRC commented on the power to detect subtle effects in the admittedly large human studies (NRC,
2000, pp. 266-267). They noted that it is possible that the differences in response between the Faroes
study and the SCDS could be due to between-sample variability in the expression of neurotoxicity at low
doses. NRC remarked that even large samples can have insufficient power to detect adverse effects if a
relatively small number of subjects are exposed in the upper ranges of the exposure distributions, where
those effects will presumably be found.
NRC said that the magnitude of the associations found in the methylmercury studies resembles that
reported for other environmental contaminants, such as low-dose lead and PCBs. If the magnitude of an
association is,not large, it is not likely that it would be detected in every cohort studied. NRC noted by
comparison that it is well established in the scientific community that a blood lead concentration in
excess of 10 [ig/dL places a child at increased risk of poor developmental outcomes. However, not all
lead studies have found an association between exposure at this level and decreased performance, and
substantial variability exists in the magnitudes of the reported effects (Bellinger, 1995). NRC noted for
the SCDS, "the evidence consistent with such effects found in the pilot phase, coupled with the
suggestion of unusual developmental robustness in the main study, suggest that the failure to detect
apparent adverse effects in the main study could be due to the substantial sample-to-sample variation
expected when trying to identify weak associations in an inherently 'noisy' system of complex, multi-
determined neurobehavioral endpoints" (NRC, 2000, p. 267).
In another comment on power, NRC says that power analyses based on total sample size can be
misleading if adverse effects occur primarily among the most heavily exposed individuals, who typically
constitute a small proportion of the sample. They note that of 700 children in the SCDS, only about 35
were exposed at levels concordant with maternal hair mercury of 15 ppm or higher. Because multiple-
regression analysis examines associations that are averaged across the entire distribution of exposure,
associations that hold only for the most highly exposed children can be difficult to detect. "Thus, if
adverse effects of prenatal MeHg exposure occur primarily in the upper range, the power to detect them
will be limited, and it would not be surprising if associations found in one Seychelles cohort (the pilot
study) were not detected in the next cohort (the main study)" (NRC, 2000, p. 267).
In this context it should be noted that Grandjean et al. (1997) published an analysis of their
neuropsychological test data on 7-year-old children, wherein they excluded all scores from children born
to mothers with 10 ppm or higher hair mercury. This decreased the number of observations by 15%. In
4-46 Methylmercury Water Quality Criterion 1/3/01
-------
the multiple-regression analyses, regression coefficients and p values were very similar to those obtained
when data on the full cohort were used. This indicates that in this study population, adverse effects of
mercury were detectable at exposures below 10 ppm maternal hair mercury.
4.2.2.9 Selection of Study
There is a large database on potential neurodevelopmental effects of methylmercury. In particular,
three large, well-designed, prospective longitudinal studies have been peer reviewed and intensively
analyzed. Some results from these studies of large populations are in apparent conflict. The previous
sections reviewed some of the factors that have been suggested to account for the finding of adverse
outcomes associated with in utero mercury exposure in the Faroes and New Zealand and the lack of this
association in the SCDS. None of these factors represents a critical flaw in study design or execution.
None of the factors adequately explains the differences in the study outcomes.
The NRC (NRC, 2000, p. 221) suggests that the finding of a low-dose methylmercury effect in a
culturally and genetically heterogenous population in New Zealand study decreases the importance of
population sensitivity issues in comparing the Seychelles and Faroes studies. The New Zealand study
had a higher baseline rate of abnormal and questionable DDST scores in the test (8%-17% in controls)
than did the Seychelles study (8% hi the complete pilot cohort, 1.9% of the complete main cohort). This
observation is consistent with the suggestion that the lack of effects in the Seychelles population is
related to its relatively higher level of neurological performance at critical early life stages. Another
possibility is that the manner in which the tests were given in the Seychelles led to better test
performance, resulting in a less sensitive measure (i.e., an easier test for children to pass). The SCDS
may also have had reduced power because of the small number of maternal-child pairs with
methylmercury over 15 ppm. A comparison of the numbers in the relatively high-exposure range is
instructive. If one uses 10 ppm maternal hair mercury as the high-exposure cutoff, there are about 150
Faroes subjects, at least 100 Seychelles subjects, and only 16 New Zealand subjects in this category (see
Fig. 5-6, p. 166, NRC report).
One strength of the New Zealand study is that an effect was shown in an ethnically heterogeneous
sample; another advantage was that the study used developmental endpoints with predictive validity.
However, FJPA acknowledges and shares the NRC reservations about using the New Zealand study as the
basis for the methylmercury RfD. The New Zealand study is relatively small, with 237 subjects, by
comparison with the population of up to 900 for the Faroes tests. Moreover, the New Zealand data have
Methylmercury Water Quality Criterion 1/3/01
4-47
-------
not had the exhaustive scientific scrutiny that have been applied to the SCDS and Faroes study. The
advantages of the Faroes study include these:
• large sample size;
• good statistical power as calculated by conventional means;
• the use of two different biomarkers of exposure;
• comprehensive and focused neuropsychological assessment;
• assessment at an age and state of development when effects on complex neuropsychological
functions are most likely to be detectable;
• statistically significant observations that remain after adjusting for potential PCB effects; and
• extensive scrutiny in the epidemiological literature.
The Faroes data have also undergone extensive reanalyses in response to questions raised by panelists in
the NffiHS (1999) workshop and by NRC (2000). The SCDS shares many strengths of the Faroes study.
However, EPA agrees with NRC that a positive study, one that shows statistically significant associations
between prenatal mercury exposure and adverse outcomes, is the strongest public health basis for an RfD
(NRC, 2000, p. 6). Moreover, although one can model the nonpositive results of the SCDS, the resulting
estimates of no effect level are difficult to interpret.
The study selected by EPA as the basis of the methylmercury RfD is the report of developmental
neurotoxicity in 7-year-old children in the Faroes. The next section discusses issues in choice of
endpoint for the RfD calculation. Many of the arguments in study selection pertain to choice of endpoint
as well.
4.2.3 Choice of Critical Effect (endpoint)
EPA considered recommendations of NRC and the external peer reviewers in making the choice of
a critical effect or endpoint from the Faroese data on neuropsychological effects in children. Rather than
choosing a single measure-for the RfD critical endpoint, EPA considers that this RfD is based on several
Faroese test scores. These test scores are all indications of neuropsychological processes that are
involved with the ability of a child to learn and process information. The issues and decision points in
coming to this choice are described in the following sections.
4-48
Methylmercury Water Quality Criterion 1/3/01
-------
4.2.3.1 Endpoints Suitable for RfD Derivation
Several studies have reported significant associations between increased numbers of combined
abnormal and questionable scores on standardized neurological examinations. NRC opined that the
functional importance of these effects is uncertain. There is little evidence that relatively low-dose, long-
term exposure has any significant effect on language or motor-skill developmental milestones. There is
some evidence of an association between in utero mercury exposure and deficits on the DDST. The
NRC put forth the opinion that this screening test is not as useful as others in developmental
neurotoxicological testing.
As is shown in Table 4-3, the tests used in the Seychelles and New Zealand studies in general were
apical tests, assessing broad functional categories. These tests are widely used clinically and have been
validated and normed for the U.S. population (but not the populations in which they were used). In
contrast, the tests used in the Faroe Islands study were chosen to assess specific behavioral domains. The
global clinical instruments such as the McCarthy, WISC-R, and CBCL have manuals that describe the
tests and domains assessed, as well as the predictive validity of scores on these instruments to "real-
world" behavior such as school performance. For the tasks used in Faroe Islands, finger tapping is a
commonly used assessment of motor speed (Letz, 1990), and the Bender is a standardized test of
childhood development. The other three endpoints also have demonstrated clinical relevance and
predictive value As outlined in the table, most of these endpoints are predictive of ability in various
academic skills, and therefore school performance. These tests, whether designed to be relatively global
or domain-specific, were adversely affected by methylmercury exposure in the Faroe Islands and New
Zealand, but not the Seychelles Islands, studies. In addition, motor performance was adversely affected
in both New Zealand and the Faroe Islands. The only study that assessed social and adaptive behavior
was the SCDS. BMD analysis performed by the NRC committee identified adverse effects on the CBCL
at maternal hair levels comparable to those at which effects were observed in the Faroe Islands study
(NRC, 2000, Table 7-5, p. 291). As concluded by the NRC (NRC, 2000, p. 325), the deficits observed in
the New Zealand and Faroe Islands study can be considered predictive of problems in cognitive and
academic performance associated with methylmercury exposure.
NRC presented BMDs and BMDLs for several endpoints in the positive Faroes and New Zealand
studies as well as for the nonpositive Seychelles study (the next section discusses choices of model and
choices made in BMDL calculation). Reproduced below is Table 7-2 from the NRC report (here as
Table 4-4), which compares BMDs from the three studies in terms of maternal hair mercury. Included in
this table are the New Zealand BMDs calculated after exclusion of the data from the highest exposed
Methylmercury Water Quality Criterion 1/3/01 4-49
-------
individual. NRC suggested that this hair mercury concentration of 86 ppm is not plausible. The text
reads:
a hair Hg concentration of 86 ppm is more than 4 times the next highest hair Hg concentration in the study. If
the one-compartment pharmacokinetic model and EPA's standard default input assumption are used, it can be
estimated that a 60-kg woman would have to eat an average of 0.5 pounds (227 g) offish containing 2.2 ppm
of Hg to reach a hair Hg concentration of 86 ppm. Consistent exposure at such a dose seems unlikely when
the mean Hg concentration in fish from fish-and-chips shops, a principal source of exposure in New Zealand
(KjellstrSm et al., 1986), is 0.72 ppm (Mitchell et al., 1982). On the basis of those considerations, the
committee concluded that analyzing the New Zealand data without the data from that individual is appropriate.
(NRC, 2000, p. 282).
The range of BMDL values is relatively small (4 to 25 ppm maternal hair mercury). Inspection of
this table shows that all the BMDs (and corresponding BMDLs) from the New Zealand study are lower
than those from the other positive study in the Faroes. Often the most sensitive adverse endpoint is
selected as the critical effect for calculation of a RfD. The most common surrogate for "most sensitive"
is the lowest BMDL or bounded NOAEL (that is, NOAEL from a study wherein an effect was observed).
The lowest BMDL is 4 ppm maternal hair mercury for the McCarthy Perceptual Performance Test
calculated by Crump et al. (1998, 2000) on the New Zealand data (Kjellstrom et al., 1986). NRC had
reservations about using the Kjellstrom (1986) data as the basis for the methylmercury RfD, with which
EPA agreed (see Section 4.2.2.9). In this instance the choice is not of the lowest BMDL, but will be
made from among the measures in the Faroese data.
Grandjean and colleagues reported significant associations between either maternal hair mercury or
cord-blood mercury and decrements in several neuropsychological measures in 7-year-old Faroese
children:
• Finger tapping—preferred hand (p = 0.05)
• Continuous Performance Test—first year of data collection
- false negatives—(p = 0.02)
— mean reaction time—(p = 0.001)
WISC-R Digit Span (p = 0.05)
• Boston Naming Test
- no cues (p = 0.0003)
- with cues (p = 0.0001)
4-50
Methylmercury Water Quality Criterion 1/3/01
-------
o
1
1
8
en
*73
•Si
CD
4— >
O
•o
CS
-a"
C/3
00
rrt
CO
I
+3
o
1
I
£
•a
CO
*o
•o
1
CO
CO
$
.
1
jj
3
03
H
vance
JO
c8
•3
0
•o
&
00
(O
CO
<
C
0
VB
0
e
o
Q
1
>-.
•o
S
CO
GO C3
O .O
O *-«
1 § 8 £
to O C3 ^ " ».
o c & *s rv
O ?•< G Kl ^ V— *
,j5 c rr iS i— < HH
lilt 1 1 1 §
^
! ifh! ill! 11 ill
§ 8 „, 1 1 1 1 1 § I "S J »• l I |i|
S M S j^ *7s -&* G^^S'4"' **O O O "^^ fi t2
| 1 ! | -1 1 1 8 | 1 1 1 1 1 1 ||l
S, ^ 1 1 1 §> ? 1 I $ S z .i M M ? * !
^J-a.^l^sS^^ggS £ g °a'>
ii^^j^iiiisi^ j a 1. 1|
o -8
"S U a
8, H 6
to a. t_
60 -O CJ tS
C bJO £3 ...
'S3 ceo IS ^
I 111 2 S
t- co o a 0 § §e^§-— esc
||o 1 >, g)'^.20'|^ 1 « S
O O ^ O cf *-2 cd Wi C P-) -^ W) O T3 j
r^ -C< M c3 ^ ^ S En ^2*o^ O ""* ^
"X c o S *"*"^ "S c .-o CT* ^P o *• ^ j^ fc2i?
O S^ pj *p* M C ^j ^ ^^ (U O ^ ^^ ^^ ^S
73O''o"c1:a'g)|^— > :ohtoM-d'>^ J -3" ry
1 1 1 1 | 1 t i 1 1 & 1 i 1 i 1 1 1 1 e I .
1 1 1 1 1 i i I 1 1 i 1 1 1 M I 1 1 ?
^(g^^S^^^^tSlfi^ g ,£ Sill
> CO
4> iJ1 i
"S ci ^3
y J(2 S P^r)
1 1 o t 1
d S O ^5 S *S-
1 i H ,ljif i ill!
ft! liiilHh e I ills
fo< oS-"ea^*^\;>— los! Ci •£ "S •§ " "3
S" kJ-StiaS^^oHQcj rj cs caOF^u
paJ§^Ucu^'[SUMPQUH^ ^ S S -i-Jc
tP O o
!5 r • W
^ r~i .•<
.. C e « S
eg C3 O ^ ^
*— ' cs •«-; E^ |"™(
1 « [S .2 M S3
»*-3 fli I-V| ^ r" ^5
>, 0 & ^ '1 J3
U C8 <1> ^ fa 0
CO B< ' £ S § 4»
< J ^
to
c
o
I
I
-------
Table 4-4. Benchmark dose calculations (ppm MeHg in maternal hair) from various studies and for
various endpoints (NRC, 2000)
Study
Seychelles'"
Faroe Islands'1
New Zealand"
Endpoint
Bender Copying Errors
Child Behavior Checklist
McCarthy General Cognitive
Preschool Language Scale
WJ Applied Problems
WJ Letter/Word Recognition
Finger Tapping
CPT Reaction Time
Bender Copying Errors
Boston Naming Test
CVLT: Delayed Recall
TOLD Language Development
WISC-R:PIQ
WISC-R:FSIQ
McCarthy Perceptual Performance
McCarthy Motor Test
BMD"
***c
21
***
***
***
***
20
17
28
15
27
12
12
13
8
13
BMDL
25
17
23
23
22
22
12
10
15
10
•14
6
6
6
4
6
"BMDs are calculated from the K-power model under the assumption that 5% of the responses will be abnormal in
unexposed subjects (Po= 0.05), assuming a 5% excess risk (BMR = 0.05).
'Data from Crump et al. (1998,2000). "Extended" covariates.
c *** indicates value exceeds 100.
dData from Budtz-J0rgensen et al. (1999).
"Data from Crump et al. (1998, 2000).
Abbreviations: WJ, Woodcock-Johnson Tests of Achievement; CPT, Continuous Performance Test; CVLT,
California Verbal Learning Test; TOLD, Test of Language Development; WISC-R:PIQ, Wechsler Intelligence Scale
for Children-Revised Performance IQ; WTSC-R:FSIQ, Wechsler Intelligence Scale for Children-Revised Full-Scale
IQ.
• California Verbal Learning Test
- short-term reproduction (p = 0.02)
- long-term reproduction (p = 0.05)
When an alternative approach to adjusting for covariates was used (Peters-Belson method) was
used, two more measures showed significant associations:
WISC-R Block Design (p = 0.05)
• Bender Gestalt Test errors (p = 0.05)
4-52
Methylmercury Water Quality Criterion 1/3/01
-------
More endpoints were significantly associated with cord-blood mercury than with maternal hair mercury.
Table 7-3 from the NRC report is reproduced below as Table 4-5; this presents calculations, in terms of
cord-blood mercury concentrations, of BMDs and BMDLs for five Faroese endpoints.
4.2.3.2 Comparison of Endpoints
Boston Naming Test (BNT)
The BNT was the endpoint of choice of the NRC panel (NRC, 2000, p. 327). This test assesses
word retrieval and formulation abilities in children, adults, and brain-injured patients. In the test, 60 line
drawings are shown to the subject one at a time, and the subject is asked to name each of them.
Familiarity (frequency of occurrence of the target names) decreases as the test progresses. Responses of
the patient are scored for latency and correctness. When the subject misses an item, two kinds of cues
may be given. A "stimulus cue" is a short phrase that gives additional information about the target item
(e.g., something to eat). A "phonetic cue" is the first sound of the target word. Scores are summarized
according to the number of spontaneously given correct responses, the number of correct responses
following stimulus cues, and the number of correct responses following phonetic cues. The number of
stimulus cues and the number of phonetic cues given by the examiner also is recorded. The peer-review
panel noted that there is not much normative data on the BNT but that it is often used by child clinical
neuropsychologists because of its documented validity in various child studies (EPA, 2000e). The BNT
Table 4-5. Benchmark dose calculations (ppb methylmercury in cord blood) from the
Faroe Islands Study for various endpoints
Endpoint
Finger Tapping
CPT Reaction Time
Bender Copying Errors
Boston Naming Test
CVLT: Delayed Recall
BMD"
140
72
242
85
246
BMDL
79
46
104
58
103
aBMDs are calculated from the K-power model under the assumption that 5% of
the responses will be abnormal in unexposed subjects (P0= 0.05), assuming a
5% excess risk (BMR = 0.05).
CPT, Continuous Performance Test; CVLT, California Verbal Learning Test.
Source: NRC (2000); data from Budtz-J0rgensen et al. (1999).
Methylmercury Water Quality Criterion 1/3/01
4-53
-------
has been useful as a measure of confrontation naming and word retrieval skills and can be used to
differentiate between children with and without language-based learning disabilities; moreover, it is a
predictor of related cognitive and academic skills, especially reading achievement (Yeates, 1994, as
quoted in U.S. EPA 2000e).
Continuous Performance Test (CPT)
The endpoint from the Faroe Islands study that yielded the lowest BMDL in the NRC analysis was
the CPT reaction time. This test was modified from the Neurobehavioral Evaluation System (NES)
version, which is a standardized battery used mainly in occupational settings in adults. In the Faroe
Islands study, the child was required to respond as quickly as possible when a silhouette of a cat appeared
on a computer screen, but not when the silhouettes of other animals (number not specified) appeared
(Grandjean et al., 1997). Dependent variables included number of missed responses (omission errors)
and average reaction time for the last 3 minutes of a 4-minute task. False positives (errors of commis-
sion) apparently were not analyzed. Reaction time in a task that includes decision making (respond to
cat, don't respond to others) is a measure of the speed of information processing. The investigators
found an increase in reaction time correlated with cord blood using all data; this correlation was still seen
when only data were used from children whose mothers had hair concentrations below 10 ppm (low-level
exposure). In addition, there was an association between cord blood mercury levels and an increase in
omission errors hi the full group and low-level exposure group. This finding indicates poorer attention to
the task as a function of methylmercury exposure.
Speed of information processing as measured by reaction time is highly correlated with IQ in
humans (Jensen and Munro, 1979; Matthews and Dom, 1989; Vernon, 1983; Vernon et al., 1985;
Western and Long, 1996). It has been argued that speed of information processing is a measure of g, the
highest order common factor in all tests of cognitive ability (Jensen, 1993b). Reaction time in complex
reaction time tasks is consistently observed to be correlated with psychometric g in studies in several
cultural groups (Buckhalt and Jensen, 1989; Ja-Song and Lynn, 1992; Lynn et al., 1991; Lynn and
Wilson, 1990; Shigehisa and Lynn, 1991). Generally, the association between g and decision reaction
time increases with increasing task complexity (Beh et al., 1994; Jensen, 1987). It is estimated that the
correlation between reaction time and g-loaded psychometric tasks is 0.3-0.5, whereas the correlation
based on several reaction time and psychometric tasks approaches 0.7 (Jensen, 1993a; Vernon, 1989),
which is similar to the correlation among different IQ tests (Jensen, 1993). Reaction time tasks also
discriminate between brain-injured and other individuals (Western and Long, 1996) and identify children
with attention deficits (Zahn et al., 1991).
4-54
Methylmercury Water Quality Criterion 1/3/01
-------
The NRC chose not to rely on CPT reaction time as the critical endpoint because results were from
only half the cohort. The Faroe investigators reported that effects on CPT reaction time were significant
for the first year of testing but not the second, with combined effects for the 2 years significant atp =
0.01. The authors stated that "[b]ecause supervision was stringent only during the first year, these data
were chosen for development of the final regression model" (Grandjean et al., 1997, pp. 422-423). The
NRC felt that measures from the full cohort would be more reliable than those based on half the cohort;
their report did not state any concerns regarding elimination of the second year data per se (NRC, 2000,
p.286).
Advantages of the choice of the CPT reaction time as the critical endpoint would be that there was
no evidence of an effect of PCBs on this measure, and the correlation of complex reaction time with
measures of intelligence such as IQ. The disadvantage is that the analysis is based on half the cohort.
However, this limitation also holds true for the BNT corrected for PCB exposure. Therefore, there is
little or no reason to choose one over the other in this regard.
California Verbal Learning Test for Children (CVLT)
The California Verbal Learning Test for Children is a word-list-learning task that measures
acquisition of information following repeated exposure to verbal stimuli. Of principal interest are the
variables of learning, delayed recall, and perseveration. The test has good test-retest reliability as well as
internal consistency. The theoretical foundations of the CVLT are based upon several decades of
cognitive science research in brain/behavior relationships. The test discriminates clinical groups such as
those with hyperactivity/attention deficit disorders, children with learning disabilities, and children
suffering prenatal insults such as fetal alcohol syndrome.
4.2.3.3 Consideration of Potential PCB effect
EPA agrees with NRC that analyses of the Faroese test results show that there are real mercury-
related adverse effects that cannot be attributed to concomitant PCB exposure. This was noted in Section
4.2.2.2. The external peer review panel for the methylmercury RfD agreed with that conclusion.
However, they disagreed with the NRC choice of the BNT results from the full cohort because of the
potential effect of PCB exposure. They thought that the BNT results were the most sensitive to PCB
influence of any evaluated in the Faroe Islands. The peer review panel pointed to the analyses presented
by NRC (reproduced in this document as Table 4-6) as presenting an opportunity to consider the use of
benchmark estimates corrected for any potential PCB influence. The Faroes investigators calculated a
Methylmercury Water Quality Criterion 1/3/01 4-55
-------
PCB-adjusted BMD and BMDL for the BNT using cord blood as the exposure biomarker; these were
considerably greater than the BMD/BMDL for either the full cohort without PCB adjustment or that from
the low-PCB tertile. Similar increases after adjusting for PCBs were not seen for finger tapping, CPT
reaction time, or CVLT delayed recall tests, when cord blood was the exposure metric. NRC noted that
the PCB measurements were done on cords from only about one-half of the Faroese cohort (about 450
children) and that the use of data from only the low-PCB tertile further reduces n to about 150 children.
NRC reported that the reduced sample sizes in these analyses increased the variability in the results.
They saw no clear pattern as to how the PCB-adjusted analyses differed from the original results. The
NRC concentrated its focus on the low-PCB subset BMDs and BMDLs. They compared results from
two tests with no PCB effect (CPT and finger tapping) with those with potential for PCB influence (BNT
and CVLT). They reported that the BMDs for the low-PCB subset for the BNT and CVLT did not differ
from the BMDs for the whole cohort any more than did the BMDs for the two tests with no influence of
PCBs. The NRC authors felt that the variability seen in Table 4-6 is no more than that which would be
expected by chance alone (NRC, 2000, p. 288).
Table 4-6. BMD (BMDL) Estimates from the Faroe Islands Study With and Without Adjustment for
PCBs and in the Subset of Low PCB-Exposed Children (calculated using the K-power model)
Exposure
Hair
Cord
Blood
Endpoint
Finger tapping
CPT Reaction Time
Boston Naming Test
CVLT: Delayed Recall
Finger tapping
CPT Reaction Time
Boston Naming Test
CVLT: Delayed Recall
Full Cohort
BMD (BMDL)3
20 (12)
18 (10)
15 (10)
27 (14)
140 (79)
72 (46)
85 (58)
246 (103)
Adjusted for PCBs
BMD (BMDL)
17(9)
27(11)
24 (10)
39 (12)
149 (66)
83 (49)
184(71)
224 (78)
Low-PCB subset
BMD (BMDL)
7(4)
13(5)
21(6)
32(7)
41(24)
53 (28)
127 (40)
393 (52)
11 BMDs are calculated under the assumption that 5% of the responses will be abnormal in unexposed subjects (P0=
0.05), assuming a 5% excess risk (BMR = 0.05).
Source: E. Budtz-J0rgensen, Copenhagen University, N. Keiding, Copenhagen University, and P. Grandjean,
University of Southern Denmark, unpublished material, April 28, 2000, in Table 7-4, p. 289, NRC 2000.
4-56
Methylmercury Water Quality Criterion 1/3/01
-------
4.2.3.4 Supporting Studies
A second Faroese cohort was recruited from children born between 1994 and 1995. In the study
reported by Steurwald et al. (2000), decreases in neurologic optimality score (NOS) were associated with
increasing cord blood mercury. This association remained statistically significant after adjustment for
confounders (including cord and maternal serum PCB levels). Inspection of data plotted in the paper
indicate that a decrease in NOS was observed in the two highest quartiles; that is, at cord blood mercury
levels greater than 20 ppb. This indicates a dose-dependent effect at levels as low as (or lower than)
those for which neuropsychological deficits were reported in the main study of 7-year-old children
(Grandjean et al, 1997). The size of this study is rather small (N = 182) and involves subtle changes at a
very early developmental period, the clinical implications of which are less clear than the changes found
in the main study of 7-year-olds.
NRC conducted an analysis that combined results from the SCDS, New Zealand, and Faroes studies
(NRC, 2000, pp. 290-294). Their approach was to use a hierarchical random-effects model that followed
a method proposed by Dominici et al. (in press). To inform their analyses, NRC plotted BMDs and
BMDLs (as ppm mercury in maternal hair) for measures from all three studies. For outcomes in the
SCDS for which there were no BMDs, the analysis used an arbitrary value of 150. They concluded from
the plot (Figure 7-3, NRC, 2000, p. 285) that study-to-study variability was large relative to outcome-to-
outcome variability. NRC felt that use of a hierarchical model would allow one to borrow strength from
the different studies to achieve greater precision in BMD and BMDL estimates. The NRC results are
seen in their Table 7-5 (NRC, 2000, p. 291). They present what they refer to as smoothed results, which
reflect reduced random variability. For the Faroes data, the BMDL estimates are not much changed from
the original values; the unsmoothed range of BMDLs is 10 to 15 ppm mercury in maternal hair, while the
smoothed results range from 12 to 15 ppm. The NRC notes that all smoothed BMDLs are closer to their
BMDs; they also concluded that the hierarchical modeling reduced much variability among outcomes but
not among studies.
NRC estimated a central tendency measure, equivalent to a BMD, across all three studies and all
endpoints. They also determined a lower limit based on a theoretical distribution of BMDs, which is the
logical equivalent of a BMDL. These values as well as other estimates derived from the Faroes and New
Zealand studies are in Table 4-7.
Methylmercury Water Quality Criterion 1/3/01
4-57
-------
Table 4-7. Central tendency estimates, ppm mercury in maternal hair3
Approach Original
BMD
Most sensitive endpoint from New Zealand 8
Median endpoint from New Zealand 12
Mean of endpoints from New Zealand 12
Most sensitive endpoint from Faroes 15
Median endpoint from Faroes . 20
Mean of endpoints from Faroes 22
Mean of all endpoints
Integrative analysis
a Source: Table 7-6, NRC 2000, p. 294.
bLogically equivalent to a BMD.
c Logically equivalent to a BMDL.
I values Smoothed values
(BMDL) BMD (BMDL)
(4) 12 (7)
(6) 13 (8)
(6) 13 (8)
(10) 17 (12)
(12) 20 (13)
(12) 21 (13)
(14) (15)
21" (8°)
The external review panel for the methylmercury RfD suggested that a reasonable alternative to
using a single test result as the basis for the RfD would be to develop a composite index from several test
outcomes. Their recommendation was to evaluate mercury-associated endpoints for any potential PCB
effect. The next step would be to use either PCB-adjusted results or only those results with no PCB
effect in some compositing approach to provide a multiendpoint BMDL. The most appropriate
compositing approach would be one with a weighting scheme to account for different sample sizes for
the individual tests.
A second way to proceed would be to use factor analysis to create a composite factor that accounts
for the majority of the variance among the individual test results. The resulting estimate would serve as
the basis for RfD calculation. The peer review panel that suggested this approach noted that it is novel
and would require substantial effort to reanalyze the data (U.S. EPA, 2000f).
EPA has decided that the two suggestions have a great deal of merit. We will pursue some of these
analyses for the extant Faroes and New Zealand data and for the SCDS data on 7-year-old children as
they become available. We felt, however, that the integrative analysis reported by NRC serves as
substantial support for the choice of an endpoint from the Faroese test data. We felt that at this time the
use of NRC's integrated BMD /BMDL or one derived form the suggested alternatives as the sole basis
for an RfD would introduce an unacceptable degree of model uncertainty into the RfD.
4-58
Methylmercury Water Quality Criterion 1/3/01
-------
4.2.3.5 Choice ofEndpoint
The lowest of the BMDLs from the Faroese tests is 46 |o,g/L mercury in cord blood for the CPT
reaction time scores. NRC recommended a different choice. They remarked that in a neuropsychological
test battery, the reliability of the individual endpoints can be highly variable, so the most sensitive
endpoint may not be the most appropriate choice. The Faroes investigators reported difficulties in
administering the CPT. The data from the second half of the cohort were discarded for the analysis of
this endpoint; thus the n was about half that for the other tests. The NRC panel suggested that a more
appropriate choice would be to select the second most sensitive endpoint, the BNT BMDL of 58 ppb
mercury in cord blood (NRC, 2000, p. 300). Interestingly, the BNT had the lowest BMDL in the
analyses based on maternal hair mercury.
The external peer reviewers of the methylmercury RfD disagreed with the NRC choice. The felt
that the use of a single neuropsychological endpoint to form the basis for making a risk assessment is
problematic. They felt that the use of the BNT data from the whole Faroese cohort was not warranted, as
the BMDL thus derived could reflect an effect of PCB exposure. The peer reviewers preferred the BNT
BMDL adjusted for PCB exposure of 71 ppb mercury in cord blood. In their report they noted that the
adverse effect of methylmercury reflected in the BNT scores is not isolated, but rather occurs at levels
not far removed from effects on other neuropsychological tests, providing some assurance of its
credibility. A difficulty with the use of the PCB-adjusted BMDL is that this BMDL is based on scores
from only about one-half of the total cohort. As noted in Section 4.2.3.3, NRC felt it was more
appropriate to use the BMDL from analyses with the larger n.
The peer review panel described three other options for RfD derivation. One option would be to use
the BMDL from the CVLT. The panel noted the clinical relevance and predictive value of this test as the
well as likelihood that there is no influence of PCB exposure on this measure. The major drawback to
this choice is that the BMDL from this test for the full cohort is the highest (103 ppb mercury in cord
blood or 14 ppm mercury in maternal hair) of those listed in Table 4-6. One could easily argue that the
RfD based on this measure is not public health protective. In the light of analyses that indicate that
mercury correlations with test measures remain when the highest exposure subset is eliminated (10 ppm
or more mercury in maternal hair), this would seem a poor choice.
A third option would be to develop a composite index across several measures in the Faroes study.
The peer reviewers suggested that the BMDLs from the statistically significant tests could be developed,
evaluated for effects of PCBs, and composited in some way, such as a geometric mean. The compositing
Methylmercury Water Quality Criterion 1/3/01 4-59
-------
method should consider a weighting scheme to deal with varying sample sizes for the different tests.
NRC essentially did a composite measure with the integrative analysis; for all endpoints in all three large
studies, the BMDL is 8 ppm mercury maternal hair, or 32 ppb cord blood mercury (Table 4-7).
Geometric means for the Faroese measures are in Table 4-8 below. These were calculated separately for
the whole cohort, PCB-adjusted BMDLs, and lowest PCB subset. EPA will pursue the suggestion of a
weighted composite index at a future time.
A final longer term option of the peer review panel was to devise a within-study integrative
multivariate approach using factor analysis for analytical derivation of a composite factor that combines
results across tests with overlapping functional domains. The panel acknowledged that this would
require some statistical methodology development.
EPA prepared a comparison of the NRC and peer-reviewer-recommended approaches, which also
includes the BMDLs from the NRC integrative analysis and geometric means of four scores from the
Faroes. Table 4-8 presents BMDLs in terms of cord blood mercury. These are converted (using a one-
compartment model as in Section 4.4.2) to an ingested dose of methylmercury that would result in the
cord blood level. The last column of Table 4-8 shows the corresponding RfD from application of a UF of
10 (see Section 4.5.6). The calculated RfD values converge at the same point: 0.1 (jg/kg/day. Among all
the endpoints listed, there are few deviations from 0.1 ng/kg/day: 0.2 jxg/kg/day for the CVLT entire
cohort and 0.05 ng/kg/day for CPT and Finger Tapping, lowest PCB subset. For comparative purposes
several measures from the New Zealand data analyses were also included hi Table 4-8; the median
BMDL from the New Zealand study would give an RfD of 0.05 (ig/kg/day. If one were to use the NRC
integrative analysis BMDL equivalent value, the resulting RfD would be 0.05 u.g/kg/day.
Rather than choosing a single measure for the RfD critical endpoint, EPA considers that this RfD is
based on several scores from the Faroes measures. These test scores are all indications of
neuropsychological processes involved with a child's ability of a child to learn and process information.
The BMDLs for these scores are all within a relatively close range. In subsequent sections, one endpoint
is carried through the dose conversion and application of the UF to calculation of the RfD; namely, the
NRC-recommended BMDL of 58 ppb mercury in cord blood from the BNT.
4-60
Methylmercury Water Quality Criterion 1/3/01
-------
Table 4-8. Comparison of BMDLs-endpoint from Faroes, New Zealand and NRC Integrative Analysis3
Test"
BNT Faroes
Whole cohort
PCB adjusted
Lowest PCB
CPT Faroes
Whole cohort
PCB adjusted
Lowest PCB
CVLT Faroes
Whole cohort
PCB adjusted
Lowest PCB
Finger Tap Faroes
Whole cohort
PCB adjusted
Lowest PCB
Geometric mean
Whole cohort
PCB adjusted
Lowest PCB
Median values
Faroes
New Zealand
Smoothed values
BNT Faroes
CPT Faroes
CVLT Faroes
Finger Tap Faroes
MCCPP New
MCMTNew
Integrative
All endpoints
BMDL ppb mercury cord blood
58
71
40
46
49
28
103
78
52
79
66
24
68
65
34
48
24
48
48
60
52
28
32
32
Ingested dose us/kg bw /dayc
1.081
1.323
0.745
0.857
0.913
0.522
1.920
1.454
0.969
1.472
1.230
0.447
1.268
1.212
0.634
0.895
0.447
0.895
0.895
1.118
0.969
0.522
0.596
0.596
RfD ug/kg bw /davd
0.1
0.1
0.1
0.1
0.1
0.05
0.2
0.1
0.1
0.1
0.1
0.05
0.1
0.1
0.1
0.1
0.05
0.1
0.1
0.1
0.1
0.05
0.1
0.1
aBMDLs from NRC (2000), Tables 7-4, 7-5,7-6. Hair mercury was converted to blood mercury using a 250:1 ratio and an
assumption of equivalent maternal and cord levels.
••Abbreviations: BNT, Boston Naming Test; CPT, Continuous Performance Test; CVLT, California Verbal Learning Test;
MCCPP, McCarthy Perceived Performance; MCMT, McCarthy Motor Test.
"Calculated using a one-compartment model as in Section 4.4.2.4.
"Calculated using an UF of 10 as in Section 4.5.6.
Methylmercury Water Quality Criterion 1/3/01
4-61
-------
4.3 CHOICE OF DOSE-RESPONSE APPROACH
4.3.1 Benchmark Versus NOAEL
In recent years, EPA has been moving to use of BMDs versus experimental NOAELs as the
departure point for calculation of RfDs. The Agency is preparing guidance for application of this
methodology. Guidance has been published in the Technical Support Document on Risk Assessment,
Human Health Methodology for Ambient Water Quality Criteria (U.S. EPA, 2000g).
NRC also made comments on the applicability or preference for BMDs over NOAELs (NRC, 2000,
pp. 272-273). They cite comments by several risk assessment scientists on statistical drawbacks to
NOAELs. The NOAEL, for example, must correspond to one of the experimental doses; it can vary
considerably across different experiments. In calculating an RfD, there is no statistical or other treatment
of the data to adjust for the choice of dose groups by different experimenters. NRC notes that the
identification of a no-effect dose group is based on statistical comparisons between exposed and controls;
thus, larger studies have higher power to detect small changes and tend to produce lower NOAELs.
Furthermore, because NOAELs are identified as a consequence of pairwise comparisons, there is no
widely accepted procedure for calculating a NOAEL in settings where exposure is measured on a
relatively continuous scale.
In its guidance documents EPA lists some other advantages of BMD over the LOAEL/NOAEL
approach. The traditional method does not incorporate information on the shape of the dose-response
curve, but rather uses only a single point (NOAEL or LOAEL). This point depends on the number of
doses and spacing of those doses in the experiment. The possible LOAEL/NOAELs are limited to the
discrete values of the experimental doses, whereas the "real" value of the NOAEL could be any value
between the experimental NOAEL and the LOAEL.
The determination of a NOAEL is dependent on the background incidence of the effect in controls.
Statistically significant differences between treatment groups and controls are more difficult to detect if
background incidence is relatively high, even if biologically significant effects are noted.
4-62
Methylmercury Water Quality Criterion 1/3/01
-------
The peer reviewers of the methylmercury RfD provided comment on the appropriateness of the
BMD methodology for the methylmercury human data:
Derivation of LOAELs and NQAELs from the data would require disaggregation of the data based upon
artificial outpoints (e.g., quartiles) to determine which range of exposure appears to be different from the
baseline group. While this approach provides a useful profile of effect with dose (e.g. Fig. 1 of the 1997
Faroes paper), it uses a grouping of the data that makes specifying the threshold less exact than with the more
statistically robust and inclusive benchmark dose approach. The LOAEL/NOAEL approach also does not
factor variability into the estimation of the threshold dose in the health protective way that the BMDL concept
accomplishes. In the LOAEL/NOAEL approach, the more variable the data the higher the LOAELs and
NOAELs tend to become because it is more difficult to define a statistical difference from the control group.
In contrast, greater variability will tend to drive down the estimate of the BMDL since it is the lower 95%
confidence limit estimate on the BMD. (G. Ginzberg in U.S. EPA, 2000f)
NRC recommended and EPA concurred with the use of a BMD approach to calculate the
methylmercury RfD.
4.3.2 Choice of Exposure Metric
NRC discussed at length in its Chapter 4 the suitability of both hair and blood mercury as
biomarkers of exposure. The measurement of mercury exposure in the study population serves two
purposes when applied to risk assessment. The biomarker serves as the surrogate for the methylmercury
dose to the target tissue, in this case fetal brain. As such, the biomarker is one of the coordinates of
inputs to the dose-response models. From this perspective, the ideal biomarker is one that is closest
pharmacokinetically to the target. Of the measurements available, cord blood represents a compartment
closer to fetal brain than does hair, which is an excretion compartment.
The other use of biomarker in this risk assessment is as a surrogate for ingested dose-, the unit in
which an RfD is expressed. The ideal biomarker for this stage is closest pharmacokinetically or has the
best correlation with ingested dose. Maternal hair or blood may be more suitable from this point of view.
Another point to consider in biomarker choice is temporality: is the biomarker an adequate
indicator of exposure during critical developmental windows? NRC noted that cord-blood mercury tends
to reflect exposure in the later stages of pregnancy, whereas hair mercury can be used to determine
exposure at any point in pregnancy, given the appropriate sample. The NRC panel noted that for most
assessment of hair mercury there will be significant uncertainty when attempting to relate a particular
Methylmercury Water Quality Criterion 1/3/01 4-63
-------
hair level to a time-specific dose to the fetal brain. In addition, there is no information on differential
effects of methylmercury at'different periods of gestation; it is in no way certain when critical
developmental windows occur. Considering the information (or lack thereof) on time of exposure
offered by each biomarker, there is no compelling reason to consider one more appropriate than the other.
NRC provided a table (Table 6-1, NRC, 2000, p. 253) that compares test performance associated
with mercury concentration as a function of either cord-blood or maternal hair measurement. This
comparison suggests that the cord-blood measure explains more of the variability in more of the
outcomes than does maternal hair mercury.
In selecting the exposure metric, the above factors were considered. Cord blood is the biomarker
most closely linked (at least conceptually) to the target organ. Cord blood is the marker for which there
are the most associated adverse effects in the Faroes study. Neither cord-blood nor maternal hair
mercury (as generally measured) provides a clear advantage in assessing exposure during putative critical
developmental windows. Maternal hair mercury is conceptually closer to maternal ingested dose than is
the cord-blood compartment. However, sensitivity analyses indicate that the maternal hairrmaternal
blood ratio is a key contributor to variability in calculations of ingested dose (Stern, 1997; Clewell et al.,
1999). On balance, the best choice for exposure metric for RfD calculation is cord-blood mercury.
4.3.3 Choice of BMD
In applying a BMD approach to data that are continuous in effect, there are several interdependent
steps as defined by Gaylor and Slikker (1992). The first is to fit a regression model that characterizes the
mean of the set of outcome measurements as a function of dose; the assumption of a normal distribution
is made. (Choice of model is described in Section 4.3.4). The second step is to define the cutoff for
normal versus abnormal response. This cutoff point (*„) is defined statistically. In the third step, the
dose-specific probability of falling into the abnormal category is determined (P0). One chooses a specific
increase in the frequency of abnormal responses by comparison to background probability; this specific
risk above background risk is the benchmark response, or BMR. The dose at which the BMR is reached
is the BMD. In other words, the BMD is the dose that results in an increased probability of an abnormal
test performance by a benchmark response; that is, from P0 for an unexposed person to P0 + BMR for a
person exposed to the BMD. The last step is to calculate the BMDL or 95% lower limit on the BMD.
Choices for P0 and BMR are described below.
4-64
Methylmercury Water Quality Criterion 1/3/01
-------
One could set P0 based on clinical definitions of adverse response or other information. For
example, long experience with birth weight in a population could prompt a choice of 2500 g as a cutoff
for normal. Alternatively P0 can be set as a fixed percentile of performance in the unexposed population.
For a linear model and random error normally distributed with variance, this has the effect of setting P0 at
a specified number of standard deviations below the mean for the unexposed group. Generally the larger
the P0, the lower the BMD. For the analysis of the behavioral data, including the Faroe study, the NRC
panel (NRC, 2000, p. 298) recommended that P0 = 0.05: that is, that the cutoff for abnormal response be
set at the lowest 5% (5th percentile) of children. This means that the cutoff point (x0) is defined by a
probability of 5% in an unexposed population. It should be noted that specification of P0 for the Faroese
data (or the other human methylmercury studies) is somewhat problematic because there are no subjects
with true zero exposure. The mean response rate at zero is not actually based on observed data but is
extrapolated from the fitted model (Budtz-J0rgensen et al., 1999). Support for P0 of 0.05 is found in
Cramp et al. (2000); the authors note that this choice is "suggested by the convention of considering 95%
of the clinical responses in healthy individuals to define the normal range." EPA agrees that P0 = 0.05 is a
reasonable choice.
BMR is the benchmark response, the specific risk above background risk. In other risk assessments
(mostly on quantal data) it has been set at 0.1, 0.05, or 0.01. In the MSRC, BMDs and BMDLs were
calculated for BMRs of 0.1, 0.05, or 0.01. EPA chose to apply a BMR of 0.1 to the Iraqi data (MSRC
volume V, pp. 6-27-6-28; U.S. EPA, 1997e). This was based on publications by Allen et al. (1994) that
indicated that a 10% risk level roughly correlated with a NOAEL for developmental toxicity data from
controlled animal studies. For a methylmercury RfD based on the Faroese data, NRC recommended that
the BMR be set to 0.05, which would result in a doubling of the number of children with a response at
the 5th percentile of an unexposed population (NRC, 2000, pp. 283, 298).
The NRC panel felt that their choice of a P0 of 0.05 and a BMR of 0.05 was justifiable in terms of
being sufficiently protective of public health. The committee recognized, however, that the choice of P0
and BMR is at the interface of science and policy and should be a science-informed policy judgment.
EPA at this time has no established policy on an acceptable risk level for the effects reported in the
Faroese children. EPA is in the process of publishing guidance on benchmark dose methodology and
processes. Most of the experience that supports this guidance comes from assessment of lexicological
Methylmercury Water Quality Criterion 1/3/01
4-65
-------
(animal) data. The guidance acknowledges that choices of model, and inputs such as P0 and BMR,
should be informed by a consideration of the type of data and the ancillary information on which the
assessment is based. Our decision in the specific case of methylmercury is influenced by the public
health conclusions that NRC articulated: the measured effects in the human studies are sentinels of
adverse outcomes in children, related to their ability to learn and achieve success in educational settings.
Thus, EPA accepts the NRC recommendation to set P0 = 0.05 and BMR = 0.05 in this instance.
4.3.4 Choice of Model
A report prepared for EPA and subsequently published by Budtz-J0rgensen (1999) provided
calculations of BMD and BMDL using square root and log transformations as well as calculations for K-
power models. NRC used these results and similar calculations for the New Zealand and Seychelles
studies to make some assessments of model suitability. They noted great variability in calculated BMDs
and BMDLs as a function of model. This was so despite the inability of standard statistical assessments
of model adequacy to distinguish between models. In response to NRC, Budtz-J0rgensen and colleagues
provided some additional analyses. These were sensitivity analyses that repeated the regression models
after omitting some of the highest observations (E. Budtz-J0rgensen, Copenhagen University, N.
Keiding, Copenhagen University, and P. Grandjean, University of Southern Denmark, unpublished
material, April 28, 2000, quoted in NRC, 2000, p. 293). Their results suggested that the influence of the
extreme observations did not explain the model-to-model variability (NRC, 2000, p. 293).
NRC concluded that the most reliable and defensible results for the purpose of risk assessment are
those based on the K-power model. (NRC, 2000, pp. 293-298). This model takes the following form, as
presented in Budtz-J0rgensen et al. (2000):
where d is the child's mercury dose and K and p are parameters to be estimated. The K-power model
was fit under the constraint that K s 1, so that supralinear models were ruled out. A power of 1 generally
provided the best fit to the Faroese data (Budtz-J0rgensen et al., 2000). With K = 1, the above model is
linear.
NRC observed that in situations where there are no internal controls (i.e., no unexposed
individuals) and where the dose response is relatively flat, the data will often be fit equally well by
4-66
Methylmercury Water Quality Criterion 1/3/01
-------
linear, square-root, and log models. The models can yield very different results for BMD calculations,
however, because these calculations necessitate extrapolating to estimate the mean response at zero
exposure level. Both the square-root and the log models take on a supralinear shape at low doses, leading
to lower estimates of the BMD than do linear or K-power models. The mechanisms by which
methylmercury exerts its neurotoxic effects in developing systems are speculative. However, no likely
mode of action for methylmercury leads one to expect a supralinear dose-response at low dose. Thus,
from a lexicological perspective, the K-power model has greater biological plausibility, because it allows
for the dose-response to take on a sublinear form, if appropriate.
NRC pointed out that the model sensitivity for BMD from the Faroes data appears in conflict with
the concept, put forward by Crump and others, that by estimating risks at moderate levels, such as 5% or
10%, the BMD should be relatively robust to model specification. Budtz-J0rgensen et al. (2000)
responded that this model dependence is a consequence of the lack of true controls (subjects with zero
exposure). The majority of exposures in the Faroes resulted in hair mercury concentrations exceeding 5
ppm (or 24 ppb cord blood). The interquartile range for hair mercury was 3 to 8 ppm (13 to 40 ppb for
cord blood) (Grandjean et al., 1992). Models fit to the Faroese data are in effect capturing the shape of
the dose-response in this middle range of exposure. The NRC report Figure 7-5, taken from Budtz-
J0rgensen et al. (1999), shows dose-response curves fitted to hair mercury data for the linear, square-root,
and log transformations. Budtz-J0rgensen et al. (2000) provided some information on model fit. They
did not present goodness-of-fit statistics per se, but rather tested each model against an expanded model
that included both the linear and logarithmic term. The authors observed that for P0 = 0.05, and with
cord blood as the exposure metric, the logarithmic transformation tended to show a better fit than the
linear model for the following tests: CPT, BNT, and CVLT. There was no difference in fit for the
Finger Tapping and Bender Gestalt test or for any of the five tests when maternal hair mercury was the
biomarker. The NRC notes that variations in estimated BMDs are not explained by differences in how
well the models fit the bulk of the data, but rather by what the models predict for the mean response for
unexposed individuals.
In reaching its conclusion on model choice, NRC concluded that biologically based arguments were
needed. The argument was as follows:
One useful way to think of differences between the various models is that the linear model implicitly assumes
an additive effect of Hg exposure, the log model assumes a multiplicative effect, and the square root lies
somewhere in between. All three models fit essentially equally well to data that for the most part correspond
to concentrations between 2 and 20 ppm in hair. However, the models differ fairly dramatically with regard to
Methylmercury Water Quality Criterion 1/3/01
4-67
-------
how they extrapolate to values below those levels. The linear model would predict that the change in mean
outcome as MeHg concentration goes from 0 to 10 ppm in hair should be the same as the change observed in
the mean outcome as concentration increases from 10 to 20 ppm. In contrast, the log model would predict that
the change in mean outcome associated with any doubling of MeHg concentration should be the same as the
change observed in the mean outcome as concentration increases from 10 to 20 ppm. Thus, the log model
would predict that the same magnitude change in outcome would be expected as the concentration goes from 1
to 2 ppm or from 4 to 8 ppm as that observed for the concentration going from 10 to 20 ppm—that is, the
extrapolation down to zero exposure will predict a very steep slope at low doses. Given the relative absence
of exposures at very low levels, a decision should be made on biological grounds regarding which model
makes the most sense for risk assessment. The committee believes that an additive (linear) or perhaps
sublinear model is the most justifiable from a biological perspective, thus ruling out square-root and log-
transformed models. For MeHg, the committee believes that a good argument can be made for the use of a K-
power model with K constrained to be greater than or equal to 1 (NRC, 2000 p. 297).
4.3.6 Selection of the Point of Departure for the RfD
Based on all considerations in the preceding sections, the following is selected as the basis for the
RfD. Our choice is a benchmark approach using the results of the Faroese tests with significant
associations with cord-blood mercury. As an example, the BNT results for the whole cohort are used.
The K-power model (K ^ 1 to eliminate supralinearity) is the model choice, with P0 = 0.05 and BMR =
0.05. Consistent with other uses of BMD, the 95% lower limit or BMDL is used as the point of
departure for the RfD.
The result for the example calculation is a BMD of 85 ppb and a BMDL of 58 ppb; other BMDs
and BMDLs are given in Table 4-8.
4.4 DOSE CONVERSION
The biomarker of choice for the Faroes data was cord blood and the BMDLs were presented in
units of ppb mercury in cord blood. In order to calculate an RfD, it is necessary to convert this figure to
an ingested daily amount that would result in exposure to the developing fetus at the BMDL level in
terms of ppb mercury in blood. NRC (2000) offered advice on the use of these dose-conversion
procedures.
4-68
Methylmercury Water Quality Criterion 1/3/01
-------
4.4.1 PBPK Models Versus One-Compartment Model
In estimating the 1995 RfD, EPA used a one-compartment model. Since publication of the MSEC,
there have been evaluations of the use of this model and the parameter inputs as well as the discussion of
PBPK models for methylmercury. None of the existing models deal specifically with young children, nor
are there data on methylmercury pharmacokinetics in children.
NRC briefly discussed the PBPK model published by Clewell et al. (1999). This model includes
several fetal compartments that could be considered fetal submodels. NRC noted that this model is
conceptually more accurate and flexible than the one-compartment model. The report also notes that the
complexity of the model makes evaluation of it more problematic (NRC, 2000, p. 84). Moreover, given
the state of the data on methylmercury exposure, it would be necessary to use default values for some
model inputs. These factors add to the overall uncertainty in the use of this or any of the other available
PBPK models for methylmercury. EPA has chosen to use the one-compartment model for dose
conversion for this RfD. This model has shown reasonably good fit to data on mercury blood level
changes in human subjects during and after consumption of methylmercury-contaminated fish (Ginsberg
and Toal, 2000). It has been used by other public health agencies such as WHO and ATSDR (1999).
4.4.2 One-Compartment Model for Methylmercury
4.4.2.1 Description of Model
The model is described by the formula below:
d \iglday =
c x b x y
A x
where
d = daily dietary intake (expressed as ng of methylmercury)
c = concentration in blood (expressed as p.g/L)
b = elimination constant (expressed as days"1)
V = volume of blood in the body (expressed as liters)
A = absorption factor (expressed as a unitless decimal fraction)
f = fraction of daily intake taken up by blood (unitless).
Methylmercury Water Quality Criterion 1/3/01
4-69
-------
The following form of the equation expresses d in units of u.g/kg body weight/day.
d =
c x
where
bw = body weight (expressed in kg).
In this one-compartment model, all maternal compartments are compressed to one: namely, blood.
It is assumed that the blood methylmercury concentration is at steady state. This assumption constitutes
an area of uncertainty with the use of this model. One could either assume that the methylmercury
concentrations of fetal blood and maternal blood are the same or adjust the cord-blood concentration to
maternal levels using an empirically derived factor. There are some published indications that mercury
in cord blood is higher than in maternal blood (for example, Dennis and Fehr, 1975: Pitkin et al., 1976;
Kuhnert et al., 1981). Other publications show that there is no difference in concentration (for example,
Fujita and Takabatake, 1977; Sikorski et al., 1989). EPA has chosen to assume that maternal blood
mercury is at the same level as fetal or cord blood and acknowledges that this is an additional area of
uncertainty in the dose conversion. This is discussed in Section 4.5.4.1.
4.4.2.2 Choice of Parameter Inputs—Distributions Versus Point Estimates
NRC presents an analysis of uncertainty and variability in the values to be used in the equation
above (NRC, 2000, pp. 83-95). Although there are data from human studies that form the basis of the
parameter estimates, it is clear that there is variability (and uncertainty) in these estimates. NRC notes
that each of the model parameters is a random variable best described by a probability distribution. The
ingested methylmercury concentration that leads to the benchmark cord-blood concentration is also a
probability distribution determined by the combination of the distributions of the individual parameters.
NRC cited two analyses of the variability and uncertainty in the ingested dose estimates based on the
one-compartment model applied to maternal hair (Stern, 1997; Swartout and Rice, 2000) as well as
similar analysis of a PBPK model (Clewell et al., 1999). Table 4-9 reproduces NRC's compilation of
those analyses. In this table NRC also presented results of analyses that took maternal blood as the
starting point, rather than maternal hair as was done in the published papers.
4-70
Methylmercury Water Quality Criterion 1/3/01
-------
In 1995, EPA used central tendency estimates (or point estimates intended to reflect central
tendency estimates) for all parameter inputs in the RfD dose conversion. Although this is a reasonable
approach, it does not encompass the range of likely parameter values or the range of estimated ingestion
values. The RfD is not intended to protect only the mid-part of a population, but the whole population
including sensitive subgroups. Thus, if one chooses to use central tendency or point estimates in the dose
Table 4-9. Comparison of Results from Three Analyses of the Interindividual Variability in the Ingested
Dose of MeHg Corresponding to a Given Maternal-Hair or Blood Hg Concentration
Study
Stern (1997)
Swartout and Rice
(2000)
Clewell et al. (1999)
Maternal
medium
Hair
Blood
Hair
1
Bloodc
Hair
Bloodf
50th percentUe"
(ng/kg-d)
0.03-0.05d
(mean = 0.04)
0.01
0.08
0.02
0.08
0.07
50th percentile/
5thb percentile
1.8-2.4
(mean = 2.1)
1.5-2.2
(mean= 1.8)
2.2
2.1
1.5
1.4
50th percentile/
1st percentilec
2.3-3.3
(mean = 2.7)
1.7-3.0
(mean = 2.4)
Data not reported
2.8
1.8
1.7
"Predicted 50th percentile of the ingested dose of methylmercury that corresponds to 1 ppm Hg in hair or 1 ppb in
blood.
bRatio of 50th percentile of ingested dose of methylmercury that corresponds to 1 ppm Hg in hair or 1 ppb in blood
to the 5th percentile.
°Ratio of 50th percentile of ingested dose of methylmercury that corresponds to 1 ppm Hg in hair or 1 ppb in blood
to the 1st percentile.
dRange reflects minimum and maximum values among eight alternative analyses.
"Data from J. Swartout, U.S. Environmental Protection Agency, personal commun.; June 9,2000.
^ata from HJ. Clewell, ICF Consulting, personal commun.; April 19, 2000.
conversion, it is necessary to include a UF in the final RfD calculation to ensure that pharmacokinetic
variability is appropriately factored into the consideration of sensitive subgroups.
The choice of UF can be informed by the analyses of variability presented by NRC. In general, all
three analyses found similar ranges of variability due to pharmacokinetic factors. The ratios of estimated
ingested doses at the 50th percentile/99th percentile ranged from 1.7 to 3.3. If one considers only the
estimates using maternal blood as the starting point, then the range for all three studies is 1.7 to 3.0.
NRC noted that variability was higher when maternal hair, rather than blood mercury was the biomarker
used. In 1997, EPA identified the hair-to-blood ratio as a major contributor to the variability (and thus
uncertainty) in estimating the ingested dose and in the RfD based on it. This provides an additional
rationale for use of the cord-blood-based BMD.
Methylmercury Water Quality Criterion 1/3/01
4-71
-------
In determining the methylmercury RfD, EPA chooses to use point estimates, rather than
distributions, in the dose conversion and to account for uncertainty by application of a numerical UF.
This UF considers the probability distribution that relates biomarker concentration and ingested dose (see
Section 4.5). This approach was recommended in the NRC report. NRC notes that use of parameter
distributions and an ingested dose distribution (the "direct approach") does not eliminate uncertainty. In
the direct approach, one would select an ingested dose corresponding to a BMD blood mercury
concentration for the percentile of the population variability that is to be accounted for; that is, one would
select the 95th or 99th (or some other suitable) percentile. The choice must be made among probability
distributions predicted by analyses such as those done by Stern (1997) and Swartout and Rice (2000).
NRC said that "the differences in the analyses are due to the use of different data sets for parameter
estimates, and there is no clear basis for choosing one data set over another. Even when central-tendency
estimates and uncertainty factors are used, the most appropriate value for each model parameter must be
selected. Selection of different values for model parameters could underlie differences in the modeling
results" (NRC, 2000, pp. 94-95).
EPA chooses to make explicit choices for each dose-conversion parameter and to deal with both the
uncertainty and variability implicit in those choices by the application of a UF in the calculation of the
RfD.
4.4.2.3 Choice of Parameter Inputs—Values for One-Compartment Model Terms
NRC recommended (NRC, 2000, p. 95), that in choices of point estimates EPA should consider the
information and analyses in three publications: Stern (1997), Swartout and Rice (2000), and Clewell et
al. (1999). All are recent contributions to the peer-reviewed literature. In addition, Swartout and Rice
(2000) largely comprises analyses that received extensive scientific review as part of the MSRC (U.S.
EPA, 1997e). EPA found little in Clewell et al. (1999) that could be used directly to make parameter
estimates, but rather used data and analyses from the other two papers. The rationales for use of specific
values for equation parameters follow.
Concentration in blood (c)
The concentration in blood is that corresponding to the BMDL (58 ppb in the example). As noted
above, no numerical change is made to account for any potential differences between maternal blood
mercury level and cord-blood concentration.
4-72
Methylmercury Water Quality Criterion 1/3/01
-------
Fraction of mercury in diet that is absorbed (A)
After administration of radiolabeled methylmercuric nitrate in water to three healthy volunteers,
uptake was reported to be >95% (Aberg et al., 1969). This value is supported by experiments in human
volunteers conducted by Miettinen et al. (1971). These researchers incubated fish liver homogenate with
radiolabeled methylmercury nitrate to produce methylmercury proteinate. The proteinate was then fed to
fish for a week; the fish were killed, cooked, and fed to volunteers after confirmation of methylmercury
concentration. The authors reported that the fraction of the administered dose not excreted in the feces
within 3 to 4 days ranged from 91.2% to 97.0% with a mean of 94%. This fraction was assumed to be
the amount absorbed; it probably includes some inorganic mercury formed from the ingested
methylmercury and subsequently excreted. Stern (1997) noted that this method is most likely to result in
an underestimate. It is generally felt that absorption of ingested methylmercury is high and not likely to
vary a great deal. Use of an absorption factor of 0.95 as was done in the MSRC is reasonable.
Fraction of the absorbed dose that is found in the blood (f)
The MSRC notes that in 1995 EPA used data from Kershaw et al. (1980), Miettinen et al. (1971),
and Sherlock et al. (1984) as the basis for the choice of a value of 0.05 (U.S. EPA, 1997e).
There are currently four published reports of the fraction of absorbed methylmercury dose
distributed to blood volume in humans. Kershaw et al. (1980) reported an average fraction of 5.9% of
absorbed dose in total blood volume, based on a study of five adult male subjects who ingested
methylmercury-contaminated tuna. In a group of nine male and six female volunteers who had received
203Hg-methylmercury in fish, approximately 10% of the total mercury body burden was present in 1 L of
blood in the first few days after exposure; this dropped to approximately 5% over the first 100 days
(Miettinen et al., 1971). In another study, an average value of 1.14% for the percentage of absorbed dose
per kg of blood was derived from data on subjects who consumed a known amount of methylmercury in
fish over a 3-month period (Sherlock et al., 1984). Average daily intake in the study ranged from 43 to
233 ng/day, and there was a dose-related effect on percentage of absorbed dose that ranged from 1.03%
to 1.26% in 1 L of blood. Smith et al. (1994) administered radiolabeled methylmercury to seven
subjects. The paper presented published modeled data rather than observations; the mean fraction of
absorbed dose in blood was 7.7% (SD, 0.88%).
Stern (1997) noted that although the Smith et al. (1994) and Kershaw et al. (1980) data could be fit
by a log-normal distribution, the data sets were too small for a reasonable determination of the
Methylmercury Water Quality Criterion 1/3/01
4-73
-------
underlying distributions. Stern used the mean and standard deviation of those two data sets for average
parameter values as inputs to the log-normal distribution; the average of the means is 0.067. Swartout
and Rice (2000) used the observations published by Kershaw et al. (1980), Miettinen et al. (1971), and
Sherlock et al. (1984) as adjusted for 5 L of blood as inputs with a log-triangular distribution. The
median value was 5.9% or 0.059, close to the values of 0.05 used in the MSRC and by other groups (e.g.,
Berglund et al., 1971, and WHO, 1990).
ATSDR (1999) used a factor of 0.05. They noted that estimates of / for the 6 women from the
study by Sherlock et al. (1984) had an average value of 0.048, as compared with the value of 0.059 for
the 14 men in the same study. ATSDR offered the opinion that these data suggest/may be lower for
women than men. Apparently the study by Miettinen et al. (1971) included six female volunteers (in
addition to nine males), though ATSDR did not comment on whether these data similarly provided any
indication that the fraction daily intake taken up by blood was lower for females. It is not likely that any
of the female subjects were pregnant. Sherlock et al. (1984) published a negative correlation between/
and body weight; thus, if this is generalizable, one would expect/to decrease (as V increases) throughout
pregnancy.
EPA chooses to use the median value of 0.059 published by Swartout and Rice (2000) for/in the
dose conversion.
Elimination constant (b)
Currently, five studies report clearance half-times for methylmercury from blood or hair: Miettinen
et al. (1971), Kershaw et al. (1980), Al-Shahristani et al. (1974), Sherlock et al. (1984), and Smith et al.
(1994). The clearance half-lives for blood in these reports are quite variable, ranging from 32 to 189
days. In the Al-Shahristani et al. (1974) study, 10% of the sample population had mercury half-lives of
110 to 120 days. Average mercury half-lives from the five publications are 45 to 70 days. The MSRC
(U.S. EPA, 1997e) used an average elimination constant from four of the studies (data from Smith et al.
[1994] were not used). The corresponding elimination constant of 0.014 was also noted to be the average
of individual values reported for 20 volunteers ingesting from 42 to 233 |ig mercury/day in fish for 3
months (Sherlock et al., 1982).
Swartout and Rice (2000) applied a log-triangular distribution to the data from the five extant
studies. They note that the distribution is highly skewed and that the median is 53 days; the
corresponding elimination constant is 0.013.
4-74 Methylmercury Water Quality Criterion 1/3/01
-------
Stern (1997) discussed the variability in the data sets. His analysis of variance indicated significant
differences among the sets, which were eliminated when the Al-Shahristani data were removed. The
author observed that the half-lives reported by Al-Shahristani are larger than those observed in the other
studies. Stern offers the opinion that this may be due to the relatively large size of the Al-Shahristani
data set by comparison to the others. Stern says that an alternative explanation is that the Al-Shahristani
data reflect a genetic polymorphism in the metabolism occurring with higher frequency in the Iraqi
population, which was the subject of this study. In his analyses, Stem (1997) treated the Al-Shahristani
data both separately and in combination with the data from the other four studies. He reports a mean
elimination constant of 0.011 for Al-Shahristani data alone; the combined data set mean elimination
constant is 0.014.
The decision to select point estimates for dose conversion parameters was done with the
acknowledgment that some of the variability around these parameters would be truncated. This is being
compensated for by the use of a pharmacokinetic uncertainty factor. Nevertheless, it does not seem
prudent to select a point estimate, which is meant to be reflective of population central tendency, from
one data set only. The two central tendency estimates of Swartout and Rice (2000) and Stem (1997) are
very close in value (0.013 versus 0.014); the differences are presumably due to the application of
different distribution types. The value of 0.014 is used for b in the dose conversion.
Volume of blood in the body (V)
In the MSRC (U.S. EPA, 1997e), blood volume was estimated, as there were no data from the study
population (the 81 pregnant women exposed in the poisoning episode in Iraq). It was noted then that
blood volume is 7% of body weight, as determined by various experimental methods. MSRC assumed an
increase of 20% to 30% (to about 8.5% to 9%) during pregnancy on the basis of the publication by Best
(1961). Specific data for the body weight of Iraqi women were not found. Assuming an average body
weight of 58 kg and a blood volume increase of 9% during pregnancy, a blood volume of 5.22 L was
derived and was rounded to 5 L for the dose conversion.
Stern (1997) cited three studies (Brown et al., 1962; Retzlaff et al., 1969; Huff and Feller, 1956)
wherein correlation of body weight and blood volume were demonstrated. All studies were of U.S.
women, presumably not pregnant at the time of the study. The mean blood volumes for each study were
3.58 L, 3.76 L, and 3.49 L, respectively; the mean of the combined data set is 3.61 L. If one assumes a
30% increase in blood volume with pregnancy, this would be 4.67 L.
Methylmercury Water Quality Criterion 1/3/01
4-75
-------
In their analysis, Swartout and Rice (2000) used data from a cohort of 20 pregnant Nigerian women
(Harrison, 1966). Whole-blood volumes in the third trimester ranged from 4 to 6 L; the mean and median
were both 5 L. Although 5 L is somewhat higher than the blood volume estimated from three studies of
U.S. women, it is a reasonable value to use for V.
Body weight (bw)
The MSRC found no data on body weight for the study population and used a default value of 60
kg (rounded from 58) for an adult female (U.S. EPA, 1997e). Swartout and Rice (2000) in their
distributional analysis used the body weight data collected on the cohort of 20 pregnant Nigerian women
(Harrison, 1966); this was the data set that they used for blood volume. Body weight during the third
trimester of pregnancy ranged from 49.5 kg to 73.9 kg, with a geometric mean of 55 kg. Stern (1997)
used the Third National Health and Nutritional Survey (NHANES JH) data for women 18 to 40 years old
(National Center for Health Statistics, 1995). The mean weight was 66.6 kg and the 50th percentile value
was 62.8 kg. The EPA Methodology for Deriving Ambient Water Quality Criteria for the Protection of
Human Health (U.S. EPA, 2000a) also cites NHANES HI data; in the Agency document, women of
childbearing age were considered to be between the ages of 15 and 44 years old. The median body
weight in this group was 63.2 kg and the mean was 67.3 kg. EPA also cites the earlier analyses of
Ershow and Canter (1989); they do not state the age range but give a median of 64.4 kg and a mean of
65.8 kg. The recommendation in the EPA Methodology was to use a body weight value of 67 kg for a
pregnant woman on the basis of the relatively current data from NHANES ffl. This is the value used for
body weight in the dose conversion.
4.4.2.4 Dose Conversion Using the One-Compartment Model
The parameter values are as follows:
c
b
V
A
f
bw
concentration in blood (expressed as 58 |ig/L)
elimination constant (expressed as 0.014 days"1)
volume of blood in the body (expressed as 5 L)
absorption factor (expressed as 0.95, unitless decimal fraction)
fraction of daily intake taken up by blood (0.059, unitless)
body weight (expressed as 67 kg)
4-76
Methylmercury Water Quality Criterion 1/3/01
-------
d =
c x b x V
A x/x bw
_i
d = 58 \iglL x Q.Q14 days' x 5Z
0.95 x 0.059 x 67 kg
d = 1.081
rounded to 1.0 ng/kg/day. Other BMDLs expressed as ingested maternal dose can be found in Table 4-8.
4.5 CHOICE OF UNCERTAINTY FACTOR
4.5.1 Background
The RfD can be considered a threshold for a population at which it is unlikely that adverse effects
will be observed. In estimating this level from either a NOAEL or a BMD, the risk assessor applies
uncertainty factors; these are used to deal with both experimental and population variability and with
lack of information that results in uncertainty in the risk estimate. For a discussion of uncertainty factors,
refer to the Technical Support Document for Risk Assessment, Human Health Methodology for Ambient
Water Quality Criteria (U.S. EPA, 2000g).
In the MSRC, EPA published qualitative discussions and quantitative analyses of uncertainty and
variability in the RfD based on the Iraqi data (U.S. EPA, 1997e,g). Major sources of uncertainty
identified were these: variability in susceptibility within the study cohort, variability in pharmacokinetic
parameters for methylmercury (particularly biological half-life of methylmercury and the hair-to-blood
ratio for mercury), response classification error, and lack of data on long term sequelae of in utero
exposure. At that time a composite UF of 10 was applied to account for these factors and the EPA policy
choice to use a UF in the absence of a two-generation reproductive bioassay.
NRC considered areas of uncertainty and variability relevant to the generation of an RfD based on
data from the Faroes population and given the current state of the databases on both pharmacokinetics
and effects of methylmercury. The panel concluded that not all sources of uncertainty or variability
require addition of numerical UFs. NRC (NRC, 2000, p. 319) suggests that given the state of the human
data on methylmercury, UFs be considered for two reasons:
Methylmercury Water Quality Criterion 1/3/01
4-77
-------
• If the uncertainty could result in underestimation of the adverse effects of methylmercury exposure
on human health.
• If there is reason to suspect that the U.S. population is more sensitive than the study populations to
the adverse effects of methylmercury.
NRC's recommendation was that a UF of at least 10 be applied to a BMD calculated from the BNT
results from the Faroe Islands study (NRC, 2000, pp. 321-322). EPA is in general agreement with NRC's
conclusions and recommendations and considered them in the choice of the numerical UF. EPA's choice
is to consider the RfD to be based on the group of Faroese neuropsychological measures associated with
cord-blood mercury; the areas of uncertainty and variability are the same for the choice of one test result
(e.g., BNT whole cohort) or the group of test results. Descriptions of areas of uncertainty and variability
and choice of UF are in the following sections.
4.5.2 Toxicodynamics
Individual response to methylmercury can vary as a function of many factors: age, gender, genetic
makeup, health status, nutritional influences (including interaction among dietary components), and
general individual toxicodynamic variability. Individual sensitivity has been noted in the published
human studies; NRC cited the example of members of the Iraqi population who seemed insensitive to
high levels of mercury exposure. EPA believes there are insufficient data to conclude that the U.S.
population is more or less sensitive than the reported human study populations. The U.S. population is
extraordinarily diverse by any measures listed above, certainly by comparison to the Faroese population.
The Faroese population is northern Caucasian, has been relatively isolated, and is thought to be
descended from a small number of so-called founders who settled the islands many generations ago. In
the heterogeneous U.S. population, it is entirely likely that there are individuals both more and less
sensitive to methylmercury toxicity than the cohort studied in the Faroes. As the RfD must be calculated
to include sensitive subpopulations, variability in response to mercury is a consideration. EPA believes
there are insufficient data to support a quantitative analysis of this area of variability and uncertainty for
methylmercury, but that toxicodynamic variability must be considered in the determination of the overall
uncertainty factor.
4-78
Methylmercury Water Quality Criterion 1/3/01
-------
4.5.3 Exposure Estimation as an Area of Uncertainty
Limitations in evaluation of exposure can be an additional source of uncertainty. As the RfD is
based on a developmental outcome, there is particular concern for uncertainty in the linkage between
time and intensity of exposure and critical periods of brain development. As noted before, cord-blood
mercury generally reflects mercury exposure during late pregnancy and does not reflect temporal
variability in exposure level. Use of any biomarker of methylmercury exposure can result in
misclassification of exposure. Generally, exposure misclassification presents a bias to the null; that is,
this source of error leads to decreased ability to detect a real effect. To the degree that there is exposure
misclassification in the critical study, it would be expected to result in underestimation of the
methylmercury effect. At this time there are not data to support a quantitative determination of this area
of uncertainty.
4.5.4 Pharmacokinetic Variability
4.5.4.1 Cord:Maternal Blood Ratios
In its use of the one-compartment model for dose conversion, EPA chose to make no adjustment for
potential differences between fetal and maternal blood mercury levels. Investigators have found that the
placenta is not a barrier to the transfer of methylmercury from the mother to the developing fetus.
Typically, there is a strong correlation between maternal blood mercury concentrations and fetal blood
mercury concentrations, as shown by cord blood.
Review of the literature identified 21 studies that reported cord blood mercury and maternal blood
mercury data (Amin-Zaki et al., 1974; Baglan et al., 1974; Dennis and Fehr, 1975; Pitkin et al., 1976;
Kuhnert et al., 1981; Nishima et al., 1977; Lauwerys et al., 1978; Fujita and Takabatake, 1977; Kuntz et
al., 1982; Tsuchiya et al., 1984; Truska et al., 1989; Sikiorski et al., 1989; Hansen et al., 1990; Soong et
al., 1991; Soria et al., 1992; Ong et al., 1993; Akagi et al., 1997; Yang et al 1997; Ramirez et al., 2000;
Bjerregaard and Hansen, 2000; Vahter et al., 2000). Twenty of the studies provided data in a format that
could be compared with one another. The exception is Truska et al. (1989), whose published data were
based on erythrocyte mercury concentrations without reported hematocrit values. Absence of these
values precluded expressing mercury concentration on a ng/L or ppb whole-blood basis.
Data from 18 of the 20 studies (with a combined total of 2,676 maternal and 2,522 cord-blood
samples) indicated that cord-blood mercury concentration exceeded maternal-blood mercury
Methylmercury Water Quality Criterion 1/3/01 4-79
-------
concentration. Mean values ranged from a ratio of 1.04 (Fujita and Takabatake, 1977) to 2.63 (Amin-
Zaki et al., 1974); the average of mean ratios was 1.55. Two studies reported cord:maternal blood ratios
equal to or less than 1. Kuntz et al. (1982) (based on 57 maternal-cord blood pairs) and Sikorski et al.
(1989) (based on 56 maternal-cord blood pairs) reported cord/maternal blood mercury concentration of
1.0 and 0.83, respectively.
Speciated mercury measurements were performed in 9 studies that included 550 maternal and 526
cord-blood samples. This permitted calculation of the ratios of cord blood methylmercury:maternal blood
methylmercury that are presented in Table 4-10. In all nine studies, the mean values for methylmercury
concentration was higher for cord blood than maternal blood. The number of subjects in these 9 studies
ranged from 9 to 226 pregnant woman-fetal pairs. To deal with this variation in n, Table 4-10 reports
both a simple average of mean ratios (cord methylmercury:maternal methylmercury = 1.68) and the
mean ratio weighted by the number of subjects in the study (ratio =1.73).
Overall, these data indicate that cord-blood mercury is higher than maternal-blood mercury. The
composite ratio from the studies reporting methylmercury concentrations indicates that the cord
blood:maternal blood ratio is around 1.7. These values are ratios of means and do not reflect the full
range of variability in the individual mother-fetal pairs. Vahter et al. (2000 reported the 5* and 95th
percentiles of cord:maternal Hg to be 0.88 and 3.1. Individual data were available from Fujita and
Takabatake (1997); ratios calculated from these data ranged from 0.78 to 4.36.
As indicated in Section 4.4.2.1, EPA chooses not to make a numerical adjustment between cord- .
blood and maternal-blood mercury. Such an adjustment factor would best be calculated after evaluation
of data quality and variability within and between studies. EPA feels that this analysis would be an
important contribution to reducing uncertainty in the RfD. At this time the relationship between cord
blood and maternal-blood mercury is considered an area of uncertainty to be included in the
determination of the UF.
4-80
Methylmercury Water Quality Criterion 1/3/01
-------
Table 4-10. Ratio of Cord to Maternal Blood Methylmercury
Investigator
Nishimaetal., 1977
Kuhnert et al., 1981
Tsuchiya et al., 1984
Hansen et al., 1990
Soriaetal., 1992
Ong et al., 1993
Akagi et al., 1997
Yang etal., 1997
Vahter et al., 2000
Number of Subjects
49 maternal, 49 fetal
29 maternal, 29 fetal
226 maternal, 226 fetal
37 maternal, 37 fetal
19 maternal, 19 fetal
29 maternal, 29 fetal
21 maternal, 21 fetal
9 maternal controls, 9 fetal controls; 9 •
occupationally exposed mothers, 9
occupationally exposed fetuses.
1 12 maternal (gestation week 36), 98
fetal
Arithmetic mean of average ratios of cord:matemal methylmercury
Mean weighted by number of subjects for cord:maternal blood methylmercury
Ratio of CordiMaternal Blood
2.17
1.34
1.60
2.11
1.08
1.65
1.75
1.67 - controls
1.39 - occupationally exposed
1.92
1.68
1.73
4.5.4.2 Other Areas of Pharmacokinetic Variability
There is no specific evidence of genetic polymorphisms that affect methylmercury metabolism or
excretion. Human studies have established, however, that there is great variability in some of the factors
affecting the delivery of ingested methylmercury to target organs. The MSRC sensitivity analysis and
the publication by Swartout and Rice (2000) noted that the greatest variability resided in the hainblood
ratio (not a factor in the current dose conversion), the fraction of absorbed methylmercury found in blood
(/), and the half-life of methylmercury in blood (the reciprocal, b, in the current dose conversion).
NRC presented an analysis of methods of ingested dose reconstruction from biomarker
measurements. NRC noted that cord-blood mercury is closely linked kinetically to the fetal brain
compartment but less closely linked to ingested dose. As described in Section 4.4.2 of this document,
EPA chose a one-compartment model and measures of cord-blood mercury for back-calculation of the
ingested dose of mercury. EPA also chose to use central tendency estimates for the parameters of the
one-compartment model, rather than introduce an additional degree of uncertainty inherent in making
choices of distribution shapes and the portion of the distribution that represents a sensitive population.
Methylmercury Water Quality Criterion 1/3/01
4-81
-------
NRC presented analyses of uncertainty around dose-conversion estimates, which are summarized in
Table 4-9 in Section 4.5.2.2. NRC discussed three independent analyses to characterize toxicokinetic
variability in estimates of ingested dose corresponding to a BMD level in a particular biomarker, whether
maternal hair or cord blood (NRC, 2000, pp. 91-95). These analyses were published by Stern (1997),
Swartout and Rice (2000, after their work on EPA 1997), and Clewell et al. (1999). Each analysis used
Monte Carlo simulation to combine probability distributions for each parameter of the model. For Stern
(1997) and Swartout and Rice (2000), this was the one-compartment model shown in Section 4.4.2.1.
Clewell et al. (1999) used a PBPK model with a fetal submodel. The analyses of the one-compartment
model were done in a similar fashion; distributions for model parameters were determined from the
published literature, and shapes of the distributions were set by the authors. Both analyses assumed
correlations between some model parameters. Stern (1997) assumed that blood volume and body weight
were correlated. Swartout and Rice (2000) made that assumption, as well as these correlations: hair-to-
blood ratio and elimination rate constant, and fraction of absorbed dose in blood and body weight. The
analysis based on the PBPK model also used parameter distribution values from the literature but
included many more parameters than the one-compartment model (and more default distributions for
model parameters).
The three published analyses all took maternal hair mercury as their starting point. NRC asked all
three sets of authors to provide analyses of variability that used maternal blood as the starting point (as a
surrogate for cord blood). These analyses were done by removing the hainblood ratio from the model
and running the Monte Carlo simulations.
Table 4-9 presents median estimates of ingested dose corresponding to 1 ppm maternal hair or 1
ppb maternal blood. Useful points of comparison are the ratios between the 50th percentile estimates and
those at the end of the distribution (5th and 1st percentiles). Table 4-9 shows that using maternal blood
as a starting point, the ratios of 50th percentile: 1st percentile estimates ranges from 1.7 to 3.0. EPA's
interpretation is that a factor of 3 will cover the toxicokinetic variability of 99% of the population. The
uncertainty introduced by assuming cord-blood mercury is equivalent to maternal mercury provides
additional justification for a toxicokinetic UF of 3. The choice of a factor of 3 is consistent with the
standard EPA practice of a using a half-log to account for toxicokinetic variability.
4.5.5 Uncertainty in Choice of Critical Effect
Another critical area discussed by NRC is uncertainty around choice of a critical effect. NRC notes
that developmental neurotoxicity is a sensitive indicator of methylmercury toxicity but that there is some
4-82 Methylmercury Water Quality Criterion 1/3/01
-------
uncertainty as to the likelihood of other effects occurring at even lower levels of exposure. They cite
indications of cardiovascular effects as well as neurotoxic effects uncovered later in life.
EPA agrees that there is a degree of uncertainty in our choice of critical effect; EPA believes this is
not currently amenable to quantitative estimation but must be considered in the setting of the uncertainty
factor. Summarized below are observations that support a concern that developmental neurotoxicity may
not be the most sensitive indicator of methylmercury effects.
4.5.5.1 Cardiovascular Effects
There are some human data linking cardiovascular effects with exposure to elemental, inorganic,
and organic forms of mercury. In addition, there are two recently published studies that show an
association between low-level methylmercury exposure and cardiovascular effects. S0rensen et al.
(1999) reported that in a study of 1,000 7-year-old Faroese children, diastolic and systolic blood
pressures increased by 13.9 and 14.6 mm Hg, respectively, as the cord-blood mercury increased from 1 to
10 p.g/L. They also reported a 47% decrease in heart rate variability (an indication of cardiac autonomic
control) for the same increase in cord-blood mercury. Salonen et al. (1995) reported effects in adults
from a study of 1,833 Finnish men. Over the 7-year observation period, men with hair mercury in the
highest tertile (2 ppm or higher) had a 2.0 times greater risk of acute myocardial infarction than the rest
of the study population.
As indicated by the Salonen (1995) study, the relatively subtle effects of methylmercury on
cardiovascular indices can have public health implications. There is an analogous situation with lead
exposure. Pirkle et al. (1985) reported on analyses of NHANES II data comparing the relationship
between systolic and diastolic blood pressure to blood lead levels. They included in their model the 37%
decrease in mean blood lead levels that was observed in white adult males between 1976 and 1980.
Their calculation predicted a 4.7% decrease in the incidence of fatal and nonfatal myocardial infarction
over 10 years, a 6.7% decrease in the incidence of fatal and nonfatal strokes over 10 years, and a 5.5%
decrease in the incidence of death from all causes over 11.5 years.
4.5.5.2 Persistent and Delayed Neurotoxicity
Another area of concern is the onset or exacerbation of neurological deficits in aging populations
exposed in utero or as children. There are indications of this in the followup studies of the Minamata
population. These present evidence that neurological dysfunction among people who have been exposed
Methylmercury Water Quality Criterion 1/3/01 4-83
-------
to methylmercury becomes more pronounced with aging. This heightened diminution of function is
greater than that attributable to either age or methylmercury exposure alone. Specifically, Kinjo et al.
(1993) surveyed 1,144 current patients with Minamata disease (MD) aged 40 or over and an equal
number of neighbor controls matched by age and sex. MD patients have symptoms of sensory
disturbance at a high prevalence rate (e.g., hypoesthesia of mouth, -20% to 29% of subjects;
hypoesthesia of limbs, -66% to 90% of subjects; dysesthesia of limbs, -83% to 93%; weakness, -75% to
84%), but these problems did not systematically increase with age. However, the MD patients did show,
as a function of age, increased difficulties in speaking, tremor, stumbling, and difficulties with buttoning,
clothing, or hearing. Although such changes also occurred among controls, evaluation of odds ratios
showed that the MD patients had higher prevalence rates than the controls for 18 separate problems
including those specifically listed above. Also evaluated were "acts of daily living" (ADL) that included
the abilities to independently eat, bathe, wash, dress, and use the toilet. Among subjects under age 60
there were no significant differences in ADL abilities between MD patients and controls. However,
among patients aged 60 or greater there were significantly lower ADL abilities among MD patients than
among age-matched controls. A conclusion of the Kinjo et al. study is that the prevalence of deficits was
relatively greater in cases compared with controls as a function of increasing age. In other words,
exposure to methylmercury three decades earlier accelerated the aging process in aged individuals
relative to younger ones.
There has also been evaluation of the health status of people living in methylmercury-polluted areas
who were not designated as MD patients. Later followup by Fukuda et al. (1999) evaluated 1,304 adults
who lived in a methylmercury-polluted area near Minamata City in Kumamoto Prefecture in Japan (but
were not designated MD patients) and 446 age-matched adults in a non-mercury-polluted area of Japan.
All subjects were older than 40 years of age. A questionnaire survey evaluated 64 complaints that could
be grouped as nonspecific, sensory, arthritic, and muscular. Complaints identified among male and
female subjects that were significantly higher in methylmercury-contaminated areas included heart
palpitation, dysesthesia, staggering when standing, resting and intention tremor in the hands, dizziness
(especially when standing), low-tone tinnitus, low pain sensation in hands and legs, and (among women
only) loss of touch sensations in hands and legs.
Animal studies lend support to the conclusion that methylmercury can have delayed effects that are
uncovered with age. Spyker (1975) exposed mice during gestation and lactation to methylmercury.
Offspring noted to be normal at birth developed deficits in exploratory behavior and swimming ability at
1 month; neuromuscular and immune effects were noted as the animals reached 1 year of age. Rice
(1989a) exposed monkeys to 50 [ig/kg/day methylmercury for the first 7 years of life. The animals were
4-84 Methylmercury Water Quality Criterion 1/3/01
-------
observed with motor incoordination only when they reached the age of 14; subsequent testing showed
effects on somatosensory functioning (Rice and Gilbert, 1995). Rice (1998) also exposed monkeys in
utero and for the first 4 years. Exposure to 10 to 50 |xg/kg/day was observed to result in decreased
auditory function compared with controls when the animals were tested at 11 and 19 years. The deficit at
19 years was relatively greater than at 11 years, providing evidence for an interaction of aging and
methylmercury exposure on auditory impairment. Rats exposed to methylmercury in utero through 16
days of age exhibited a decline in performance in a task that required a substantial motor output at an
earlier age than did control rats; high-dose rats exhibited a decline in performance at about 500 days of
age compared with 950 days for controls (Newland and Rasmussen, 2000), with no differences between
groups in survival time. All of these observations are consistent with a hypothesis that early life or in
utero exposure to methylmercury can have adverse long-term sequelae that may not be detected in
childhood.
4.5.5.3 Reproductive Effects
EPA has a concern for potential reproductive effects of methylmercury. There are no studies of
reproductive deficits in humans exposed to low-dose methylmercury. Bakir et al. (1973) did comment on
the low number of pregnant women in the Iraqi population exposed to methylmercury in treated grain.
They noted that among the 6,350 cases admitted to the hospital for toxicity, they would have expected
150 pregnancies; only 31 were reported. There are no two-generation reproductive assays for
methylmercury. Shorter term studies in rodents and guinea pigs have reported effects including low
sperm counts, testicular tubule atrophy, reduced litter size, decreased fetal survival, resorptions, and fetal
malformations (Khera, 1973; Lee and Han, 1995; Hughes and Annau, 1976; Fuyuta et al., 1978, 1979;
Hirano et al., 1986; Mitsumori et al., 1990; Inouye and Kajiwara, 1988). Burbacher et al. (1988) reported
decreased conception rates, early abortions, and stillbirths in Macaco, fascicularis monkeys treated with
methylmercury hydroxide; the NOAEL for this study was 0.05 mg/kg/day. In a study of male Macaco.
fascicularis (Mohamed et al., 1987), a LOAEL for sperm abnormalities was 0.05 mg/kg/day.
The MSRC did an evaluation of the potential for methylmercury to be a germ-cell mutagen.
Methylmercury is clastogenic but does not appear to cause point mutations. Methylmercury is widely
distributed in the body, crossing both blood-brain and placental barriers in humans. Data indicate that
methylmercury administered intraperitoneally. reaches germ cells and may produce adverse effects.
When Suter (1975) mated female mice to treated males, he observed a slight reduction in both numbers
of implantations and viable embryos; this was true for one mouse strain but not for another tested at the
Methylmercury Water Quality Criterion 1/3/01
4-85
-------
same time. When Syrian hamsters were treated intraperitoneally with methylmercury, aneuploidy but not
chromosomal aberrations was seen in oocytes (Mailhes, 1983). Sex-linked recessive lethal mutations
were increased in Drosophila melanogaster given dietary methylmercury (Ramel, 1972). Watanabe et al.
(1982) noted some decrease in ovulation in hamsters treated subcutaneously with methylmercury, further
indication that methylmercury is distributed to female gonadal tissue. Studies have reported increased
incidence of chromosome aberrations (Skerfving et al., 1970,1974) or sister chromatid exchange (Wulf
et al., 1986) in lymphocytes of humans ingesting mercury-contaminated fish or meat. Chromosome
aberrations have been reported in cats treated in vivo and in cultured human lymphocytes in vitro.
Evidence of DNA damage has been shown in a number of in vitro systems. The MSRC (U.S. EPA
1997e) concluded that because there are data for mammalian germ-cell chromosome aberrations and
limited data from a heritable mutation study, methylmercury is placed in a group of high concern for
potential human germ-cell mutagenicity. The only factor keeping methylmercury from the highest level
of concern is lack of positive results in a heritable mutation assay.
In summary, there is increasing weight of evidence for effects other than neurodevelopmental that
may be associated with low-dose methylmercury exposure.
4.5.6 Choice of Uncertainty Factor
For this methylmercury RfD the two major areas of uncertainty that can be addressed with a UF are
interindividual toxicokinetic variability in ingested dose estimation and pharmacodynamic variability and
uncertainty. For the former, EPA relied in part on the NRC analyses of variability in the pharmacokinetic
factors underlying the conversion of a biomarker level of methylmercury to an ingested daily dose of
methylmercury that corresponds to that level. We chose not to make a numerical adjustment in the dose
conversion for the potential differences in cord vs. maternal blood mercury level, but rather consider this
an additional area of toxicokinetic uncertainty. A quantitative uncertainty analysis was not feasible for
toxicodynamics. A.common practice is to apply a threefold UF for toxicodynamic variability and
uncertainty.
In the calculation of this methylmercury RfD, a composite UF of 10 is used. This is to account for
the following factors:
• Pharmacokinetic variability and uncertainty in estimating an ingested mercury dose from cord
blood. A factor of 3 is applied for this area.
4-86 Methylmercury Water Quality Criterion 1/3/01
-------
• Pharmacodynamic variability and uncertainty. A factor of 3 is applied for this area.
There are additional areas of concern in this risk estimate that lend support to an overall factor of
10. These include the following: inability to quantify long-term sequelae, lack of a two-generation
reproductive effects assay, and issues on selection of critical effect (concern that there may be observable
methylmercury effects at exposures below the BMDL). Section 4.5.5 discusses some of the concerns on
selection of the critical effect. In this context one must also consider the analyses of the Faroese
neuropsychological data wherein the observations in the most highly exposed subgroup were excluded
from the model. Associations remained significant when the part of the cohort with maternal hair
mercury concentrations greater than 10 ppm was excluded from the analyses. This indicates that it would
be reasonable to expect some percentage of the population to show effects at or below 10 ppm hair
mercury or at levels at or below 40 ppb cord blood. Given the overall robustness of the methylmercury
database, but in consideration of the above areas of uncertainty, a composite factor of 10 is warranted.
4.6 CALCULATION OF THE RfD
The critical endpoint is drawn from the series of neuropsychological test results reported from the
Faroese cohort. The BMDLs calculated on these endpoints are in Table 4-8. The ingested doses in
Hg/kg bw/day that correspond to the BMDLs range from 0.447 to 1.92. The ingested dose for the BNT
whole-cohort BMDL is 1.081 (J-g/kg bw/day, rounded to 1.0 p-g/kg bw/day.
For methylmercury, the RfD is calculated as follows:
RfD =
BMD
- 0.1 fj.g/kg/day.
UF x MF
1.0 \ig/kg- day
10
= 1 x 10" mg/kg-day
Methylmercury Water Quality Criterion 1/3/01
4-87
-------
As shown in Table 4-5, an RfD of 0.1 jig/kg bw/day reflects the range of neuropsychological test
results in the Faroese children exposed in utero. These test scores are all indications of
neuropsychological processes that are involved with the ability of a child to learn and process
information. In the studies so far published on subtle neuropsychological effects in children, there has
been no definitive separation of prenatal and postnatal exposure that would permit dose-response
modeling. That is, there are currently no data that would support the derivation of a child (vs. general
population) RfD. This RfD is applicable to lifetime daily exposure for all populations including sensitive
subgroups. It is not a developmental RfD per se, and its use is not restricted to pregnancy or
developmental periods.
4-88
Methylmercury Water Quality Criterion 1/3/01
-------
5.0 EXPOSURE ASSESSMENT
5.1 OVERVIEW OF RELATIVE SOURCE CONTRIBUTION ANALYSIS
When a water quality criterion is based on noncarcinogenic effects, anticipated exposures from
sources other than drinking water and fish ingestion are taken into account so that the entire RfD is not
attributed to drinking water and freshwater/estuarine fish consumption alone. The amount of exposure
attributed to each source compared with total exposure is called the relative source contribution (RSC)
analysis. The RfD used in calculating the criterion incorporates the RSC to ensure that the criterion is
protective enough, given the other anticipated sources of exposure. The method of accounting for
nonwater exposure sources is described in more detail in the revised 2000 Human Health Methodology
(U.S. EPA, 2000a).
The method of determining the RSC differs depending on several factors, including (1) the
magnitude of total exposure compared with the RfD, (2) the adequacy of the exposure data available, (3)
whether more than one guidance or criterion is to be set for a contaminant, and (4) whether there is more
than one significant exposure source for the chemical and population of concern. The population of
concern for methylmercury is discussed in Section 5.2. The sources of exposure to methylmercury and
estimates of exposure used to determine the RSC for the identified population are discussed in Sections
5.3 through 5.4. Section 5.5 summarizes the exposure uncertainties based on data adequacy. Finally,
Section 5.6 provides the RSC estimates for methylmercury.
5.2 POPULATION OF CONCERN
Methylmercury is a highly toxic contaminant that can cause a variety of adverse health effects.
Toxicity has been observed in adults exposed through consumption of contaminated food. Toxic effects
and subtle neuropsychological effects have been seen in children exposed in utero when their mothers
consumed contaminated food while pregnant. The RfD (see section 4) is based on changes in
neuropsychological measures in children exposed in utero. The choice was made to use a developmental
endpoint, as this appeared to be the most sensitive indicator of a methylmercury effect. As discussed in
section 4, there is concern that other less-studied effects may occur at lower doses. There is also concern
(based on recent reports on the Minamata, Japan, population) that exposure in utero or in childhood
could result in subtle impairments that would not be detectable until middle age or older.
Methylmercury Water Quality Criterion 1/3/01
5-1
-------
The RfD for methylmercury was not calculated to be a developmental RfD only. It is intended to
serve as a level of exposure without expectation of adverse effects when that exposure is encountered on
a daily basis for a lifetime.
In the studies on subtle neuropsychological effects in children published so far, there has been no
definitive separation of prenatal and postnatal exposure that would permit dose-response modeling. That
is, there are currently no data that would support the derivation of a child RfD versus a general
population RfD.
Therefore, the population at risk evaluated for the methylmercury criterion is adults in the general
population, not only the developing fetus or child.
5.3 OVERVIEW OF POTENTIAL FOR EXPOSURE
The sources and fate of methylmercury are discussed in detail in Volume in of the Mercury Study
Report to Congress (MSRC) (U.S. EPA, 1997b). The MSRC exposure assessment is in Volume IV (U.S.
EPA, 1997c). A brief summary of the information in that document is presented here. Methylmercury
occurs naturally in the environment. It is readily produced from inorganic mercury in fresh and marine
surface waters and sediments through the methylating action of certain microorganisms. Bacterial
methylation rates appear to increase under anaerobic conditions, elevated temperatures, and low pH.
Methylmercury generally constitutes no more than 25% of the total mercury in surface water; typically,
less than 10% is observed (U.S. EPA, 1997b). According to the MSRC, mercury cycles in the
environment as a result of natural and anthropogenic activities. Most of the mercury in the atmosphere is
elemental mercury vapor, which can remain there for as much as 1 year and, due to atmospheric
mobilization, can be widely dispersed and transported thousands of miles from likely sources of emission
(U.S. EPA, 1997b). However, the MSRC also clearly states that methylmercury is the chemical species
of concern due to its fate and transport to waterbodies and sediments, and its subsequent bioaccumulation
in the aquatic food web.
Because the source of most mercury is deposition from atmospheric mercury emissions, ingestion is
an indirect route of exposure. The MSRC included numerous computer-simulated estimates of mercury
exposure for selected population scenarios, based on fate and transport models (see U.S. EPA, 1997b,c).
These are summarized throughout this chapter in the Predicted Concentrations subsections. Further
exposure assessment information is presented in Volumes HI and IV of the MSRC (U.S. EPA, 1997b,c)
5-2 Methylmercury Water Quality Criterion 1/3/01
-------
and a characterization of human health from methylmercury exposure is discussed in detail in Volume
VIE (U.S. EPA, 1997g). That exposure assessment information is summarized throughout this chapter.
The primary source of human exposure to methylmercury is through consumption of contaminated fish
and seafood. This reflects the tendency of aquatic organisms to rapidly absorb methylmercury and to
store it for long periods of time in their muscle tissue, thus accumulating it to levels that are potentially
toxic to humans who eat fish and shellfish. The concentrations of methylmercury in fish tissue are highly
variable across water bodies. Within a water body, methylmercury concentration generally increases
with fish size and trophic level.
Derivation of the water quality criterion requires that intake of methylmercury from other sources
of exposure be evaluated for comparison with intake from water and/or freshwater and estuarine fish. In
addition to its occurrence in water and freshwater and estuarine fish, methylmercury occurs in soil, air,
marine fish and other seafood, and nonfish foods. Intake of these media thus represent potential
pathways for exposure. Other potential routes include occupational exposure and erosion of dental
amalgams. Estimates of intake from these sources are presented in Section 5.4 below. Assessment of
these sources of methylmercury clearly indicates that substantially all exposure to methylmercury occurs
from the ingestion of contaminated fish. The other sources of exposure (water, nonfish foods, air, and
soil) are all several orders of magnitude less than exposures from fish consumption.
5.4 ESTIMATES OF OCCURRENCE AND EXPOSURE FROM ENVIRONMENTAL MEDIA
This section reports data available for the estimation of methylmercury intake from relevant
exposure sources. Exposure may occur from several environmental sources including soil, sediment,
ambient surface water, drinking water, food products, and air. Human exposures are estimated by
combining information on the occurrence of methylmercury in environmental media with intake rates for
these media. Information on intake assumptions, environmental concentrations, and estimated exposure
are reported by medium below.
Methylmercury Water Quality Criterion 1/3/01
5-3
-------
Table 5-1. Exposure parameters used in derivation of the water quality criterion
Parameter
Body Weight, kg
Drinking Water Intake, L/day
Freshwater/Estuarine Fish Intake, gin/day
Inhalation, nfYday
Soil Ingestion, g/day
Mean Marine Fish Intake, kg/day
Median Marine Fish Intake, kg/day
90th Percentile Marine Fish Intake,
g/day
Population
Children
(0-14 years)
30
1.0
156.3"
10.4
0.0001, 0.01"
74.9"
59.71"
152.29"
Women of
Childbearing
Age
(15-44 years)
67
2.0
165.5"
11
0.00005
91.04"
75.48"
188.35"
Adults in the
General
Population
70
2.0
17.5C
20
0.00005
12.46C
Oc
49.16=
Source
U.S. EPA
(2000a)
U.S. EPA
(2000a)
U.S. EPA
(2000a)
U.S. EPA
(1994, 1997h)d
U.S. EPA
(1997h)
U.S. EPA
(2000b)
U.S. EPA
(2000b)
U.S. EPA
(2000b)
"Pica child soil ingestion
"For children and women of childbearing age, intake rates are estimates of "consumers only" data (as described in
U.S. EPA, 2000b).
Tor adults in the general population, intake rates are estimates of all survey respondents to derive an estimate of
long-term consumption (U.S. EPA).
•"Inhalation rates for children and women of childbearing age from U.S. EPA, 1997h. Inhalation rates for adults in
the general population from U.S. EPA (1994).
5.4.1 Exposure Intake Parameters
Exposure parameters selected for derivation of the water quality criterion should reflect the
population to be protected. Default values for most exposure parameters are provided in the 2000
Human Health Methodology (U.S. EPA, 2000a). Where necessary, values for parameters not specified in
the Methodology were obtained from the Exposure Factors Handbook (U.S. EPA, 1997h). Parameter
values used to estimate intake of methylmercury by children aged 0-14 years, women of childbearing age,
and adults in the general population are summarized in Table 5-1.
5-4
Methylmercury Water Quality Criterion 1/3/01
-------
5.4.2 Intake from Drinking Water/Ambient Water
In cases where the water quality criterion is based on fish intake only, drinking water intake is
accounted for as a separate exposure. In these instances, information on treated drinking water, if
available, is the relevant information to use when accounting for other sources of exposure. Measured
concentrations for methylmercury in drinking water and raw surface and ground source waters have been
reported in the MSRC (U.S. EPA, 1997c). Predicted concentrations and ingestion rates summarized in
this section are based on computer simulation models described in Volume IV of the MSRC (U.S. EPA,
1997c).
5.4.2.1 Measured Concentrations in Water
Raw Surface Water. Studies in the United States and Europe suggest that the concentrations of
methylmercury in raw surface water are highly variable (U.S. EPA, 1997b). Properties reported to
influence the levels of methylmercury in water bodies include proximity to a point source of mercury,
pH, anoxia, dissolved organic carbon, and the presence of wetlands (U.S. EPA, 1997b). Estimates of the
percent of total mercury in surface waters that exists as methylmercury are available from a number of
studies. The available data suggest that methylmercury generally constitutes less than 20% of the total
mercury in the water column (Kudo et al., 1982; Parks et al., 1989; Bloom and Effler, 1990; Watras et al.,
1995a). In lakes without point source discharges, methylmercury frequently constitutes 10% or less of
total mercury in the water column (Lee and Hultberg, 1990; Bloom et al., 1991; Lindqvist, 1991;
Porcella et al., 1991; Watras and Bloom, 1992; Driscoll et al., 1994, 1995; Watras et al., 1995b). U.S.
EPA (1997b) reported the use of Monte Carlo simulation to derive a point estimate of 0.078 for the
fraction of total mercury present as methylmercury in the epilimnion (water column above the
thermocline) of lakes for the purpose of estimating a bioaccumulation factor (BAF) for trophic level 4.
Speciation data used as input for the simulation are shown in Table 5-2.
Data for measured concentrations of methylmercury and total mercury in ambient water as
presented in the MSRC (U.S. EPA, 1997b) are summarized in Table 5-3. Since publication of the
MSRC, Krabbenhoft et al. (1999) reported concentrations of total mercury and methylmercury in surface
water samples collected as part of a U.S. Geological Survey (USGS) national scale pilot study to examine
relations for total mercury and methylmercury in water, sediment, and fish. Water samples were
collected in the summer and fall of 1998 at 106 sites from 21 basins across the United States, including
Alaska and Hawaii. The sampling sites spanned the dominant east-to-west mercury deposition gradient
Methylmercury Water Quality Criterion 1/3/01
5-5
-------
Table 5-2. Data Used in the Monte Carlo Simulation to Estimate the Fraction of Total Dissolved
Mercury in the Epilimnion Present as Methylmercury
Fraction of Total Mercury
Present as Methylmercury
0.046
0.054
0.059
0.089
0.089
0.092
0.15
Location
Pallette Lake, WI
Oregon Pond, NY
Lake Michigan
Clear Lake, CA
Onondaga Lake, NY
Iso Valkjarvi, Finland
22 lake aggregate, WI
Reference
Bloom etal. (1991)
Driscoll et al. (1995)
Mason and Sullivan (1997)
Suchanek et al. (1993)
Henry et al. (1995)
Rask and Verta (1995)
Watras et al. (1995a,b)
Source: U.S. EPA (1997c, Appendix D)
and represented a wide range of environmental settings. The study authors reported that most (number
not reported) samples were collected from streams. Total mercury was measured using U.S. EPA
Method 1631 with detection by cold vapor atomic fluorescence spectroscopy (CVAFS). Methylmercury
was analyzed by distillation and aqueous phase ethylation, with detection by CVAFS. The detection
limits for total mercury and methylmercury were 0.04 ng/L and 0.025 ng/L, respectively (Olson and
DeWild, 1999). Of the 106 total sites, 21 were classified as background or reference sites. The mean
concentration for methylmercury at background sites was 0.13 ng/L, which represented 3.4% of the mean
total mercury concentration. When all sites were considered, the mean methylmercury concentration
(104 sites) was 0.15 ± 0.26 ng/L (range 0.01 to 1.481 ng/L). The median value was 0.06 ng/L. The
difference in mean and median values was attributed to high mercury concentrations at sites impacted by
mining activities, which resulted in a skewed distribution. Methylmercury constituted 1% to 11% of total
mercury concentration in the 21 study basins.
Other measured concentrations of total mercury and methylmercury in fresh water as reported in
the MSRC (U.S. EPA, 1997b) are summarized in Table 5-3. Reported values for methylmercury
measured at two sites in the United States ranged from less than 0.004 ng/L to 0.06 ng/L. The New
Jersey Department of Environmental Protection and Energy (NJDEPE) (1993) reported total mercury
concentrations for lakes of 0.04 to 74 ng/L and values of 1 to 7 ng/L for rivers and streams. Based on the
5-6
Methylmercury Water Quality Criterion 1/3/01
-------
Table 5-3. Measured Methylmercury Concentrations in Surface Fresh Water
Study Description
Lake Crescent, WA
Little Rock Lake
(reference basin)
Lake Michigan
(total)
Lake Champlain
Lakes
Rivers and Streams
USGS National
Mercury Pilot Study
(predominately
streams)
Total Mercury
(ng/L)
0.163
1.0-1.2
7.2 microlayer
8.0 at 0.3m
6.3 at 10m
(filtered). 3.4
microlayer
3.2 at 0.3m
2.2 at 15m
0.04 - 74
1-7
3.43 Background
16.6 All sites
Methylmercury
(ng/L)
<0.004
0.045-0.06
NA
NA
NA
0.13 Background
0.15 All sites
Methylmercury
% of Total
<2.5
mean of 5
NA
NA
NA
3.4
1-11
Reference
Bloom and Watras
(1989)a
Watras and Bloom
(1992)a
Cleckneretal. (1995)a
Cleckner et al. (1995)a
NJDEPE (1993)a
Krabbenhoft et al. (1999)
' As reported in U.S. EPA (1997c)
NA Not available
U.S. EPA (1997b) Monte Carlo estimate for speciation (0.078), these values would correspond to
approximate methylmercury concentrations of 0.003 to 6 ng/L for lakes and 0.078 to 0.55 ng/L for rivers
and streams. The MSRC did not indicate whether the NJDEPE (1993) data represented measures of
central tendency.
Ground Water. Nationally aggregated data for mercury or methylmercury concentrations in ground
water were not reported in the MSRC (U.S. EPA, 1997b). Local estimates of concentration are available
from three studies. Krabbenhoft and Babiarz (1992) reported mercury levels of 2 to 4 ng/L in near-
surface ground water in remote areas of Wisconsin, with a maximum of 0.3 ng/L (roughly 7.5% to 15%
of total mercury concentration) occurring as methylmercury. Bloom et al. (1989) reported a value of 0.3
ng/L for total mercury in a Washington state well. In contrast to these comparatively low concentrations,
Dooley (1992) reported total mercury levels up to and exceeding 2,000 ng/L in southern New Jersey
domestic wells.
Methylmercury Water Quality Criterion 1/3/01
5-7
-------
Drinking Water. Much of the data reported for total mercury concentration in drinking water is
below the detection limit of 100 ng/L associated with older methods of analysis (U.S. EPA, 1997b).
Lindqvist and Rodhe (1985) estimated that the concentration range of mercury in drinking water is the
same as rain, with an average level of total mercury in drinking water of 25 ng/L. NJDEPE (1993)
reported,a range of 0.3 to 25 ng/L for total mercury in U.S. drinking and tap water. Speciation data for
mercury in drinking water are not available, but may be similar to those observed for rain water (U.S.
EPA, 1997c). The percentage of total mercury that is methylmercury in rain water ranged from 0.1% to
6.3% in two studies reported by Lee and Iverfeldt (1991) and Fitzgerald et al. (1991). The high end of
this range approaches the point estimate of 7.8% derived for the fraction of methylmercury in the water
column of lakes using Monte Carlo simulation (U.S. EPA, 1997b). Assuming that 7.8% of the total
mercury is methylmercury (U.S. EPA, 1997b), these data suggest a crude estimate of methylmercury
concentration in drinking and tap water ranging from 0.023 ng/L to 1.95 ng/L.
5.4.2.2 Predicted Concentrations in Water
U.S. EPA (1997b) reported the results of watershed fate and transport modeling conducted to
predict the background concentration of mercury in water bodies. Atmospheric concentrations and
deposition rates were used as inputs to the IEM-2M model. The IEM-2M model is composed of two
integrated models that simulate mercury fate using mass balance equations that describe processes in
watershed soils and a shallow lake. Using this approach, background levels of total dissolved mercury
concentrations in the water column of 0.9 and 0.2 ng/L were predicted for hypothetical Eastern and
Western U.S. sites, respectively. More than 80% of the total mercury in the water column was predicted
to occur as the inorganic divalent species. As indicated above, the fraction of the predicted background
concentration occurring as methylmercury was 7.8% (U.S. EPA, 1997b).
In the MSRC, the background values reported above were used as inputs to a localized model
analysis that examined the impact of a variety of anthropogenic emission sources (municipal waste
combustors, hospital medical waste incinerators, utility boilers, chlor-alkali plant) on methylmercury
concentrations in the water column at distances of 2.5,10, or 25 km from the source. This effort was
undertaken because some monitoring studies suggest that measured mercury concentrations may be
higher in areas adjacent to stationary industrial and combustion sources known to emit mercury (U.S.
EPA, 1997b). Results of this analysis are of relevance to derivation of the water quality criterion because
they include data specifically for predicted methylmercury concentrations, and thus permit comparison
with measured concentrations.
5-8
Methylmercury Water Quality Criterion 1/3/01
-------
The Industrial Source Code air dispersion model (ISC3) was used for simulation. Hypothetical
facilities were defined to represent actual emissions from existing industrial processes and combustion
sources; these were situated in hypothetical locations intended to simulate a site in either the Western or
Eastern United States (U.S. EPA, 1997b). Input values for air concentrations consisted of simulated
concentration results (50th and 90th percentile values) obtained using the regional Lagrangian model of
air pollution (RELMAP). The assumptions and inputs utilized in this modeling effort are described in
detail in Volume m of the MSRC (U.S. EPA, 1997b).
Results for predicted methylmercury concentrations in water are illustrated in Table 5-4. Predicted
concentrations for dissolved methylmercury in water across all scenarios ranged from 0.014 to 1.0 ng/L.
The highest predicted concentrations occurred at a location 2.5 km from a chlor-alkali plant. The
predicted contribution of the hypothetical emission sources to methylmercury concentration ranged from
0 to 99% across all modeling scenarios. Although these results are meant to describe events on a local
(adjacent to emission source) rather than nationwide scale, they provide a general frame of reference for
comparison with measured values. The predicted range compares to the measured concentration range
of 0.01 to 1.481 ng/L reported by Krabbenhoft et al. (1999) for 104 surface water samples collected at
sites across the United States. The range of predicted concentrations overlapped the methylmercury
concentrations in ground water (less than or equal to 0.3 ng/L, based on one study) and drinking water
(0.023 to 1.95 ng/L) estimated from measurement data presented in Section 5.4.2.1.
5.4.2.3 Intake Estimates for Drinking Water and Ambient Water
Using the methylmercury concentration data in treated drinking water, and in ambient water it is
possible to estimate exposure from water ingestion. For methylmercury, data on measured
concentrations in ground and treated drinking water are limited. The database for surface water is
somewhat more extensive. Estimates of intake based on ingestion of drinking water and ambient water
are provided below.
Ambient Surface Water
A central tendency value for methylmercury in ambient surface water based on national data is
available from a pilot study conducted by the U.S. Geological Survey (Krabbenhoft et al., 1999).
Concentrations of methylmercury in ambient surface water ranged from a mean background level of 0.13
ng/L (or 1.3 x 10"7 mg/L) to a mean concentration for all sites of 0.15 ng/L (or 1.5 x 10"7 mg/L).
Methylmercury Water Quality Criterion 1/3/01 5-9
-------
Combining the mean for methylmercury concentrations at all sites with default exposure assumptions of
a 30 kg child aged 0 to 14 years who consumes 1 L/day of ambient surface water yields an estimated
exposure of 5.0 x 10'9 mg/kg-day. Combining the mean value for methylmercury concentrations at all
sites with default exposure assumptions of 2 L/day for water ingestion rate and 67 kg for body weight
yields an exposure estimate of 4.5 x 10'9 mg/kg-day for a woman of childbearing age (15-44 years old).
Adults in the general population have an estimated exposure value of 4.3 x 10"9 mg/kg-day, based on a
default body weight and water intake rate of 70 kg and 2 L/day, respectively. These values are
summarized in Table 5-5.
Table 5-4. Range of Predicted Dissolved Methylmercury Concentrations in Water for Hypothetical
Emissions Scenarios
Site
Eastern
Eastern
Western
Western
RELMAP
Percentile
50
90
50
90
Methylmercury (ng/L)
Min
0.077
0.11
0.014
0.034
Max
1.0
1.0
1.0
1.0
Scenario
Min
Large hospital incinerator,
25km
Multiple scenarios
Multiple scenarios
Multiple scenarios
Max
Chlor-alkali plant, 2.5 km
Chlor-alkali plant, 2.5 km
Chlor-alkali plant, 2.5 km
Chlor-alkali plant, 2.5 km
Source: U.S. EPA (1997c)
Table 5-5. Ambient Surface Water Intake Assumptions and Estimates
Population of
Concern
Children
(0-14 yr)
Childbearing
Women
Adults in the
General
Population
Methylmercury in
Ambient Surface
Water3
(mg/L)
1.5 x 10-7
1.5 x lO'7
1.5 x lO'7
Ingestion
Rate"
(L/day)
1.0
2.0
2.0
Body
Weight"
(kg)
30
67
70
Daily Exposure Estimate
(mg/kg-day)
5.0 xlO'9
4.5 xlO'9
4.3 xlO'9
1 Methylmercury concentration is the mean for all sites in the national pilot study as reported in Krabbenhoft et al.
(1999)
b U.S. EPA (2000a)
5-10
Methylmercury Water Quality Criterion 1/3/01
-------
Drinking Water
Although drinking water concentrations can be calculated based on surface water and ground-water
concentrations (U.S. EPA, 2000a), the available ground-water data were not adequate for this purpose.
Therefore, exposure from drinking water was roughly estimated for women of childbearing age, children
aged 0-14 years, and adults in the general population based on existing drinking and tapwater
concentration data (NJDEPE, 1993). For the purpose of this estimate, it was assumed that the reported
data reflected contributions from both ground water and surface water. Combining the estimated range
for methylmercury concentrations in drinking water (0.0234 to 1.95 ng/L, or 2.34 x 10"8 to 1.95 x 10"6
mg/L) with default values for a 30 kg child aged 0 to 14 years consuming 1 L/day of drinking water
yields an exposure estimate ranging from 7.8 x 10'10 to 6.5 x 10'8 mg/kg-day. Combining the estimated
range for methylmercury concentrations in drinking water with default values of 2 L/day for drinking
water intake and 67 kg for body weight yields an exposure estimate that ranges from 7.0 x 10"10 to 5.8 x
10"8 mg/kg-day for a woman of childbearing age (15-44 years old). Exposure estimates from ingesting
drinking water by adults in the general population range from 6.7 x 10"10 to 5.6 x 10"8 mg/kg-day, based
on a default body weight and water intake rate of 70 kg and 2 L/day, respectively. These values and
intake assumptions are summarized below in Table 5-6.
Table 5-6. Drinking Water Intake Assumptions and Estimates
Population of
Concern
Children
(0-14 yr)
Childbearing
Women
Adults in the
General
Population
Methylmercury in
Drinking Water
(mg/L)
2.3xlO-8tol.9xlO-6
2.3x10*101.9x10*
2.3 x 10'8 to 1.9 x 10-6
Ingestion
Rate3
(L/day)
1.0
2.0
2.0
Body
Weight3
(kg)
30
67
70
Daily Exposure Estimate
(mg/kg-day)
7.8 xlO-10 to 6.5 xlO'8
7.0 xlO'10 to 5.8 xlO'8
6.7 xlO'10 to 5.6 xlO'8
'U.S.EPA(2000a)
Methylmercury Water Quality Criterion 1/3/01
5-11
-------
5.43 Nonfish Dietary Exposures
5.4.3.1 Measured Concentrations in Food Other Than Fish
Historically, measurements of mercury have not been speciated in food items other than fish,
primarily because of the lack of adequate methodology (Madson and Thompson, 1998). However, the
limited data available suggest that nonfish foods such as dairy products, fruits, and vegetables may
potentially contribute to intake of methylmercury. Furthermore, it is possible that the agricultural
practice of using fishmeal in animal feeds may result in increased levels of methylmercury in nonfish
foods (ATSDR, 1999). This section examines the available data on mercury and methylmercury
concentrations in nonfish human food items.
Information on the concentration of total mercury in dietary items is available from the Total Diet
Study (TDS) conducted by the U.S. Food and Drug Administration (U.S. FDA). The TDS is an on-going
nationwide program that determines the levels of nutrients and selected contaminants in foods for the
purpose of estimating intakes of these substances by the U.S. population. A total of 839 samples for 47
food items were collected and analyzed for total mercury during the period from 1991 to 1996 (U.S.
FDA, 1999). Of the reported results, 756 (90%) were below the detection limit for mercury (0.01 to 0.02
mg/kg depending on food item) and 30 (3.6%) were considered to contain trace amounts of mercury.
These trace values represent the best estimates of those who analyzed the data, but in all cases are below
the nominal limit of quantitation.
Examination of the data for the 41 nonfish dietary items analyzed (6 items were fish) indicates that
the total mercury concentration was below the detection limit for most samples. These samples were
assigned a concentration of zero for statistical analysis (U.S. FDA, 1999). Trace amounts of total
mercury were found in one sample each (out of 18 total samples for each item) of fried beef liver, cooked
oatmeal, and boiled spinach. The maximum detected concentration of mercury in nonfish dietary items
was 0.03 mg/kg in fried beef liver. The reported median concentrations for total mercury in all
individual nonfish dietary categories were zero. Based on these data, the central tendency estimate for
methylmercury intake from nonfish dietary items is zero. For comparison, the mean mercury
concentration from all 47 food categories (containing both fish and nonfish dietary items) was 0.006
mgVkg (U.S. FDA, 1999).
5-12
Methylmercury Water Quality Criterion 1/3/01
-------
The MSRC (U.S. EPA, 1997b) also summarized data for methyhnercury concentrations reported in
local studies. Measured concentrations of methylmercury in garden produce and crops are summarized
in Table 5-7. Because the database for methylmercury content in these foods is limited, information is
also presented from studies that report total mercury concentrations. In general, the level of
methylmercury in agricultural produce is low, with the highest concentration (30 ng/g dry weight)
observed in leafy vegetables. Plants grown in the presence of elevated soil or atmospheric concentrations
of mercury are reported to contain elevated concentrations of total mercury (U.S. EPA, 1997b). Temple
and Linzon (1977) sampled the mercury content of fresh fruits and vegetables around a large chlor-alkali
plant in an urban-residential neighborhood. Among garden produce, leafy crops accumulated the highest
levels of mercury. One lettuce sample contained 99 ng/g wet weight of mercury (background: <0.6 ng/g),
and a sample of beet greens contained 37 ng/g wet weight (background: 3 ng/g). Tomatoes and
cucumbers within 400 m of the chlor-alkali plant averaged 2 and 4.5 ng/g wet weight of mercury,
respectively, compared with measured background levels of 1 ng/g.
Because the mercury content in plants tends to be low, livestock typically accumulate little mercury
from forage or silage (U.S. EPA, 1997b). However, use of fishmeal as food for poultry and other
livestock may result in increased mercury levels in these animals (ATSDR, 1999). Measured
concentrations of mercury and methylmercury in meat products are summarized in Table 5-8. Although
the database is limited, the available data suggest that methylmercury concentrations in meats are
generally low in comparison with levels observed in fish (U.S. EPA, 1997b).
Pedersen et al. (1994) monitored the level of mercury in wine, beer, soft drinks, and various juices.
Total mercury levels in these beverages were at or below the detection limit of 6 |ig/L in all samples
tested.
Infant postnatal exposure to methylmercury through ingestion of breast milk is a pathway of
potential concern. As noted in Section 3.4, methylmercury is excreted in breast milk (Bakir et al., 1973;
Sundberg and Oskarsson, 1992). The ratio of mercury hi breast milk to mercury in whole blood was
approximately 1:20 in women exposed to methylmercury via contaminated grain in Iraq between 1971
and 1972 (Bakir et al., 1973). Skerfving (1988) found that 16% of mercury in human breast milk is
methylmercury. Note that the MSRC found the data on breast milk to be insufficient to support
estimation of exposure by this route.
Methylmercury Water Quality Criterion 1/3/01
5-13
-------
rence
!
9
u
o>
S
t
0)
a 1
Sb
p§ M
i»,^
4j S
s ?
u p
S b
s-s
t-s
w ^g.
is
E?G"
tal Mercu
g/g dry w
0 B
H ^
Study Description
* — v
£5
0\
o
§*
U
CO
>n
V— <
en
o
*— 1
X
o
CO
1
o
X
-------
Table 5-8. Measured Mercury Concentration in Meats
Study Description
Saginaw River, MI
"Roaster" Ducks
(n=6)
Wild Deer
(Northern
Wisconsin)
Beef: Raw
Beef: Lunch Meat
Beef: Frank
Beef Muscle:
Control Group
Beef Muscle:
Exposed Group
Beef Liver:
Control Group
Beef Liver:
Exposed Group
Pork: Raw and
Sausage
Chicken: Raw and
Lunch Meat
Turkey: Lunch
Meat
Total Mercury
(ng/g wet weight)
48
5-14
<1
21
<1
2-3
1-4
3000 - 7000
9000 - 26000
<1
< 1 to 29
<1
Approx.
Total Mercury
(ng/g mercury
dry weight)1
124.7
13-36
<2.6
54.5
<2.6
5.2 - 7.8
2.6 - 10.4
7800-18000
23400- 67000
<2.6
< 2.6 to 75.4
<2.6
Approx.
Total Mercury
(mg/kg mercury
dry weight)
124.7 x ID'3
13xlO-3-36xlO-3
<2.6 x ID'3
54.5 x lO'3
<2.6 x ID'3
5.2 x 10-' - 7.8 x 10"3
2.6 x ID'3 - 10.4 x 1C'3
7.8 - 18.0
23.4 - 67.0
<2.6 x 10'3
<2.6 x ID'3 - 75.4 x
10'3
<2.6 x 10'3
%
Methyl-
mercury
NA
11-57%
>10%
4%
>60%
NA
NA
NA
NA
0-70%
20-67%
>20%
Reference
U.S. EPA
(1992a)
Bloom and
Kuhn (1994)
Vreman et al.
(1986)*
Bloom and
Kuhn (1994)
* Exposed animals received 1.7 mg mercury/day as mercury acetate; intake for controls was approximately 0.2 mg mercury/day.
1 Based on an assumed water content of 0.615, which is average for beef (Bass et al., 1984).
Source: U.S. EPA (1997c)
5.4.3.2 Predicted Concentrations in Foods Other than Fish
U.S. EPA (1997d) reported predicted concentrations in fruits, vegetables, beef, pork, poultry, dairy
products, and eggs. As described in previous sections on predicted concentrations in various media, this
effort was undertaken because some monitoring studies suggest that measured mercury concentrations
may be higher in areas adjacent to stationary industrial and combustion sources known to emit mercury
(U.S. EPA, 1997b). Results of this local study are of relevance to derivation of the water quality
criterion because they include data specifically for predicted methylmercury concentrations, and thus
permit comparison with measured concentrations.
Methylmercury Water Quality Criterion 1/3/01
5-15
-------
The Industrial Source Code air dispersion model (ISC3) was used for the computer simulation to
estimate nonfish dietary exposure. Model plants (defined as hypothetical facilities which were developed
to represent actual emissions from existing industrial processes and combustion sources), were situated in
hypothetical locations intended to simulate a site in either the Western or Eastern United States (U.S.
EPA, 1997b). Input values for air concentrations consisted of simulated concentration results (50th and
90th percentile values) obtained using the regional Lagrangian model of air pollution (RELMAP). The
assumptions and inputs utilized in this modeling effort are described in detail in Volume HI of the MSRC
(U.S. EPA, 1997b).
Predicted concentrations in a variety of nonfish foods are reported in Table 5-9. Because the
computer models used to generate these concentrations incorporated a point source for mercury
emissions, these predictions likely approach a worst-case scenario for methylmercury levels in foods.
Based on a large hospital waste incinerator scenario in the Eastern United States (50th percentile),
concentrations of methylmercury (expressed on a dry-weight basis) ranged from 0.095 ng/g to 7.1 ng/g in
fruits and vegetables, with the highest concentration observed in leafy vegetables. Concentrations of
methylmercury animal products ranged from 0.0013 ng/g to 4.2 ng/g, with the highest concentrations
observed in beef and dairy products. The hypothetical facility was considered to contribute less than
10% to the total plant mercury concentration (U.S. EPA, 1997b). The local source was considered to
contribute 7% to 11% of the total mercury in beef, dairy products, and pork and 41% of total mercury in
poultry and eggs (U.S. EPA, 1997b).
5.4.3.3 Intake Estimates for Food Other Than Fish
Data from the U.S. FDA TDS (described in Section 5.4.3.1) suggest that nonfish dietary items
generally account for a very small fraction of total mercury intake. For the purpose of estimating
methylmercury intake from nonfish foods, the central tendency estimate of methylmercury concentration
is assumed to be zero. Thus, the average daily intake is zero mg/kg-day for adults in the general
5-16
Methylmercury Water Quality Criterion 1/3/01
-------
Table 5-9. Predicted Methylmercury Concentrations in Produce and Animal Products Based on a Large
Hospital Waste Incinerator Scenario
Item
Total Mercury
(ng/g dry wt.)
% Methylmercury
Methylmercury
(ng/g dry wt.)
Produce
Root vegetables
Fruits
Fruiting vegetables
Leafy vegetables
1.9
35
35
34
5
5
5
21
0.095
1.7
1.7
7.1
Animal Products
Beef
Beef liver
Dairy
Pork
Poultry
Eggs
Lamb
8.6
22
11
0.007
0.12
0.12
3.9
19
19
19
18
3
3
19
1.6
4.2
2.1
0.0013
0.0036
0.0036
0.74
"Data based on ISC simulation for receptors at a humid site 2.5 km from a large hospital hazardous materials incinerator (HMI)
and input from RELMAP (East 50"1 Percentile).
Source: U.S. EPA (1997b)
population, children, and women of childbearing age. This estimate is in agreement with WHO (1990),
which reported that nonfish foods accounted for 0% of average daily intake of methylmercury.
Methylmercury intake from animal products and produce has been estimated by computer model
simulation for four hypothetical high-end exposure scenarios: rural subsistence farmer (adult and child),
rural home gardener (adult and child), urban high-end adult, and high-end fisher (adult and child) (U.S.
EPA, 1997c). These predicted methylmercury intakes are presented in Table 5-10. Methylmercury
intake from animal products was estimated only for the rural subsistence farmer. Intake from animal
products and produce was not considered in the remaining scenarios. The subsistence farmer was
anticipated to represent a very high-end exposure scenario. Simulation of intake for these scenarios
employed a body-weight exposure assumption for children (i.e., 17 kg) that differs from the currently
recommended value (i.e., 30 kg) for derivation of water quality criterion values (see Table 5-1).
Estimated exposure from produce for several high-end scenarios ranged from 2.3 x 10'7 mg/kg-day for the
Methylmercury Water Quality Criterion 1/3/01
5-17
-------
high-end urban adult to 5.8 x 10~5 mg/kg-day for the adult high-end fisher. Estimated exposures from
animal products for the rural subsistence farmer scenario were 2.1 x 10"6 mg/kg-day and 5.3 x 10"6 mg/kg-
day for an adult and child, respectively. These model-predicted estimates support the finding of
generally low methylmercury intake from nonfish foods indicated by measurement data from the TDS
(U.S. FDA, 1999) and the conclusion in the MSRC that substantially all exposure to methylmercury is
from fish consumption.
5.4.4 Fish Consumption Estimates
The MSRC concluded that most human exposure to methylmercury is from food and that it is
primarily from fish consumption (U.S. EPA, 1997g). Ihgestion of contaminated fish is also reported by
many other authors to be the only significant source of methylmercury exposure to the general human
population (Stern, 1993; Swedish EPA, 1991; WHO, 1990). This conclusion is based on the observation
that in many nonfish foods, the mercury content is typically near detection limits and is comprised mainly
of inorganic species (WHO, 1990). In contrast, most of the mercury in fish is methylated.
This section provides information on measured and predicted tissue concentrations of
methylmercury in freshwater fish and marine fish, and estimates of intake for several target populations.
The MSRC presented data for freshwater fish and marine fish. The MSRC did not include a separate
evaluation of estuarine fish, although the data on marine species presented in the MSRC (from the
National Marine Fisheries Service) include some estuarine species. Sections 5.4.4.1 and 5.4.4.2, below,
summarize the major studies presented in the MSRC for freshwater fish. Section 5.4.4.3 presents an
estimate of intake for both freshwater and estuarine species. Although the intake estimate is based on the
freshwater fish methylmercury concentrations only, EPA believes that the freshwater fish concentrations
are similar to the concentrations in these estuarine species presented in the MSRC. EPA, therefore,
believes that calculating an intake estimate using the freshwater/estuarine default consumption rates
provides a reasonable approximation of combined freshwater/estuarine fish methylmercury exposure. A
more accurate estimate of marine fish methylmercury intake has been made (Section 5.4.4.7) since this
source of exposure is included in the RSC estimate that is factored into the final water quality criterion
calculation.
5-18
Methylmercury Water Quality Criterion 1/3/01
-------
00
I
13
K
§
"S
CA
ca
I
"o
«
1
CO
o
c.
W
c
"ca 'bb
u c
0 <1
Pi
u
CO
E
•O
w
•g,
£
5
«>
S *
0 S
= 1
«1
U4 ™
<*
I
m u<
'53 g
^ c
co £
1
1
2
S
o
*j
1
•o
•o S
S3 V
6 "S
£
2 1
IH S3
rj >
<
T3
3 *
"O JS
<.£?
DC
0
"a g1
3 S
"^ <
^3
IS
u
_;..
1
1
1
Parameter
O
5^
o
^
"-•
•-
o
l>
o
t—
Ju
®
t-
R
1 Body Weight (kg)
S
o\
os
OS
cs
cs
\o
vo
vo
2
o
Fraction of Total Mercury
From All Sources'1 That Is
Methylmercury0 (%)
vo
10
s
CJ
vO
1-11
o
*
VO
1
cs
»H
9
cs
CO
t~-
9
c-i
OS
9
o
•*'
B
pi
00
f^.
s
<
VO
9
f;
<— <
Methylmercury Intake
From Produce
(mg/kg-day)
Z
<
Z
o
^
Z
§
Z
^
z
o
o
VO
o
W
CO
10
VO
o
H
oi
Methylmercury Intake
From Animal Products
(mg/kg-day)
S
1
o
to
CO
c3
s — /
<3
1
i
.,_,
3
a.
.5
•o
i
K
^
0
0
1
CO
CO
*«
u
s >»
^ 3
•S 4)
S2 S3" .S .S .S
g ts .S3 .2 .3
w 1C >» >-i S^>
£ £ s s s
e o a a a
c j- u a o
"Data based on ISC simulatil
bAll sources includes intake
'Predicted fraction of total m
Predicted fraction of total rr
"Predicted fraction of total m
NA Not available
Source: U.S. EPA (1997c)
0\.
T 1
-------
5.4.4.1 Measured Concentrations in Freshwater Fish
Data for mercury concentrations in freshwater fish have been previously compiled and evaluated by
EPA in Volume IV of the MSRC (U.S. EPA, 1997c). The discussion below provides information on the
national studies considered and the database selected by U.S. EPA after careful consideration of data
quality issues to provide concentration data for estimating human exposure to methylmercury (U.S. EPA,
1997c).
Two national studies were considered by U.S. EPA (1997c) for estimation of mercury
concentrations in freshwater finfish populations. Lowe et al. (1985) reported mercury concentrations in
fish from the National Contaminant Biomonitoring Program. The freshwater fish data were collected
between 1978-1981 at 112 stations located across the United States. Mercury was measured by a
flameless cold vapor technique, with a detection limit of 0.01 u.g/g wet weight. Most of the sampled fish
were taken from rivers (93 of the 112 sample sites were rivers); the other 19 sites included larger lakes,
canals, and streams. Fish weights and lengths were consistently recorded. The mercury concentrations
measured in this study are shown in Table 5-11. Several varieties of fish were sampled. Carp, large
mouth bass, and white sucker were most common. The geometric mean mercury concentration of all
sampled fish was 0.11 u.g/g wet weight; the minimum and maximum concentrations reported were 0.01
and 0.77 p.g/g wet weight, respectively. The highest reported mercury concentrations (0.77 u.g/g wet
weight) occurred in a northern squawfish collected from the Columbia River. Mean mercury
concentrations (whether geometric or arithmetic mean not specified) by species are reported in the
MSRC (U.S. EPA, 1997c).
A national study of chemical residues in freshwater fish was conducted by U.S. EPA (1992b) and
also reported by Bahnick et al. (1994). As reported in the MSRC (U.S. EPA, 1997c), five bottom-
feeding species (e.g., carp) and five game fish species (e.g., bass) were sampled at each of the 314
sampling sites in the United States. These sites were selected based on proximity to either point or
nonpoint pollution sources. Thirty-five "remote" sites among the 314 total sites were included to provide
nonimpacted background pollutant concentrations. The study primarily targeted sites that were expected
to be impacted by increased dioxin levels. The point sources proximate to sites offish collection included
the following: pulp and paper mills, Superfund sites, publicly owned treatment works (POTWs), and
other industrial sites. Data describing fish age, weight, and sex were not consistently collected. Whole
body mercury concentrations were determined for bottom feeders, and mercury concentrations in fillets
were analyzed for the game fish. Total mercury levels were analyzed using flameless atomic absorption,
5-20
Methylmercury Water Quality Criterion 1/3/01
-------
with reported detection limits of 0.05 [ig/g early in the study (465 samples analyzed prior to 1990) and
0.0013 [ig/g later in the study (195 samples), as the analytical technique improved. Nondetects were
reported as a zero value and averaged as zeros. The estimated standard deviation for replicate samples
was 0.047 \igfg in the concentration range of 0.08 to 1.79 u.g/g. Mercury was detected in fish collected
from 92% of the sample sites. Concentration data are provided in Table 5-12. The maximum mercury
level detected was 1.8 u.g/g, and the mean concentration in 669 fish samples across all sites was 0.26
u.g/g. The highest measurements occurred in walleye, largemouth bass, and carp. The mercury
concentrations in measured in fish around POTWs were the highest among all point source data; the
median value for mercury concentration was 0.61 u.g/g.
The intake estimates presented in this document, similar to the MSRC, are based on the mean
concentration values from the studies described above; that is, the fish mercury concentration data based
on the Bahnick et al. (1994) and Lowe et al. (1985) studies were used for the estimates. However, the
MSRC also includes summary data from numerous other studies that indicate significantly higher levels
of methylmercury in freshwater fish. For example, concentrations of methylmercury in bass, crappie,
northern pike, and trout of 2.0, 1.39, 1 .7 1 , and 1.19 [ig/g, respectively, represent a few of the higher
species concentrations reported (see U.S. EPA, 1997d, Table 4-48).
Measurements of elevated levels of mercury in fish have been reported elsewhere. For example,
the North East States Coordinated Air Use Management (NESCAUM) summarized data from New
England's freshwater fish in the "Mercury Study: A Framework for Action" by the Northeast States and
Eastern Canadian Provinces (1998) (see Table 5-11).
Additional data are available for New York State (Simonin and Meyer, 1998). In New York State,
maximum mercury concentrations over 2 ppm were seen for the following species: walleye (3.2 ppm),
striped bass (5.4 ppm), white perch (3.2 ppm) Northern pike (2.1 ppm), smallmouth bass (3.34 ppm),
largemouth bass (2.39 ppm), rock bass (2.7 ppm), drum (1.4 ppm), channel catfish (2.0 ppm), sunfish
(1.2 ppm), American eel (1.6 ppm), Lake trout (2.7 ppm), white sucker (1.2 ppm), black crappie (1.4
ppm), and carp (5.8 ppm).
5.4.4.2 Predicted Concentrations in Freshwater Fish
As previously indicated, the MSRC included numerous computer-simulated estimates of mercury
exposure for selected population scenarios (U.S. EPA, 1997c). These included predicted concentrations
Methylmercury Water Quality Criterion 1/3/01
5-21
-------
in Tier 4 (predatory) fish based on exposure modeling. The Industrial Source Code air dispersion model
(ISC3) was used for simulation of methylmercury concentrations in water and biota near mercury
emissions sources. Model plants (large and small municipal waste combustors, large and small
hazardous materials incinerators, coal and oil-fired utility boilers, chlor-alkali plant), defined as
hypothetical facilities which were developed to represent actual emissions from existing industrial
processes and combustion sources, were situated in hypothetical locations intended to simulate a site in
either the Western or Eastern United States (U.S. EPA, 1997b). Input values for air concentrations
consisted of simulated concentration results (50th and 90th percentile values) obtained using the regional
Lagrangian model of air pollution (RELMAP). The assumptions and inputs utilized in this modeling
effort are described in detail in Volume JTL of the MSRC (U.S. EPA, 1997b).
Fish tissue methylmercury concentrations of 5.3 x 10'1 |o.g/g and 9.7 x 10'2 |o,g/g were predicted for
the simulated Eastern and Western sites, respectively, in scenarios where the hypothetical emission
sources had zero percent impact on local mercury levels (i.e., the predicted concentration resulted only
from background levels of mercury in the environment and regional anthropogenic sources). These
levels are of the same order of magnitude as the mean measured values of 0.11 and 0.26 u.g/g (1.1 x 10'1
and 2.6 x 10'1 p,g/g) reported by Lowe et al. (1985) and Bahnick et al. (1994) respectively. The
maximum predicted tissue concentration of 68 ng/g was associated with the Eastern site chlor-alkali plant
scenario.
5.4.4.3 Intake Estimates from Freshwater/Estuarine Fish
The mercury concentration data reported in U.S. EPA (1992b) and Bahnick et al. (1994) were
selected to determine a rough estimate of methylmercury intake from freshwater and estuarine fish. In
contrast to the data reported by Lowe et al. (1985), the selected study provides an arithmetic mean as a
measure of central tendency. These data have previously been used by U.S. EPA (1997d) to calculate
methylmercury intake estimates under different fish ingestion scenarios. In this section, new estimates of
methylmercury intake are calculated in accordance with technical guidance provided in the 2000 Human
Health Methodology (U.S. EPA, 2000a). Using the mean mercury concentration of 0.26 fig mercury/g
fish wet weight (or mg/kg) reported by U.S. EPA (1992b) and Bahnick et al. (1994), and assuming that
approximately 100 percent is methylmercury (U.S. EPA, 1997d), the average estimated methylmercury
concentration in freshwater/estuarine fish is 0.26 mg/kg.
5-22
Methylmercury Water Quality Criterion 1/3/01
-------
Table 5-11. Freshwater Fish Mercury Concentrations from Lowe et al. (1985) and Northeast States and
Eastern Canadian Provinces (1998)
LoweetaL (1985)
Fish Species
Bass
Bloater
Bluegill
Smallmouth Buffalo
Carp, Common
Catfish (channel, largemouth, rock, striped, white)
Crappie (black, white)
Freshwater Drum
Northern Squawfish
Northern Pike
Perch (white and yellow)
Sauger
Sucker (bridgelip, carpsucker, klamath, largescale, longnose,
rivercarpsucker, tahoe)
Trout (brown, lake, rainbow)
Walleye
Mean of Measured Fish
Mean Mercury Concentration
(fjg/g Wet Wt)
0.157
0.093
0.033
0.096
0.093
0.088
0.114
0.117
0.33
0.127
0.11
0.23
0.114
0.149
0.1
0.11"
Northeast States and Eastern Canadian Provinces (1998)
Fish Species
Largemouth bass
Smallmouth bass
Yellow perch
Chain pickerel
Lake trout
Walleye '
Brown bullhead
Brook trout
Maximum Mercury Concentration in
ppm
8.94
5.0
3.15
2.81
2.70
2.04
1.10
0.98
"Geometric mean; U.S. EPA (1997c) did not specify whether means for individual species or species categories were
geometric or arithmetic means.
Source: U.S. EPA (1997c), Northeast States and Eastern Canadian Provinces (1998).
Methylmercury Water Quality Criterion 1/3/01
5-23
-------
To estimate daily exposure from methylmercury in freshwater/estuarine fish, average body weights
and high-end fish ingestion rates (90th percentile) for the populations of concern are estimated, as
recommended in the 2000 Human Health Methodology. Default intake values for fish intake by children,
women of child-bearing age, and adults in the general population are provided in U.S. EPA (2000a).
These intake values were estimated from the on-going, nationally based Continuing Survey of Food
Intake for Individuals (CSFIT) conducted by the U.S. Department of Agriculture. The CSFn is conducted
annually, and dietary data from all 50 States are collected (U.S. EPA, 2000a). The estimates of intake
based on CSFn incorporated data for both consumers and nonconsumers of fish, and represent intake of
all fish whether store-bought or sport-caught (U.S. EPA, 2000a). The freshwater/estuarine fish ingestion
rates for children, women of child-bearing age, and adults in the general population are estimated to be
156.3 g/day, 165.5 g/day, and 17.5 g/day, respectively (U.S. EPA, 2000a). Note that the estimates for
both children and women of childbearing age are based on short-term consumption, whereas the estimate
for adults in the general population is based on average long-term consumption.
Table 5-12. Freshwater Fish Mercury Concentrations from Bahnick et al. (1994).
Species
Carp
Sucker (white, redhorse, spotter)
Catfish (channel and flathead)
Bass (white, largemouth, smallmouth)
Walleye
Northern Pike
Crappie
Brown Trout •
Mean of Measured Fish
Mean Mercury Concentration
(p,g/g Wet Wt)
0.11
0.167
0.16
0.38
0.52
0.31
0.22
0.14
0.26
Source: U.S. EPA (1997c)
5-24
Methylmercury Water Quality Criterion 1/3/01
-------
The recommended body weights for children 0 to 14 years, women of childbearing age, and adults
in the general population are 30 kg, 67 kg, and 70 kg, respectively (U.S. EPA, 2000a). Based on these
exposure assumptions, the daily exposure estimates of methylmercury intake from ingestion of
freshwater/estuarine fish for children, women of childbearing age, and adults in the general population
are 1.4 x 10'3 mg/kg-day, 6.4 x 10"4 mg/kg-day, and 6.5 xlO"5 mg/kg-day, respectively. Input assumptions
and calculated daily exposure estimates for freshwater/estuarine fish are summarized in Table 5-13.
5.4.4.4 Measured Concentrations in Marine Fish and Shellfish
The MSRC (U.S. EPA, 1997b,c) has summarized data on concentrations of total mercury and
methylmercury in marine fish and shellfish. Analyses of total mercury concentrations in marine fish and
shellfish have been carried out over the past two to three decades. Data describing methylmercury
concentrations in marine fish are predominantly based on the National Marine Fisheries Service (NMFS)
database, the largest publicly available database on mercury
Table 5-13. Freshwater/Estuarine Fish Intake Assumptions and Estimates
Population of
Concern
Children
Women of
Childbearing
Age
Adults in the
General
Population
Mercury
in Fish3
(mg/kg)
0.26
0.26
0.26
Methyl-
mercury/
Mercury
in Fish"
100
100
100
Methyl-
mercury
in Fish
(mg/kg)
0.26
0.26
0.26
Ingestion
Ratec
(kg/day)
0.1563
0.1655
0.0175
Body
Weightc
(kg)
30
67
70
Daily
Exposure
Estimate
(mg/kg-day)
1.4xlO-3
6.4 x 10-4
6.5 x 10-5
1 U.S. EPA (1992b) and Bahnick et al. (1994)
bU.S.EPA(1997c)
cU.S.EPA(2000a)
Methylmercury Water Quality Criterion 1/3/01
5-25
-------
concentrations in marine fish. In the early 1970s, the NMFS conducted testing for total mercury in more
than 200 seafood species of commercial and recreational interest (Hall et al., 1978). The determination
of mercury in fish was based on flameless (cold vapor) atomic absorption spectrophotometry following
chemical digestion of the fish sample. These analytical methods are described in Hall et al. (1978).
The NMFS Report provides data on number of samples, the number of samples where mercury was
not detected ("nondetects"), and mean, standard deviation, minimum, and maximum detected mercury
levels (in parts per million wet weight) for 1,333 combinations offish/shellfish species, variety, location
caught, and tissue (Hall et al., 1978). This database consists of 177 fish/shellfish species for which
mercury concentration data are provided. This represents 5,707 analyses of fish and shellfish tissues for
total mercury, of which 1,467 or 26%, were reported at nondetectable levels. A discussion of the issues
associated with evaluation and use of nondetect data for methylmercury in the NMFS database is
provided in the MSRC (U.S. EPA, 1997c). A summary of NMFS concentration data is provided in Table
5-14.
Two additional databases for mercury concentration in marine fish and shellfish are cited in the
MSRC (U.S. EPA, 1997d). These are the Report on the Chance of U.S. Seafood Consumers Exceeding
"The Current Daily Intake for Mercury and Recommended Controls " (U.S. FDA, 1978) and a report by
Stern et al. (1996) that examined exposure of New Jersey residents to mercury via fish consumption.
Although concentration data from these databases are reported in the MSRC (U.S. EPA, 1997c), detailed
descriptions and evaluations of study quality are not provided.
The intake estimates presented in this document, similar to the MSRC, are based on the mean
concentration values from the studies described above; that is, the fish mercury concentration data based
on the NMFS, Stern et al., and U.S. FDA studies were used for the estimates. However, the MSRC also
includes summary data from numerous other studies that indicate significantly higher levels of
methylmercury in marine fish. For example, concentrations of methylmercury in mackerel, pompano,
shark, snapper, and swordfish of 2.9, 8.42,4.53, 2.17, and 2.72 u.g/g, respectively, represent a few of the
higher species concentrations reported (see U.S. EPA, 1997c).
5.4.4.5 Other Measured Concentration Data for Marine Fish and Shellfish
Additional national-scope information on methylmercury in marine biota is available from Project
Mussel Watch. Project Mussel Watch measures concentrations of organic and trace metal contaminants
5-26
Methylmercury Water Quality Criterion 1/3/01
-------
in fresh, whole soft-parts of bivalve mollusks (i.e., mussels and oysters) at more than 240 coastal and
estuarine sites. Data are currently available from 1986 through 1993 and are summarized in the MSRC
(U.S. EPA, 1997b). Average concentrations along the North Atlantic, Eastern Gulf, and Pacific coasts
(0.15, 0.14, and 0.11 p,g/g dry weight, respectively) are higher than those collected along the Middle
Atlantic, South Atlantic, and Western Gulf coasts (0.06, 0.09, and 0.08 \ig/g dry weight, respectively).
The highest concentrations exceeded 1.0 [ig/g dry weight and were collected along the Western Gulf and
Pacific coasts (1.80 and 1.01 ng/g dry weight, respectively).
Annual Mussel Watch data on mercury concentrations in bivalve mollusks at specific sites have
been aggregated to national geometric means for the purpose of analyzing temporal trends (O'Conner and
Beliaeff, 1995). The national means do not show any temporal trend in mercury concentrations in
mussels and oysters for the period 1986-1993. Temporal trend analysis was also conducted on a site-by-
site basis for 154 Mussel Watch sites for which there were data for at least 6 years during the period of
1986-1993 (O'Conner and Beliaeff, 1995). Seven sites exhibited an increasing trend in mercury
concentrations, and eight sites exhibited a decreasing trend in mercury concentrations, with 95%
statistical confidence.
5.4.4.6 Predicted Concentrations in Marine Fish and Shellfish
The computer simulations conducted by EPA and reported in the MSRC (U.S. EPA, 1997c) did not
provide predictions for methylmercury concentrations in marine fish or shellfish.
5.4.4.7 Intake Estimates from Marine Fish and Shellfish
In accord with technical guidance provided in U.S. EPA (2000a), mean, median, and 90th percentile
concentrations of methylmercury in marine fish were used to estimate daily exposure from
methylmercury in marine fish. Species-specific mean concentrations of mercury in marine fish from the
National Marine Fisheries Service (NMFS, 1978) are presented in EPA's MSRC (U.S. EPA, 1997c).
These data are summarized in Table 5-14. For species where concentration was not reported in NMFS
(1978), concentrations were estimated from data reported by Stern et al. (1996), U.S. FDA Compliance
Testing data, or U.S. FDA (1978) as cited in U.S. EPA (1997c).
Methylmercury Water Quality Criterion 1/3/01
5-27
-------
Table 5-14. Average Mercury Concentrations in Marine Fish and Shellfish
Species
Concentration3
(u.gHg/gWetWt)
Species
Concentration
(ligHg/gWetWt)
Finfish
Anchovy
Barracuda, Pacific
Cod*
Croaker, Atlantic
Eel, American
Flounder*'6
Haddock*
Hake
Halibut*
Herring
Kingfish
Mackerel*
Mullet
Ocean Perch*
Pollock*
0.047
0.177
0.121
0.125
0.213
0.092
0.089
0.145
0.25
0.013
0.10
0.081
0.009
0.116
0.15
Pompano*
Porgy*
Ray
Salmon*
Sardines*
Sea Bass*
Shark*
Skate
Smelt, Rainbow*
Snapper*
Sturgeon
Swordfish*
Tuna*
Whiting (silver hake)*
Whitefish*
0.104
0.522"
0.176
0.035
0.1
0.135
1.327
0.176
0.1
0.25
0.235
0.95C
0.206
0.041
0.054"
Shellfish
Abalone
Clam*
Crab*
Lobster*
0.016
0.023
0.117
0.232
Oysters
Scallop*
Shrimp
Other shellfish*
0.023
0.042
0.047
0.012b
Molluscan Cephalopods
Octopus*
0.029
Squid*
0.026
Source: U.S. EPA (1997c).
*Denotes species used in calculation of methylmercury intake from marine fish for one or more populations of
concern, based on existence of data for consumption in the CSFII (U.S. EPA, 2000b).
* Mercury concentrations are from NMFS (1978) as reported in U.S. EPA (1997d) unless otherwise noted, measured
as u.g of total mercury per gram wet weight of fish tissue.
b Mercury concentration data are from Stern et al. (1996) as cited in U.S. EPA (1997c).
e Mercury concentration data are from U.S. FDA Compliance Testing as cited in U.S. EPA (1997c).
dMercury concentration data are from U.S. FDA (1978) as cited in U.S. EPA (1997c).
0 Mercury data for flounder were used as an estimate of mercury concentration in marine flatfish in marine intake
calculations
5-28
Methylmercury Water Quality Criterion 1/3/01
-------
A consumption-weighted mean concentration of mercury for all marine fish was calculated as
follows. Each of the marine species selected for inclusion in the analysis was weighted based on species-
specific U.S. population intake rates among the three populations of concern (U.S. EPA, 2000b). This
weighting system accounts for variability of consumption among different species and across different
populations of concern. The consumption weighting factor for each of the selected marine species was
calculated as follows. The consumption rates for individual marine species were summed to give a total
consumption rate for a particular population of concern. The weighting factor was then calculated as the
quotient of the species-specific consumption rate divided by the total consumption rate:
Species A consumption rate (g/day)
Weighting factor for species A =
Sum of consumption rates for all selected species
(g/day)
For each population of concern, the average mercury concentration for each species was multiplied
by its consumption weighting factor. This product was then summed across all selected marine species
to estimate the mean concentration of mercury in all marine fish for that particular population of concern:
Mean conc(p,g/g) = £ [species-specific conc(p,g/g) x species-specific weighting factor]
Assuming that approximately 100% of the mercury in marine fish is present as methylmercury
(U.S. EPA, 1997c), the weighted-average methylmercury concentrations in marine fish consumed by
each of the populations of concern are 0.167 mg/kg, 0.147 mg/kg, and 0.157 mg/kg for children (aged 0-
14 years), women of childbearing age, and adults in the general population, respectively.
Specific body weights and several fish ingestion rates (arithmetic mean, median and 90th percentile)
for the populations of concern were used to estimate daily exposure from methylmercury in marine fish.
Marine fish intake values for children, women of childbearing age, and adults in the general population
are provided in U.S. EPA (2000b). For children and women of childbearing age, these intake values
were estimated using 3 years of "consumers only" data (1994-1996) from the on-going, nationally based
Continuing Survey of Food Intake for Individuals (CSFH) conducted by the U.S. Department of
Agriculture. Intake values for adults in the general population were obtained using all survey
respondents to derive an estimate of long-term consumption. The marine fish ingestion rates for
children, women of childbearing age, and adults in the general population are presented in Table 5-15.
Methylmercury Water Quality Criterion 1/3/01
5-29
-------
The current default body weights for children 0 to 14 years, women of childbearing age, and adults
in the general population are 30 kg, 67 kg, and 70 kg, respectively (U.S. EPA, 2000a). Based on these
exposure assumptions, the mean daily exposure estimates of methylmercury intake from ingestion of
marine fish for children, women of childbearing age, and adults in the general population are 4.1 x 10"4
mg/kg-day, 2.0 x 10"4 mg/kg-day, and 2.7 x 10'5 mg/kg-day, respectively. The median daily exposure
estimates of methylmercury intake from ingestion of marine fish for children, women of childbearing
age, and adults in the general population are 3.2 x 10"4 mg/kg-day, 1.6 x 10"4 mg/kg-day, and 0 mg/kg-
day, respectively. In addition, the 90th percentile daily exposure estimates of methylmercury intake from
ingestion of marine fish for children, women of childbearing age, and adults in the general population are
8.5 x W4 mg/kg-day, 4.1 x 10"4 mg/kg-day, and 1.1 x 10"4 mg/kg-day, respectively. Input assumptions
and calculated daily exposure estimates for marine fish are summarized in Table 5-16.
5.4.5 Respiratory Exposures
5.4.5.1 Measured Concentrations in Air
Outdoor Air. Vapor-phase elemental mercury is the predominant form of mercury in the
atmosphere and constitutes up to 98% of the total mercury concentration (U.S. EPA, 1997b). Increased
Table 5-15. Marine Fish Ingestion Rates
Population of Concern
Children
Women of Childbearing Age
Adults in the General Population
Mean Intake
(kg/day)
0.07490
0.09104
0.01246
Median Intake
(kg/day)
0.05971
0.07548
0
90th Percentile Intake
(kg/day)
0.15229
0.18835
0.04916
Source: U.S. EPA (2000b)
_
5-30
Methylmercury Water Quality Criterion 1/3/01
-------
concentrations of the divalent form of mercury may be present near emission sources. Small fractions of
paniculate mercury and methylmercury may also be present. Atmospheric mercury concentrations in the
United States are generally very low (U.S. EPA, 1997b). U.S. EPA (1993) as cited in the MSRC
summarized information on total mercury concentrations in outdoor air and reported ranges of 1 to 4
ng/m3 for rural areas and 10 to 170 ng/m3 for urban areas. Methylmercury concentrations from these
samples constituted 0% to 21% of the total mercury concentration, with percentage values reported to
generally be on the low end of this range. A measure of central tendency was not provided with this
estimate. Particulate mercury typically constituted less than 4% of total atmospheric mercury in rural
areas, although this fraction was increased in urban areas. The current background mercury
concentration over the Northern Hemisphere is considered to be between 1.5 and 2.0 ng/m3 (Expert Panel
on Mercury Atmospheric Processes, 1994). A background concentration of 1.6 ng/m3 was reported by
Fitzgerald (1994). This value was subsequently used by U.S. EPA (1997b) to model mercury fate in
watershed soils and surface waters.
Bloom and Fitzgerald (1988) measured vapor-phase mercury concentrations in outdoor air samples
collected from Long Island Sound, CT. Total mercury concentrations ranged from 1.4 to 5.3 ng/m3. The
fraction of total mercury present as methylmercury was estimated to be 0% to 1%. During the month of
October, the mean methylmercury concentration was 12 pg/m3 (range 4 to 38 pg/m3). This concentration
represented 0.7% of the total gaseous mercury concentration. During the month of November, the
measured methylmercury concentration Was less than 10 pg/m3 and from December through August, the
concentration was below the detection limit of 5 pg/m3.
Indoor Air. No data were identified for indoor air concentrations of methylmercury.
5.4.5.2 Predicted Concentrations in Air
EPA has modeled mercury air concentrations for the continental United States using RELMAP
simulation, meteorological data for the year 1989, and current mercury emission data. The background
level of mercury in the atmosphere was assumed to be 1.6 ng/m3. The results of this simulation are
reported in (U.S. EPA, 1997b). Predicted concentrations for total mercury are given in Table 5-17. The
predicted total mercury concentrations ranged from approximately 1.6 to 1.9 ng/m3, with the highest
concentrations predicted for the Eastern United States. The tabulated results indicate that total
Methylmercury Water Quality Criterion 1/3/01
5-31
-------
Table 5-16. Intake Assumptions and Estimates for Marine Fish
Population of
Concern"
Children
Women of
Childbearing Age
Adults in the
General Population
Mercury
in Marine
Fish
(mg/kg)
1.67E-01
1.47E-01
1.57E-01
Methylmercury/
Mercury in
Marine Fish
. %
100%
100%
100%
Methylmercur
y in Marine
Fish
(mg/kg)
1.67E-01
1.47E-01
1.57E-01
Body
Wt.
(kg)
30
67
70
Mean
Daily
Exposure
Estimate
(mg-kg-
day)
4.1E-04
• 2.0E-04
2.7E-05
Median
Daily
Exposure
Estimate
(mg-kg-
day)
3.2E-04
1.6E-04
O.OE+00
90th%Daily
Exposure
Estimate
(mg-kg-
day)
8.3E-04
4.1E-04
1.1E-04
• Marine fish intake assumptions for the populations of concern from U.S. EPA (2000b) are summarized in
Table 5-15.
Table 5-17. Percentile Analysis of RELMAP Predicted Total Mercury Concentration Results (ng/m3) for
the Continental United States
Region
Continental U.S.
East of 90° W longitude
West of 90° W longitude
Min
1.602
1.616
1.602
10th
1.607
1.640
1.606
50th
1.624
1.668
1.616
90th
1.685
1.720
1.642
Max
1.995
1.995
1.743
Source: U.S. EPA (1997b)
mercury concentration never exceeded the background level by a large percentage (25% maximum)
under the conditions of this simulation. Methylmercury concentration estimates were not provided in the
model output as reported in the MSRC (U.S. EPA, 1997b) but, again, is presumed to be present
predominantly as elemental mercury.
5.4.5.3 Intake Estimates for Air
The primary species of mercury to which humans are exposed through inhalation is vapor-phase
elemental mercury (U.S. EPA, 1997g). Thus, inhalation exposure to methylmercury is not expected to be
a significant route of concern when compared to intake via fish consumption.
5-32
Methylmercury Water Quality Criterion 1/3/01
-------
Assuming the background mercury concentration of 0.0016 ng/m3 (or 1.6 ng/m3) reported by
Fitzgerald (1994), of which approximately one percent is methylmercury (Bloom and Fitzgerald, 1988),
the average methylmercury concentration in air is 0.000016 fig/m3 (or 1.6 x 10"8 mg/m3). Estimates of
daily exposure from methylmercury in air were calculated using inhalation rates and body weights
specific to the populations of concern. The long-term inhalation rate based on a time-weighted average
for children 0 to 14 years is estimated to be 10.4 nrVday (U.S. EPA, 1997h). The average, long-term
inhalation rates for women of childbearing age and adults in the general population are estimated to be 11
nrVday and 20 nrYday, respectively (U.S. EPA, 1994, 1997h). The recommended body weights for
children 0 to 14 years, women of childbearing age, and adults in the general population are 30 kg, 67 kg,
and 70 kg, respectively (U.S. EPA, 2000a). Based on these exposure input assumptions, the daily
exposure estimates from methylmercury in air for children 0 to 14 years, women of childbearing age, and
adults in the general population are 5.5 xlO'9 mg/kg-day, 2.6 xlO"9 mg/kg-day, and 4.6 xlO"9 mg/kg-day,
respectively. These input assumptions and calculated daily exposure estimates for air are presented in
Table 5-18.
U.S. EPA (1997c) reported inhalation exposure estimates based on ISC simulation for a humid site
2.5 km from a large hospital medical waste incinerator (HMT) and input from RELMAP (Eastern U.S.,
50th percentile) (Table 5-19). The inhalation parameters used in the simulation for children (16 nrYday)
differed from the rate adopted from U.S. EPA (1997h) for calculation of inhalation intake from measured
concentrations (see Table 15-1). Estimated intake for all five exposure scenarios was zero mg/kg-day.
This prediction supports the finding of low methylmercury intake via inhalation as calculated from
measured concentrations.
Methylmercury Water Quality Criterion 1/3/01
5-33
-------
Table 5-18. Inhalation Exposure Intake Assumptions and Estimates
Population of
Concern
Children
(0-14 yr)
Women of
Childbearing
Age
Adults in the
General
Population
Mercury
in Air*
(mg/m3)
1.6 x lO'6
1.6 x lO'6
1.6 x lO'6
Methyl-
mercury/
Mercury
inAirb
1
1
1
Methyl-
mercury
in Air
(mg/m3)
1.6xlO'8
1.6xlO-8
1.6xlO-8
Inhalation
Rate0
(nrYday)
10.4
11
20
Body
Weightd
(kg)
30
67
70
Daily
Exposure
Estimate
(mg/kg-day)
5.5 xlO'9
2.6 xlO"9
4.6 xlO'9
a Fitzgerald (1994) as cited in U.S. EPA (1997b).
b Bloom and Fitzgerald (1988) as cited in U.S. EPA (1997b).
c Inhalation rates from U.S. EPA (1994,1997h).
d Current default body weight values from U.S. EPA (2000a).
Table 5-19. Predicted Methylmercury Intake from Air for Five Hypothetical High-End Exposure
Scenarios
Parameter
Inhalation Rate
(nrYday)
Contact Rate for
Inhalation
(hr/day)
Body Weight (kg)
Methylmercury
Intake
(mg/kg-day)
Rural
Subsistence
Farmer
Adult
20
24
70
0
Child
16
24
17
0
Exposure Scenario3
Rural Home
Gardener
Adult
20
24
70
0
Child
16
24
17
0
Urban
Adult
Aver-
age
20
24
70
0
Adult
High-
end
20
16
70
0
Child
Aver-
age
16
24
17
0
Child
High-
end
16
24
17
0
High End
Fisher
Adult
20
24
70
0
Child
16
24
17
0
Recrea-
tional
Angler
Adult
20
24 •
70
0
"Data based on ISC simulation for a receptors at a humid site 2.5 km from a large Hospital medical waste incinerator
(HMI) and input from RELMAP (East 50th Percentile).
Source: U.S. EPA (1997c)
5-34
Methylmercury Water Quality Criterion 1/3/01
-------
5.4.6 Soil/Sediment Exposures
5.4.6.1 Measured Concentrations in Soil/Sediment
The available data for measured methylmercury and total mercury concentrations in soils and
sediments are summarized in Table 5-20, including a small number of studies that provide some data that
are national in scope. In general, soil mercury levels are usually less than 200 ng/g in the top soil layer,
but values exceeding this level are not uncommon, especially in areas affected by anthropogenic
activities (U.S. EPA, 1997b). Soil mercury levels vary greatly with depth, with nearly all the mercury
found in the top 20 cm of soil. Mercury levels are positively correlated with the percentage of organic
matter in soil (Nriagu, 1979).
Some information is available on estimated typical or background levels of total mercury in U.S.
soils and may be used with speciation data to estimate soil methylmercury concentrations. The MSRC
(U.S. EPA, 1997b) states that approximately 1 to 3% of the total mercury in surface soil is
methylmercury. The other 97% to 99% of total soil mercury can be considered to be largely Hg(H)
complexes, although a small fraction of mercury in typical soil will be Hg° (Revis et al., 1990). The
methylmercury percentage has been observed to exceed 3% in garden soil with high organic content
under slightly acidic conditions (Cappon, 1987). Computer simulations of mercury fate and transport
predict that methylmercury constitutes 2% of the total mercury in watershed soils (U.S. EPA, 1997b).
Davis et al. (1997) reported a range of 50 to 200 ng/g for total mercury concentrations in
nonmercuriferous soils and sediments in background areas not directly impacted by volcanic emissions or
anthropogenic releases. The authors stated that methylmercury typically constitutes 0.01% to 2 % of the
total mercury concentration. Supporting information on the derivation of this estimate was not provided
by the authors. The MSRC (U.S. EPA, 1997b) cited data from NJDEPE (1993) that indicates that typical
U.S. soils contain 8 to 117 ng/g of total mercury. Neither an estimate of mean mercury concentration nor
speciation data were provided in the description of this study as summarized in the MSRC. Assuming
that approximately 2% of the total mercury concentration is present as methylmercury, these data suggest
that typical U.S. soils contain 0.16 to 2.3 ng/g as methylmercury.
Shacklette and Boerngen (1984) reported mean concentrations, geometric standard deviations, and
ranges for total mercury in soils and other surficial materials based on samples collected at 1318 sites
across the conterminous United States. The geometric mean concentration for the conterminous United
States was 58 ± 2,520 ng/g (ppb), and the estimated arithmetic mean was 89 ng/g. Additional data
Methylmercury Water Quality Criterion 1/3/01 5-35
-------
indicate that the mean concentration of mercury in soils varies by region. In soils from the Western
United States (west of the 96th meridian), the geometric mean concentration was 46 ± 2,330 ng/g (range
<10 to 4,600 ng/g) and the estimated arithmetic mean was 65 ng/g. In soils from the Eastern United
States (east of the 96th meridian), the geometric mean concentration was 81 ± 2,520 ng/g (range 10 to
3,400 ng/g), with an estimated arithmetic mean of 120 ng/g. Speciation data were not reported by these
authors. Assuming that methylmercury constitutes approximately 2% of the total mercury concentration,
the geometric and arithmetic mean levels of mercury present as methylmercury in soils in the
conterminous United States would be approximately 1.2 ng/g and 1.8 ng/g, respectively.
\
Additional data are available on soil mercury and methylmercury concentrations for sites in the
United States. As reported in the MSRC (U.S. EPA, 1997b), methylmercury concentrations in soil
samples at locations in New York and Washington ranged from 0.3 to 22.9 ng/g dry weight and
constituted 0.5% to 5.3% of the total soil mercury content. No other information on these studies was
provided.
As characterized in the MSRC (U.S. EPA, 1997b), sediment mercury levels are typically higher
than soil levels, and concentrations exceeding 200 ng/g are not unusual. Sediment mercury levels follow
the same trends as soil in regards to depth, humic matter, and methylmercury percentage. There is some
evidence suggesting that the methylmercury percentage increases with increasing total mercury
contamination (Parks et al., 1989). Concentrations of mercury and (where available) methylmercury are
tabulated in Table 5-20.
5-36
Methylmercury Water Quality Criterion 1/3/01
-------
Table 5-20. Concentrations of Total Mercury and Methylmercury in Soil and Freshwater Aquatic
Sediments
Location
Total Mercury
(ng/g dry wt)
Methylmercury
(ng/g dry wt)
% Methylmercury
Reference
Soils
Discovery Park,
Seattle, WA
Wallace Falls,
Cascades, WA
Control Soil
New York State
Compost
New York State
Garden Soil
New York State
Soil and Other
Surficial
Materials in
Conterminous
U.S.
Typical U.S.
Soils
Typical
background
levels in
nonmercurifer-
ous soils
29-133
155 - 244
117
213
406
Conterminous
U.S.
58 (geo mean)
89 (arith mean)
Western U.S.
46 (geo mean)
65 (arith mean)
Eastern U.S.
81 (geo mean)
120 (arith mean)
8-117
50 - 200
0.3 - 1.3
1.0-2.6
4.9
7.3
22.9
NA
NA
0.01 - 2
0.6 - 1.5
0.5-1.2
4.2
3.3
5.3
NA
NA
NA
Lindqvistetal. (1991)"
Lindqvistetal. (1991)"
Cappon, (198 l)a
Cappon, (1987)a
Cappon, (1987)a
Shacklette and
Boerngen (1984)
NJDEPE (1993)"
Davis et al. (1997)
Freshwater Aquatic Sediments
80 Minnesota
Lakes
North Central
Wisconsin lakes
Little Rock Lake,
Wisconsin
U.S. Lake
sediment mean
ranges
34 -753
mean 174
90 -190
10 - 170
70-310
NA
NA
NA
NA
NA
NA
NA
NA
Sorenson et al. (1990)a
Radaetal. (1989)a
Wiener etal. (1990)a
NJDEPE (1993)a
Methylmercury Water Quality Criterion 1/3/01
5-37
-------
Location
U.S. GS National
Pilot Study
Total Mercury
(ng/g dry wt)
105 Background
211 All sites
Methylmercury
(ng/g dry wt)
2.1 Background
1.87 All sites
% Methylmercury
0.1
1
Reference
Krabbenhoft et al.
(1999)
" As cited in U.S. EPA (1997b)
5.4.6.2 Predicted Concentrations in Soil
The MSRC (U.S. EPA, 1997b) reported the results of watershed fate and transport modeling
conducted to predict the concentration of mercury in watershed soils. Atmospheric concentrations and
deposition rates were used as inputs to the IEM-2M model. The IEM-2M model is composed of two
integrated models that simulate mercury fate using mass balance equations which describe processes in
watershed soils and a shallow lake. Using this approach, total mercury concentrations of 47 and 8 ng/g
were predicted for soils at hypothetical Eastern and Western U.S. sites, respectively. These predicted
concentrations for total mercury in soils are lower than the measured concentrations reported by
Shacklette and Boergen (1984) for conterminous and regional U.S. soils. More than 90% of the total
mercury in soil was predicted to occur as the inorganic divalent species. The fraction of the predicted
background concentration occurring as methylmercury was 2% for the Eastern site (U.S. EPA, 1997c),
suggesting a soil methylmercury concentration of 0.9 ng/g based on modeling predictions for speciation.
Corresponding speciation data was not reported for the Western site.
5.4.6.3 Intake Estimates for Soil/Sediment
The primary species of mercury in soil is largely considered to be Hg(IT) complexes, although a
small fraction of mercury in typical soil will be Hg° (Revis et al., 1990). Thus, ingestion exposure to
methylmercury in soil is not expected to be a significant route of concern when compared to exposure via
fish ingestion.
Assuming the background mercury arithmetic mean concentration of 89 ng/g (or 0.089 mg/kg)
reported by Shacklette and Boerngen (1984), of which approximately 2% is methylmercury (U.S. EPA,
1997b,c; Cappon, 1987; Davis et al., 1997), the average estimated methylmercury concentration in soil is
1.78 ng/g (or 0.00178 mg/kg). To estimate daily exposure from methylmercury in soil, ingestion rates
and body weights for populations of concerns must also be estimated. The average incidental soil
ingestion rate for children is estimated to be 1 x 10"4 kg/day (U.S. EPA, 1997h). In addition, the average
soil ingestion rate for pica children is estimated to be 1 x 10"2 kg/day (U.S. EPA, 1997h). The average
soil ingestion rates for women of child-bearing age and the general adult population are both estimated to
5-38
Methylmercury Water Quality Criterion 1/3/01
-------
be 5 x 10'5 kg/day (U.S. EPA, 1997h). The default body weights for children 0 to 14 years, women of
child-bearing age, and adults in the general population are 30 kg, 67 kg, and 70 kg, respectively (U.S.
EPA, 2000a). Based on these exposure input assumptions, the daily exposure estimates from
methylmercury in soil for children, pica children, women of child-bearing age, and adults in the general
population are 5.9 xlO"9 mg/kg-day, 5.9 x 10'7 mg/kg-day, 1.3 xlO'9 mg/kg-day, and 1.3 xlO"9 mg/kg-day,
respectively. These input assumptions and calculated daily exposure estimates for soil are presented in •
Table 5-21.
Table 5-21. Summary of Soil Ingestion Intake Assumptions and Estimates
Population of
Concern
Children
Pica Children
Women of
Childbearing
Age
Adults in the
General
Population
Mercury
in Soil3
(mg/kg)
0.089
0.089
0.089
0.089
Methyl-
mercury/
Mercury
in Soil"
(%)
2
2
2
2
Methyl-
mercury
in Soil
(mg/kg)
0.00178
0.00178
0.00178
0.00178
Ingestion
Ratec
(kg/day)
0.0001
0.01
0.00005
0.00005
Body
Weightd
(kg)
30
30
67
70
Daily
Exposure
Estimate
(mg/kg-day)
5.9 xlO'9
5.9 xlO'7
1.3 xlO'9
1.3 xlO'9
a Shacklette and Boerngen for the conterminous U.S. (1984).
b U.S. EPA (1997b,c); Cappon (1987) as cited in U.S. EPA (1997b); Davis et al. (1997).
c U.S. EPA (1997h).
" U.S. EPA (2000a).
Methylmercury Water Quality Criterion 1/3/01
5-39
-------
Estimates of soil ingestion based on exposure modeling reported in the MSRC (U.S. EPA, 1997c)
are summarized in Table 5-22. Predicted exposures are based on an ISC model simulation for a receptors
at a humid site 2.5 km from a large hospital medical waste incinerator (EMI) and input from RELMAP
(East 50th Percentile). Soil intake among the hypothetical receptors was highest for the urban pica child
(1.2 x 10"6 mg/kg-day). The remaining estimates ranged from 3 x 10'9 to 2.4 x 10'8 mg/kg-day. These
approximations are comparable to exposure estimates based on measured concentrations of mercury in
soils in Table 5-21 when the twofold difference in assumed soil ingestion rate is considered.
5.4.7 Occupational and Other Exposures
Occupational Exposure. Occupational exposures are not routinely factored into the derivation of
water quality criterion but may be considered on a chemical-specific basis. Information on occupational
exposure to mercury has been summarized in the MSRC (U.S. EPA, 1997c). OSHA (1975) estimated
that approximately 150,000 U.S. workers are exposed to mercury in at least 56 occupations. More
recently, Campbell et al. (1992) reported that about 70,000 workers are annually exposed to mercury.
Occupational settings in which exposure to mercury may occur include chemical and drug synthesis,
hospitals, laboratories, dental practices, instrument manufacture, and battery manufacture (NIOSH,
1977). Jobs and processes involving mercury exposure include manufacture of measuring instruments
(barometers, thermometers, etc.), mercury arc lamps, mercury switches, fluorescent lamps, mercury
broilers, mirrors, electric rectifiers, electrolysis cathodes, pulp and paper, zinc carbon and mercury cell
batteries, dental amalgams, antifouling paints, explosives, photographs, disinfectants, and fur processing.
Inorganic mercury accounts for nearly all occupational exposures (U.S. EPA, 1997c). Airborne
elemental mercury vapor is the main pathway of concern, particularly in those industries with the greatest
number of mercury exposures. Occupational exposure to methylmercury appears to be insignificant or
rare. Thus, occupational exposures are not considered relevant to the derivation of ambient water criteria
for methylmercury.
5-40
Methylmercury Water Quality Criterion 1/3/01
-------
°o
i
CO
i
iS
s — >-
h B "Si
U o pu
sg^-S
•s
1
fa
^
i
sS
F*H
B
£
1
o>
1
o
s
o
a
"3
^
ctf
S
.2 Jg
•^ S
CJD ra
es
1
—
3
^
•a
1
i
^3
2
j=
U
•M
I
**
g
si
1
^
•§>"§
3 g
g>
b
S*
^
•o
1
w
''w
^4
§
9
T3
!
(2
^
(N
C5
» — i
O
S
^t
0
0
"i
^
c5
f^
cs
o
C3
0
Soil Ingestion Ral
(g/day)
o
"•<
o
^
^:
o
o
— '
^
T— <
°
1
oo
1
•o
o
t^
s
T— 1
vo
o
CS
1—1
[^.
§
^i
^^
9
VO
g
s
r— (
i
CS
g
w
o
cs
g
MJ
s
PQ
*O
*— i
O
CS
— '
o
w
1O
1
Total Mercury Int
(mg/kg/day)
CS
cs
c-l
cs
cs
cs
d
cs
cs
cs
cs
s-
Fraction of Total
Mercury That Is
Methylmercury (9
0\
s
0
CO
oo
^
cs'
o\
o
^3
CO
g
w
CS
oo
o
CO
§
s
§
w
o
^
oo
^
cs'
ON
o
w
o
CO
oo
^
cs
ON
o
s
CO
1
1 Methylmercury In
(mg/kg/day)
c
o
- '•§
^ So
« .S
cs ~
B -s
n. OT
r3 o
^ is
^ e
B o
M-J 1^-J
PH O
<£ «§
C3 OT
ess >
ffi 5
^ fib
0 <
'E. .2
js S;
« 2
£? o
S e
S I
o ^
p to
5 i B
^ fe u
Q> T3 ^3
'« So
•0 0 1-.
's § •*=
3 "S !C
"S = -o
r; -a e
« a u
M 2 JS
o «3 -£P
•^ g J3
O -^3 ^
^~< ' &0
| 11
Q »— • O
"c3 3s-
3 1- M (^
c tP S3 c^
'S3 ^ -O ON
aData based on IS
Percentile).
*Soil ingestion ra
rates for the rural
Source U.S. EPA
o
S
§
1
3
I
-------
Exposure from Dental Amalgam. Gradual erosion of dental amalgam represents a pathway by
which many people are routinely exposed to extremely small amounts of mercury. Dental amalgam
fillings contain approximately 50% mercury by weight. The mercury in the amalgam is continuously
released over time. Speciation data indicate that release occurs primarily as elemental mercury vapor
(Begerow et al., 1994). Exposure to methylmercury via this route is thus expected to be insignificant.
Therefore, exposure to methylmercury via this pathway is not considered relevant to RSC analysis for
derivation of the water quality criterion.
5.5 EXPOSURE DATA ADEQUACY AND ESTIMATE UNCERTAINTIES
After identifying relevant exposure pathways and obtaining available data for quantifying exposure
via each pathway, it is important to consider whether the data are adequate to describe exposure
estimates for each exposure medium. The adequacy of the contaminant concentration data, in part,
determines the specific method with which the RSC estimates will be determined. Important factors
include sample size, accurate representation of the sample (e.g., whether sample selection was biased and
whether data are current), the accuracy in the sample analysis procedures (i.e., whether errors occurred
during measurement), and the sensitivity of the measurement relative to the environmental levels of
concern (i.e., whether detection limits are low enough such that the concentration can be detected in most
samples within a data set). Additional discussion on data adequacy is provided in the 2000 Human Health
Methodology (U.S. EPA, 2000a).
5.5.1 Adequacy of Intake Estimate for Drinking Water
Ground water. Nationally distributed data for methylmercury or total mercury in ground water
were not located. The MSRC (U.S. EPA, 1997b) reports data from three-local studies in the United
States. However, supporting information on sample size, detection limits, analytical methodology, and
other information relevant to data adequacy are not provided in the MSRC. Therefore, these data (as
presented in the MSRC) do not satisfy the adequacy requirements of the 2000 Human Health
Methodology.
Drinking Water. The MSRC (U.S. EPA, 1997b) cited a typical level of 25 ng/L for total mercury
concentration in drinking and tap water (Lindqvist and Rodhe, 1985). A range of 0.3 to 25 ng/L for total
mercury in drinking water was also reported (NJDEPE, 1993). The presentation of these data in the
MSRC did not provide information on the composition of this water (e.g., fraction from ground water
and surface water) or treatment status. Furthermore, the presentation of data in the MSRC did not
5-42 Methylmercury Water Quality Criterion 1/3/01
-------
provide information on the method of calculation or a detailed description of data quality (including
source of data, sample size, detection limits, arid analysis procedures) for this estimate. Thus, the data
for drinking water (as presented in the MSRC) are considered sufficient only for a rough estimate of
intake. Yet, using the higher-end value of 25 ng/L results in an estimate within the range estimated for
surface water.
Raw surface water. National data for surface water concentrations (primarily stream data) are
available from the U.S. Geological Survey National Pilot Study of Mercury Contamination (Krabbenhoft
et al., 1999). Water samples were collected in the summer and fall of 1998 and thus are representative of
current concentrations. Sampling occurred at 106 sites clustered in 21 basins across the United States,
including Alaska and Hawaii. Data from 104 sites were used to determine values for mean, median,
maximum, and minimum methylmercury concentrations. The sampling sites spanned the dominant east-
to-west mercury deposition gradient and represented a wide range of environmental settings. Total
mercury and methylmercury were measured using sensitive analytical methodology (U.S. EPA Method
1631). The detection limits for total mercury and methylmercury were reported in a separate document
(Olson and DeWild, 1999) referenced in the report. Some samples were collected at sites impacted by
mining activity. The high concentration of mercury in samples collected at those sites resulted in a
positively skewed distribution, and this is reflected in the difference between the arithmetic mean and
median values for samples collected at all sites (0.15 ± 0.26 ng/L vs. 0.06 ng/L, respectively). The
measures of central tendency from this study compare favorably to a methylmercury concentration of
0.07 ng/L in surface water predicted by IEM-2M corhputer simulation (U.S. EPA, 1997b). The data
reported by Krabbenhoft et al. (1999) are therefore considered to be adequate to estimate intake from
surface water.
5.5.2 Intake from Nonfish Dietary Sources
Data for measured methylmercury concentrations in nonfish foods are available from several local
studies and one national study. Estimates of methylmercury concentration in selected produce and
animal products are also available from computer simulations (U.S. EPA, 1997c). Data from the local
studies provide supporting information on methylmercury speciation and concentration in a variety of
foods, but are considered too limited in scope for estimation of intakes for use in RSC analysis.
Information on mercury content offish and nonfish foods is available from the Total Diet Study (1991-
1997) conducted by U.S. FDA (1999). This is an on-going, nationally based study conducted for
determining intake of nutrients and contaminants by the U.S. population. Based on data adequacy
requirements of the 2000 Human Health Methodology (U.S. EPA, 2000a), the sample size of the U.S.
Methylmercury Water Quality Criterion 1/3/01 5-43
-------
EPA study is sufficient for calculation of central tendency and 90th percentile values. Detection limits
and the number of samples with mercury concentrations below detection the limit are reported by food
item. The procedure for treating these samples for statistical analysis is reported. These data are thus
considered adequate to estimate central tendency and high-end intakes from nonfish food items.
5.5.3 Intake From Fish
The MSRC (U.S. EPA, 1997c) assessed data sources for estimates of both freshwater and marine
fish intake. Reliable mercury concentration data are available from databases maintained for marine fish
and shellfish by the National Marine Fisheries Service (NMFS, 1978) and two databases for freshwater
fish (Lowe et al., 1985; Bahnick et al., 1994). These studies are national in scope, in contrast to many
studies that have a local or regional focus. In addition, the studies were not initiated in response to
specific incidents of mercury contamination, and thus may avoid potential bias toward high values.
Results in these studies are reported as total mercury. However, the MSRC concluded, based on research
conducted by Bloom (1992) and Morgan et al. (1994), that over 90% of the mercury present in fish and
seafood is methylmercury. Thus, total mercury concentrations are considered appropriate for evaluation
of methylmercury exposure in human populations. Detailed information on mercury concentration by
species and statistical considerations in use of the available data are presented in U.S. EPA (1997c).
Issues relating to data adequacy for methylmercury concentrations in marine fish and shellfish have
been addressed in the MSRC (U.S. EPA, 1997c). Although the NMFS data were initially compiled
beginning in the 1970s, comparisons of the mercury concentrations identified in the NMFS database with
compliance samples obtained by the U.S. FDA indicate that the NMFS data are appropriate to use in
estimating intake of mercury from marine fish at the national level of data aggregation. Cramer (1994)
reported on Exposure of U.S. Consumers to Methylmercury from Fish and noted that recent information
from NMFS indicated that the fish mercury concentrations reported in the 1978 report do not appear to
have changed significantly. The U.S. FDA also monitors methylmercury concentration in seafood.
Cramer (1994) observed that results of recent U.S. FDA surveys indicate results parallel to earlier
findings by U.S. FDA and NMFS. The National Academy of Sciences' National Research Council's
Subcommittee on Seafood Safety (1991) also assessed the applicability of the NMFS 1978 database to
current estimates of mercury concentrations in fish. This subcommittee similarly concluded that the
mercury concentrations in the 1978 database differed little in from the U.S. FDA compliance samples
estimating mercury concentrations in fish. An assessment of the NMFS database by persons with
expertise in analytical chemistry and patterns of mercury contamination in the environment indicates that
temporal patterns of mercury concentrations in fish do not preclude use of this database in current risk
5-44 Methylmercury Water Quality Criterion 1/3/01
-------
assessment activities (EPA's Science Advisory Board's ad hoc Mercury Subcommittee; Literagency Peer
Review Group, External Peer Review Group).
An issue raised by some reviewers of the MSRC (U.S. EPA, 1997rc) concerned use of data in the
NMFS database where mercury concentration was below the analytical detection limit. A detailed
analysis of the methods for reporting and analyzing nondetect data (U.S. EPA, 1997c, Appendix C)
indicated that differences among methods used to handle nondetect samples had negligible impact on the
reported mean concentrations in marine fish tissue. Additional information on analytical and statistical
considerations in use of the NMFS data is available in EPA's MSRC (U.S. EPA, 1997d). Overall, EPA
finds that these data are adequate for estimating exposure from marine fish for derivation of the
methylmercury water quality criterion.
Two compilations of data on mercury concentrations in freshwater fish were considered for use in
development of the water quality criterion for methylmercury. The strengths and weaknesses of these
studies have been evaluated and reported in the MSRC (U.S. EPA, 1997c). The studies reported by
Lowe et al. (1985) and by Bahnick et al. (1994) appear to be systematic, national collections offish
pollutant concentration data. However, higher mercury concentrations in fish have been detected in other
studies, and the values obtained in the Lowe et al. (1985) and Bahnick et al. (1994) studies should be
interpreted as approximations of the mean concentrations in freshwater finfish (U.S. EPA, 1997c). The
mean mercury concentrations for each study in all fish sampled vary by a factor of two. The mean
mercury concentration reported by Lowe et al. (1985) was 0.11 ng/g, whereas the mean mercury
concentration reported by Bahnick et al. (1994) was 0.26 \ig/g. The basis for these differences hi
methylmercury concentrations is unknown. Differences in sampling offish by trophic position, size, or
age might have been responsible for the differences in mean mercury concentrations reported in the two
studies. Older and larger fish, which occupy higher trophic positions in the aquatic food chain, would be
expected to have higher mercury concentrations. The type of water body from which fish were collected
may also influence fish mercury concentrations. Most of the fish collected by Lowe et al. (1985) were
from rivers. The fate and transport of mercury hi river systems is not as well characterized as in small
lakes. In comparison, most of the data reported by Bahnick et al. (1994) were collected with a bias
toward more contaminated/industrialized sites, although sampled sites were not specifically contaminated
with mercury. Thus, it is possible that there is more mercury available to the aquatic food chains at the
sites sampled by Bahnick et al. (1994). Another possibility is that the higher mercury concentrations
reported by Bahnick et al. (1994) when compared with those reported by Lowe et al. (1985) reflect
increases in mercury contamination over the time period between the studies. Trend data for
methylmercury concentrations in freshwater fish over time do not exist, although there are data for fish
Methylmercury Water Quality Criterion 1/3/01 5-45
-------
collected from coastal and estuarine sites (U.S. EPA, 1997c) as discussed above and in Section 5.4.4.5.
Those data suggest that there are no clear temporal trends in tissue mercury concentrations in fish and
shellfish over the past two decades. Overall, the data from either study were considered adequate for
calculating central tendency and high-end estimates of methylmercury intake from freshwater fish.
5.5.4 Intake from Air
The MSRC (U.S. EPA, 1997V) reported concentration ranges for mercury in urban and rural air.
Information on geographic location, sample sizes, and detection limits were not provided. A range of 0
to 21% for methylmercury speciation was presented without an estimate of central tendency. Thus, these
data as presented in the MSRC do not satisfy the adequacy requirements of the 2000 Human Health
Methodology. A value of 1.6 ng/m3 was presented in the MSRC as representative of national background
levels for total mercury. Details on the derivation of this concentration were not provided; however, this
value was considered of sufficient reliability to be used as input for fate and transport modeling reported
in the MSRC (U.S. EPA, 1997b,c). Concentration measurements and exposure modeling data presented
in the MSRC (U.S. EPA, 1997c) were also evaluated as an alternative estimate of methylmercury
concentration in air. Many factors (including selection of modeling equations, input assumptions, and
source data) in the modeling analysis affect the predicted concentrations and resulting exposures. These
factors are summarized and discussed in U.S. EPA (1997b,c,g). No data were located for methylmercury
concentrations in indoor air. Thus, this potential source of exposure was not considered in the estimate
of intake from air.
The information available on both measured and predicted air concentrations of methylmercury
from the MSRC is insufficient to fully determine data adequacy for estimating central tendency and high-
end exposures to methylmercury via inhalation. Estimates of inhalation exposure are presented, although
they are considered to represent rough approximations of actual (or likely) intake. Yet, the available data
summarized in the MSRC (including the computer-simulated estimates) indicate that exposure to
methylmercury in ambient air is negligible.
5.5.5 Intake From Soil
Three studies report aggregate values for measured soil mercury concentration. Shacklette and
Boemgen (1984) reported arithmetic and geometric mean concentrations, geometric standard deviations,
and ranges for total mercury in soils and other surficial materials based on samples collected at 1,318
sites across the conterminous United States. Sample size for these estimates is adequate, and the data are
5-46 Methylmercury Water Quality Criterion 1/3/01
-------
representative of concentrations in the United States, although detailed information on analytical
methodology, detection limit, and the number and statistical treatment of samples below detection limit
was not provided.
Davis et al. (1997) reported a range of 50 to 200 ng/g for total mercury concentration and an
estimate of the percent present as methylmercury in nonmercuriferous soils and sediments in background
areas not directly impacted by volcanic emissions or anthropogenic releases. However, supporting
information on the derivation of this estimate was not provided by the authors. The MSRC (U.S. EPA,
1997b) cited data from NJDEPE (1993) which indicates that typical U.S. soils contain 8 to 117 ng/g of
total mercury. Information necessary for assessment of data adequacy was not provided in the summary
of this study.
Additional data are available on soil mercury and methylmercury concentrations for sites in the
United States. The MSRC (U.S. EPA, 1997b) summarized two reports on methylmercury speciation in
soils collected at sites in New York and Washington state. Because each of these studies addressed soil
concentrations in only one state, they were not considered adequate for estimating methylmercury
exposure from soil.
Computer simulation data for predicted soil concentration, methylmercury speciation, and exposure
estimates are available for comparison to measured values. Predicted concentrations were calculated on
a regional (Eastern and Western U.S.) basis. As noted by U.S. EPA (1997b,c,g), many factors in the
simulation analysis (including modeling equations, input assumptions, and source data) potentially affect
the predicted concentrations.
Overall, the currently available soil concentration data are considered adequate to obtain central
tendency and high-end estimates of exposure. Although some information was not readily available from
the summarized studies in the MSRC (e.g., detection limits), the estimates of exposure from soil
ingestion presented in this document are considered adequate given the sampling size (especially the
Shacklette and Boerngen study) and geographic representativeness. There is also a clear indication from
all available studies that the amount of methylmercury in soil that is methylmercury is approximately 2%.
5.6 TOTAL EXPOSURE ESTIMATES
Total exposure (calculated as the sum of exposure from water, freshwater and estuarine fish, marine
fish, nonfish foods, air, and soil) for the three population groups in comparison to the RfD is shown in
Methylmercury Water Quality Criterion 1/3/01 5-47
-------
Table 5-23. To evaluate potential differences in exposure from ambient water and drinking water, total
exposure was calculated using methylmercury exposure estimates for each source. Because the
contribution of ambient water or drinking water intake to total exposure is negligible in comparison to
the sum of intake from other sources, there is no difference in the total exposure estimated using these
two alternatives.
The contribution of exposure from different media as a percentage of total exposure for three types
of individuals is summarized in Tables 5-24 through 5-26. Daily exposure estimates on a mg/kg-day
basis are presented in Tables 5-27 through 5-29. The information in these tables reflects use of three
different intake assumptions for consumption of marine fish: mean, median and 90th percentile.
5-48
Methylmercury Water Quality Criterion 1/3/01
-------
€
"fe *-<
tf, •« '•— *
o ^ i*
s P •?
0 M 81
-2 O
£
g
'? fe
lit
o 'S K1
I'll
"s -^
^
2
2
t2
a
i.
3
g
C.
X
w
e «
'5 ^
"
c c
it
i °c
'i I
s
»
II
*sJ ^
a^
"§ ^
ic
5^
'S 1
§
a ~
* S S e "Sb
-5,
g a^js?
S "S .2 ^ 5
§ ««,
j||||
s fle
B^.
iff
"*§ «
*3 "***
a-
lllll
felfa>s*
|ll|
Q ^
ih
"S
1 g
3 §
§•§
Cu U
0
X
oo
^
tn
X
tn
'fe
X
S
^
X
00
'^
o
X
V£3
'S
X
s
VO
ON
s
o'
0
o
to
o
C3
C5
O
o
C-l
tn
o
o
o
g
Adults in the
General
Population
to
*
^
to
X
0
od
to
X
od
2
X
-*
to
X
0
00
to
X
od
<
lg»
s § a
*" O O
-
O «
s"^
18
G
O
I
"S
S.i a
M O "S
O O, C
CX X ^ *^ *^2
Oi
-------
*e<
•H
S
t
o
.to
to
*«-
'1
S
1
1
fa
&
1
o
Q<
m
-2
5
s a Percent oi
GS
2
3
8
a.
M
W
1
0
1
3
o
s a Percent of T<
Exposure a;
i|
2 "*
2 o
u «
ao
«w e5
Women o
Childbeari
Age
«
43 a
** TS O
.S 2 '•§
42 a "3
"3 O o*
^ fc
1^
ca ^
•a o
CD
Women of
Childbearing
Age
1- §
••« aj "ea
III
^ ^
1
1
i
S*
a
CM
p
C°-
00
oo
s
1
d
r-
CO
8
o
d
VO
d
00
o
p
o
en
P
c?
8
o
a\
vq -
«
PL(
hwater/Estuarine
CO
a
UH
1
§
en
CO
00
en
T— 1
en
OJ
f-
o\
en
CM
en
o\
o
oo
en
en
o\
CM
JS
E
1
O
O
o
0
o
o
CO
•o
O
«S
_co
1
s
en
0
o
o
d
en
o
d
0
o
d
8
o
o
d
en
o
o
o
d
o
d
*
en
o
o
o
d
CM
O
d
*— i
§
d
en
d
1
0
d
o
o
d
00
C
_O
CO
H>
eo
c
CO
2
o
C3
'S.
"O
3
1
o .0
d "S
Table 5-1) and do
Ish and water crite
' criterion.
weight values (
tes used in the i
I in the fish onlj
•& § »
u cj w
J> -| s
O en .5
ares are based upon
mt surface water ex]
ng water exposure e
to w •*-
O 3 e
g-e-s
W < Q
« J3 O
-------
I
I
oo
I
<*_»
§
I
CO
O
P-,
CO
I
I
•a
CD
bfi
C
3
|
CO
O
fi
u-
§
•a
42
•a
o
t
o
"i
b
^o
?Ti
^^
•»•*
O
1
es
•a
J3
.
Exposure
"3
-2
r~
CM
^
•4^
1
£
0*
03
CA
03
2
s
1
1
0
H
[T
O
a Percent of T
w
C3
i
i
f
Children
^.ge 0-14 Years
^
M
11 a
||«!
^ §
,a t!
Adults in f
General
Populatio
2
si
IS
ll
<
(•N
g
oo
CO
ON
ON
ON
1
ON
ON
ON
c>
co
f-
o
oo
o
o
1
C>
OO
ON
ON
ON
r-
g
ON
ON
ON
ON
-C
CO
E
c
'i
water/Esti
JS
CO
gj
fe
Ul
1
esi
ON
<— i
ON
ON
ON
^_^
O
ON
CO
ON
oo
ON
ON
ON
ON
O
CO
CD
1
O
O
o
o
o
o
"M*
•o
o
JZ
I
•23
Q
8
o
o
8
o
o
o
o
o
8
o
o
0
CO
§
o
o
0
o
-l
s
o
o
0
g
o
o
o
1
o
o
o
o
o
0
o
o
o
o
o
'o
00
1
1
(50
o
2
IS
o
03
o
'a,
o
1
*H C3
o o
s 'S
able 5-1) and do
ih and water crite
;riterion.
b« ^
tn O c
dy weight value
mates used in th
sed in the fish o;
0 rs 3
lit
^5 eo .S
•^ §,1
O (D 2
OH !_, 3
3 aj g
•o « 2
es are base
t surface w
g water exj
3 § .S
sjJT C "S
w < Q
o .0 o
p
en
o
1
Uf
O
3
I
3
i
-------
S
1
I
I
<
S
•2
o
H
CM
O
Percent
OS
C3
1 Exposure
£• V
-a "»
. 53 ^
ao
CM e1
Women o
Childbeari
Age
u _
s 1— •
"s "g ••»
•"" 3 R
^ §s
5^0*
•<
>2
C3
fi 0}
'•& 0
u «
So
en
U-l
o
0
o
ON
s
o
vo
a>
Q
cn
o
c>
04
O
o
o
o
VO
0
o
cn
o
o
o
o
o
VO
C5
*
cn
0
c
0
7 — I
o
o
(-
o
o
o
cn
o
o
0
o
o
o
o
o
o
o
C3
'o
OT3
c
_O
S
60
child soil in
S
'a,
•8
3
"o
c
™ £3
) and do no
ater criterio
— ' g C
1 "^ Q
-------
§
U
CO
I
I
(D
I
CO
'S
(D
t>0
I
KS
I
I
dl
3
Q
«M rt
O °
a
e
t«
1
•S SB
4* cfi
60,
3|
o^
a
3
I
'S
a
£
"s
X
f-
o ®^
.=< en
x o
ON p
O
10 o
to
O oo
— 0\
o
og
p
o
p
o
H a
-I
6^
s
2
U
x
oo
X
vq
tN
X oo
oo p
u-> C5
x 5
o o
en p
•^J- O
Population
oto/ Exposu
O
1
en
s
_o
«
s
"«
U
•o
-------
I
a
15
CO
O
'i
I
I
C/3
13
3
,2
J
•s
•a
I
I?
g
I
•3
O
"
^
Q
co
«
IS
£
II
5 g-
a
'3
C/3
E
£>
2 S
| |
Population of Concern
X
VO
O\ p
W) O
l-H CO
in p
>n o
CM
°l
O
O
,-H co
O 0
Children Age 0-14 Y
%To
o
f"M
X
p
06
co p
—' o
-H CO
vo O
IN 0
.
CO
CO
0
o
0
oo p
V) O
p
O
tl
60"^
Women of Childbea
VO
~ 0
xS
CO O
vo O
^f O
%
0.0
o
_, CD
O CD
O
O
VO p
>n o
VO
53
o
1
"8
%
a
S
H S
ure param
oncentrat
e-1
o
jp
o
si
-------
(D
I
I
I
U
Q
•S
CS
I
O
ON
W)
I
I
I
I
i
o
o °
W
%
T3
M
1-
£
Exp
'reshwate
Estuarine
Fish
S 2 I
2 SS
5 3
II
S
I
o
o
a,
o
x
"-1 S
xg
cy\ o
in o
p
CJ
>n
o
0
p
ci
0 0
i
t
14 Y
%To
o
x o
CO O
^ o
xg
\o o
fe
O
vd
o
0
o
o
p
o
Idbearing Age
Total Expos
%
6
§
I
X
00
o
o
o
1-3 o
xg
\o o
•
T~H V
fa
O
oo
n
xg
Population
Total Expos
1
IT)
o
'S
5
i
•i*
i
ei
-------
5.7 RELATIVE SOURCE CONTRIBUTION (RSC) ESTIMATES
5.7.1 RSC Policy Summary
As described in Section 5.1, water quality criteria for noncarcinogens account for anticipated
exposures from sources other than drinking water and freshwater/estuarine fish ingestion. These
exposures can include other dietary intakes, air, and soil. By accounting for other exposures, the entire
RfD is not attributed to drinking water and freshwater/estuarine fish consumption alone. The relative
source contribution (RSC) approach apportions the RfD to ensure that the water quality criterion is
sufficiently protective, given the other anticipated sources of exposure. Thus, accounting for nonwater
exposure sources results in a more stringent water quality criterion than if those sources were not
considered. Details of the RSC approach (the Exposure Decision Tree) are described in more detail in
the 2000 Human Health Methodology (U.S. EPA, 2000a).
!_
The RSC determination differs from chemical to chemical depending on several factors: (a) the
magnitude of total exposure compared with the RfD; (b) the adequacy of data available; (c) whether more
than one guidance or criterion is to be set for the chemical in question; and (d) whether there is more than
one significant exposure source for the chemical and population of concern. The target population for
this methylmercury criterion is discussed in Section 5.2; the sources of methylmercury exposure,
exposure estimates, and data adequacy are discussed in Sections 5.3 through 5.5.
5.7.2 Target Population for RSC/Rationale for Approach to Methylmercury
The target population for the RSC estimate is the general population. The health risk measure, the
RfD, is intended to be protective of the whole population, including (but not restricted to) sensitive
subpopulations. This is not a developmental RfD per se. Even though the critical endpoint was
neurotoxic effects observed in children exposed in utero, application of the RfD is not restricted to
pregnancy only, or to developmental periods only.
As discussed in the 2000 Human Health Methodology, the RSC policy approach allows for use of a
subtraction method to account for other exposures when one health-based criterion is relevant for the
chemical in question. In this circumstance, other sources of exposure can be considered "background"
and can be subtracted from the RfD. Such is the case with methylmercury; that is, there are no health-
based criteria, pesticide tolerances, or other regulatory activities to warrant apportionment using the
alternate percentage method.
5-56 Methylmercury Water Quality Criterion 1/3/01
-------
5.7.3 Data Adequacy for RSC Estimate
Section 5.4 describes information on levels of occurrence and provides estimates of exposure to
methylmercury in ambient surface water, drinking water, fish, nonfish foods, air, soil, and sediment. The
information in Section 5.4 indicates that, for almost all media sources, the sampling data meet the
adequacy requirements (e.g., sample sizes, representativeness) for describing both central tendency and
high-end concentrations for those sources (Box 3 of the Methodology Decision Tree approach [U.S.
EPA, 2000a]). Thus, the data summarized for ambient surface water concentrations, nonfish dietary
concentrations, marine fish concentrations, and soil concentrations are adequate to use for estimating
overall exposure and RSC. Available data on methylmercury in ground water and estimates of
methylmercury in drinking water are not as adequate, as defined by the data adequacy requirements in the
2000 Human Health Methodology. However, the estimates made for both ground water and drinking
water in Section 5.4.2.3 indicate levels no higher in magnitude than the surface water estimates, even
when using most high-end values. Information on ambient air concentrations summarized from the
MSRC failed to indicate sample sizes, geographic representativeness, or detection limits and, thus, are
not considered adequate in terms of the Methodology's Decision Tree (Box 3) requirements. However,
98% of mercury in ambient air occurs in the form of vapor-phase elemental mercury, according to the
MSRC. Therefore, exposures to methylmercury in ambient air are probably negligible. This assumption
is supported by the estimates presented in Section 5.4.5, including the MSRC model simulations
predicting exposures of zero near a waste incinerator.
5.7.4 RSC Estimate/Apportionment of the RFD
Once it has been determined that the data are adequate to describe exposure intakes for relevant
exposure sources and that there are no other health-based criteria to apportion, exposure intakes from
sources other than the source addressed by the criterion are subtracted from the RfD (Box 12 of the
Decision Tree, see U.S. EPA, 2000a). Based on the available data, human exposures to methylmercury
from all media sources except freshwater/estuarine and marine fish are negligible, both in comparison to
exposures from fish and compared to the RfD. Estimated exposure from ambient water, drinking water,
nonfish dietary foods, air, and soil are all, on average, at least several orders of magnitude less than those
from freshwater/estuarine fish intakes. Nonfish sources of intake are in the range of 10"s to 10"9 jig
methylmercury/kg body weight-day for adults in the general population. The combined methylmercury
exposure intakes from water ingestion, (nonfish) diet, air, and soil represent approximately 0.07% of total
estimated exposure to methylmercury (and less than 1/100 of 1% of the RfD) for adults in the general
Methylmercury Water Quality Criterion 1/3/01
5-57
-------
population. Therefore, these exposures are not factored into the RSC because they will not quantitatively
affect the final criterion value.
Ingestion of marine fish is a significant contributor to total methylmercury exposure. The MSRC
(U.S. EPA, 1997c) indicates that in the general population offish consumers, those that consume
freshwater/estuarine species offish are also consumers of marine species offish. EPA has, therefore,
made the assumption in the derivation of the methylmercury fish tissue criterion. In making this
assumption, EPA does not believe that, by and large, the high-end consumer of freshwater/estuarine fish
is also a high-end consumer of marine fish. The Agency believes that it is more appropriate, and a
reasonably conservative assumption, to use the average intake rate (approximately 12.5 g/day) for the
marine fish component of the RSC estimate.
The marine fish exposure source is estimated using species-specific mean methylmercury fish
tissue data from NMFS (see Section 5.4.4.4) and calculating species-weighted intakes from the CSFn
consumption rates (see Section 5.4.4.7). Following the MSRC (U.S. EPA, 1997c), nearly 100% of the
mercury in marine fish was assumed to be present as methylmercury. The RSC estimate from marine
fish has been calculated with an overall assumed average intake of 12.46 g/day of marine fish based on
the CSFn, for all respondents aged 18 and over. The estimated weighted-average methylmercury
concentration in marine fish is 0.157 mg methylmercury/kg fish, and the estimated average exposure to
methylmercury from marine fish is 2.7 x 10'5 mg methylmercury/kg body weight-day. This exposure
represents 27% of the RfD.
All exposure intake values estimated for methylmercury are presented in Table 5-30. The RSC
factor in this case is determined by adding the estimated intakes that are quantitatively relevant for
methylmercury; that is, only the intake from marine fish consumption of 2.7 x 10'5 mg/kg-day has any
affect on the calculation. This amount is subtracted from the RfD of 0.1 ng methylmercury/kg body
weight-day or 1.0 x 10^ mg methylmercury/kg body weight-day. The remainder of the RfD is used to
calculate the fish tissue residue concentration in terms of the assumed body weight and
freshwater/estuarine fish ingestion. This results in an amount of methylmercury that is allowable in
freshwater/estuarine fish and that will not exceed the RfD, considering the additional exposure from
marine fish consumption.
5-58
Methylmercury Water Quality Criterion 1/3/01
-------
Table 5-30. Exposure estimates for methylmercury and percent of total exposure based on adults in the
general population
Exposure Source
Ambient water intake
Drinking water intakea
Nonfish dietary intake
Marine fish intake
Air intake
Soil Intake
Total intake
Exposure Estimate
(mg/kg-day)
4.3 x 10'9
5.6 x lO'8
0
2.7 x 10'5
4.6 x lO'9
1.3 x ID'9
9.2 xlO'5
Percent of Total
Exposure
0.0047
0.0605
0
29.33
0.005
0.0014
100
Percent of
RfD
0.004
0.006
0
27
0.005
0.001
27.01
This represents the high-end of the range of estimates. Because the contribution of ambient water or drinking water intake to
total exposure is so negligible in comparison to the sum of intake from other sources, there is no difference in the total exposure
estimated using either of these two alternatives.
Methylmercury Water Quality Criterion 1/3/01
5-59
-------
-------
6.0 METHYLMERCURY BIOACCUMULATION
6.1 INTRODUCTION
Aquatic organisms can accumulate and retain certain chemicals in their bodies when exposed to
these chemicals through water, their diet and other sources. This process is called bioaccumulation. In
order to prevent harmful exposures to waterborne pollutants through the consumption of contaminated
fish and shellfish, national 304(a) water quality criteria for,the protection of human health must address
the process of chemical bioaccumulation hi aquatic organisms. For deriving national 304(a) ambient
water column criteria to protect human health, EPA accounts for potential bioaccumulation of pollutants
in fish and shellfish through the use of national bioaccumulation factors (BAFs). A national BAF is a
ratio (in L/kg) which relates the concentration of a chemical in water to its expected concentration in
commonly consumed aquatic organisms in a specified trophic level. The magnitude of bioaccumulation
by aquatic organisms varies widely depending on the chemical but can be extremely high for some highly
persistent and hydrophobic chemicals. For such highly bioaccumulative chemicals, concentrations in
aquatic organisms may pose unacceptable human health risks from fish and shellfish consumption even
when concentrations in water are too low to cause unacceptable health risks from drinking water
consumption alone. These chemicals may also biomagnify in aquatic food webs, a process whereby
chemical concentrations increase in aquatic organisms of each successive trophic level due to increasing
dietary exposures (e.g., increasing concentrations from algae, to zooplankton, to forage fish, to predator
fish). Methylmercury is a chemical that bioaccumulates and biomagnifies to a relatively high extent.
Methylmercury BAFs for upper trophic level freshwater and estuarine fish and shellfish typically
consumed by humans generally range between 500,000 and 10,000,000 (Glass et al., 1999; Lores et al.,
1998; Miles and Fink, 1998; Monson and Brezonik, 1998; Watras et al., 1998; Mason and Sullivan,
1997).
6.2 ISSUES IN DEVELOPING METHYLMERCURY BAFS
The fates of mercury and methylmercury in the environment are complex processes affected by
numerous biotic and abiotic factors that are subjects of ongoing research by various government, private,
and academic groups around the world. Methylation of mercury is a key step in the entrance of mercury
into food chains. The biotransformation of inorganic mercury species to methylated organic species in
water bodies can occur in the sediment and the water column. Inorganic mercury can be absorbed by
aquatic organisms but is generally taken up at a slower rate and with lower efficiency than is
Methylmercury Water Quality Criterion 1/3/01
6-1
-------
methylmercury. Methylmercury continues to accumulate in fish as they age. Predatory organisms at the
top of aquatic and terrestrial food webs generally have higher methylmercury concentrations because
methylmercury is typically not completely eliminated by organisms and is transferred up the food chain
when predators feed on prey; for example, when a largemouth bass feeds on a bluegill sunfish, which fed
on aquatic insects and smaller fish, all of the prey could contain some amount of methylmercury that gets
transferred to the predator. Nearly 100% of the mercury that bioaccumulates in upper trophic level fish
(predator) tissue is methylmercury (Bloom, 1992; Akagi, 1995; Kim, 1995; Becker and Bigham, 1995).
Numerous factors can influence the bioaccumulation of mercury in aquatic biota. These include,
but are not limited to, the acidity (pH) of the water, length of the aquatic food chain, temperature, and
dissolved organic material. Physical and chemical characteristics of a watershed, such as soil type and
erosion or proportion of area that is wetlands, can affect the amount of mercury that is transported from
soils to water bodies. Interrelationships among these factors are poorly understood and are likely to be
site-specific. No single factor (including pH) has been correlated with extent of mercury
bioaccumulation in all cases examined. Two lakes that are similar biologically, physically, and
chemically can have different methylmercury concentrations hi water, fish, and other aquatic organisms
(Cope et al., 1990; Grieb et al., 1990; Jackson, 1991; Lange et al., 1993). For more in-depth discussions
about the chemical, physical, and biological interactions affecting methylmercury bioaccumulation in
aquatic organisms see the Mercury Study Report to Congress (MSRC), Volume HI and Volume III
Appendix D (U.S. EPA, 1997c), and the compilation of papers in Mercury Pollution: Integration and
Synthesis (Watras and Huckabee, 1994).
To derive section 304(a) water quality criteria for the protection of human health, EPA needs to
conduct a human health risk assessment on the pollutant in question and to gather information on the
target population's exposure to the pollutant. Traditionally, EPA has expressed its section 304(a) water
quality criteria guidance to protect human health in the form of pollutant concentrations in ambient
surface water. To account for human exposure through the aquatic food pathway when deriving a water
column-based water quality criterion, EPA uses national BAFs (U.S. EPA 2000). A BAF is a ratio (in
L/kg) that relates the concentration of a chemical in water to its expected concentration in commonly
consumed aquatic organisms in a specified trophic level (U.S. EPA 2000). A national BAF is meant to
be broadly applicable to all waters in the United States, whereas a site-specific BAF is based on local
data and integrates local spacial and temporal factors that can influence bioaccumulation. For pollutants
that biomagnify, such as methylmercury, EPA's preferred approach for deriving national BAFs for use in
deriving section 304(a) water quality criteria is to use empirical field data collected in the natural
6-2
Methylmercury Water Quality Criterion 1/3/01
-------
environment. EPA prefers this approach because BAFs derived with field data integrate the chemical,
biological, and physical factors that can affect bioaccumulation in fish and shellfish. With this
preference in mind, EPA explored the feasibility of developing field-derived national methylmercury
BAFs for each trophic level of the aquatic food chain consumed by humans (i.e., trophic levels 2-4).
Using Agency guidance on BAFs contained in the 2000 Human Health Methodology and procedures
outlined in Volume HI, Appendix D of the peer-reviewed MSRC (U.S. EPA, 1997c), EPA empirically
derived draft national methylmercury BAFs for each trophic level of the aquatic food chain. The draft
national BAFs were single value trophic level-specific BAFs calculated as the geometric mean of field
data collected across the United States and reported in the open literature as well as other publically
available reports. These draft methylmercury BAFs were compiled in a draft internal report and
submitted to a panel of external scientific experts for peer review. The Appendix contains a summary of
the internal BAF report and BAF peer review report. The entire internal draft methylmercury BAF report
and peer review report can be obtained from the Water Docket W-00-20.
Within any given trophic level, the individual empirically derived draft methylmercury BAFs
generally ranged up to two orders of magnitude. This range in BAFs reflects the various biotic factors
(such as food chain interactions and fish age/size) and abiotic factors (such as pH and dissolved organic
carbon). The large range in the individual empirically derived draft methylmercury BAFs results in
uncertainty as to the ability of single trophic level-specific national methylmercury BAFs to accurately
predict bioaccumulation of methylmercury in general across the waters of the United States. Presently, it
is EPA's understanding that the mechanisms that underlie many of the influencing factors are not well
understood and can not be accurately predicted. As the science of methylmercury improves, in the future
it may be possible predict or model these processes and use such information to more accurately predict
bioaccumulation. Until such time, EPA is unable to improve the predictive power of the methylmercury
BAFs by universally accounting for influencing factors. This is not the case for other highly
bioaccumulative pollutants; for example polychlorinated biphenyls (PCBs). For such pollutants, EPA
has methods that improve the predictive capability of empirically derived or model predicted BAFs (such
as normalizing fish tissue concentrations to lipid and normalizing ambient water concentrations to
dissolved and particulate organic carbon). EPA is actively involved in, and will continue to support,
various types of research aimed at better understanding the fate of mercury in the environment and the
processes that underlie methylmercury bioaccumulation. EPA hopes that results of new research will
enable better predictions of methylmercury bioaccumulation.
Methylmercury Water Quality Criterion 1/3/01
6-3
-------
The BAF peer reviewers recognized the need for methylmercury BAFs and were supportive of most
aspects of the methodology used to derive the draft national methylmercury BAFs. The peer reviewers
did have issues with certain data used to derive the methylmercury BAFs and certain assumptions about
food chain relationships. Overall, most of the peer reviewers believed that derivation of single-value
trophic level-specific national BAFs for methylmercury that would be generally applicable to all waters
of the United States under all conditions is difficult at best, and perhaps impossible. This opinion was
based on consideration of the highly site-specific nature of methylmercury bioaccumulation in aquatic
environments and the large range in the empirically derived draft methylmercury BAFs. These peer
reviewers recommended developing methylmercury BAFs on a more local or regional scale, if not on a
site-specific basis. Although EPA generally agrees with this suggestion, the data needed to derive BAFs
at more localized scales across the U.S. are not available. See Appendix A for a summary of the internal
BAF report and the BAF peer review report.
6.3 CONSIDERATION OF A FISH TISSUE RESIDUE CRITERION
After considering the various issues about mercury fate in the environment, the recent report by the
National Research Council (NRC, 2000) on the lexicological effects of mercury, and the methylmercury •
BAF peer review comments, EPA concluded that it is more appropriate at this time to derive a fish tissue
(including shellfish) residue water quality criterion for methylmercury rather than a water column-based
water quality criterion. EPA believes a fish tissue residue water quality criterion for methylmercury is
appropriate for many reasons. A fish tissue residue water quality criterion integrates spatial and temporal
complexity that occurs in aquatic systems and that affect methylmercury bioaccumulation. A fish tissue
residue water quality criterion in this instance is more closely tied to the CWA goal of protecting the
public health because it is based directly on the dominant human exposure route for methylmercury. The
concentration of methylmercury is also generally easier to quantify in fish tissue than in water and is less
variable in fish and shellfish tissue over the time periods in which water quality standards are typically
implemented in water quality-based controls, such as NPDES permits. Thus, the data used in permitting
activities can be based on a more consistent and measurable endpoint. Finally, this approach is
consistent with the way in which fish advisories are issued. Fish advisories for mercury are also based on
the amount of methylmercury in fish tissue that is considered acceptable, although such advisories are
usually issued for a certain fish or shellfish species in terms of a meal size. A fish tissue residue water
quality criterion should enhance harmonization between these two approaches for protecting the public
health.
6-4
Methylmercury Water Quality Criterion 1/3/01
-------
Because EPA did not use national, empirically derived methylmercury BAFs to establish today's
section 304(a) recommended methylmercury water quality criterion, EPA has deferred further efforts to
derive national BAFs for methylmercury at this time. EPA notes, however, that there may be adequate
field data for some waterbodies or geographical regions on which to base accurate predictive, site-
specific methylmercury BAFs. EPA may reconsider developing national methylmercury BAFs in the
future once more field data is available for a broader range of species and aquatic ecosystems, or once
more information is available describing the mechanisms that affect bioaccumulation. Such information
could enable EPA to more accurately predict methylmercury bioaccumulation on a broader scale given a
certain total mercury concentration in water.
Methylmercury Water Quality Criterion 1/3/01
6-5
-------
-------
7.0 WATER QUALITY CRITERION CALCULATION
7.1 EQUATION FOR TISSUE RESIDUE CONCENTRATION AND PARAMETERS USED
The equation for calculating the methylmercury fish tissue residue criterion is:
BWx(RfD - RSC)
iKC =
Where:
TRC
RfD
RSC
BW
FI
Fish tissue residue criterion (mg methylmercury/kg fish) for freshwater and
estuarine fish
Reference dose (based on noncancer human health effects) of 0.0001 mg
methylmercury/kg body weight-day
Relative source contribution (subtracted from the RfD to account for marine fish
consumption) estimated to be 2.7 x 10"5 mg methylmercury/kg body weight-day
Human body weight default value of 70 kg (for adults)
Fish intake at trophic level (TL) i (i = 2, 3,4); total default intake is 0.0175 kg
fish/day for general adult population. Trophic level breakouts for the general
population are: TL2 = 0.0038 kg fish/day; TL3 = 0.0080 kg fish/day; and TL4 =
0.0057 kg fish/day.
This yields a methylmercury TRC value of 0.3 mg methylmercury/kg fish (rounded to one significant
digit from 0.288 mg methylmercury/kg fish).
This equation is essentially the same equation used in the 2000 Human Health Methodology to
calculate a water quality criterion, but is rearranged to solve for a protective concentration in fish tissue
rather than in water. Thus, it does not include a BAF or drinking water intake value (as discussed above,
exposure from drinking water is negligible). The TRC of 0.3 mg methylmercury/kg fish is the
concentration in fish tissue that should not be exceeded based on a total consumption of 0.0175 kg
fish/day.
Methylmercury Water Quality Criterion 1/3/01
7-1
-------
7.2 SITE-SPECIFIC OR REGIONAL ADJUSTMENTS TO CRITERIA
Several parameters in the Water Quality Criterion equation can be adjusted on a site-specific or
regional basis to reflect regional or local conditions and/or specific populations of concern. These
include the fish consumption rates and the RSC estimate. States and authorized Tribes can also choose to
apportion an intake rate to the highest trophic level consumed for their population or modify EPA's
default intake rate based on local or regional consumption patterns. EPA strongly encourages States and
authorized Tribes to consider developing a criterion using local or regional data over the default values if
they believe that they would be more appropriate for their target population. States and authorized
Tribes are encouraged to make such adjustments using the guidance provided in the 2000 Human Health
Methodology (U.S. EPA, 2000a).
7-2
Methylmercury Water Quality Criterion 1/3/01
-------
8.0 REFERENCES
Aaseth J., A. Wannag, and T. Norseth. 1976. The effect of N-acetylated DL-penicillamine and DL-
homocysteine thiolactone on the mercury distribution in adult rats, rat foetuses and Macaca monkeys
after exposure to methyl mercuric chloride. Acta Pharmacol. Toxicol. 39:302-311 (as cited in Luecke et
al., 1997).
Aberg, B., L. Ekman, R. Falk, U. Greitz, G. Persson, and J. Snihs. 1969. Metabolism of methyl mercury
(Hg) compounds in man: excretion and distribution. Arch. Environ. Health 19:478-484.
Akagi, H., O. Malm, Y. Kinjo, M. Harada, F.J.P. Branches, W.C. Pfeiffer, and H. Kato. 1995T
Methylmercury pollution in the Amazon, Brazil. Sci. Total Environ. 175:85-95.
Akagi-H, I. Kanoka, and K. Kaneko. 1997. J. Jpn. Soc. Obstet. Gynecol. Neonat. Hematol. 7(2):S112-
S113.
Al-Shahristani, H., and K.M. Shihab. 1974. Variation of biological half-life of methyl mercury in man.
Arch. Environ. Health 28:342-344.
Allen, B.C., R.J. Kavlock, C.A. Kimmel, and E.M. Faustman. 1994. Dose-response assessment for
developmental toxicity. n. Comparison of generic benchmark dose estimates with no observed adverse
effect levels. Fundam. Appl. Toxicol. 23(4):487-495.
Altmann, L., K. Sveinsson, U. Kramer, et al. 1998. Visual functions in 6-year-old children in relation to
lead and mercury levels. Neurotoxicol. Teratol. 20(1):9-17.
Amin-Zaki, L., S. Elhassani, M.A. Majeed, T.W. Clarkson, R.A. Doherty, and M. Greenwood. 1974.
Intra-uterine methylmercury poisoning in Iraq. Pediatrics 54:587-595.
Amin-Zaki, L., S. Elhassani, M. Majeed, T. Clarkson, R. Doherty, and M. Greenwood. 1976. Perinatal
methylmercury poisoning in Iraq. Am. J. Dis. Child 130:1070-1076.
Amin-Zaki, L., M. Majeed, S. Elhassani, T. Clarkson, M. Greenwood, and R. Doherty. 1979. Prenatal
methylmercury poisoning. Am. J. Dis. Child 133:172-177.
Amin-Zaki, L., M. Majeed, M. Greendow, et al. 1981. Methylmercury poisoning in the Iraqi suckling
infant: A longitudinal study over five years. J. Appl. Toxicol. 1:210-214.
Andersen, M.E., HJ. Clewell, and K. Krishnan. 1995. Tissue dosimetry, pharmacokinetic modeling, and
interspecies scaling factors. Risk Anal. 15:533-537.
Arito, H., and M. Takahashi. 1991. Effect of methyl mercury on sleep patterns in the rat. In: Suzuki, T.,
N. Imura, and T.W. Clarkson, eds. Advances in mercury toxicology. New York: Plenum Press, 381-394.
Aschner, M., and J.L. Aschner. 1990. Mercury neurotoxicity: mechanisms of blood-brain barrier
transport. Neurosci. Biobehav. Rev. 14(2): 169-176.
ATSDR (Agency for Toxic Substances Disease Registry). 1999. Toxicological profile for mercury.
Update. Atlanta, GA: ATSDR.
Methylmercury Water Quality Criterion 1/3/01
R-l
-------
Axtell, C.D., GJ. Myers, P.W. Davidson, A.L. Choi, E. Cernichiari, J. Sloane-Reeves, C. Cox, C.
Shamlaye, and T.W. Clarkson. 1998. Semiparametric modeling of age at achieving developmental
milestones after prenatal exposure to methylmercury in the Seychelles child development study. Environ.
Health Perspect. 106(9):559-564.
Axtell, C.D., C. Cox, GJ. Myers, P.W. Davidson, A.L. Choi, E. Chernichiari, J. Sloane-Reeves, C.F.
Shamlaye, and T.W. Clarkson. 2000. Association between methylmercury exposure from fish
consumption and child development at five and a half years pf age in the Seychelles child development
study: an evaluation of nonlinear relationships. Environ. Res. Section A. 84:71-80.
Baglan R.J., A.B. Brill, A. Schulert, D. Wilson, K. Larsen, N. Dyer, M. Mansour, W. Schaffner, L.
Hoffman, and J. Davies. 1974. Utility of placental tissue as an indicator of trace element exposure to
adult and fetus. Environ. Res. 8:64-70.
Bahnick, D., C. Sauer, B. Butterworth, and D. Kuehl. 1994. A national study of mercury contamination
offish. Chemosphere 29:537-546.
BaMr, F., S. Damluji, L. Amin-Zaki, et al. 1973. Methylmercury poisoning in Iraq. Science 181:230-241.
Baldi, F., and M. Filippelli. 1991. New method for detecting methylmercury by its enzymatic conversion
to methane. Environ. Sci. Technol. 25(2):302-305.
Ballatori, N., and T. Clarkson. 1982. Developmental changes hi the biliary excretion of methyl mercury
and glutathione. Science 216(2):61-63.
Becker, D.S., and G.N. Bigham. 1995. Distribution of mercury in the aquatic food web of Onondaga
Lake, New York. Water Air Soil Pollut. 80:563-571.
Beh, H.C., R.D. Roberts, and A. Pritchard-Levy. 1994. The relationship between intelligence and choice
reaction time within the framework of an extended model of Hick's Law: a preliminary report. Person.
Individ. Diff. 16:891-897.
Bellinger, D. 1995. Interpreting the literature on lead and child development: the neglected role of the
"experimental system." Neurotoxicol. Teratol. 17(3):201-212.
Berglund, F., M. Berlin, G. Birke, R. Cederlof, U. von Euler, L. Friberg, B. Holmsteadt, E. Jonsson, et al.
1971. Methyl mercury in fish: a toxicologic-epidemiologic evaluation of risks. Report from an expert
group. Nordisk Hygienisk Tidskrift, Stockholm Suppl. 4:19-364.
Bernard, S., and P. Purdue. 1984. Metabolic models for methyl and inorganic mercury. Health Phys.
46(3):695-699.
Bemaudin, J. F., E. Druet, P. Druet, et al. 1981. Inhalation or ingestion of organic or inorganic
mercurials produces auto-immune disease in rats. Clin. Immunol. Immunopathol. 20:129-135.
Best, C. H. 1961. The Physiological Basis of Medical Practice. Baltimore, p. 19 and 29.
Betti, C., T. Davini, and R. Barale. 1992. Genotoxic activity of methyl mercury chloride and dimethyl
mercury in human lymphocytes. Mutat. Res. 281(4):255-260.
R-2
Methylmercury Water Quality Criterion 1/3/01
-------
Bidone, E.D., Z.C. Castilhos, T.M. Cid de Souza, et al. 1997. Fish contamination and human exposure to
mercury in the Tapajos River Basin, Para State, Amazon, Brazil: a screening approach. Bull. Environ.
Contam. Toxicol. 59(2): 194-201.
Birke, G., G. Johnels, L-O Plantin, B. Sjostrand, S. Skerfving, and T. Westermark. 1972. Studies on
humans exposed to methyl mercury through fish consumption. Arch. Environ. Health 25:77.
Bjerregaard P., and J.C. Hansen. 2000. Organochlorines and heavy metals in pregnant women from the
Disko Bay area in Greenland. Sci. Total Environ. 17:245(1-3): 195-202.
Blakley, B. R. 1984. Enhancement of urethane-induced adenoma formation hi Swiss mice exposed to
methylmercury. Can. J. Comp. Med. 48:299-302.
Blakley, B.R., C.S. Sisodia, and T.K. Mukkur. 1980. The effect of methyl mercury, tetraethyl lead, and
sodium arsenate on the humoral immune response in mice. Toxicol. Appl. Pharmacol. 52:245-254.
Bloom, N.S. 1992. On the chemical form of mercury in edible fish and marine invertebrate tissue. Can. J.
Fish. Aquat. Sci. 49:1010-1017.
Bloom, N.S., and S.W. Effler. 1990. Seasonal variability in the mercury speciation of Onpndaga Lake
(New York). Water Air Soil Pollut. 56:477-491.
Bloom, N., and W.F. Fitzgerald. 1988. Determination of volatile mercury species at the picogram level
by low-temperature gas chromatography with cold-vapor atomic fluorescence detection. Analytica
Chimica Acta, 208:151-161.
Bloom, N.S., and E. Kuhn. 1994. Mercury speciation in meat products, personal communication. October
1, 1994.
Bloom, N.S., and C.J. Watras. 1989. Observations of methylmercury in precipitation. Sci. Total
Environ. 87/88:199-207. •
Bloom, N.S., CJ. Watras, and J.P. Hurley. 1991. Impact of acidification on the methylmercury cycle of
remote seepage lakes. Water Air Soil Pollut. 56:477-491.
Bonthius, D.J., and J.R. West. 1990. Alcohol-induced neuronal loss in developing rats: increased brain
damage with binge exposure. Alcohol Clin. Exp. Res. 14(1): 107-118.
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.
Borum, D. 2000. Personal communication via e-mail. May 25.
Brown, E., J. Hopper, Jr., J.L. Hodges, Jr., B. Bradley, R. Wennesland, and H. Yamuchi. 1962. Red cell,
plasma, and blood volume in healthy women measured by radiochromium cell-labeling and hematocrit. J.
Clin. Invest. 41:2182-2190.
Buckhalt, J.A., and A.R. Jensen. 1989. The British Ability Scales speed of information processing
subtest: what does it measure? Br. J. Educ. Psychol. 59:100-107.
Methylmercury Water Quality Criterion 1/3/01
R-3
-------
Budtz-J0rgensen, R, P. Grandjean, N. Keiding, et al. 2000. Benchmark dose calculations of
methylmercury-associated neurobehavioral deficits. Toxicol. Lett. 112-113:193-199.
Budtz-J0rgensen, E., N. Keiding, and P. Grandjean. 1999. Benchmark modeling of the Faroese
methylmercury data. Final Report to U.S. EPA. Research Report 99/5. Department of Biostatistics,
University of Copenhagen.
Burbacher, T. M., and K.S. Grant. 2000. Methods for studying nonhuman primates in neurobehavioral
toxicology and teratology. Neurotoxicol. Teratol. 22(4):475-86
Burbacher, T.M., K.S. Grant, and N.K. Mottet. 1986. Retarded object permanence development in
methylmercury exposed Macacafascicularis infants. Dev. Psychobiol. 22:771-776.
Burbacher, T.M., M.K. Mohamed, and N.K. Mottett. 1988. Methyl mercury effects on reproduction and
offspring size at birth. Reprod. Toxicol. 1:267-278.
Burbacher, T.M., P.M. Rodier, and B. Weiss. 1990a. Methylmercury developmental neurotoxicity: a
comparison of the effects in humans and animals. Neurotoxicol. Teratol. 12:191-202.
Burbacher, T.M., G.P. Sackett, and N.K. Mottet. 1990b. Methylmercury effects on the social behvaior of
Macacafascicularis infants. Neurotoxicol. Teratol. 12:65-71.
Byrne, A. R., and L. Kosta. 1974. Simultaneous neutron-activation determination of selenium and
mercury in biological samples by volatilization. Talanta. 21:1083-1090.
Campbell, D., M. Gonzales, and J. B. Sullivan. 1992. Mercury. In: Hazardous materials toxicology,
clinical principles of environmental health. Sullivan, J.B., and G.R. Krieger, eds. Baltimore, MD:
Williams and Wilkins, pp. 824-833.
Cappon, CJ. 1981. Mercury and selenium content and chemical form in vegetable crops grown on
sludge-amended soil. Arch. Environ. Contain. Toxicol. 10:673-689.
Cappon, C.J. 1987. Uptake and speciation of mercury and selenium in vegetable crops grown on
compost-treated soil. Water Air Soil Pollut. 34:353-361.
Cemichiari, E., R. Brewer, GJ. Myers, D.O. Marsh, L.W. Lapham, C. Cox, C.F. Shamlaye, M. Berlin,
P.W. Davidson, and T.W. Clarkson. 1995. Monitoring methylmercury during pregnancy: maternal hair
predicts fetal brain exposure. NeuroToxicology 16:705-710.
Chang, L.W., S. Yamaguchi, and J.A.W. Dudley. 1974. Neurological changes in cats following long-term
diet of mercury contaminated tuna. Acta. Neuropathol. (Berlin) 27:171-176.
Charbonneau, S.M., I. Munro, and E. Nera. 1976. Chronic toxicity of methyl mercury in the adult cat.
Toxicology 5:337-340.
Charleston J.S., R.P. Bolender, R.L. Body, T.M. Burbacher, M.E. Vahter, and N.K. Mottet. 1994.
Methylmercury induced cell population changes at specific brain sites of the monkey Macaca
fascicularis. Toxicologist 14:259.
Clarkson, T.W. 1972. The pharmacology of mercury compounds. Ann. Rev. Pharmacol. 12:375-406.
R-4
Methylmercury Water Quality Criterion 1/3/01
-------
Clarkson, T. W. 1993. Molecular and ionic mimicry of toxic metals. Annu. Rev. Pharmacol. Toxicol
32:545-571.
Clarkson, T.W., L. Amin-Zaki, and S. Al-Tikriti. 1976. An outbreak of methyl mercury poisoning due to
consumption of contaminated grain. Fed. Proc. 35:2395-2399.
Clarkson, T.W., J.B. Hursh, P.R. Sager, et al. 1988. Biological monitoring of toxic metals. New York:
Plenum Press, p. 199-246.
Cleckner, L.B., E.S. Esseks, P.O. Meier, and G.J. Keeler. 1995. Mercury concentrations in two great
waters. Water Air Soil Pollut. In press. [Note: as listed in MSRC]
Clewell, H.J. 1995. The application of physiologically based pharmacokinetic modeling in human health
risk assessment of hazardous substances. Toxicol Lett 79:207-217.
Clewell, HJ. 2000. Personal communication, April 19, 2000. ICF Consulting.
Clewell, H.J., and M.E. Andersen. 1985. Risk assessment extrapolations and physiological modeling.
Toxicol Ind Health 1(4): 111-131.
Clewell H.J., and M.E. Andersen. 1989. Biologically motivated models for chemical risk assessment.
Health Phys. 57(Suppl 1): 129-137.
Clewell, HJ., J.M. Gearhart, P.R. Gentry, T.R. Covington, C.B. Van Landingham, K.S. Crump, and
A.M. Shipp. 1999. Evaluation of the uncertainty in an oral reference dose for methylmercury due to
interindividual variability in pharmacokinetics. Risk Anal. 19:547-558.
Clewell, HJ., P.R. Gentry, and A.M. Shipp. 1998. Determination of a site-specific reference dose for
methylmercury for fish-eating populations. Peer reviewed report for the Toxicology Excellence in Risk
Assessment (TERA). International Toxicity Estimates for Risk (TTER) Database TERA, Cincinnati, OH,
February 1998. http://www.tera.org/iter/.
Coccini, T., G. Randine, S. Candura, R. Nappi, L. Prockop, and L. Manzo. 2000. Low-level exposure to
methylmercury modifies muscarinic cholinergic receptor binding characteristics in rat brain and
lymphocytes: Physiological implications and new opportunities in biologic monitoring. Environ. Health
Perspect. 108(l):29-33.
Cope, W.G., J.G. Wiener, and R.G. Rada. 1990. Mercury accumulation in yellow perch in Wisconsin
seepage lakes: Relation to lake characteristics. Environ. Toxicol. Chem. 9:931-940.
Cordier, S., and M. Garel. 1999. Neurotoxic risks in children related to exposure to methylmercury in
French Guiana. INSERT U170 and U149-Study financed by the Health Monitoring Institute (RNSP).
National Institute of Health and Medical Research.
Costa, M., N. T. Christie, O. Cantoni, et al. 1991. DNA damage by mercury compounds: An overview.
In: Advances in Mercury Toxicology, T. Suzuki, N. Imura, T.W. Clarkson, Eds. New York: Plenum
Press, pp. 255-273.
Methylmercury Water Quality Criterion 1/3/01
R-5
-------
Counter, S.A., L.H. Buchanan, G. Laurell, and F. Ortega. 1998. Blood mercury and auditory neuro-
sensory responses in children and adults in the Nambija gold mining area of Ecuador. Neurotoxicology
19(2): 185-196.
Cox, C., T. W. Clarkson, D. E. Marsh, et al. 1989. Dose-response analysis of infants prenatally exposed
to methyl mercury: An application of a single compartment model to single-strand hair analysis.
Environ. Res. 49(2):318-332.
Cox, C., D. Marsh, G. Myers, and T. Clarkson. 1995. Analysis of data on delayed development from the
1971-72 outbreak of methylmercury poisoning in Iraq: assessment of influential points. Neurotoxicology
16(4):727-730.
Cramer, G.M. 1994. Exposure of U. S. consumers to methylmercury from fish. Presented at the
DOE/FDA/EPA Workshop on Methylmercury and Human Health, Bethesda, MD. March 22-23,1994.
Crump, K., T. Kjellstrom, A. Shipp, A. Silvers, and A. Stewart. 1998. Influence of prenatal mercury
exposure upon scholastic and psychological test performance: statistical analysis of a New Zealand
cohort. Risk Anal. 18:701-713.
Crump, K., C. Landingham, C. Shamlaye, C. Cox, P. Davidson, G. Myers, and T. Clarkson. 2000.
Benchmark concentrations for methylmercury obtained from the Seychelles child development study.
Environ. Health Perspect. 108:257-263.
Crump, K., J. Viren, A. Silvers, H. Clewell, J. Gearhart, and A. Shipp. 1995. Reanalysis of dose-response
data from the Iraqi methylmercury poisoning episode. Risk Anal. 15:523-532.
Dahl, R., R.F. White, P. Weihe, N. Sorensen, R. Letz, H.K. Hudnell, D.A. Otto, and P. Grandjean. 1996.
Feasibility and validity of three computer-assisted neurobehavioral tests in 7-year old children.
Neurotoxicol. Teratol. 19(4):413-419.
Davidson, P., G. Myers, C. Cox, C. Shamlaye, D. Marsh, M. Tanner, M. Berlin, J. Sloane-Reeves, E.
Cernichiari, O. Choisy, A. Choi, and T. Clarkson. 1995. Longitudinal neurodevelopmental study of
Seychellois children following in utero exposure to methylmercury from maternal fish ingestion:
outcomes at 19 and 29 months. NeuroToxicology 16:677-688.
Davidson, P.W., GJ. Myers, C. Cox, C. Axtell, C. Shamlaye, J. Sloane-Reeves, E. Cernichiari, L.
Needham, A. Choi, Y. Yang, M. Berlin, and T.W. Clarkson. 1998. Effects of prenatal and postnatal
methylmercury exposure from fish consumption on neurodevelopment: Outcomes at 66 months of age in
the Seychelles child development study. JAMA 280:701-707.
Davidson, P.W, D. Palumbo, GJ. Myers, C. Cox, C.F. Shamlaye, J. Sloane-Reeves, E. Chernichiari, G.E.
Wilding, and T.W. Clarkson. 2000. Neurodevelopmental outcomes of Seychellois children from the pilot
cohort at 108 months following prenatal exposure to methylmercury from a maternal fish diet. Environ
Res. Section A 84:1-11.
Davis, A., N.S. Bloom, and S.S. Que Hee. 1997. The environmental geochemistry and bioaccessibility of
mercury in soils and sediments: a review. Risk Anal. 17:557-569.
Dennis, C.A., and F. Fehr. 1975. The relationship between mercury levels in maternal and cord blood.
Sci. Total Environ. 3(3):275-277.
R-6
Methylmercury Water Quality Criterion 1/3/01
-------
Dolbec, J., D. Mergler, C.J. Sousa Passes, S. Sousa de Morals, and J. Lebel. 1998. Methylmercury
exposure and neurotoxic effects in the Brazilian Amazon. Methylmercury Workshop. Raleigh, NC Nov
18-20, 1998.
Dolbec J., D. Mergler, C. J. Sousa Passos, S. Sousa de Morais, and J. Lebel. 2000. Methylmercury
exposure affects motor performance of a riverine population of the Tapajos river, Brazilian Amazon. Int
Arch Occup Environ Health 73(3): 195-203.
Dooley, J.H. 1992. Natural sources of mercury in the Kirkwood-Cohansey aquifer system of the New
Jersey Coastal Plain. New Jersey Geological Survey, Report 27.
Driscoll, C.T., V. Blette, C. Yan, C.L. Schofield, R. Munson, and J. Holsapple. 1995. The role of
dissolved organic carbon in the chemistry and bioavailability of mercury in remote Adirondack lakes.
Water Air Soil Pollut. 80:499-508.
Driscoll, C.T., C. Yan, C.L. Schofield, R. Munson, and J. Holsapple. 1994. The mercury cycle and fish in
the Adirondack lakes. Environ. Sci. Technol. 28:136A-143A.
Ershow, A.G., and K.P. Canter. 1989. Total water and tapwater intake in the United States: population-
based estimates of quantities and sources. Life Sciences Research Office, Federation of American
Societies for Experimental Biology, Bethesda, MD. (Prepared under NCI #263-MD-810264.)
Fang, S.C. 1980. Comparative study of uptake and tissue distribution of methyl mercury in female rats by
inhalation and oral routes of administration. Bull. Environ. Contam. Toxicol. 24:65-72.
Fangstrom, B., M. Athanasiadou, A. Bergman, P. Grandjean, and P. Weihe. 2000. Levels of
PCBs and hydroxylated PCB metabolites in blood from pregnant Faroe Island women. Hum.
Exposure 48:21-24.
Farris, F.F., R.L. Dedrick, P.V. Allen, J.C. Smith. 1993. Physiological model for the pharmacokinetics of
methyl mercury in the growing rat. Toxicol. Appl. Pharmacol. 119:74-90.
Fiskesjo, G. 1979. Two organic mercury compounds tested for mutagenicity in mammalian cells by use
of the cell line V 79-4. Hereditas 90:103-110.
Fitzgerald, W.F. 1994. Global biogeochemical cycling of mercury. Presented at the DOE/FDA/EPA
Workshop on Methylmercury and Human Health, Bethesda, MD, March 22-23, 1994.
Fitzgerald, W.F., R.P. Mason, and G.M. Vandal. 1991. Atmospheric cycling and air-water exchange of
mercury over mid-continental lacustrine regions. Water Air Soil Pollut. 56:745-767.
Francis, P., W. Birge, B. Roberts, et al. 1982. Mercury content of human hair: a survey of dental
personnel. J. Toxicol. Environ. Health. 10:667-672.
Franchi, E., G. Loprieno, M. Ballardin, L. Petrozzi, and L. Migliore. 1994. Cytogenetic monitoring of
fishermen with environmental mercury exposure. Mutat. Res. 320:23-29.
Fujita, M., and E. Takabatake. 1977. Mercury levels in human maternal and neonatal blood, hair and
milk. Bull. Environ. Contam. Toxicol. 18(2):205-209.
Methylmercury Water Quality Criterion 1/3/01
R-7
-------
Fukuda, Y., K. Ushijima, T. Kitano, M. Sakamoto, and M. Futatsuka. 1999. An analysis of subjective
complaints in a population living in a methylmercury-polluted area. Environ. Res. 81:100-107.
Fredriksson, A., L. Dencker, T. Archer, Danielsson. 1996. Prenatal coexposure to metallic mercury
vapour and methylmercury produce interactive behavioural changes in adult rats. Neurotoxicol. Teratol.
18(2): 129-134.
Futatsuka M., T. Kitano, M. Shono, Y. Fukuda, K. Ushijima, T. Inaoka, M. Nagano, J. Wakamiya, and
K. Miyamoto. 2000. Health surveillance in the population living in a methylmercury-polluted area over a
long period. Environ Res 83(2):83-92.
Fuyuta, M., T. Fujimoto, and S. Hirata. 1978. Embryotoxic effects of methylmercuric chloride
administered to mice and rats during organogenesis. Teratology 18(3):353-366.
Fuyuta, M., T. Fujimoto, and E. Kiyofuji. 1979. Teratogenic effects of a single oral administration of
methylmercuric chloride in mice. Acta Anat. (Basel) 104(3):356-362.
Ganther, H.E. 1978. Modification of methyl mercury toxicity and metabolism by selenium and vitamin E:
possible mechanisms. Environ. Health Perspect. 25:71-76.
Gaylor, D.W., and W. Slikker. 1992. Risk assessment for neurotoxicants. In: Neurotoxicology. Tilson,
H., and C. Mitchell, eds. New York: Raven Press, pp. 331-343.
Gearhart, J., H. Clewell, K. Crump, A. Shipp, and A. Silvers. 1995. Pharmacokinetic dose estimates of
mercury in children and dose-response curves of performance tests in a large epidemiological study. In:
Mercury as a global pollutant. Porcella, D.B., J.W. Huckabee, and B. Wheatley, Eds. Boston: Kluwer
Academic Publishers, pp. 49-58.
Gilbert, S.G. and Grant-Webster, K.S. 1995. Neurobehavioral effects of developmental methylmercury
exposure. Environ. Health Perspect. 103 Suppl. 6: 135-142.
Ginsberg, G.L. and B. F. Toal. 2000. Development of a single-meal fish consumption advisory for
methylmercury. Risk Analysis. 20:41-47.
Glass, G., J. A. Sorenson, K. W. Schmidt, and G. R. Rapp, Jr. 1990. New source identification of
mercury contamination in the Great Lakes. Environ. Sci Technol. 24:1059-1068.
Glass, G., J.A. Sorenson, and G.R. Rapp, Jr. 1999. Mercury deposition and lake quality trends. Final
report Project: 1-11/1-15, Legislative Commission on Minnesota Resources, St Paul, MN.
Goodlett, C.R., SJ. Kelly, and J.R. West. 1987. Early postnatal alcohol exposure that produces high
blood alcohol levels impairs development of spatial navigation learning. Psychobiology 15(l):64-74.
Grandjean, P., P. Weihe, P.J. Jorgensen, T. Clarkson, E. Cernichiari, and T. Videro. 1992. Impact of
maternal seafood diet on fetal exposure to mercury, selenium, and lead. Arch. Environ. Health 47:185-
195.
Grandjean, P., P.J. Jorgensen, P. Weihe. 1994. Human milk as a source of methylmercury exposure in
infants. Environ. Health Perspect. 102:74-77.
R-8
Methylmercury Water Quality Criterion 1/3/01
-------
Grandjean, P., P. Weihe, L.L. Needham, V.W. Burse, D.G. Patterson, Jr., E.J. Sampson, P. J. J0rgensen,
and M. Vater. 1995a. Relation of a seafood diet to mercury, selenium, arsenic, and polychlorinated
biphenyl and other organochlorine concentrations in human milk. Environ. Res. 71:29-38.
Grandjean, P., P. Weihe, and R. White. 1995b. Milestone development in infants exposed to
methylmercury from human milk. NeuroToxicology 16:27-34.
Grandjean, P., P. Weihe, R. White, F. Debes, S. Arak, K. Yokoyama, K. Murata, N. Sorensen, R. Dahl,
and P. Jorgensen. 1997. Cognitive deficit in 7-year-old children with prenatal exposure to
methylmercury. Neurotoxicol. Teratol. 20:1-12.
Grandjean, P., P. Weihe, R.F. White, N. Keiding, E., Budtz-J0rgensen, K. Murato, and L. Needham.
1998. Prenatal exposure to methylmercury in the Faroe Islands and neurobehavioral performance at age
seven years. Response to workgroup questions for presentation on 18-20 November 1998. In: Scientific
issues relevant to assessment of health effects from exposure to methylmercury. Appendix H-B.- Faroe
Islands Studies. National Institute for Environmental Health Sciences. [Online]. Available:
http://ntp-server.mehs.mh.govMain_Pages/PUBS/MethMercWkshpRpt.html.
Grandjean, P., R.F. White, A. Nielsen, D. Cleary, and E. de Oliveira Santos. 1999. Methylmercury
neurotoxicity in Amazonian children downstream from gold mining. Environ. Health Perspect. 107:587-
591.
Gray, D.G. 1995. A physiologically based pharmacokinetic model for methylmercury in the pregnant rat
and fetus. Toxicol. Appl. Pharmacol. 132:91-103.
Greenwood, M.R., T.W. Clarkson, R.A. Doherty, et al. 1978. Blood clearance half-times in lactating and
nonlactating members of a population exposed to methyl mercury. Environ. Res. 16:48-54.
Grieb, T.M., C.T. Driscoll, S.P. Gloss, C.L. Schofield, G.L. Bowie, and D.B. Porcella. 1990. Factors
affecting mercury accumulation in fish hi the upper Michigan peninsula. Environ. Toxicol. Chem. 9:919-
930.
Gunderson, V., K. Grant, T. Burbacher, J. Pagan, and N. Mottet. 1986. The effect of low-level prenatal
methylmercury exposure on visual recognition memory in infant crab-eating macaques. Child Dev.
57:1076-1083.
Gunderson, V.M., K.S. Grant-Webster, T.M. Burbacher, and N.K. Mottet, 1988. Visual recognition
memory deficits in methylmercury-exposed Macaco fascicularis infants. Neurotoxicol. Teratol. 10:373-
379.
Hall, R.A., E.G. Zook, and G.M. Meaburn. 1978. National Marine Fisheries Survey of trace
Elements in the fishery resource. NOAA Technical Report NMFS SSRF-721, U.S. Department of
Commerce, Washington, DC.
Hansen, J. 1988. Blood mercury concentrations in birth giving Greenlandic women. Arctic. Med. Res.
47(1): 175-178.
Hansen, J., E. Reske-Nielsen, O. Thorlacius-Ussing, et al. 1989. Distribution of dietary mercury in a dog.
Quantitation and localization of total mercury in organs and central nervous system. Sci. Total Environ.
78:23-43.
Methylmercury Water Quality Criterion 1/3/01
R-9
-------
Hansen, J.E., U. Tarp, and J. Bohm. 1990. Prenatal exposure to methyl mercury among Greenlandic Polar
Inuits. Arch. Environ. Health 45:355-358.
Harada, M. 1995. Minamata disease: methylmercury poisoning in Japan caused by environmental
pollution. Crit. Rev. Toxicol. 25(1): 1-24.
Harada, M., H. Akagi, T. Tsuda, T. Kizaki, and H. Ohno. 1999. Methylmercury level in umbilical cords
from patients with congenital Minamata disease. Sci. Total Environ. 234(l-3):59-62.
Harada, M., J. Nakanishi, S. Konuma, K. Ohno, T. Kimura, H. Yamaguchi, K. Tsuruta, T. Kizaki, T.
Ookawara, and H. Ohno. 1998. The present mercury contents of scalp hair and clinical symptoms in
inhabitants of the Minamata area. Environ. Res. Section A 77:160-164.
Harrison, K.A. 1966. Blood volume changes in normal pregnant Nigerian women. J. Obstet. Gynaec. Br.
Cwlth. 73:717-723.
Hatch, W.R., and W.L. Ott. 1968. Determination of sub-microgram quantities of mercury by atomic
absorption spectrophotometry. Anal. Chem. 40(14):2085-2087.
Heddle, J. R., and W. R. Bruce. 1977. Comparison of the micronucleus and sperm assay for
mutagenicity with the carcinogenic activities of 61 different agents. In: Origins of Human Cancer, H.H.
Hiatt, J.D. Watson, J.A. Winsten, Eds. Vol. 4. Cold Spring Harbor Conferences.
Henry, E.A., LJ. Dodge-Murphy, G.N. Bigham, and S.M. Klein. 1995. Modeling the transport and fate
of mercury in an urban lake (Onondaga Lake, NY). Water Air Soil Pollut. 80:489-498.
Hirano, M., K. Mitsumori, K. Maita, and Y. Shirasu. 1986. Further carcinogenicity study on
methylmercury chloride in ICR mice. Nipon Juigaku Zasshi (Jpn. J. Vet. Sci.) 48(1):127-135.
Hislop J., T. Collier, G. White, et al. 1983. The use of keratinized tissues to monitor the detailed
exposure of man to methyl mercury from fish. Chemical Toxicology and Clinical Chemistry of Metals.
Published by IUPAC. pp. 145-148.
Hollins, J., R. Willes, F. Bryce, et al. 1975. The whole body retention and tissue distribution of
[203Hg]methyl mercury in adult cats. Toxicol. Appl. Pharmacol. 33:438-449.
H66k, O., K-D Lundgren, and A. Swensson. 1954. On alkyl mercury poisoning. Acta. Med. Scand.
150:131-137.
Huff, R.L., and D.D. Feller. 1956. Relation of circulating red cell volume to body density and obesity. J.
Clin. Invest. 35:1-10.
Hughes, J.A., and Z. Annau. 1976. Postnatal behavioral effects in mice after prenatal exposure to
methylmercury. Pharmacol. Biochem. Behav. 4(4):385-391.
Hultman, P., and H. Hansson-Georgiadis. 1999. Methyl mercury-induced autoimmunity in mice. Toxicol.
Appl. Pharmacol. 154:203-211.
Hytten, F.E., I. Leitch. 1971. The physiology of human pregnancy. 2nd ed. Oxford: Blackwell Scientific
Publications.
R-10
Methylmercury Water Quality Criterion 1/3/01
-------
Hback, N.G. 1991. Effects of methyl mercury exposure on spleen and blood natural-killer (NK) cell-
activity in the mouse. Toxicology 67(1): 117-124.
Inouye, M., and U. Murakami. 1975. Teratogenic effect of orally administered methylmercuric chloride
in rats and mice. Congenital Anom. 15:1-9.
Inouye, M., and Y. Kajiwara. 1988. Developmental disturbances of the fetal brain in guinea pigs caused
by methylmercury. Arch. Toxicol. 62(1): 15-21.
IPCS (International Programme on Chemical Safety). 1990. Environmental Health Criteria Document
101: Methylmercury. Geneva. World Health Organization.
Ja-Song, M., and R. Lynn. 1992. Reaction times and intelligence in Korean children. J. Psychol. 126:421-
428.
Jackson, T.A. 1991. Biological and environmental control of mercury accumulation by fish in lakes and
reservoirs of northern Manitoba, Canada. Can. J. Fish. Aquat. Sci. 48:2449-2470.
Jacobson, J.L., S.W. Jacobson, and H.E.B. Humphrey. 1990. Effect of in utero exposure to
polychlorinated biphenyls and related contaminants on cognitive functioning in young children. J.
Pediatr. 116:38-45.
Jacobson, J. L., and S. W. Jacobson. 1991. Assessment of teratogenic effects on cognitive and
behavioral development in infancy and childhood. In: Methodological Issues in Controlled Studies on
Effects of Prenatal Exposure to Drugs of Abuse, Research Monograph 114. M.M. Kilbey and K.
Asghar, Eds. Rockville, MD: National Institute on Drug Abuse, pp. 248-261
Jacobson, J.L., and S.W. Jacobson. 1996. Intellectual impairment in children exposed to polychlorinated
biphenyls in utero. N. Engl. J. Med. 335(11):783-789.
Jensen, A.R. 1987. Process differences and individual differences in some cognitive tasks. Intelligence
11:107-136.
Jensen, A.R. 1993a. Spearman's hypothesis tested with chronometric information-processing tasks.
Intelligence 17:47-77.
Jensen, A.R. 1993b. Why is reaction time correlated with psychometric g? Curr. Dir. Psychol. Science
2:53-56.
Jensen, A.R., Munro, E. 1979. Reaction time, movement time, and intelligence. Intelligence 3:121-126.
Jenssen, O., and C. Ramel. 1980. The micronucleus test as part of a short-term mutagem'city test
program for the prediction of carcinogenicity evaluated by 143 agents tested. Mutat. Res. 75:191-202.
Kalamegham, R., and K. O. Ash. 1992. A simple ICP-Ms procedure for the determination of total
mercury in whole blood and urine. J. Clin. Lab. Anal. 6(4): 190-193.
Kanematsu, N., M. Kara, and T. Kada. 1980. Rec assay and mutagenicity studies on metal compounds.
Mutat Res. 77:109-116.
Methylmercury Water Quality Criterion 1/3/01
R-ll
-------
Kaufman, H.C. 1969. Handbook of organometallic compounds. Princeton, NJ: Van Nostrand Co., Inc.
Kawasaki Y, Y. Eceda, T. Yamamoto, and K. Ikeda. 1986. Long-term toxicity study of methylmercury
chloride in monkeys. J. Food Hyg. Soc. Jpn. 27:528-552.
Kerper, L.E., N. Ballatori, and T.W. Clarkson. 1992. Methyl mercury transport across the blood-brain
barrier by an amino acid carrier. Am. J. Physiol. 262(5):R761-R765.
Kershaw, T.G., T.W. Clarkson, and P.H. Dhahir. 1980. The relationship between blood levels and dose
of methyl mercury in man. Arch. Environ. Health 35:28-36.
Khera, K.S. 1973. Reproductive capability of male rats and mice treated with methylmercury. Toxicol.
Appl. Pharmacol. 24(2): 167-177.
Kim, J.P. 1995. Methylmercury in rainbow trout [Oncorhynchus mykiss± from Lakes Okareka, Okaro,
Rotmahana, Rotorua and Tarawera, North Island, New Zealand._Sci. Total Environ. 164:209-219.
Kinjo, Y., H. Higashi, A. Nakano, M. Sakamoto, and R. Sakai. 1993. Profile of subjective complaints and
activities of daily living among current patients with Minamata disease after 3 decades. Environ. Res.
63(2):241-251.
Kitamura S., K. Sumino, K. Hayakawa, and T. Shibata. 1976. Dose-response relationship of
methylmercury. In: Effects and dose-response relationships of toxic metals. Nordberg, G.F., ed.
Amsterdam: Elsevier Scientific Publishing Company, pp. 262-272 (as cited in Luecke et al., 1997).
Kjellstrom, T., P. Kennedy, S. Wallis, and C. Mantell. 1986a. Physical and mental development of
children with prenatal exposure to mercury from fish. Stage 1: preliminary tests at age 4. Report 3080.
Solna, Sweden: National Swedish Environmental Protection Board.
Kjellstrom, T., P. Kennedy, S. Wallis, et al. 1986b. Physical and mental development of children with
prenatal exposure to mercury from fish. Stage 2: interviews and psychological tests at age 6. Solna,
Sweden: National Swedish Environmental Protection Board, Report 3642.
Kjellstrom, T., P. Kennedy, S. Wallis, A. Stewart, L. Friberg, B. Lind, T. Wutherspoon, and C. Mantell.
1989. Physical and mental development of children with prenatal exposure to mercury from fish. Stage 2:
interviews and psychological tests at age 6. Report 3642. Solna, Sweden: National Swedish
Environmental Protection Board, p. 112.
Koller, L.D., J.H. Exon, and B. Arbogast. 1977. Methyl mercury: Effect on serum enzymes and humoral
antibody. J. Toxicol. Environ. Health 2:1115-1123.
Korogi, Y., M. Takahashi, T. Hirai, I. Ikushima, M. Kitajima, T. Sugahara, Y. Shigematsu, T. Okajima,
and K. Mukuno. 1997. Representation of the visual field in the striate cortex: comparison of MR findings
with visual field deficits in organic mercury poisoning (Minamata Disease). AJNR Am J. Neuroradiol.
18:1127-1130.
Krabbenhoft, D.P., and C.L. Babiarz. 1992. The role of groundwater transport in aquatic mercury
cycling. Water Resour. Res. 28:3119-3128. (as cited in ATSDR, 1999).
R-12
Methylmercury Water Quality Criterion 1/3/01
-------
Krabbenhoft, D.P., J.G. Wiener, W.G. Brumbaugh, M.L. Olson, J.F. DeWild, and T.J. Sabin. 1999. A
national pilot study of mercury contamination of aquatic ecosystems along multiple gradients. U. S.
Geological Survey Toxic Substances Hydrology Program: Proceedings of the Technical Meeting.
Charleston, South Carolina, March 8-12, 1999. Volume 2 of 3: contamination of hydrologic systems and
related ecosystems, water-resources investigation report 99-4018B. Http://toxics.usgs.gov./pubs/wri99-
4018/Volume2/sectionB/2301_&abbenhoft/index.html
Kudo, A., H. Nagase, and Y. Ose. 1982. Proportion of methylmercury to the total mercury in river waters
of Canada and Japan. Water Res. 16:1011-1015.
Kuhnert, P.M., B.R. Kuhnert, and P. Erhard. 1981. Comparison of mercury levels in maternal blood, fetal
cord blood, and placental tissues. Am. J. Obstet. Gynecol. 139(2):209-213.
Kuntz, W.D., R.M. Pitkin, A. Bostrom, and M.S. Hughes. 1982. Maternal and cord blood background
mercury levels: a longitudinal surveillance. Am. J. Obstet. Gynecol. 143:440-443.
Lange, T.R., H.E. Royals, and L.L. Connor. 1993. Influence of water chemistry on mercury concentration
in largemouth bass from Florida lakes. Trans. Am. Fish. Soc. 122:74-84.
Lange, T.R., H.E. Royals, and L.L. Connor 1994. Mercury accumulation in largemouth bass
(Micropterus salmoides) in a Florida lake. Arch. Environ. Contain. Toxicol. 27:466-471.
Lanting, C.I., et al. 1998. Determinants of polychorinate diphenyl levels in plasma from 42-month-old
children. Arch. Environ. Contam. Toxicol. 35:135-139.
Lauwerys, R., J.P. Buchet, H. Roels, and G. Hubermont. 1978. Placental transfer of lead, mercury,
cadmium, and carbon monoxide in women. I. Comparison of the frequency distributions of the biological
indices in maternal and umbilical cord blood. Environ. Res. 15(2):278-289.
Lebel, J., D. Mergler, F. Branches, M. Lucotte, M. Amorim, F. Larribe, and J. Dolbec. 1998. Neurotoxic
effects of low-level methylmercury contamination in the Amazonian Basin. Environ. Res. 79(l):20-32.
Lebel, J., D. Mergler, M. Mucotte, et al. 1996. Evidence of early nervous system dysfunction in
Amazonian populations exposed to low-levels of methylmercury. NeuroToxicology 17:157-168.
Lee, J.H., and D.H. Han. 1995. Maternal and fetal toxicity of methylmercuric chloride administered to
pregnant Fischer 344 rats. J. Toxicol. Environ. Health 45(4):415-425.
Lee, Y. H. and H. Hultberg. 1990. Methylmercury in some Swedish surface waters. Environ. Sci.
Technol. 9:833-841.
Lee, Y. and A. Iverfeldt. 1991. Measurement of methylmercury and mercury in run-off, lake and rain
waters. Water Air Soil Pollut. 56:309-321.
Leisenring, W., and L. Ryan. 1992. Statistical properties of the NOAEL. Regul. Toxicol. Pharmacol.
15(2 Pt. 1):161-171.
Letz, R. 1990. The neurobehavioral evaluation system (NES): An international effort. In: Advances In
Neurobehavioral Toxicology: Applications in Environmental and Occupational Health. B.L. Johnson,
W.K. Anger, A. Durao, and C. Xintaras, Eds. Chelsea: Lewis Publishers, pp. 189-202.
Methylmercury Water Quality Criterion 1/3/01
R-13
-------
Leyshon, K., and A. J. Morgan. 1991. An integrated study of the morphological and gross-elemental
consequences of methyl mercury intoxication in rats, with particular attention on the cerebellum.
Scanning Microsc. 5:895-904.
Lind, B., L. Friberg, and M. Nylander. 1988. Preliminary studies on methylmercury biotransformation
and clearance in the brain of primates: n. Demethylation of mercury in brain. J. Trace Elem. Exp. Med
1:49-56.
Lindqvist, 0.1991. Mercury in the Swedish environment. Recent research on causes, consequences and
corrective measures. Water Air Soil Pollut. 55:1-261.
Lindqvist, O., and H. Rodhe. 1985. Atmospheric mercury: a review. Tellus 376:136-159.
Liu, K. Z., Q. G. Wu, and H. I. Liu. 1990. Application of a Nafion-Schiff-base modified electrode in
anodic-stripping voltammetry for the determination of trace amounts of mercury. Analyst 115(6):835-
837.
Lok, E. 1983. The effect of weaning on blood, hair, fecal and urinary mercury after chronic ingestion of
methylmercuric chloride by infant monkeys. Toxicol. Lett. 15:147-152.
Lores, E.M., J. Macauley, L.R. Goodman, R.G. Smith, and D.M. Wells. 1998. Factors affecting
bioavailability of methylmercury in Florida Bay. Soc. Environ. Toxicol. Chem. 19th Annual Meeting.
Charlotte, NC. Abstr. No. 468. p. 101.
Lowe, T.P., T.W. May, W.G. Brumbaugh, and D.A. Kane. 1985. National contaminant biomonitoring
program: concentrations of seven elements in fresh-water fish, 1978-1981. Arch. Environ. Contam.
Toxicol. 14:363-388.
Luecke, R.H., W.D. Wosilait, B.A. Pearce, J.F. Young. 1994. A physiologically based pharmacokinetic
computer model for human pregnancy. Teratology 49:90-103.
Luecke, R.H., W.D. Wosilait, B.A. Pearce, and J.F. Young. 1997. A computer model and program for
xenobiotic disposition during pregnancy. Comp. Meth. Prog. Biomed. 53:201-224.
Lutz, R.J., R.L. Dedrick, H.B. Matthews, T.E. Eling, and M.W. Anderson. 1977. A preliminary
pharmacokinetic model for several chlorinated biphenyls in the rat. Drug Metab. Dispos. 5:386-396 (as
cited in Farris et al., 1993).
Lynn, R., and R.G. Wilson. 1990. Reaction times, movement times and intelligence among Irish nine
year olds. Irish J. Psychol. 11:329-341.
Lynn, R., J.W.C. Chan, and H.J. Eysenck. 1991. Reaction times and intelligence in Chinese and British
children. Percept. Motor Skills 72:443-452.
MacDonald, J.S., and R.D. Harbison. 1977. Methyl mercury-induced encephalopathy in mice. Toxicol.
Appl. Pharmacol. 39:195-205.
Madson, M., and R. Thompson. 1998. Determination of methylmercury in food commodities by gas-
liquid chromatography with atomic emission detection. J. AOAC Intl. 81(4):808-816.
R-14
Methylmercury Water Quality Criterion 1/3/01
-------
Magos, L. 1987. The absorption, distribution, and excretion of methyl mercury. In: Eccles, C.U., and Z.
Annau, eds. The Toxicity of Methyl Mercury. Baltimore, MD: The Johns Hopkins University Press (as
cited in Gray, 1995).
Magos, L., A.W. Brown, S. Sparrow, et al. 1985. The comparative toxicology of ethyl and
methylmercury. Arch. Toxicol. 57:260-267.
Magos, L., and A. A. Cernik. 1969. A rapid method for estimating mercury in undigested biological
samples. Br. J. rnd. Med. 26(2):144-149.
Magos, L., and T.W. Clarkson. 1972. Atomic absorption and determination of total, inorganic, and
organic mercury in blood. J. AOAC 55(5):966-971.
Mahaffey, K.R. 1998. Methylmercury exposure and neurotoxicity. JAMA 280:737-738.
Mailhes, J.B. 1983. Methyl mercury effects on Syrian hamster metaphase H oocyte chromosomes.
Environ. Mutagen. 5:679-686.
Malm, O., W.C. Pfeiffer, C.M.M. Souza, and R. Reuther. 1990. Mercury pollution due to gold-mining in
the Madera River Basin, Brazil. Ambio 19:1 Irl5.
Marsh, D., T. Clarkson, C. Cox, G. Myers, L. Amin-Zaki, and S. Al-Tikriti. 1987. Fetal methylmercury
poisoning. Relationship between concentration in single strands of maternal hair and child effects. Arch.
Neurol. 44:1017-1022.
Marsh, D., T. Clarkson, G. Myers, P. Davidson, C. Cox, E. Cernichiari, M. Tanner, W. Lednar, C.
Shamlaye, O. Choisy, C. Horareau, and M. Berlin. 1995a. The Seychelles study of fetal methylmercury
exposure and child development: introduction. NeuroToxicology 16:583-596.
Marsh, D.O, M.D. Turner, J.C. Smith, P. Allen, and N. Richdale. 1995. Fetal methylmercury study in a
Peruvian fish-eating population. Neurotoxicology 16(4):717-726.
Marsh, D., G. Myers, T. Clarkson, L. Amin-Zaki, S. Al-Tikriti, and M. Majeff. 1980. Fetal
methylmercury poisoning: Clinical and lexicological data on 29 cases. Ann. Neurol. 7:348-353.
Marsh, D., G. Myers, T. Clarkson, et al. 1981. Dose-response relationship for human fetal exposure to
methylmercury. Clin. Toxicol. 18:1311-1318.
Mason, R.P., and K.A. Sullivan. 1997. Mercury in Lake Michigan. Environ. Sci. Technol. 31:942-947.
Mason, R. P., and K. A. Sullivan. 1998. Mercury and methylmercury transport through an urban water
shed. Water Res. 32:321-330.
Mason, R.P., W.F. Fitzgerald, and F.M.M. Morel. 1994. The biogeochemical cycling of elemental
mercury: anthropogenic influences. Geochim. Cosmochim. Acta. 58(15):3191-3198.
Matthews, G., and L. Dorn. 1989. IQ and choice reaction time: an information processing analysis.
Intelligence 13:229-317.
Methylmercury Water Quality Criterion 1/3/01
R-15
-------
McKeown-Eyssen, G., and J. Ruedy. 1983a. Prevalence of neurologic abnormality in Cree Indians
exposed to methylmercury in Northern Quebec. Clin. Invest Med. 6:161-169.
McKeown-Eyssen, G., and J. Ruedy. 1983b. Methyl mercury exposure in northern Quebec. I. Neurologic
findings in adults. Am. J. Epidemiol. 118:461-469.
McKeown-Eyssen, G., J. Ruedy, and A. Neims. 1983c. Methyl mercury exposure in northern Quebec. H
Neurologic findings in children. Am. J. Epidemiol. 118:470-479.
MDEQ (Michigan Department of Environmental Quality). 1996. Michigan default metals translators.
Staff Report, June 1996. MI/DEQ/SWQ-95/085. Ann Arbor, MI.
Mergler, D., and J. Dolbec. 1998. Neurotoxic effects of low-level methylmercury contamination in the
Amazonian Basin. Environ. Res. 79(1): 20-32.
Miettinen, J.K., T. Rahola, T. Hattula, K. Rissanen, andM. Tillander. 1971. Elimination of 203Hg-
methylmercury in man. Ann. Clin. Res. 3:116-122.
Miles, C.J., and L.E. Fink. 1998. Monitoring and mass budget for mercury in the Everglades nutrient
removal project. Arch. Environ. Contam. Toxicol. 35:549-557.
Miller, C. T., Z. Zawidska, E. Nagy, et al. 1979. Indicators of genetic toxicity in leukocytes and
granulocytic precursors after chronic methyl mercury ingestion by cats. Bull. Environ. Contam. Toxicol.
21:296-303.
Mitchell, J.W., T.E.U. Kjellstrom, and L. Reeves. 1982. Mercury in takeaway fish in New Zealand. N. Z.
Med. J. 95(702): 112-114.
Mitsumori, K., K. Maita, and Y. Shirasu. 1984. Chronic toxicity of methyl mercury chloride in rats:
Pathological study. Jpn. J. Vet. Sci. 46(4):549-557.
Mitsumori, K., K. Takahashi, O. Matano, S. Goto, and Y. Shirasu. 1983. Chronic toxicity of methyl
mercury chloride in rats: Clinical study and chemical analysis. Jpn. J. Vet. Sci. 45(6):747-757.
Mitsumori, K., M. Hirano, H. Ueda, K. Maita, and Y. Shirasu. 1990. Chronic toxicity and carcinogenicity
of methylmercury chloride in B6C3F1 mice. Fundam. Appl. Toxicol. 14:179-190.
Mohamed, M., T. Burbacher, and N. Mottet. 1987. Effects of methylmercury on testicular functions in
Macaca fascicularis monkeys. Pharmacol. Toxicol. 60(l):29-36.
Monson, B.A., and P.L. Brezonik. 1998. Seasonal patterns of mercury species in water and plankton from
softwater lakes in Northeastern Minnesota. Biogeochemistry 40:147-162.
Morgan, J.N., M.R. Berry, Jr., and R.L. Graves. 1994. Effects of Native American cooking practices on
total mercury concentrations in walleye. Presented at ISEE/ISEA Joint Conference, September 18-21,
1994.
Morimoto, K., S. Jijima, and A. Koizumi. 1982. Selenite prevents the induction of sister-chromatid
exchanges by methyl mercury and mercuric chloride in human whole-blood cultures. Mutat. Res.
102:183-192.
R-16
Methylmercury Water Quality Criterion 1/3/01
-------
Moszczynski, P., J. Lisiewicz, R. Bartus, et al. 1990. The serum immunoglobulins in workers after
prolonged occupational exposure to the mercury vapors. Rev. Roum. Med. Intern. 28(1):25-30.
Mottet, N.K., R.L Body, V. Wilkens, and T.M. Burbacher. 1987. Biologic variables in the hair uptake of
methylmercury from blood in the Macaque monkey. Environ. Res. 42:509-523.
Munro, I., E. Nera, S. Charbonneau, B. Junkins, and Z. Zawidzka. 1980. Chronic toxicity of
methylmercury in the rat. J. Environ. Pathol. Toxicol. 3:437-447.
Murata, K., P. Weihe, S. Araki, E. Budtz-Jorgensen, and P. Grandjean. 1999a. Evoked potentials in
Faroese children prenatally exposed to methylmercury. Neurotoxicol. Teratol. 21:471-472.
Murata, K., P. Weihe, A. Renzoni, F. Debes, R. Vasconcelos, F. Zino, S. Araki, P. Jorgensen, R. White,
and P. Grandjean. 1999b. Delayed evoked potentials in children exposed to methylmercury from seafood.
Neurotoxicol. Teratol. 21:343-348.
Myers, G., D. Marsh, P. Davidson, C. Cox, C. Shamlaye, M. Tanner, A. Choi, E. Cernichiari, O. Choisy,
and T. Clarkson. 1995a. Main neurodevelopmental study of Seychellois children following in utero
exposure to methylmercury from a maternal fish diet: outcome at six months. Neurotoxicology 16:653-
664.
Myers, G.J., D.O. Marsh, C. Cox, P.W. Davidson, C.F. Shamlaye, M.A. Tanner, A. Choi, E.
Chernichiari, O. Choisy, and T.W. Clarkson. 1995b. A pilot neurodevelopmental study of Seychellois
children following in utero exposure to methylmercury from a maternal fish diet. Neurotoxicology
16:629-638.
Myers, G.J., P.W. Davidson, C. Cox, C.F. Shamlaye, M.A. Tanner, O. Choisy, J. Sloane-Reeves, D.O.
Marsh, E. Cernichiari, A. Choi, M. Berlin, and T.W. Clarkson. 1995c. Neurodevelopmental outcomes of
Seychellois children sixty-six months after in utero exposure to methylmercury from a maternal fish diet:
pilot study. Neurotoxicology 16:639-652.
Myers, G.J., P.W. Davidson, and C.F. Shamlaye. 1998. A review of methylmercury and child
development. Neurotoxicology 19(2):313-328.
Myers, G.J., P.W. Davidson, D. Palumbo, C. Shamlaye, C. Cox, E. Chernichiari, and T.W. Clarkson.
2000. Secondary analysis from the Seychelles child development study: the child behavior checklist.
Environ. Res. Section A 84: 12-19.
Nakai, S., and I. Machida. 1973. Genetic effect of organic mercury on yeast. Mutat. Res. 21:348.
Nakamura, I., K. Hosokawa, H. Tamra, et al. 1977. Reduced mercury excretion with feces in germfree
mice after oral administration of methyl mercury chloride. Bull. Environ. Contam. Toxicol. 17:528-533.
Nakazawa, N., F. Makino, and S. Okada. 1975. Acute effects of mercuric compounds on cultured
mammalian cells. Biochem. Pharmacol. 24:489-493.
NAS (National Academy of Sciences). 1991. Methyl mercury: FDA risk assessment and current
regulations. In: Seafood Safety. Committee on Evaluation of the Safety of Fishery Products, National
Academy Press, Washington, DC. p. 196-221.
Methylmercury Water Quality Criterion 1/3/01
R-17
-------
NCHS (National Center for Health Statistics). 1995. [Section 4.2.3, body weight]
Newland, C.M., and KB. Rasmussen. 2000. Aging unmasks adverse effects of gestational exposure to
methylmercury in rats. Neurotoxicol. Teratol. 22: in press.
NIEHS (National Institute of Environmental Health Sciences). 1999. Scientific issues relevant to
assessment of health effects from exposure to methylmercury. Workshop organized by Committee on
Environmental and Natural Resources (CENR) Office of Science and Technology Policy (OSTP), The
White House, November 18-20, 1998, Raleigh, NC.
Nielsen, J.B., and O. Andersen. 1992. Transplacental passage and fetal deposition of mercury after low-
level exposure to methylmercury-effect of seleno-L-methionine. J. Trace Elem. Electrolyt. Health Dis. 6:
227-232.
NIOSH (National Institute for Occupational Safety and Health). 1977. A recommended standard for
occupational exposure to inorganic mercury.
Nishima, T., S. Dceda, T. Tada, H. Yagyu, and I. Mizoguchi. 1977. Mercury content levels in mother and
newborn and their interrelation. Ann. Rep. Tokyo Metro Res. Lab. PH 28:215-220.
NJDEPE (New Jersey Department of Environmental Protection and Energy). 1993. Final report on
municipal solid waste incineration. Volume It Environmental and health issues.
NMFS (National Marine Fisheries.Service). 1995. The current publicly available National Marine
Fisheries Service database was supplied to U.S. EPA via fax from Malcolm Meaburn (Charleston
Laboratory/Southeast Fisheries Science Center/National Marine Fisheries Service/National Oceanic and
Atmospheric Administration/U.S. Dept. Of Commerce) to Kathryn Mahaffey (Environmental Criteria
and Assessment Office-Cincinnati,OH/Office of Health and Environmental Assessment/Office of
Research and Development/ U.S. Environmental Protection Agency). February 23, 1995. [Note: cited as
NMFS, 1978 in text of MSRC].
Nordberg, G.F., and P. Strangert. 1976. Estimations of a dose-response curve for long-term exposure to
methylmercuric compounds in human being taking into account availability of critical organ
concentration and biological half-time: a preliminary communication. In: Effects and dose-response
relationships of toxic metals. Nordberg, G.F., ed. Amsterdam: Elsevier, pp. 273-282.
Nordenhall, K., L. Dock, and M. Vahter. 1988. Cross-fostering study of methyl mercury retention,
demethylation and excretion in the neonatal hamster. Pharmacol. Toxicol. 81:132-136.
Norseth T., and T.W. Clarkson. 1970. Studies on the biotransformation of 203Hg-labeled methyl mercury
chloride in rats. Arch. Environ. Health 21:717-727 (as cited in Gray, 1995).
Norseth T., and T.W. Clarkson. 1971. Intestinal transport of 203Hg-labeled methyl mercury chloride.
Arch. Environ. Health 22:568-577 (as cited in Gray, 1995).
Northeast States and Eastern Canadian Provinces. 1998. Mercury study: a framework for action. Boston.
NRC (National Research Council). 2000. Toxicological effects of methylmercury. Committee on the
Toxicological Effects of Methylmercury, Board on Environmental Studies and Toxicology, Commission
on Life Sciences, National Research Council. Washington, DC: National Academy Press.
R-18
Methylmercury Water Quality Criterion 1/3/01
-------
Nriagu, J.O. 1979. The biogeochemistry of mercury in the environment. Elsevier/North Holland. New
York: Biomedical Press.
O'Conner, T.P., and B. Beliaeff. 1995. Recent trends in coastal environmental quality: results from the
Mussel Watch Project. 1986 to 1993. U.S. Department of Commerce, National Oceanic and Atmospheric
Administration, National Ocean Service, Office of Ocean Resources Conservation and Assessment,
Silver Spring, MD.
Ohi, G., M. Fukuda, H. Seto, et al. 1976. Effect of methyl mercury on humoral immune responses in mice
under conditions simulated to practical situations. Bull. Environ. Contain. Toxicol. 15:175-180.
Olson, M.L., and J.F. DeWild. 1999. Low-level collection techniques and species-specific analytical
methods for mercury in water, sediment, and biota. In: U.S. Geological Survey Toxic Substances
Hydrology program - Proceedings of the Technical meeting, Charleston, SC, March 8-12, 1999. Volume
2. Morgenwalp, D.W., and T.H. Buxton, eds. Contamination of Hydrologic Systems and Related
Ecosystems: U.S. Geological Survey Water Resources Investigations Report 99-4018B.
Ong, C.N., S.E. Chia, S.C. Foo, H.Y. Ong, M. Tsakok, and P. Liouw. 1993. Concentrations of heavy
metals in maternal and umbilical cord blood. Biometals 6:61-66.
OSHA (Occupational Safety and Health Administration). 1975. Mercury. Job Health Hazards Series.
OSHA report 2234.
Ostlund, K. 1969. Studies on the metabolism of methylmercury in mice. Acta Pharmacol. Toxicol.
27(Suppl.l): 1-132.
Palumbo, D.R., C. Cox, P.W. Davidson, G.J. Myers, A. Choi, C. Shamlaye, J. Sloane-Reeves, E.
Chernichiari, and T.W. Clarkson. 2000. Association between prenatal exposure to methylmercury and
cognitive functioning in Seychellois children : a reanalysis of the McCarthy Scales of Children's Ability
from the main cohort study. Environ. Res. Section A 84:81-88.
Pankow, J.F., and S.W. McKenzie. 1991. Parameterizing the equilibrium distribution of chemicals
between the dissolved, solid particulate matter, and colloidal matter compartments in aqueous systems.
Environ. Sci. Technol. 25:2046-2053.
Parks, J.W., A. Lutz, and J.A. Sutton. 1989. Water column methylmercury in the Wabigoon/English
River-Lake system: factors controlling concentrations, speciation, and net production. Can. J. Fish.
Aquat. Sci. 46:2184-2202.
Patandin, S., C.I. Lanting, P.O. Mulder, E.R. Boersma, PJ. Sauer, and N. Weisglas-Kuperus. 1999a.
Effects of environmental exposure to polychlorinated biphenyls and dioxins on cognitive abilities in
Dutch children at 42 months of age. J. Pediatr. 134:33-41.
Patandin, S., J. Veenstra, P.G.H. Mulder, A. Sewnaik, P.J.J. Sauer, and N. Weisglas-Kuperus. 1999b.
Attention and activity in 42-month-old Dutch children with environmental exposure to polychlorinated
biphenyls and dioxins. In: S. Patandin, ed. Effects of Environmental Exposure to Polychlorinated
Biphenyls and Dioxins on Growth and Development in Young Children. Ph.D. thesis, Erasmus
University, Amsterdam, pp. 124-142.
Methylmercury Water Quality Criterion 1/3/01
R-19
-------
Pedersen, G.A., G.K. Mortensen, and E.H. Larsen. 1994. Beverages as a source of toxic trace element
intake. Food Addit. Contam. 11:351-363.
Pfeiffer, W.C., L.D. Lacerda, O. Malm, C.M.M. Souza, E.J. Silviera, and W.R. Bastos. 1989. Mercury
concentration in inland waters of gold-mining areas in Rondonia, Brazil. Sci. Tot. Environ. 87:233-240.
Phelps, R., T. Clarkson, T. Kershaw, et al. 1980. Interrelationships of blood and hair mercury
concentrations in a North American population exposed to methylmercury. Arch. Environ. Health
35:161-168.
Pirkle, J.L., J. Schwartz, J.R. Landis, and W.R. Harlan. 1985. The relationship between blood lead levels
and blood pressure and its cardiovascular risk implications. Am. J. Epidemiol. 121:246-258.
Pitkin, R.M., J.A. Bahns, L.J. Filer, Jr., and W.A. Reynolds. 1976. Mercury in human maternal and cord
blood, placenta, and milk. Proc. Soc. Exp. Biol. Med. 151(3):565-567.
Porcella, D. B. 1994. Mercury in the environment, in Mercury Pollution: Integration and Synthesis, CJ.
Watras and J.W. Huckabee, Eds. New York: Lewis Publishers. 727 pp.
Porcella, D.B., CJ. Watras, and N.S. Bloom. 1991. Mercury species in lake water. In: Verry, S., and SJ.
Vermette. The deposition and fate of trace metals in our environment. Gen. Tech. Rep. NC-150. St. Paul,
MN: U.S. Dept. Agric., Forest Service, North Central Forest Exp. Station, pp. 127-138.
Post, E.M., M.G. Yang, J.A. King, et al. 1973. Behavioral changes of young rats force-fed methyl
mercury chloride (37480). Proc. Soc. Exp. Biol. Med. 143:1113-1116.
Prager, J.C. 1997. Environmental contaminant reference databook, vol. EL New York: Van Nostrand
Reinhold, Inc.
Queiroz, M. L., and D. C. Dantas. 1997. B lymphocytes in mercury-exposed workers. Pharmacol.
Toxicol. 81(3):130-133.
Rada, R., J. Wiener, M. Winfrey, and D. Powell. 1989. Recent increases in atmospheric deposition of
mercury to north-central Wisconsin lakes inferred from sediment analysis. Arch. Environ. Contam.
Toxicol. 18:175-181.
Ramel, C. 1972. Genetic effects. In: Mercury in the environment. Friberg, L., and J. Vostal, eds.
Cleveland: CRC Press, pp. 169-181.
Ramirez, G.B., M.C.V. Cruz, O. Pagulayan, E. Ostrea, and C. Dalisay. 2000. The Tagum Study: I.
Analysis and clinical correlates of mercury in maternal and cord blood, breast milk, meconium, and
infants' hair. Pediatrics 106:774-781.
Ramussen, E. B., and M. C. Newland. 1999. Acquisition of a Multiple DRH Extinction Schedule of
Reinforcement in Rats Exposed during Development to Methylmercury. No. 697. p. 149. SOT 1999
Annual Meeting.
Rask, M., and M. Verta. 1995. Concentrations and amounts of methylmercury in water and fish in the
limed and acid basins of a small lake. Water Air Soil Pollut. 80:577-580.
R-20
Methylmercury Water Quality Criterion 1/3/01
-------
Retzlaff, J.A., W.N. Tauxe, J.M. Khely, and C.F. Strobel. 1969. Erythrocyte volume, plasma volume, and
lean body mass in adult men and women. Blood 33:649-^664.
Revis, N.W., T.R. Osborne, G. Holdsworth, and C. Hadden. 1990. Mercury in soil: a method for
assessing acceptable limits. Arch. Environ. Contam. Toxicol. 19:221-226.
Rice, D.C. 1996. Sensory and cognitive effects of developmental methylmercury exposure in monkeys,
and a comparison to effects in rodents. Neurotoxicology 17:139-154.
Rice, D.C. 1998. Age-related increase in auditory impairment in monkeys exposed in utero plus
postnatally to methylmercury. Toxicol. Sci. 44(2): 191-196.
Rice, D.C. 1989a. Delayed neurotoxicity in monkeys exposed developmentally to methylmercury.
Neurotoxicology 10:645-650.
Rice, D.C. 1989b. Brain and tissue levels of mercury after chronic methyl mercury exposure in the
monkey. J. Toxicol. Environ. Health 27:189-198.
Rice, D.C. 1989c. Blood mercury concentrations following methyl mercury exposure in adult and infant
monkeys. Environ. Res. 49:115-126.
Rice, D.C., and S.G. Gilbert. 1982. Early chronic low-level methylmercury poisoning in monkeys impairs
spatial vision. Science 216:759-761.
Rice, D.C., and S.G. Gilbert. 1990. Effects of developmental exposure to methylmercury on spatial and
temporal visual function in monkeys. Toxicol. App'l. Pharmacol. 102:151-163.
Rice, D.C., and Gilbert, S.G. 1992. Exposure to methylmercury from birth to adulthood impairs high-
frequency hearing in monkeys. Toxicol. Appl. Pharmacol. 102:151-163.
Rice, D.C., and S.G. Gilbert. 1995. Effects of developmental methylmercury exposure or lifetime lead
exposure on vibration sensitivity function in monkeys. Toxicol. Appl. Pharmacol. 134(1): 161-169.
Rice, D.C., D. Krewski, B.T. Collins, and R.F. Willes. 1989. Pharmacokinetics of methylmercury in the
blood of monkeys (Macaco, fascicularis). Fundam. Appl. Toxicol. 12:23-33.
Rowland, I., M. Davies, and J. Evans. 1980. Tissue content of mercury in rats given methyl mercury
chloride orally: influence of intestinal flora. Arch. Environ. Health. 35:155-160.
Rowland, I, M. Davies, and P. Grasso. 1977. Biosynthesis of methylmercury compounds by the
intestinal flora of the rat. Arch. Environ. Health. 32(l):24-28.
Rustam, H., and T. Hamdi. 1974. Methyl mercury poisoning in Iraq: a neurological study. Brain 97-499-
510.
Salonen, J.T., K. Seppanen, K. Nyyssonen, H. Korpela, J. Kauhanen, M. Kantola, J. Tuomilehto, H.
Esterbauer, F. Tatzber, and R. Salonen. 1995. Intake of mercury from fish, lipid peroxidation, and the
risk of myocardial infarction and coronary, cardiovascular, and any death in Eastern Finnish men
Circulation 91(3):645-655.
Methylmercury Water Quality Criterion 1/3/01
R-21
-------
Sato, T., and F. JJcuta. 1975. Long-term studies on the neurotoxicity of small amounts of methyl mercury
in monkeys (first report). In: Tsubaki, T., ed. Studies on the health effects of alkylmercury in Japan.
Japan: Environmental Agency, pp. 63-70.
Schwartz, J.G., T.E. Snider, and M.M. Montiel. 1992. Toxicity of a family from vacuumed mercury. Am.
J. Emerg. Med. 10(3):258-261.
Seppanen, K., R. Laatikainen, J.T. Salonen, M. Kantola, S. Lotjonen, M. Hani, L. Nurminen, J.
Kaikkonen, and K. Nyyssonen. 1998. Mercury-binding capacity of organic and inorganic selenium in rat
blood and liver. Biol. Trace Element Res. 65:197-210.
Shacklette, H.T., and J.G. Boerngen. 1984. Element concentrations in soils and other surficial materials
of the conterminous United States. U.S. Geological Survey Paper 1270. Washington, DC: United States
Government Printing Office.
Shamlaye, C., D. Marsh, G. Myers, et al. 1995. The Seychelles child development study on
neurodevelopmental outcomes in children following in utero exposure to methylmercury from a maternal
fish diet: Background and demographics. NeuroToxicology 16:597-612.
Sherlock, J., J. Hislop, D. Newton, G. Topping, and K. Whittle. 1984. Elevation of mercury in human
blood from controlled chronic ingestion of methylmercury in fish. Hum. Toxicol. 3:117-131.
Sherlock, J.C., andMJ. Quinn. 1988. Underestimation of dose-response relationship with particular
reference to the relationship between the dietary intake of mercury and its concentration in blood. Hum.
Toxicol. 7(2):129-132.
Sherlock, J.C., D.G. Lindsay, J. Hislop, W.H. Evans, and T.R. Collier. 1982. Duplication diet study on
mercury intake by fish consumers in the United Kingdom. Arch. Environ. Health 37(5):271-278.
Shigehisa, T., and R. Lynn. 1991. Reaction times and intelligence in Japanese children. Int. J. Psychol.
26:195-202.
Sikorski, R., T. Paszkowski, P. Slawinski, J. Szkoda, J. Zmudzki, and S. Skawinski. 1989. The
intrapartum content of toxic metals in maternal blood and umbilical cord blood. Ginekol. Pol.
60(3): 151-155.
Simonin, H.A., and M.W. Meyer. 1998. Mercury and other air toxics in the Adirondack region of New •
York. Environ. Sci. Policy 1:199-209.
Skerfving, S. 1974. Methyl mercury exposure, mercury levels in blood and hair, and health status in
Swedes consuming contaminated fish. Toxicology. 2:3-23.
Skerfving, S. 1988. Mercury in women exposed to methylmercury through fish consumption, and in their
newborn babies and breast milk. Bull. Environ. Contam. Toxicol. 41(4):475-482.
Skerfving, S., K. Hansson, and J. Lindsten. 1970. Chromosome breakage in humans exposed to methyl
mercury through fish consumption. Arch. Environ. Health 21(2): 133-139.
Skog, E., and J. Wahlberg. 1964. A comparative investigation of the percutaneous absorption of metal
compounds in the guinea pig by means of the radioactive isotopes: Cr, Co, Zn, Ag, Cd, Hg.
R-22
Methylmercury Water Quality Criterion 1/3/01
-------
J. Invest Dermatol. 43:187-192.
Smith, J.C., P.V. Allen, M.D. Turner, B. Most, H.L. Fisher, and L.L. Hall. 1994. The kinetics of
intravenously administered methylmercury in man. Toxicol. Appl. Pharmacol. 128:251-256.
Soong, Y-K, R. Tseng, C. Liu, and P-W Lin. 1991. Lead, cadmium, arsenic and mercury levels in
maternal and fetal cord blood. J. Formosan Med. Assoc. 90:59-65.
Sorensen, J., G. Glass, K. Schmidt, J. Huber, and G. Rapp. 1990. Airborne mercury deposition and
Watershed characteristics in relation to mercury concentrations in water, sediments, plankton and
Fish of eighty northern Minnesota lakes. Environ. Sci. Technol. 24:1716-1727.
Sorensen, N., K. Murata, E. Budtz-Jorgensen, P. Weihe, and P. Grandjean. 1999. Prenatal methylmercury
exposure as a cardiovascular risk factor at seven years of age. Epidemiology 10:370-375.
Soria, M.L., P. Sanz, D. Martinez, M, Lopez-Artiguez, R. Garrido, A. Grilo, and M. Repetto. 1992. Total
mercury and methylmercury in hair, maternal and umbilical blood, and placenta from women in the
Seville area. Bull. Environ. Contam. Toxicol. 48:494-501.
Spyker, J.M. 1975. Assessing the impact of low level chemicals on development: behavioral and latent
effects. Fed. Proc. 34(9): 1835-1844.
Stern, A.M. 1993. Re-evaluation of the reference dose for methyl mercury and assessment of current
exposure levels. Risk Anal. 13:355-364.
Stern, A.H. 1997. Estimation of the interindividual variability in the one-compartment pharmacokinetic
model for methylmercury: implications for the derivation of a reference dose. Regul. Toxicol. Pharmacol.
25:277-288. (As cited in Clewell et al., 1999).
Stern, A.H., L.R. Korn, and B.F. Ruppel. 1996. Estimation offish consumption and methylmercury
intake in the New Jersey population. J. Euro. Anal. Env. Epi. 6:503-525.
Steurwald, U., P. Weibe, P. Jorgensen, K. Bjerve, J. Brock, B. Heinzow, E. Budtz-Jorgensen, and P.
Grandjean. 2000. Maternal seafood diet, methylmercury exposure, and neonatal neurologic function. J.
Pediatr. 136(5):599-605.
Suchanek, T. H., P. J. Richerson, L. A. Woodward, D. G. Slotton, L. J. Holts, and C. E. E. Woodmansee.
1993. A survey and evaluation of mercury. In: Sediment, water, plankton, periphyton, benthic
invertebrates and fishes within the aquatic ecosystem of Clear Lake, California. Preliminary lake study
report prepared for the U.S. Environmental Protection Agency, Region 9, Superfund Program.
Suda, I., and K. Hirayama. 1992. Degradation of methyl- and ethylmercury into inorganic mercury by
hydroxyl radical produced from rat liver microsomes. Arch. Toxicol. 66(6):398-402.
Suda, L, and H. Takahashi. 1986. Enhanced and inhibited biotransformation of methyl mercury in the rat
spleen. Toxicol. Appl. Pharmacol. 82:45-52.
Sundberg, J., and A. Oskarsson. 1992. Placental and lactational transfer of mercury from rats exposed to
methyl mercury in their diet: Speciation of mercury in the offspring. J. Trace Elem. Exp. Med. 5 (1):47-
56.
Methylmercury Water Quality Criterion 1/3/01
R-23
-------
Sung, W. 1995. Some observations on surface partitioning of Cd, Cu, and Zn in estuaries. Environ. Sci.
Technol. 29:1303-1312.
Suter, K.E. 1975. Studies on the dominant lethal and fertility effects of the heavy metal compounds
methyl mercuric hydroxide, mercuric chloride, and cadmium chloride in male and female mice. Mutat.
Res. 30:365-374.
Suzuki, T. 1988. Hair and nails: advantages and pitfalls when used in biological monitoring. In:
Biological monitoring of toxic metals. Clarkson, T.W., L. Friberg, G.F. Nordberg, and P.R. Sager, eds.
New York: Plenum, pp. 623-640.
Suzuki, T., J. Yonemoto, H. Satoh, A. Naganuma, N. Imura, and T. Kigama. Normal organic and
inorganic mercury levels in the human feto-placental system. J. Appl. Toxicol. 4(5):249-252.
Swartout, J. 2000. Personal communication, June 9, 2000, U.S. Environmental Protection Agency.
Swartout, J., and G. Rice. 2000. Uncertainty analysis of the estimated ingestion rates used to derive the
methylmercury reference dose. Drug Clin. Toxicol. 23(1):293-306.
Swedish EPA. 1991. Mercury in the environment: problems and remedial measures in Sweden. ISBN 91-
620-1105-7.
Swensson, A., and U. Ulfvarson. 1968. Distribution and excretion of mercury compounds in rats over a
long period after a single injection. Acta Pharmacol. Toxicol. 26:273-283.
Szymczak, J., and H. Grajeta. 1992. Mercury concentrations in soil and plant material. Pol. J. Food Nutr.
Sci. l/42(2):31-39.
Tamashiro, H., M. Arakaki, H. Akagi, et al. 1986. Effects of ethanol on methyl mercury toxicity in rats.
J. Toxicol. Environ. Health. 18:595-605.
Tamashiro, H., M. Arakaki, M. Futatsuka, and E. S. Lee. 1986. Methylmercury exposure and mortality
in southern Japan: A close look at causes of death. J. Epidemiol. Commun. Health 40:181-185.
Tanaka, T., A. Naganuma, K. Kobayashi, et al. 1991. An explanation for strain and sex differences in
renal uptake of methyl mercury in mice. Toxicology 69:317-329.
Tanaka, T., A. Naganuma, N. Miura, et al. 1992. Role of testosterone in gamma-glutamyl transpeptidase-
dependent renal methyl mercury uptake in mice. Toxicol. Appl. Pharmacol. 112:58-63.
Takeuchi, T., and K. Eto. 1975. Minamata disease. Chronic occurrence from pathological view points.
In: Studies on the Health Effects of Alkylmercury in Japan. Tokyo, Japan Environment Agency.
Temple, P.J., and S.N. Linzon. 1977. Contamination of vegetation, soil, snow and garden crops by
atmospheric deposition of mercury from a chlor-alkali plant. In: Hemphill, D.D., ed. Trace
substances in environmental health - XI. Columbia, MO: University of Missouri, pp. 389-398.
TexaSoft. 1999. WINKS 4.6: Windows Version of KWIKSTAT Statistical Data Analysis Program.
Cedar Hill, TX.
R-24
Methylmercury Water Quality Criterion 1/3/01
-------
Thomas, D., H. Fisher, L. Hall, et al. 1982. Effects of age and sex on retention of mercury by methyl
mercury-treated rats. Toxicol. Appl. Pharmacol. 62:445-454.
Thomas, D., H. Fisher, M. Sumler, et al. 1986. Sexual differences in the distribution and retention of
organic and inorganic mercury in methylmercury-treated rats. Environ. Res. 41:219-234.
Thomas, D.J., H.L. Fisher, M.R. Sumler, et al. 1988. Distribution and retention of organic and inorganic
mercury in methyl mercury-treated neonatal rats. Environ. Res. 47:59-71.
Thomas, D., H. Fisher, M.R. Sumler, et al. 1987. Sexual differences in the excretion of organic and
inorganic mercury by methyl mercury-treated rats. Environ. Res. 43:203-216.
Trillingsgaard, A., O.N. Hansen, and I. Beese. 1985. The Bender-Gestalt Test as a neurobehavioral
measure of preclinical visual-motor integration deficits in children with low-level lead exposure. In:
WHO Environmental Health, Document 3. Neurobehavioral methods in occupational and environmental
health, Second International Symposium, Copenhagen, Denmark, Aug. 6-9, 1985. Copenhagen,
Denmark: World Health Organization, pp. 189-193.
Truska, P., I. Rosival, G. Balazova, J. Hinst, A. Rippel, O. Palusova, and J. Grunt. 1989. Placental
concentrations of cadmium, lead, and mercury in mothers and their newboms. J. Hyg. Epidemiol.
Microbiol. Immunol. 33(2): 141-147.
Tsubaki, T.K. and K. Irukayama. 1977. Minamata disease: methyl mercury poisoning in Minamata and
Niigata, Japan. New York: Elsevier, pp. 143-253.
Tsuchiya, H., K. Mitani, K. Kodama, and T. Nakata. 1984. Placental transfer of heavy metals in normal
pregnant Japanese women. Arch. Environ. Health 39(1):11-17.
Turner, M., D. March, J. Smith, J. Inglis, et al. 1980. Methylmercury in populations eating large
quantities of marine fish. Arch. Environ. Health 35:367-378.
U.S. Environmental Protection Agency (U.S. EPA). 1995. Great Lakes Water Quality Initiative Technical
Support Document for the Procedure to Determine Bioaccumulation Factors. Office of Water.
Washington, DC. EPA/820/B-95/005.
U.S. EPA. 1996. The metals translator: Guidance for calculating a total recoverable permit limit from a
dissolved criteria. June 1996. Washington, DC.
U.S. EPA. 1980. Ambient water quality criteria document for mercury. Prepared by the Office of Health
and Environmental Assessment, Environmental Criteria and Assessment Office, Cincinnati, OH, for the
Office of Water Regulation and Standards, Washington, DC. EPA/440/5-80-058. NTIS PB 81-117699.
U.S. EPA. 1992a. Assessment and Remediation of Contaminated Sediments (ARCS) Program. EPA 905-
R92-008.
U.S. EPA. 1992b. A national study of chemical residues in fish. (EPA823-R-92-008a and b.) Office of
Water Regulations and Standards. Vols. 1 and 2. September 1992.
\
U.S. EPA. 1993. Water quality guidance for the Great Lakes system and correction: Proposed Rules. Fed.
Regist. 58(72):20802-21047 (April 16, 1993).
Methylmercury Water Quality Criterion 1/3/01
R-25
-------
U.S. EPA. 1993. Memo from Martha G. Prothro, Acting Assistant Administrator for Water, to Water
Management Division Directors and Environmental Services Division Directors, titled "Office of Water
Policy and Technical Guidance on Interpretation and Implementation of Aquatic Life Metals Criteria."
October 1,1993, Washington, DC.
U.S. EPA. 1994. Methods for derivation of inhalation reference concentrations and application of
inhalation dosimetry. Office of Health and Environmental Assessment, Environmental Criteria and
Assessment Office. Research Triangle Park, NC. EPA/600/8-90/066F.
U.S. EPA. 1996. The metals translator: guidance for calculating a total recoverable permit limit from a
dissolved criteria. June 1996. Washington, D.C.
U.S. EPA. 1997a. Mercury study report to Congress. Vol. I. Executive summary. U.S. Environmental
Protection Agency. December, 1997.
U.S. EPA. 1997b. Mercury study report to Congress. Vol. ffl. Fate and transport of mercury in the
environment. U.S. Environmental Protection Agency. December 1997. EPA-452/R97-005.
U.S. EPA. 1997c. Mercury study report to Congress. Vol. IV. An assessment of exposure to mercury in
the United States. U.S. EPA, Office of Air Quality Planning and Standards and Office of Research and
Development. EPA/452/R-97-006.
U.S. EPA. 1997e. Mercury study report to Congress. Vol. V. Health effects of mercury and mercury
compounds. U.S. Environmental Protection Agency. December, 1997.
U.S. EPA. 1997f. Mercury study report to Congress. Vol. VI. An Ecological assessment for
anthropogenic mercury emissions in the United States. U.S. Environmental Protection Agency.
December 1997.
U.S. EPA. 1997g. Mercury study report to Congress. Vol. VBL Characterization of human health and
wildlife risks from mercury exposure in the United States. U.S. Environmental Protection Agency.
December 1997.
U.S. EPA. 1997h. Exposure factors handbook. Vols. I, H, and m. EPA/600/P-95/002Fa. August 1997.
U.S. EPA. 1997L The national survey of mercury concentrations in fish. Database summary. 1990-1995.
September 29, 1997.
U.S. EPA. 1998a. Federal Register notice: draft revisions to the methodology for deriving ambient water
quality criteria for the protection of human health. EPA 822-Z-98-001. August 1998.
U.S. EPA. 1998b. Ambient water quality criteria derivation methodology: human health. Technical
support document. April. EPA-822-B-98-005. July 1998.
U.S. EPA. 2000a. Methodology for Deriving ambient water quality criteria for the protection of human
health (2000). Office of Science and Technology, Office of Water. Washington, DC. EPA-822-B-00-004.
October.
R-26
Methylmercury Water Quality Criterion 1/3/01
-------
U.S. EPA. 2000b. Estimated per capita fish consumption in the united states: based on data collected by
the United States Department of Agriculture's 1994-1996 continuing survey of food intake by
individuals. Office of Science and Technology, Office of Water, Washington, DC. March.
U.S. EPA. 2000c. Peer review comments report. Peer review of EPA's National Bioaccumulation Factors
for Methylmercury. Prepared by Versar, Inc., under EPA Contract 68-C-98-189. August 23, 2000.
U;S. EPA. 2000d. Per capita fish consumption estimates in the U.S. March 2000.
U.S. EPA. 2000e. Revisions to the methodology for deriving ambient water quality criteria for the
protection of human health (2000); Notice. Fed. Regist. 65:66444.
U.S. EPA. 2000f. Peer review workshop report on reference dose (RfD) for methylmercury. Prepared by
Versar Inc., 6850 Versar Center, Springfield VA 22151, for U.S. Environmental Protection Agency,
ORD/ National Center for Environmental Assessment, Washington DC 20460.
U.S. EPA. 2000g. Methodology for deriving ambient water quality criteria for the protection of human
health (2000). Technical Support Document. Volume 1: Risk Assessment. October 2000. EPA-822-B-00-
005.
U.S. FDA (United States Food and Drug Administration). 1978. As cited in text Mercury Study Report to
Congress. Vol. IV. Reference information not listed in bibliography.
U.S. FDA. 1999. Total diet study statistics on element results, 1991-1996. Revision 0. June 15, 1999.
Urano, T., N. Imura, and A. Naganuma. 1997. Inhibitory effect of selenium on biliary secretion of methyl
mercury in rats. Biochem. Biphys. Res. Comm. 239:862-867.
Vahter, M., A. Akesson, B. Lind, U. Bjors, A. Schutz, and M. Berglund. 2000. Longitudinal study of
methylmercury and inorganic mercury hi blood and urine of pregnant and lactating women, as well as in
umbilical cord blood. Environ. Res. 84:186-194.
Vernon, P.A. 1983. Speed of information processing and general intelligence. Intelligence 7:53-70.
Vernon, P.A. 1989. The generality of g. Personal. Individ. Diff. 10:803-804.
Vernon, P.A., S. Nador, and L. Kantor. 1985. Reaction times and speed-of-processing: their relationship
to timed and untimed measures of intelligence. Intelligence 9:357-374.
Verschaeve, L., M. Kirsch-Volders, and C. Susanne. 1983. Mercury chloride- and methyl mercury
chloride-induced inhibition in NOR activity. Teratol. Carcinogen. Mutagen. 3:447-456.
Verschaeve, L., M. Kirsch-Volders, and C. Susanne. 1984. Mercury-induced segregational errors of
chromosomes in human lymphocytes and in Indian muntjac cells. Toxicol. Lett. 21:247-253.
Von Burg, R., and H. Rustam. 1974a. Electrophysiological investigations of methyl mercury intoxication
in humans: Evaluation of peripheral nerve by conduction velocity and electromyography. Electroenceph.
Clin. Neurophysiol. 37:381-392.
Methylmercury Water Quality Criterion 1/3/01
R-27
-------
Von Burg, R., and H. Rustam. 1974b. Conduction velocities in methyl mercury poisoned patients. Bull.
Environ. Contam. Toxicol. 12:81-85.
Vreman, K., N.J. van der Veen, EJ. van der Molen, and W.G. de Ruig. 1986. Transfer of cadmium, lead,
mercury and arsenic from feed into milk and various tissues of dairy cows: Chemical and pathological
data. Netherlands J. Agric. Sci. 34:129-144.
Wakita, Y. 1987. Hypertension induced by methyl mercury in rats. Toxicol. Appl. Pharmacol. 89:144-
147.
Wannag, A. 1976. The importance of organ blood mercury when comparing foetal and maternal rat organ
distribution of mercury after methylmercury exposure. Acta Pharmacol. Toxicol. 38:289-298.
Watanabe, T., T. Shimada, and A. Endo. 1982. Effects of mercury compounds on ovulation and meiotic
and mitotic chromosomes in female golden hamsters. Teratology 25(3):381-384.
Watras, C.J., and N.S. Bloom. 1992. Mercury and methylmercury in individual zooplankton: Implications
forbioaccumulation. Limnol. Oceanogr. 37:1313-1318.
Watras, C.J., and J.W. Huckabee, eds. 1994. Mercury pollution: integration and synthesis. New York:
Lewis Publishers.
Watras, C.J., K.A. Morrison, J. Host, and N.S. Bloom. 1995a. Concentration of mercury species in
relationship to other site-specific factors in the surface waters of northern Wisconsin lakes. Limnol.
Oceanogr. 40:556-565.
Watras, C.J., K.A. Morrison, and N.S. Bloom. 1995b. Mercury in remote Rocky Mountain lakes of
Glacier National Park, Montana, in comparison with other temperate North American regions. Can. J.
Fish. Aquat. Sci. 52:1220-1228.
Watras, C.J., R.C. Back, S. Halvorsen, R.J.M. Hudson, K.A. Morrison, and S.P. Wente. 1998.
Bioaccumulation of mercury in pelagic freshwater food webs. Sci. Total Environ. 219:183-208.
Western, S.L., and C.J. Long. 1996. Relationship between reaction time and neuropsychological test
performance. Arch. Clin. Neuropsychol. 11:557-571.
WHO (World Health Organization). 1976. Environmental health criteria: mercury. Geneva, Switzerland:
World Health Organization, p. 121.
WHO. 1990. Environmental health criteria 101. Methylmercury. Geneva, Switzerland: World Health
Organization.
Wiener, J., W. Fitzgerald, C. Watras, and R. Rada. 1990. Partitioning and bioavailability of mercury in an
experimentally acidified Wisconsin lake. Environ. Toxicol. Chem. 9:909-918.
Wiersma, D., BJ. van Goor, and N.G. van der Veen. 1986. Cadmium, lead, mercury and arsenic
concentrations in crops and corresponding soils in the Netherlands. J. Agric. Food Chem. 34:1067-1074.
Wild, L., H. Ortega, M. Lopez, and J. Salvaggio. 1997. Immune system alteration in the rat after indirect
exposure to methyl mercury chloride or methyl mercury sulfide. Environ. Res. 74:34-42.
R-28
Methylmercury Water Quality Criterion 1/3/01
-------
Willett, W. 1990. Nature of variation in diet. In: Nutritional epidemiology. Willett, W., ed. Monographs
in Epidemiology and Biostatistics, Vol. 15. New York/Oxford: Oxford University Press, pp. 34-51.
Wren, C. 1992. Relationship of mercury levels in sportfish with lake sediment and water quality
variables. Toronto: Ontario Environmental Research Program. Govt Reports Announcements and Index
(GRA&I) Issue 08.
Wulf, H. C., N. Kromann, N. Kousgaard, J. C. Hansen, E. Niebuhr, and K. Alboge. 1986. Sister
chromatic exchange (SCE) hi Greenlandic Eskimos. Dose-response relationship between SCE and seal
diet, smoking, and blood cadmium and mercury concentrations. Sci. Total Environ. 48(l-2):81-94.
Yang, J., Z. Jiang, Y. Wan, LA. Qureshi, and X.D. Wu. 1997. Maternal-fetal transfer of metallic mercury
via the placenta and milk. Ann. Clin. Lab. Sci. 27(2): 135-141.
Yip, R.K., and L.W. Chang. 1981. Vulnerability of dorsal root neurons and fibers toward methyl mercury
toxicity: a morphological evaluation. Environ. Res. 26:152-167.
Zahn, T.P., M. Kruesi, and J.L. Rapopbrt. 1991. Reaction time indices of attention deficits in boys with
disruptive behavior disorders. J. Abnor. Child Psychol. 19:233-252.
Zhuang, G., Y. Wang, M. Zhi, W. Zhou, J. Yin, M. Tan, and Y. Cheng. 1989. Determination of arsenic,
cadmium, mercury, copper and zinc in biological sampels by radiochemical neutron-activation analysis.
J. Radioanal. Nucl. Chem. 129(2):459-464.
Methylmercury Water Quality Criterion 1/3/01
R-29
-------
-------
APPENDIX A
SECTION I. DRAFT NATIONAL METHYLMERCURY BIOACCUMULATION FACTORS
This appendix is a brief summary of the initial effort conducted to determine the feasibility of
deriving draft National bioaccumulation factors for methylmercury. This appendix is based on the draft
bioaccumulation report. The complete version of the original draft bioaccumulation factor report, with
more in-depth discussions of the methodology, a list of the references cited, rationales for using data, and
an uncertainty discussion can be obtained from the Water Docket W-00-20.
This appendix does not reflect comments or changes suggested by the peer reviewers. No changes
were made to the draft report that served as the basis for this appendix. Data interpretations, findings, or
conclusions discussed in this appendix are preliminary and may be changed in the future.
Introduction
The methylmercury bioaccumulation factors (BAFs) were estimated using guidance presented in
the Methodology for Deriving Ambient Water Quality Criteria for the Protection of Human Health (U.S.
EPA, 2000a; hereafter "the 2000 Human Health Methodology") and supplemented with methods
presented in the Mercury Study Report to Congress (MSRC; U.S. EPA, 1997c). The generalized
equation for estimating a BAF is as follows:
BAF =
Equation-1
where:
Ct = Concentration of the chemical in the wet tissue (either whole organism or specified
tissue)
Cw = Concentration of chemical in water
Literature searches were conducted to obtain data on bioaccumulation, concentrations of different
forms of mercury in water, percent methylmercury in tissue, and mercury predator-prey data. The data
sources primarily included articles from peer reviewed journals published between 1990 and April of
1999 and publicly available reports (e.g., State, Federal, or trade/industry group reports; dissertations;
Methylmercury Water Quality Criterion 1/3/01
A-l
-------
proceedings from professional meetings). Data from a variety of aquatic ecosystems (i.e., lakes, rivers,
estuaries) and on lower trophic levels was specifically looked for since the MSRC focused only on lakes
(primarily northern oligotrophic lakes) and trophic levels 3 and 4 fish.
BAFs are used in the ambient water quality criteria (AWQC) equation to estimate human mercury
exposure from consumption of contaminated fish. Equation 2 is the generalized AWQC equation for a
noncarcinogen and shows where the BAF fits into the calculation.
AWQC=RfDxRSC
BW
Equation 2
Where:
RfD = reference dose for noncancer human health effects
RSC = relative source contribution to account for non-water sources of exposure
BW = human body weight
DI = drinking water intake
FI = fish intake
BAFj = bioaccumulation factor for chemical "i".
The methylmercury BAFs that would be used in the above equation are presented in the
accompanying table A-9, and are calculated as the geometric mean BAF of all BAFs calculated for a
given trophic level.
Attachment A at the end of this appendix also contains the general comments made by the external
peer reviewers on the draft national methylmercury BAFs.
Methods for Estimating Bioaccumulation Factors
Three approaches were used to derive draft BAFs that could be used to derive draft national
methylmercury BAFs. These are direct, indirect, and conversion (modified direct) approaches. Each of
A-2
Methylmercury Water Quality Criterion 1/3/01
-------
these approaches has its own limitations, biases, and uncertainties associate with it. These approaches
and the BAFs derived using them are summarized below.
EPA's BAF derivation guidance is based on a data hierarchical preference approach. Under the
hierarchy, the preferred method for deriving a BAF for an organometallic compound such as
methylmercury is to use field-measured data to directly calculate a BAF (i.e., the direct method). BAFs
estimated using this direct approach are calculated using the simple ratio of the chemical concentration in
tissue and water. When such field data do not exist, or if the available field data are considered
unreliable, the next preferred method in the hierarchy estimates a BAF by multiplying a bioconcentration
factor (BCF) by a food chain multiplier (FCM) (i.e., the indirect method). The FCM is a factor used to
account for food chain interactions and biomagnification. EPA has used this indirect method to estimate
BAFs to support the development of wildlife criteria values in the Great Lakes Water Quality Initiative
or GLWQI (EPA, 1993) and in the MSRC (EPA, 1997). With few exceptions, field-derived FCMs were
calculated using concentrations of methylmercury in predator and prey species using the following
equations:
FCM ^ = (BMF^3) (BMP ^
FCM M = (BMP ^) (BMP ^ (BMP ^
Equation-3
Equation-4
Equation-5
where:
FCM = Food chain multiplier for designated trophic level (TL2, TL3, or TL4)
BMP = Biomagnification factor for designated trophic level (TL2, TL3, or TL4)
The basic difference between FCMs and BMFs is that FCMs relate back to trophic level one, whereas
BMFs always relate back to the next lowest trophic level. Biomagnification factors are calculated from
methylmercury tissue residue concentrations determined in biota at a site according to the following
equations:
BMP TL2 = C,, TL2) / (C, ,TL1)
Equation-6
Methylmercury Water Quality Criterion 1/3/01
A-3
-------
BMP TU = (C,, TL3) / (Cf, TL2)
BMP™ = (C,, TL4) / (C,, TL3)
Equation-7
Equation-8
where:
C,= concentration of chemical in tissue of appropriate biota that occupy the specified trophic
level (TL2, TL3, or TL4).
With the indirect BAF approach, it is important that when either selecting predator prey field data
from the literature or when conducting a site-specific field study to obtain such data, that the feeding
relationships between predator and prey are based on functional feeding relationships. It should be
verified that a given predator is feeding on a given prey item at the location in question so that the BMFs
and FCMs reflect actual trophic transfer of the chemical as close as possible. Usually, it is not enough to
simply know that organisms are from two different trophic levels. Unfortunately, for the analyses
presented here, much of the available data obtained from the published literature were insufficient to
document functional feeding relationships. Thus, BAFs derived using the indirect approach were not
used in determining the draft national methylmercury BAFs, but are presented only for comparison
purposes.
In the MSRC, in cases where the direct empirical BAF derivation method could be used, but the
available data was for a form of mercury other than dissolved methylmercury, a modified direct approach
was also used. The modified direct approach was used when either the water data or organism tissue data
was not in the methylmercury form (e.g., total mercury, dissolved total mercury, total methylmercury) but
could be converted to methylmercury using translating factors. Data for mercury in water was converted
to dissolved methylmercury by using chemical translators (see Section n of this Appendix). Mercury in
tissue reported as total mercury was converted to methylmercury by multiplying by a factor that estimates
the fraction of total mercury present in the methylated form (i.e., fmmf translator). The fmmfs were
developed from field studies where both total mercury and methylmercury were measured in biota tissue.
Using the methods outlined above, BAFs were estimated initially by trophic level for lakes (lentic
aquatic systems), rivers and streams (lotic aquatic systems), and estuaries. An ecosystem-based approach
to deriving the BAFs was used because differences in general bioaccumulation trends would be expected
among the aquatic ecosystems due to inherent differences in methylation processes, food web dynamics,
A-4
Methylmercury Water Quality Criterion 1/3/01
-------
mercury loadings, and watershed interactions, among other factors. However, due to the lack of data in
terms of both quality and quantity, no clear differences in bioaccumulation trends were observed between
lentic and lotic ecosystems based on the available data (see Figure A-3). Based on qualitative and semi-
quantitative comparisons of the data, no significant difference was found between the lentic and lotic
BAFs. Thus, they were combined for each trophic level to obtain the trophic level-specific draft national
BAFs. A near complete lack of adequate data prohibited derivation of draft national BAFs for estuarine
systems.
Summary of BAFs for Methylmercury in Lentic Ecosystems
Table A-l compares the BAFs estimated using the two primary approaches (direct and indirect)
methods for estimating BAFs for trophic levels 2, 3, and 4 species. Although the BAFs based on the
indirect approach are not used in the national draft BAF calculations because they are not based on
verifiable functional predator-prey feeding relation ships, they are nonetheless useful for comparing and
assessing general tends in bioaccumulation. Other than the BAF2, the BAFs are within a factor of two of
one another. Both the direct and indirectly estimated BAFs show an expected increase in methylmercury
bioaccumulation with increasing trophic position. This suggests that if functional predator-prey feeding
relationships can be developed, that indirect BAFs could provide reasonably good approximations of
methylmercury bioaccumulation in organisms in the field.
Table A-l: Summary of Bioaccumulation Factors for Methylmercury Mercury in Lentic
Ecosystems
Parameter
BCF
BAF2
BAF3
BAF4
Methylmercury (1)
Direct ft/kg'1)
5.9 x 104
8.6 xlO4
1.3 xlO6
6.8 x 106
Indirect (L'kg'1)
NA
3.1 x 105
2.2 x 106
1.1 x 107
(1) All values are geometric means
Methylmercury Water Quality Criterion 1/3/01
A-5
-------
Summary of BAFs for Methylmercury in Lotic Ecosystems
Table A-2 compares the lotic BAFs estimated using the direct and indirect methods. The BAFs
based on the indkect approach are not used in the draft national BAF calculation because they are not
based on verifiable functional predator-prey feeding relation ships; they are nonetheless useful for
comparing and assessing general tends in bioaccumulation. As was the case with the lentic indirectly
estimated BAFs, the indirect lotic BAFs are close approximations of the directly estimated BAFs (within
a factor of 3 or less). Also, as was observed for lentic ecosystems, both the direct and indirectly
estimated lotic BAFs show an expected increase in methylmercury bioaccumulation with increasing
trophic position. This suggests that if functional predator-prey feeding relation ships can be developed,
Table A-2: Summary of Dissolved Methylmercury Bioaccumulation Factors for Lotic Ecosystems
Parameter
BCF
BAF2
BAF3
BAF4
Methylmercury (1)
Direct (I/kg'1)
1.2 xlO4 ,
4.4 x 10s
1.6 x 106
2.5 x 106
Indirect (I/kg'1)
NA
1.9 x 105
5.6 x 105
3.2 x 106
(1) values are geometric means
that indirect BAFs could provide reasonably good approximations of methylmercury bioaccumulation in
organisms in the field.
Methylmercury BAFs Translated from Other Mercury Forms
Converted BAFs (that is, in terms of other mercury forms) were derived for dissolved
methylmercury using translator factors (see Section n, Chemical Translators for Mercury and
Methylmercury) and by using factors to convert total mercury measured in organism tissues to
methylmercury in tissues.
A-6
Methylmercury Water Quality Criterion 1/3/01
-------
Mercury Translators
For those studies that met the data quality objectives but did not analyze or report water mercury
concentrations in the dissolved methylmercury form, the reported form of mercury was converted to the
mean fraction of dissolved methylmercury (fd MeHgj) by using one or more of the "translators" listed in
Table A-3. Section n below discusses the methodology and data used to derive the translators. Section
II of this appendix also provides partition coefficients (KD) that were not necessary for this analysis, but
that can be used along with total suspended solids information to estimate the desired fraction of mercury
in water.
Table A-3; Summary of Mercury Translators for Mercury in Water
fd value
fdHgd/Hgt
fdMeHgd/Hgt
fdMeHgrf/MeHg,
Lentic
0.600
0.032
0.613
Lotic
0.370
0.014
0.490
Conversion Factors for Mercury in Organism Tissue
Similar to the water data, if mercury in biota tissue (muscle or whole body) was reported as total
mercury then the appropriate mean (arithmetic) estimate of the fraction present in the methylated form
(fmmf) for the respective trophic level was used to convert it to methylmercury. Table A-4 summarizes
the fmmfs used to estimate converted BAFs.
Table A-4: Summary of fmmfs for Lentic and Lotic Ecosystems
Trophic Level
1
2
3
4
Lentic
0.18
0.44
1.00
1.00
Lotic
0.05
0.49
1.00
1.00
Methylmercury Water Quality Criterion 1/3/01
A-7
-------
Summary and Comparison of Converted BAFs and BCFs derived for Lentic and Lotic Ecosystems
Methylmercury translator factors (see Section n, Chemical Translators for Mercury and
Methylmercury) were used to estimate dissolved methylmercury BCFs and BAFs in lotic and lentic
ecosystems. Table A-5 summarizes the converted BAFs. The converted lentic BAFs range from
approximately 2 to 37 times greater than the converted lotic BAFs.
Figures A-l and A-2 compare the direct and converted estimates of BAFs and BCFs for lentic and
lotic ecosystems, respectively. Although the data sets are relatively small, the ranges of converted BAFs
are in agreement with BAFs directly estimated. Tables A-6 and A-7 summarize and compare the point
estimates of each data set. In lentic ecosystems, the difference between the mean directly estimated
BAFs and mean converted BAFs is generally less than a factor of two. For lotic ecosystems, the
difference is slightly larger, ranging from a factor of two to a factor of seven, with an overall mean
difference of four. This information suggests that the converted BAFs in each ecosystem are good
estimates of directly measured BAFs for all trophic levels. However, because the set of BAFs estimated
using the two different approaches are small for each ecosystem, insufficient data were available to
perform any rigorous statistical evaluation to determine if a significant difference exists between the
BAFs of each system. Nonetheless, graphically the data suggest that the direct and converted BAFs can
be combined to derive overall BAFs for each trophic level in each ecosystem. The BAFs based on the
combined data sets are presented in Table A-8.
Figure A-3 compares the combined data sets (e.g., directly-measured and converted BAFs and
BCFs) for lentic and lotic ecosystems. While the lotic BAFs clearly span a greater range than the lentic
BAFs, the differences between the mean lotic BAFs and the mean lentic BAFs for each trophic level are
fairly small (differences range between 1 and 5). To investigate if there were significant differences
between the BAFs for the two ecosystems significant, a student's T-test was performed on the combined
data for each trophic level-specific BAF and BCF using the computer software WINKS (Texasoft, 1999).
Although differences in mercury bioaccumulation between lentic and lotic ecosystems could be expected
due to differences in mercury loading characteristics, bioavailability, food web dynamics, and
methylation processes, among other factors, no significant statistical differences (p>0.05) were found
between the lentic and lotic BAFs and BCFs. Furthermore, a closer inspection of the converted lentic
BAF4 data for several Minnesota Lakes (Glass et al., 1999) suggests that, given a larger sample size, the
lower range of field-measured lentic BAF4 values could be similar to the lower range of values observed
for lotic ecosystems. Whether these observations are artifacts of the available data or trends due to real
A-8
Methylmercury Water Quality Criterion 1/3/01
-------
Table A-5: Comparison of Converted Bioaccumulation Factors for Methylmercury in Lotic and
Lentic Ecosystems
Parameter
BCF
MDBAF2
MDBAF3
vmBAF4
Lentic (I/kg'1)
4.3 x 104
1.5 x 105
1.3 x 106
4.1 x 106
Lotic (Lkg-1)
6.1 x 103
6.2 x 104
3.5 x 104
1.4xl06
Table A-6: Comparison of Direct and Converted Methylmercury BAFs and BCFs for Lentic
Ecosystems
Value1
5th
50th (GM)
95th
GSD
MoBCF
direct
12,300
58,700
281,000
2.59
converted
13,400
43,000
138,000
2.26
MoBAF2
direct
16,700
85,600
439,000
2.70
converted
47,500
150,000
474,000
2.01
MDBAF3
direct
322,000
1,260,000
4,900,000
2.29
converted
466,000
1,330,000
3,820,000
1.90
MDBAF4
direct
3,270,000
6,800,000
14,200,000
1.56
converted
3,800,000
4,080,000
4,380,000
1.04
1 GM = geometric mean; GSD = geometric standard deviation.
Table A-7: Comparison of Direct and Converted Methylmercury BAFs and BCFs for Lotic
Ecosystems
Value"
5th
SO"1
(GM)
95th
GSD
MnBCF
direct
340
5,400
85,800
5.38
converted
1,200
6,000
29,800
2.63
MoBAF2
direct
15,600
179,000
2,000,000
4.40
converted
3,400
61,900
1,130,000
3.39
MDBAF3
direct
261,800
1,640,000
10,200,000
3.05
converted
45,800
346,000
2,620,000
3.42
MDBAF<
direct
283,000
2,520,000
22,500,000
3.78
converted
55,400
1,380,000
30,300,000
6.80
" GM = geometric mean; GSD = geometric standard deviation.
processes is not distinguishable. Because the range of available BAF values for lentic and lotic systems
overlap one another, the individual BAFs for the two systems were combined in one data set to derive the
trophic level-specific draft national methylmercury BAFs.
Methylmercury Water Quality Criterion 1/3/01
A-9
-------
le+7 -
I
fa
1
u
w
I?
1
i
I
le+6 -
le+5 -
le+4 -
BCF
o
o
o
o
8
o
o
• Dkectly Calculated
O Converted
BAF2
BAF3
BAF4
Figure A-l. Comparison of direct field-measured and converted field-measured
methylmercury BCFs and BAFs for lentic ecosystems.
A-10
Methylmercury Water Quality Criterion 1/3/01
-------
le+7
le+6 -
1
i
£
i
!
I
le+5 -
le+4 -
le+3 -
• o
BCF
BAF2
O
o
8
o
o
8
o
o
•
o
Directly Calculated
Converted
BAF3
BAF4
Figure A-2. Comparison of direct field-measured and converted field-measured methylmercury
BCFs and BAFs for lotic ecosystems.
Methylmercury Water Quality Criterion 1/3/01
A-ll
-------
Table A-8: Summary of Lentic and Lotic Methylmercury BAFs and BCFs
Value(1)a (%)
5*
50th (GM)
9501
GSD
MDBCF
Lentic
13,300
45,000
153,000
2.10
Lotic
800
5,700
43,200
5.14
MoBAF2
Lentic
37,000
127,800
440,000
2.12
Lotic
8,000
105,000
1,390,000
4.80
MoBAF3
Lentic
423,000
1,115,000
2,930,000
2.02
Lotic
46,000
517,000
5,820,000
4.36
MoBAF4
Lentic
2,800,000
5,740,000
11,800,000
1.55
Lotic
73,400
1,240,000
20,900,000
5.57
(1) Values are based on combined direct and converted BAFs and BCFs.
8 GM = Geometric Mean; GSD = geometric standard deviation.
A-12
Methylmercury Water Quality Criterion 1/3/01
-------
fl
e
•s
a
I
s
I
le+7 -
le+6-
le+5 -
le+4 -
le+3 -
1
o
0
• 0
•
•
I : °
o
t 0
• ° °
• o
o
0
1 ... 1
1
0 •
• o
o
° o
i §
• 0 o
S o
o
• 8
8 o
o
o
o
8
o
o
• Lentic
O Lotic
BCF
BAF2
BAF3
BAF4
Figure A-3. Comparison of lentic and lotic methylmercury B AFs. Data includes both
direct field-measured BAFs and converted field-measured BAFs.
Methylmercury Water Quality Criterion 1/3/01
A-13
-------
Draft National Bioaccumulation Factors for Methylmercury
Based on the data presented above, and because the goal of the draft national BAFs is to be
applicable under as many circumstances and to as many water bodies as possible, the BAFs based on the
combined data sets (e.g., direct and converted, lentic and lotic) were chosen to be the empirically-derived
draft national BAFs for methylmercury. The draft National BAFs, along with the draft BCF, and their
empirical distributions are presented in Table A-9.
Table A-9: Summary of Draft National BAFs and BCF for Dissolved Methylmercury
Value"
5th percentile
SO* (GM) percentile
9Sa percentile
GSD
Draft National
Values
BCF
5,300
33,000
204,000
3.03
3.3 x 10"
BAF2
18,000
117,000
770,000
3.15
1.2 x 10s
BAF3
74,300
680,000
6,230,000
3.84
6.8 x 10s
BAF4
250,000
2,670,000
28,400,000
4.21
2.7 x 106
"GM = geometric mean; GSD = geometric standard deviation.
Discussion of Uncertainty and Variability in the BAF Estimates
The BAFs in this document were designed to estimate the central tendency of the concentration of
mercury in fish of a given trophic level from an average concentration of dissolved mercury for water
bodies located hi the continental U.S. As shown in figures A1-A3, there is at least an order of magnitude
in the variability of the individual BAF estimates for a given trophic level, which leads to uncertainty in
the overall central tendency estimate. This is further reflected hi the range of 90 percent (5th and 95th
percentiles) confidence intervals. Although the empirical range of any given 90 percent confidence
interval may largely overestimate the true extent of variability, the distributions do provide a rough
estimate of the total uncertainty in the aggregate processes and an idea of the precision (or lack thereof)
of the BAF estimates. The uncertainty in the BAF estimates is related to two basic sources. First is the
uncertainty arising from natural variability, such as size of individual fish or differences in metabolic
processes. Second is the uncertainty due to measurement error, such as error in measurements of
mercury in water and fish samples or lack of knowledge of the true variance of a process (e.g.,
methylation). These two sources of uncertainty are generally referred to as "variability" and
A-14
Methylmercury Water Quality Criterion 1/3/01
-------
"uncertainty", respectively. In this analysis, there was no distinction made between variability and
uncertainty; they are aggregated in the final BAF distributions and point estimates. Thus, it cannot be
determined where natural variability stops and uncertainty starts. However, some of the more important
sources of variability and uncertainty are highlighted below in order to assist risk managers in
understanding what the limitations are surrounding the BAFs, to see how the uncertainty in the BAF
estimates might be reduced should they derive more data, and to assist them in decisions on development
of site-specific BAFs.
Uncertainty Due to Sampling and Chemical Analysis
In many cases, water methylmercury concentrations reported in the available studies incorporated
limited or no cross-seasonal variability, incorporated little or no spacial variability, and were often based
on a single sampling event. Because fish integrate exposure of mercury over a life time, comparing fish
concentrations to a single sample or mean annual concentrations introduces bias to the estimates. The
geographic range represented by the water bodies is also limited. The available lentic data are biased
towards northern oligotrophic lakes, primarily located in the Great Lakes region. The lotic BAFs are
primarily based on data from canals of the Everglades (assumed to act as flowing aquatic ecosystems)
and from a point-source-contaminated stream in Tennessee. Because of this general lack of data, a few
studies on water bodies in other countries were included in the analysis, requiring one to assume that
biotic and abiotic processes in these lakes are similar to lakes in the continental U.S.
The same sampling and analytical methods for water and tissue samples were not used in each
acceptable study. Although all studies used met general requirements for data quality, studies with
different analytical detection limits were combined to estimate the BAFs. The range of species used in
the BAF estimates is relatively small compared to the suite offish and invertebrates consumed by the
general human population. Much of the available trophic level 4 data for both lentic and lotic
ecosystems is limited to walleye, pike, or bass. For trophic level 3 much of the data is for bluegill and
perch. For trophic level 2, most of the data was for zooplankton in lentic waters and for planktivorous
fish in lotic waters. The lack of data complicated comparisons between the two aquatic ecosystems and
introduces uncertainty into application of the BAFs.
Methylmercury Water Quality Criterion 1/3/01
A-15
-------
Uncertainty Due to Estimation Method
Each of the approaches use to estimate BAFs have their own inherent uncertainties. Both the direct
and indirect approaches assume that the underlying process and mechanisms of mercury bioaccumulation
are the same for all species in a given trophic level and for all water bodies. The indirect approach deals
with this assumption more specifically by assuming that the translators and fmmfs used to convert BAFs
are equally applicable to all ecosystems. In reality, these factors are based on a limited set of data.
Although the translators and fmmfs used in the analysis are consistent with those reported elsewhere
(Porcella, 1994), they may over- or underestimate bioavailability and bioaccumulation in specific water
bodies. Ideally, site-specific conversion factors would be used to estimate BAFs more reflective of
conditions in a given water body. The approach used here aggregates all of the species-specific BAFs
into a single trophic level-specific BAF; this also increases the over all variability in the BAF estimates.
Uncertainty Due to Biological Factors
Other than deriving BAFs based on organism trophic level, and initially by general water body type
(i.e., lentic and lotic), there were no distinctions in the BAFs as to size/age offish, water body trophic
status, or underlying mercury uptake processes. It has been shown that methylmercury bioaccumulation
for a given species can vary as a function of the ages (body size) of the organisms examined (Glass et al,
1999; Watras et al., 1998; Suchanek et al., 1993; Lange et al. 1993). As a result, it has been suggested
that to reduce some of the lake-to-lake variability seen in BAFs for a given species, comparisons between
water bodies should be made using "standardized" fish values (i.e., a value for a hypothetical 1 kg
northern pike; Glass et al., 1999). Typically such data "normalization" is derived by linear regression of
residue data collected from individuals of varying size and/or age. However, the currently available data
are too limited to perform this kind of normalization; most of the water body-specific BAFs, and
resulting trophic level distributions, are based on "opportunity" (whatever you catch, you include) and do
not report age or size of individuals sampled.
Uncertainty Due to Universal Application of BAFs
Perhaps the greatest source of variability is that of model uncertainty. That is, uncertainty
introduced by failure of the model (in this analysis a single trophic level-specific BAF) to represent
significant real-world processes that vary from water body to water body. The simple linear BAF model
relating methylmercury in fish to total mercury in water simplifies a number of nonlinear processes that
A-16 Methylmercury Water Quality Criterion 1/3/01
-------
lead to the formation of bioavailable methylmercury in the water column and subsequent accumulation.
Much of the variability in field data applicable to the estimation of mercury BAFs can be attributed to
differences in biotic factors (e.g., food chain, organism age/size, primary production,
methylation/demethylation rates), and abiotic factors (e.g., pH, organic matter, mercury loadings,
nutrients, watershed type/size) between aquatic systems. As an example, in lake surveys conducted
within a relatively restricted geographic region, large differences can exist between lakes with respect to
mercury concentrations in a given species of fish (Cope et al., 1990; Grieb et al., 1990; Sorenson et al.,
1990; Jackson, 1991; Lange et al., 1994; Glass et al., 1999). These observations have led to the
suggestion that a considerable portion of this variability is due to differences in within-lake processes
that determine the percentage of total mercury that exists as the methylated form. Limited data also
indicate that within a given water body, concentrations of methylmercury are likely to vary with depth
and season. Unfortunately, while the concentration of methylmercury in fish tissue is presumably a
function of these varying concentrations, published BAFs are generally estimated from a small number of
measured water values, whose representativeness of long-term exposure is poorly known. Furthermore,
although it is known that biotic and abiotic factors control mercury exposure and bioaccumulation, the
processes are not well understood, and the science is not yet available to accurately model
bioaccumulation on a broad scale.
Summary
Three different approaches were use to estimate methylmercury bioaccumulation factors for use in
deriving national 304(a) ambient water quality criteria for mercury. All three approaches resulted in
BAFs with central tendency point estimates in good agreement with one another. Based on data
comparability and EPA's national guidance for deriving BAFs, methylmercury BAFs estimated using
directly measured and converted field data were used as the basis for deriving the draft national BAFs.
Given the large range in the data, at this time lotic BAFs can not be distinguished from lentic BAFs,
though the data suggests slightly reduced methylmercury accumulation may occur in higher trophic level
organisms in lotic/wetland environments. The same trend is observed when BAFs are compared on a
total mercury basis. Some of this difference might be accounted for by the lower accumulation of
methylmercury at the base of the food chain in lotic/wetland ecosystems. A plausible explanation for this
difference is the observation that the bioavailability of methylmercury in lentic environments (usually a
low dissolved organic carbon content) may exceed the bioavailability of methylmercury in lotic/wetland
environments (usually a high dissolved organic carbon content). Methylmercury and mercury have a
high binding capacity to dissolved organic carbon which can affect their bioconcentration in
Methylmercury Water Quality Criterion 1/3/01
A-17
-------
phytoplankton/periphyton. Watras et al. (1998) used modeling to show that BAFs based on the
bioavailable fraction of methylmercury in water exceed BAFs based on the operationally defined
(filtered) dissolved methylmercury in water. Bioavailability is perhaps the single most important factor
affecting BAFs for mercury.
EPA fully recognizes that the approach taken to derive mercury BAFs collapses a very complicated
non-linear process, which is affected by numerous physical, chemical, and biological factors, into a
rather simplistic linear process. EPA also recognizes that uncertainty exists in applying a National BAF
universally to all water bodies of the United States. Therefore, in the revised 2000 Human Health
Methodology (EPA, 2000) we encourage and provide guidance for States, Territories, Authorized
Tribes, and other stakeholders to derive site-specific field-measured BAFs when possible. In addition,
should stakeholders believe some other type of model may better predict mercury bioaccumulation on a
site-specific basis they are encouraged to use one, provided it is scientifically justifiable and clearly
documented with sufficient data.
A-18
Methylmercury Water Quality Criterion 1/3/01
-------
SECTION H. CHEMICAL TRANSLATORS FOR MERCURY AND METHYLMERCURY
Introduction
By regulation (40 CFR 122.45(c)), the permit limit, in most instances, must be expressed as total
recoverable metal. Because chemical differences between the discharged effluent and the receiving
water are expected to result in changes in the partitioning between dissolved and adsorbed forms of
metal, an additional calculation using what is called a translator is required.
The translator is used to convert the dissolved concentration of a metal to a total metal concentration for
use in waste load limit calculations. The translator is the fraction of the total recoverable metal in the
downstream water that is dissolved, fd. The translator can be used to estimate the concentration of total
recoverable metal in a water body.
Methods
Two procedures were used to develop site-specific translators. The most straightforward approach
for translating from a dissolved water quality criterion to a total recoverable effluent concentration is to
analyze directly the dissolved and total recoverable fractions. The translator is the fraction of total
recoverable metal that is dissolved. It may be determined directly by measurements of dissolved and
total recoverable metal concentrations in water samples taken from the well mixed effluent and receiving
water (i.e., at or below the edge of the mixing zone). In this approach, a number of samples are taken
over time and an fd value is determined for each sample:
fd = Cd/Ct [Eqn. 1]
where:
Cd = the dissolved concentration, and
Ct = the total metal concentration.
The translator is then calculated as the geometric mean (GM) of the dissolved fractions.
Methylmercury Water Quality Criterion 1/3/01
A-19
-------
The second approach derives an fd from the use of a partition coefficient KD where usually the
coefficient is determined as a function of total suspended solids (TSS) (although some other basis such as
humic substances or particulate organic carbons may be used). The partition coefficient is the ratio of
the particulate-sorbed and dissolved metal species multiplied by the adsorbent concentration, i.e.
Cd + TSS *•" Cp, where Cp is the bulk particulate-sorbed concentration, and is expressed as:
KD = Cp/(Cd -TSS) [Eqn.2].
The dissolved fraction and the partition coefficient are related as shown in equation 3.
fd = (l + KD-TSS)-' [Eqn.3]
As in the first approach, numerous samples are collected over time, and the fd and TSS values found at
the site are fit to a least squares regression, the slope of which is KD. The established KD is then used to
determine the translator using Eqn. 3 with a TSS value representative of some critical condition, e.g., low
flow conditions.
Although development of site-specific translators is recommended, EPA also envisions the possible
need for national or default translators for use in translating dissolved mercury and dissolved
methylmercury criteria into total mercury and methylmercury water quality permit limitations.
Translators and/or related KD values can be generated from an acceptable existing literature-derived data
base. EPA's MSRC (U.S. EPA, 1997) contains extensive data, obtained primarily from lake systems,
that are relevant to developing translators for mercury (e.g., percent total as methylmercury, percent total
as dissolved mercury). Supplementation of these translators with additional, acceptable data from lotic
and estuarine systems and update of lentic systems provides the necessary data base for the translators.
To gather this data base, peer-reviewed literature papers from 1990 to present, were searched and
reviewed. Since awareness of the contamination problems with mercury at low levels and the existence
of analytical methods capable of accurately and precisely measuring mercury and methylmercury at low
levels are relatively recent, the literature review was not conducted for publications prior to 1990. All
data from the literature for use in developing the translators were required to meet the following criteria:
• Clean techniques, or equivalent, to reduce contamination were used in sampling and analysis.
• Adequate QA/QC procedures were used.
• Analytical methods used provided sufficiently low enough detection level.
A-20 Methylmercury Water Quality Criterion 1/3/01
-------
Draft Translators
Table A-10 summarizes the numerous tables from the EPA internal draft BAF report (see Water-
Docket W-00-20). These results are presented separately for lake, river and esruarine systems, and for
each system, where sufficient data were available, both fd and KD values were tabulated. The KD values
were calculated using Eqn. 2. The KD values could not be derived using the fd-TSS correlation approach
due to the limited data, i.e., multiple sampling events over time with measurements of both fd and TSS
were not conducted in most of the studies. The results are presented separately for both mercury and
methylmercury. Table A-10 provides a summary of the GM values calculated for each system for fd and
KD values, again for both mercury and methylmercury.
It is possible to calculate a "pseudo" KD value for the partitioning of dissolved methylmercury with
particulate total mercury using fd and KD data for a waterbody utilizing the following equation (see
Attachment B for derivation and example calculation):
"Pseudo" KD MeHgd/Hgt = KD MeHgd • MeHgt • Ratio Hgd/MeHgp • Ratio MeHgd/MeHgp
[Eqn. 4]
Table A-10; Summary of Fd and E^ Values for Lakes, Rivers, and Estuaries3
f d and KD Values
fdHg
fdMeHgd/Hgt
fdMeHgd/MeHgt
LogKDHg
LogKDMeHg
"pseudo" Log KD
MeHgd/Hgt
Lakes
0.60
0.032
0.613
5.43
5.53
6.83
Rivers
0.37
0.014
0.49
5.06
4.81
6.44
Estuaries
0.353
0.190"
0.612"
5.52
NF
NCd
a Values calculated as GM
b Only two sites
c No data found from the literature search
d Not able to calculate due to insufficient data
Methylmercury Water Quality Criterion 1/3/01
A-21
-------
The KD so derived is a "pseudo" value since dissolved methylmercury partitioning with particulate total
mercury is just a synthetic or functional type description. These values are also given in Table A-10.
The "pseudo" KD values, however, allow for direct translation of dissolved methylmercury criteria to
total mercury permit limits employing some designated TSS level. Insufficient data were found, e.g.,
K0MeHg, to allow for calculation of "pseudo" KDs for estuaries. It should be understood that all values
in Table A-10 represents values generated from the above-described literature-gleaned data base.
Insufficient data were obtained to provide either reliable fd (translator) or KD "default" values for
methylmercury for estuarine systems (only two sites). Examination of the translator values for lakes and
rivers shows that in all instances the river values for both fds and KDs are lower than the lake values. The
lower translator values can be generally explained by the generally higher TSS levels found in rivers as
compared to lakes. For example, typical TSS values for eastern Washington state lakes are 0.5 to 5
mg/L, whereas river levels can be typically 5-50 mg/L (Pankow and McKenzie, 1991). Higher TSS
levels lead to lower fd values.
The lower KD values for rivers vs. lakes are not as readily explainable. KD values are not constant
and are sensitive to environmental conditions and water chemistry (Sung, 1995). Inclusion of the
colloidal fraction in the dissolved phase that is used in determining the KD has been used to explain
variation of KD values and for deviation of the values from any true KD (Pankow and McKenzie, 1991;
Sung, 1995). Higher colloidal contents or higher DOC levels in the river samples compared with lake
samples would produce lower apparent (as measured) KD values. However, the following other factors
have been suggested to play major roles in KD determinations, and one or all of these may contribute
significantly to the reason why the river KDs are less than the lake KDs for both mercury and
methylmercury:
• Biotic or organic content of the TSS
• Dissolved organic content of the water
• Geochemistry and residual metal content of the TSS
• TSS particle size
• Pollution level existing in the waters
A-22
Methylmercury Water Quality Criterion 1/3/01
-------
Regardless of the reason(s) for the differences between the lake and river values, differences do exist and
are sufficiently significant that it is recommended that the two systems be treated separately with regard
to translator values. Until additional data are available for estuarine systems, and a satisfactory
comparison to lake and river systems can be made, it is recommended that separate values be retained for
estuaries also.
One can estimate the TSS level that is represented by the fd values for each system through the use
of Eqn. 3 and employing the default KD values provided in Table A-10. The results of calculations of
these estimated levels and an example calculation are presented in Table A-l 1. The data show the
following:
• In lakes, the fd for mercury (0.60) would reflect TSS levels of 2.5 mg/L. The fd for methylmercury
(0.032) would reflect TSS levels of 1.8 mg/L. At TSS levels lower than these values, a greater
fraction of the mercury and methylmercury would be expected to be dissolved than indicated by the
In rivers, the fd for mercury (0.37) would reflect TSS levels of 14.8 mg/L. The fd for
methylmercury (0.014) would reflect TSS levels of 16.3 mg/L. At TSS levels lower than these
values, a greater fraction of the mercury and methylmercury would be expected to be dissolved than
indicated by the fd.
In estuaries, the fd for mercury (0.35) would reflect TSS levels of 5.5 mg/L.
Existing TSS levels less than those above would, in any instance, that the dissolved fraction present in
the water could be greater than the value suggests.
Use of the partition coefficient approach may provide advantages over the dissolved fraction. EPA
suggests (EPA, 1996) that when using dynamic simulation for Waste Load Allocation (WLA) or the
Total Maximum Daily Load (TMDL) calculations and permit limit determinations, KD allows for greater
mechanistic representation of the effects that changing environmental variables have on fd (the
significance of the TSS variable has been shown in Table A-l 1 data and discussed above, and this
variable is addressed or can be handled in the KD approach).
Methylmercury Water Quality Criterion 1/3/01
A-23
-------
Table A-ll: Estimation of TSS Level at frt Values
Mercurya
Methylmercuryb
Lakes
fd
0.60
0.032
Est. TSS,
mg/L
2.5*
1.8
Rivers
fd
0.37
0.014
Est. TSS,
mg/L
14.8
16.3
Estuaries
fa
0.35
0.190
Est. TSS, .
mg/L
5.5
NCC
(a) Calculated using default KD values and equation: fd = 1/(1+KD x TSS)
(b) Calculated using default "pseudo" KD values and equation: fd = l/(Hgd/HgMed + KD x TSS)
(c) Not able to calculate; insufficient data.
* Calculation:
fd = 1/(1 + KD x TSS x 10'6) note: lO'6 used to provide TSS in mg/L units
default KoHg (lakes) = 269,153
substituting: 0.60 = 1/(1 + 269,153 x TSS x 10'6)
0.60 + 0.161 x TSS = 1
0.161 x TSS = 0.40
TSS = 2.5
Although the KD approach may be advantageous in use, employment of a default KD value has
inherent problems as does the use of a fd. For example, mercury KDs have been shown to range from
about 104 to about 106 (Watras et al., 1995). At an average KD value of about 10s (the value found for
rivers), and a critical TSS level of 10 mg/L, a translator value of 0.5 is derived from the KD approach.
However, if the site KD, for example, is close to the lower end of the KD range, the translator value
should be about 0.9. Thus the value is inaccurate at this site. Only at sites where the existing KD is 105
or greater (at 10 mg/L TSS) would the use of the default KD yield a translator value that does not
underestimate the dissolved mercury level.
An additional problem with the use of the KD approach is that even at a given site, KD values can
vary. Usually, KD values decrease at a site as TSS increases, as has been shown recently for mercury and
methylmercury in a Virginia river (Mason and Sullivan, 1998). In addition, the KD translator approach
necessitates that fd correlate with TSS. A poor correlation, however, has been found to exist for many
metals in a recent analysis of data obtained from State of Michigan surface waters (MDEQ, 1996).
A-24
Methylmercury Water Quality Criterion 1/3/01
-------
Although the KD approach has its advantages, the fd approach is the most straightforward. Both
approaches have their disadvantages, as discussed previously. The KD is derived from fd values and so
the two approaches are truly linked. Therefore, preferential recommendation of either one approach over
the other at present cannot be made.
Use of either fd or KD default values can be made as long as one recognizes the short comings of the
approach taken. Perhaps the approach taken should be the one with the stronger data base, if a clear
difference exists. As additional data appears in the literature, it is reasonable to assume that a fine-tuning
of both the fd and KD default values will result. EPA recommends that translators be derived from site-
specific studies when possible, but the values in Table A-10 could be used in absence of any site-specific
data.
Methylmercury Water Quality Criterion 1/3/01
A-25
-------
ATTACHMENT A: BAF PEER REVIEWERS' GENERAL COMMENTS
The following was excerpted from the BAF Peer Review Comments Report, August 23, 2000. See
Water Docket W-00-20 for a complete version of the peer review report.
2.0 REVIEWERS' COMMENTS
2.1 General Comments
Nicolas Bloom
Overall, I found the document quite clear and well written compared to other EPA mercury
documents that I have recently reviewed, a fact that made my job considerably easier. On the other hand,
it seems quite clear that there is insufficient data currently available for the EPA to make any more than
the broadest generalizations about methyl mercury bioaccumulation factors. The current greater than one
order of magnitude spread in estimated BAFs will not be very useful in any actual case, although it
serves to describe the situation in general terms. The EPA should be impelled to proceed by instigating
research and/or requiring site-specific bioaccumulation factors to be developed until such time that a
sufficient database is accumulated to allow some meaningful resolution between BAFs from different
water body types, climates, and trophic levels.
I oppose the general use of the confusingly similar terms "lentic" and "lotic," which although
probably clear to fish ecologists, never-the-less provide endless confusion to the rest of us. I conducted a
poll of the 51 employees of our aquatic sciences research company, and no one could define these words
correctly, although a few did say that they had heard of them back in college. Additionally, even though
physically, the term "lentic" can be used to lump together the Everglades with a swiftly moving glacial
stream, I see no logical biogeochemical reason to do so.
There is also the overwhelming sense, in the description of the trophic levels considered, that the
only valid food chain model being considered is the water to plankton to zooplankton to fish model.
However, many systems (i.e., Lavaca Bay, TX) are dominated by a sediment porewater to benthic
invertebrates to fish model, which means that sediment issues (methyl concentrations, methylation depth
profiles, redox condition, seasonally, etc.) loom way more important that water column concentrations.
A-26
Methylmercury Water Quality Criterion 1/3/01
-------
James Hurley
First and foremost, the development of a national AWQC for methylmercury must be based on
sound data with strict quality control/quality assurance to ensure that the calculation of bioaccumulation
factors (BAFs) is scientifically valid. This is a difficult task when conducting literature searches for data
that form the backbone of the report. Among the data chosen, methods must be comparable to allow
transferability. Individual investigators also apply different definitions of biological assemblages and
food chain pathways. This makes the task of synthesizing appropriate data a difficult task at best.
My overall concern with data used for determination of the national B AF is that not one study from
which data was obtained for this report was actually with the specific purpose of generating MeHg-based
BAFs through all trophic levels. I fully understand that EPA also recognizes this problem and commend
them for assembling the data presented. However, I do think that EPA should consider a research effort
designed to produce results directly related to their MeHg BAF goals. This would ensure that sample
types and methodologies were consistent with the overall goal of development of national BAFs for
methylmercury. Development of a scientifically sound BAF is a critical step in development of a
management plan for this Level I contaminant in the U.S.
In addition to developing a field effort, EPA should also consider development of dedicated
laboratory studies that address Hg and MeHg partitioning and transport in trophic levels 1 and 2.
Although EPA decided to choose an approach that incorporates field-derived BAFs, laboratory studies
using cultures of phytoplankton and zooplankton, coupled with key contrasting water chemistries, would
certainly aid in reducing the variability that is inherent hi using field-derived data on partitioning.
Results of these studies alone would avoid the ambiguity that is inherent in using the terms "seston" and
"phytoplankton" interchangeably for BCFs.
The current report divides the data into two environments (lentic and lotic) but then combines
BAFs to determine a national BAF in the final section of the report. I strongly encourage EPA to
establish a series of National BAFs that are watershed-type based, in slightly more detail than a simple
lentic/lotic division. Data from lotic systems in the report combine wetlands with flowing rivers. As a
result, the lotic grouping contains high dissolved organic carbon (DOC) systems such as wetlands, with
low DOC headwater streams. This type of grouping of sites with such disparate Hg-cycling
environments most likely accounts for both the spread of data for directly-calculated BCFs and the lack
of agreement between directly calculated and converted BCFs depicted in Figure 5-2.
Methylmercury Water Quality Criterion 1/3/01
A-27
-------
While I agree that translators are appropriate in some instances, they too should be calculated on a
more site-specific basis. Use of the translators to calculate the fraction (fj) of total Hg as MeHg should
be refined to address factors such as trophic state and watershed type. The grand mean of 3.2% for this
translator encompasses a range from 0.2% to 13.9% in lake waters. Similarly, the grand mean from
rivers of 1.4% encompasses a range from 0.2 to 5.11% in rivers. Better grouping of the data would
reduce variability for this data set. For instance, Kd's for several contaminants have been shown to
decrease with increasing DOC. The processes controlling methylation and particle partitioning are site-
specific, and the current report attempts to define complex chemical and biological processes across
gradients by the use of a simple fraction. Since this factor (the amount of inorganic Hg that is converted
to the bioaccumulative methyl form) is perhaps the most critical step in developing a BCF, a simple
default conversion factor is not the best approach.
Finally, development of an acceptable model is mentioned within the report as a future goal, but I
feel that model development and acceptance should be fast-tracked along with development of a National
MeHg BAF. Models, such as the recent revisions of the Mercury Cycling Model (MCM), that
incorporate processes such as methylation, aquatic speciation, and bioenergetics are keys to validation of
the BAFs among contrasting sites. Having worked specifically with the MCM Model, I am confident
that is has been tested on a number of contrasting environments (northern Wisconsin lakes, Everglades,
Great Lakes) and could be used to validate BAFs for differing aquatic environments.
David Krabbenhoft
Overall, I found the document to be in very good order structurally, grammatically, and was of an
appropriate length for the subject matter; my compliments to the authors. A quality manuscript makes
the reviewer's job much easier, and a better technical review results when he or she is not "put off for
having to do editorial service too. I heartily support the U.S. EPA's decision to pursue changes to the
AWQC for mercury and have methylmercury (MeHg) be the basis for such regulations. Although this
has been a long time in coming, I do recognize that the peer reviewed data for this type of proposed
change has been limited to just a few study locations until the past few years. That being said, however, I
have serious reservations as to whether enough high quality data has been made available by the
scientific community for the EPA to make an important decision like assigning "National BAF's". The
authors of this report have largely done an admirable job with what is available, but it may be slightly
ahead of its time. It may be that with the very recent release of the National Academy of Sciences report
on human health and mercury, and the proposed decision time line of the EPA to enact emissions
A-28
Methylmercury Water Quality Criterion 1/3/01
-------
regulations in the 5-year time frame, that a well-conducted, national-synoptic study to for the proper basis
for a MeHg BAF's is in order.
David Masckwitz/Edward Swain
1. An update of the mercury bioaccumulation factor (BAF) is very much needed, for the reasons cited
on page 1 of Section I. The new analytical methods that can measure ambient mercury in water at
sub-nanogram per liter levels, and the large number of recent studies that provide field measured
BAF data make the determination of a new BAF a necessity, if EPA plans to update the human
health-based mercury criterion. The BCFs/BAFs used in previous EPA mercury criteria are clearly
outdated. A new mercury BAF and criterion will be a great help to states and tribes (hereinafter,
state). The determination of a BAF is often the biggest road block to the calculation of a human
health-based water quality standard for state regulatory agencies.
2. The following comments are on National Bioaccumulation Factors for Methylmercury (Section I)
and Default Chemical Translator for Mercury and Methylmercury (Section IT). We have not
reviewed for comment the background document.
3. The overall organization of Sections I and n, is logical, straight forward and easy to follow.
4. The EPA search for both available published and unpublished BAF data uncovered a substantial
amount of new information; and, short of carrying out an independent literature search to confirm
this comment, it should be reasonably complete and current.
5. The discussion of uncertainty associated with the final recommended BAFs (beginning on page 73,
Section I), including a discussion of the limitations associated with reducing highly variable BAF
data to a single national BAF (for each trophic level), and the myriad of variables that can affect
BAFs, is appropriate. Further, EPA's rationale that, in spite of the uncertainty (actually, because of
it), the recommendation of a single default BAF for each trophic level is valid. The
recommendation that states should use local BAF data is good as well, but EPA must realize that
local BAF data is not likely to be available in many situations. Thus, the default BAFs will get
substantial use.
Methylmercury Water Quality Criterion 1/3/01
A-29
-------
6. The decision to use only the preferred, field measured, BAF data (including the converted direct
BAFs) and not use the indirectly determined BAFs (BCFs or BAFs times a FCM or BFM) is
appropriate given the quality and quantity of the former. This is consistent with the proposed new
EPA human health criteria methodology (EPA, 1998). However, including the comparison of
direct and indirect BAFs in Section I (Tables 3-10 and 4-11) is valuable information.
7. To eliminate any uncertainty about the proper application of the translators listed in Tables 5-1 and
5-2, it is suggested that EPA include in Tables 5-3 through 5-10 columns showing the translators
used in the conversions, and/or a column showing the "raw" as well as the converted BAFs. An
alternative to expanding these tables is to add to the summary information at the beginning of each
subsection (i.e., Variable, Definition, Estimate, Distribution) a section on "Translators" or
"Conversion" that shows the translator(s) and conversion calculations (this option assumes the
translators used and all the conversion calculations are the same for all the individual BCFs/BAFs).
A third, but less desirable alternative, is to provide example calculations in the introductory
discussion of converted methylmercury BAFs, beginning on page 49 of Section I.
8. Overall, we believe the final recommended BAFs (Table 5-15) are supported and a reasonable
conclusion of the data analysis.
9. The introduction to Section n (page 1) talks about EPA's policy to use dissolved analyses for trace
metals to measure compliance with the standard. This policy was developed in the context of the
toxicity of particulate and chemically bound, versus the toxicity of "dissolved" or ionic forms, of
trace metals to aquatic life. The science behind EPA's dissolved metal policy may not be as
relevant to a highly bioaccumulative metal like mercury, for which the concern is the methyl form,
and the risk is to human health through fish consumption rather than to aquatic life directly. EPA
should expand this section to discuss if and how mercury differs from non-bioaccumulative trace
metals with regard to the need or desirability of measuring dissolved metal in water.
10. EPA discusses in the "Background" part of Section n, total to dissolved metal conversion factors.
Along the lines of comment number nine, the conversion factor of 0.85 for the current mercury
criteria (CMC and CCC) are applicable to toxicity-based mercury criteria, not the human health-
based chronic criterion (Federal Register 63: 68354-68364). The conversion factor for the chronic
human health-based mercury criterion is 1.0 (see also Federal Register 60: 15392). EPA needs to
A-30
Methylmercury Water Quality Criterion 1/3/01
-------
revise their discussion of conversion factors to reflect the conversion factor for the human health
criterion, and to address the points made hi comment number nine.
11. Separate average translators and KD values for lakes, rivers and estuaries as derived in Section n
seem to be reasonable and supported by the data presented.
Darell Slotton
I found the reports to be clear in their intent and in their explanation of approaches used. I
especially appreciated the straightforward acknowledgment of the myriad sources of uncertainty and
variability. My overall response to the entire exercise is that those sources of uncertainty and variability
(geographic, water quality, water trophic status, analytical, individual organism, true trophic "level",
food web complexity, etc.) make this a very difficult if not impossible proposition. I strongly support the
development of tissue-based mercury criteria as the preferred mechanism for addressing mercury risk
assessment and regulatory concerns throughout the huge range of aquatic systems affected. That said, if
EPA has a legal charge to also develop the best predictive relationships it can as defaults, etc., the
approach being used is probably as good as can be expected. It may be significantly more useful as a
regional tool, though (e.g., northern midwestern lake systems, California rivers, Florida, etc). A truly
applicable, nation-wide set of factors may be unattainable. I strongly concur with the suggestion that
site-specific research is preferable in the event that BAFs are to be used.
Methylmercury Water Quality Criterion 1/3/01
A-31
-------
ATTACHMENT B: DERIVATION AND CALCULATION OF
'TSEUDO" KDS FOR METHYLMERCURY
Derivation
MeHgd + TSS ^ Hgp, including MeHgp
"Pseudo"KDMeHg / Hg =
HgP
MeHgo.TSS
[Eqn. A.1]
MeHgp
Equating TSS and combining Eqn. A.l. and Eqn. A.2 yields:
MeHgd MeHgd
Hg / Hg • — = KDMeHg •
Eg,
MeHgf
Rearranging:
HgP MeHgd
"Pseudo"KDMeHg/Hg= „ Tr *KMeHg« „ TT [Eqn.A.3]
MeHgd
MeHgp
Example Calculation for Lakes (see text of original draft report for source of data)
= 338,844
When HgT = 1, MeHgd = 0.032, Hgd = 0.60 and therefore Hgp = 0.40
and the ratio Hgp/MeHgd = 0.40/0.032 = 12.5
When MeHg = 1, MeHgd = 0.613, and therefore MeHgp= 0.387
and the ratio MeHgd/MeHgp = 0.613/0.387 = 1.58
Substituting the above values in Eqn. A.3 gives:
"Pseudo" KcMeHg / Hg = 12.5 • 338,844 • 1.58 = 6,692,169
Log "Pseudo" KcMeHg / Hg = 6.83
A-32
Methylmercury Water Quality Criterion 1/3/01
-------
Example Calculation for Rivers (see text of draft report for source of data)
= 64,565
When HgT = 1 , MeHgd = 0.014, Hgd = 0.37 and therefore Hgp = 0.63
and the ratio Hgj/MeHgd = 0.63/0.014 = 45.0
When MeHg = 1 , MeHgd = 0.49, and therefore MeHgp= 0.51
and the ratio MeHg,j/MeHgp = 0.49/0.51 = 0.96
Substituting the above values in Eqn. A.3 gives:
"Pseudo" KcMeHg / Hg = 45.0 • 64,565 • 0.96 = 2,789,208
Log "Pseudo" KoMeHg / Hg = 6.44
Methylmercury Water Quality Criterion 1/3/01
A-33
-------
------- |