Blood Methylmercury and Fish Consumption
Among People of Childbearing Age in the
General U.S. Population
NHANES 1999-March 2020
EPA 820-R-24-011
-------
Disclaimer
The EPA does not endorse any products or commercial services mentioned in this publication.
This document does not impose legally binding requirements on the EPA, states, authorized
tribes, other regulatory authorities, or the regulated community.
Acknowledgments
Technical support for this report was provided by Westat under EPA contracts EP-C-10-023, EP-D-
12-050, and 68HERD19D001. We would like to thank Rebecca Jeffries Birch, Karen Delia Torre, Yan
Zhuang, Angela Chen, and Xiaoshu Zhu for their capable input.
-------
EXECUTIVE SUMMARY
In the United States, exposure to methyl mercury (MeHg) in humans occurs largely through the
consumption offish (National Research Council, 2000; Rice et a/., 2000). Blood concentrations of
MeHg in women of childbearing age are of particular interest because exposure to MeHg in utero
is associated with adverse health effects, e.g., neuro-developmental deficits such as IQ and motor
function deficits in children (Mergler et a/., 2007; National Research Council, 2000).
This report documents an analysis offish consumption and blood mercury concentrations in
women aged 16-49 years in the United States using data collected by the National Health and
Nutrition Examination Survey (NHANES) from 1999 through March 2020 (10 survey releases).
NHANES is a continuous survey designed to collect data on the health and nutritional status of
the U.S. population. The NHANES reports data on chemicals, or their metabolites as measured in
blood and urine samples collected from a statistically representative sample of the U.S.
population. CDC releases NHANES data every two years.
In this study, we applied fish tissue mercury concentrations to fish species reported being
consumed by the study participants, and estimated the usual intake of mercury through fish
consumption (United States Environmental Protection Agency, 2014). We imputed blood MeHg
from blood total and blood inorganic mercury data and investigated the trend of blood MeHg
concentration over time by demographic characteristics. Additionally, we looked the association
between blood mercury concentration and demographic characteristics and the association
between blood mercury concentration and estimated usual intake of mercury. We also looked for
trends in frequency offish consumption as well as the association offish consumption and
mercury intake with demographic characteristics. Finally, we looked for geographic differences in
blood mercury.
ES-1
-------
Kev Findings
Trends in MeHg concentrations in blood: There are statistically significant decreasing trends in
blood MeHg concentrations over time (higher in 1999-2000 and lowest in 2017-March 2020) and
by demographic characteristics (such as education, income, race/ethnicity, and age over the study
period).
• The geometric mean blood MeHg concentration in 1999-2000 survey release is 1.94 times
higher than the geometric mean in 2017-March 2020 survey release, representing a 48
percent decrease between N HAN ES1999-2000 (the earliest set of data analyzed) and
2017-March 2020 (the most recent dataset).
• The percentage of women of childbearing age with blood MeHg concentrations over 5.8
[jg/L in 1999-2000 is about 3.5 times the concentration found in 2001-March 2020,
representing a 71 percent decrease. There is a statistically significant difference between
the survey releases for the percent with blood MeHg concentrations over 5.8 [jg/L (Rao-
Scott Chi-square p<0.001). No significant difference (p=0.12) is found between the survey
releases after removing survey release 1999-2000.
• The geometric mean blood MeHg concentration in women who reported their race as
"Other Race - Including Multi-Racial" (which includes Asian, Native American, Pacific and
Caribbean Islander, Alaska Native, multiracial, and unknown race) in 1999-2004 is 1.82
times higher than the geometric mean in 2017-March 2020 data, representing a 45
percent decrease between N HAN ES 1999-2004 and 2017-March 2020.
• Higher blood MeHg concentrations are observed with increasing age, ratio of family
income to poverty, education level, and among participants who reported their race as
"Other Race - Including Multi-Racial."
Predictors of MeHg concentrations in blood: T ransformed usual intake (TUI) is a significant
predictor (p<0.0001) of blood MeHg concentrations. The rate of increase in blood MeHg
concentration due to TUI varied by education, race/ethnicity, and log-transformed body weight.
Other significant predictors of blood MeHg concentrations are NHANES survey release, education,
race/ethnicity, log-transformed hematocrit concentration, and log-transformed bodyweight.
Household income is a marginally significant predictor of blood MeHg concentrations (p=0.054).
Trends in fish consumption by demographic characteristics and geography: Blood MeHg
concentrations are positively associated with the reported frequency offish consumption over the
previous 30 days. While significant differences in reported frequency offish consumption are
found across the six NHANES survey releases from 1999-2010 in a previous study (United States
ES-2
-------
Environmental Protection Agency, 2013), there are no statistically significant differences in
reported frequency offish consumption across the more recent NHANES survey releases in the
current study (2013 to March 2020).
There are statistically significant differences in blood MeHg concentrations geographically, with
higher levels among residents of coastal counties compared to residents of non-coastal counties.
The residents of the Northeast region have the highest levels of the four regions, followed by the
West, South, and the Midwest.
Overarching Interpretation
There is a decreasing trend in the geometric mean of blood MeHg concentrations. While the
estimated amounts of total fish eaten over the previous 30 days in NHANES 2013-March 2020 are
at the higher end of estimated amounts for NHANES 1999-2010, the estimated mercury intake
from total fish consumption over the previous 30 days in NHANES 2013-March 2020 are lower than
estimated intakes for NHANES 1999-2010. This suggests that women of childbearing age are
potentially choosing to eat fish that tend to have lower mercury concentrations leading to lower
estimates of mercury intake per unit body weight.
ES-3
-------
CONTENTS
LIST OF ABBREVIATIONS iv
1 BACKGROUND AND PURPOSE 1
2 DATA AND METHODS 4
2.1 NHANES and Methods Overview 4
2.2 NHANES Data 4
2.2.1 Blood Mercury Data 6
2.2.2 Dietary Recall Data 7
2.3 Fish Tissue Mercury Data 8
2.4 Statistical Methods 9
2A1 Usual Intake of Fish and Mercury 11
2A2 Imputation 13
2.5 Estimation of 30-Day Fish Consumption and Mercury Intake 14
3 RESULTS 16
3.1 Blood MeHg Summary Statistics 16
3.1.1 Time Trends in Blood Mercury Concentrations 16
3.1.2 Demographic Distributions 18
3.2 Blood MeHg Modeling 19
3.3 Trends in Fish Consumption 23
3.3.1 Trends in Frequency of Consumption 24
3.3.2 Trends in Estimated Amounts Consumed Over the Previous 30 days 28
3.3.3 Association Between Fish Consumption Frequency and Blood Mercury 29
3.3.4 Relationship Between Fish Consumption and Intake of Mercury with Demographic Factors 31
4 DISCUSSION and CONCLUSIONS 39
5 QUALITY CONTROL/QUALITY ASSURANCE 42
6 REFERENCES 44
APPENDIX A Extended Data Tables A-1
APPENDIX B Geographic Distribution of Blood MeHg in General U.S. Population Using NHANES 2009-2012 B-1
-------
List of Tables
Table 1. Survey weights and adjustment factor for combining multiple NHANES cycles 5
Table 2. Sample size and weighted percent of results below the limit of detection, by survey release 7
Table 3. Distribution of blood MeHg concentration (pg/L), by survey release 17
Tabled. Weighted percent of women 16 and 49 years with blood MeHg >5.8 pg/Land blood THg >5.8 pg/L, by survey
release 18
Table 5. Comparison of Imputed Blood MeHg concentration (pg/L) overtime, by demographic characteristics,
women aged 16-49 years 19
List of Figures
Figure 1. Distribution of log-transformed blood MeHg (pg/L), by NHANES survey releases, women aged
16-49 years 17
Figure 2. Relative blood MeHg concentrations with 95 percent confidence intervals, by demographic
characteristics (NHANES 2013-March 2020) 20
Figure 3. Slope parameter relating transformed usual fish intake of mercury and log-transformed blood MeHg
concentrations, overall and by demographic group, with 95 percent confidence intervals 22
Figure k. Estimated blood MeHg given usual intake offish mercury, by race/ethnicity 23
Figure 5a. Weighted percent of participants by 30-day fish consumption frequency, by NHANES survey
release (2013-March 2020), women aged 16-49 years 25
Figure 5b. Weighted percent of participants by 30-day fish consumption frequency, by NHANES survey
release(1999-2010), women aged 16-49 years 25
Figure 6a. Weighted percent of participants by 30-day total fish consumption frequency, by demographic
characteristics, women aged 16-49 years, NHANES 2013-March 2020 27
Figure 6b. Weighted percent of participants by 30-day total fish consumption frequency, by demographic
characteristics, women aged 16-49 years, NHANES 1999 - 2010 27
Figure 7a. Mean blood MeHg concentrations by reported frequency of total fish consumption in 30 days, women
aged 16-49 years, NHANES 2013-March 2020 (with 95% confidence intervals, median, and 90th percentile) 30
Figure 7b. Mean blood MeHg concentrations by reported frequency of total fish consumption in 30 days, women
aged 16-49 years, NHANES 1999-2010 (with 95% confidence intervals, median, and 90th percentile) 30
Figure 8. Relative ratios and 95 percent confidence intervals from models predicting fish consumption
and mercury intake variable by education, NHANES 2013-March 2020 32
Figure 9a. Relative ratios and 95 percent confidence intervals from models predicting fish consumption and
mercury intake variables by race/ethnicity, NHANES 2013-March 2020 33
ii
-------
Figure 9b. Relative ratios and 95 percent confidence intervals from models predicting fish consumption and
mercury intake variables by race/ethnicity, NHANES1999-2010 34
Figure 10. Relative ratios and 95 percent confidence intervals from models predicting fish consumption and
mercury intake variables by income, NHANES 2013-March 2020 36
Figure 11a. Relative ratios and 95 percent confidence intervals from models predicting fish consumption and
mercury intake variables by age group, NHANES 2013-March 2020 37
Figure lib. Relative ratios and 95 percent confidence intervals from models predicting fish consumption and
mercury intake variables by age group, NHANES 1999-2010 38
iii
-------
LIST OF ABBREVIATIONS
CDC
Centers for Disease Control and Prevention
DK
Don't know
DL
Detection limit
EPA
Environmental Protection Agency
FDA
Food and Drug Administration
FNDDS
Food and Nutrient Database for Dietary Studies
IHg
Inorganic mercury
MeHg
Methylmercury
NCHS
National Center for Health Studies
NCI
National Cancer Institute
NHANES
National Health and Nutrition Examination Survey
RR
Relative Ratio
THg
Total mercury
TUI
Transformed usual intake
USDA
United States Department of Agriculture
iv
-------
BACKGROUND AND PURPOSE
In the United States, exposure to methyl mercury (MeHg) in humans occurs largely through the
consumption offish (National Research Council, 2000; Rice et al., 2000). Mercury released into the
environment is converted to MeHg in sediments and in the water column and bioaccumulates
through aquatic food webs. This bioaccumulation leads to increased levels of MeHg in larger,
older, predatory fish; concentrations in fish tissue may exceed a million-fold the concentrations in
water (National Research Council, 2000). Fish and shellfish tissue contaminated by MeHg can put
human health at risk. Blood total mercury (THg) concentrations reflect exposure to organic
mercury, predominantly MeHg, from consumption offish (Bjornberg et al., 2003; Sanzo et al., 2001;
Svensson et al., 1992). MeHg exposure in utero is associated with adverse health effects, e.g.,
neurodevelopmental deficits such as IQ and motor function deficits in children (Mergler et al.,
2007; National Research Council, 2000).
In October 2021, the EPA and FDA issued Advice about Eating Fish and Shellfish (EP/^FDA, 2021),
which updated the consumer advice on mercury in fish originally issued in 2001. This update was
due, in part, to research over the past decade that has indicated that fish consumption during
pregnancy may be beneficial for the growth and brain development of the fetus and young
children (Bramante et al., 2018; Golding et al., 2016; Stratakis et al., 2020; Taylor et al., 2016). The
advice provides recommendations for pregnant and breastfeeding women, women of
childbearing age, and young children, and includes a chart with fish species considered "Best
Choices," "Good Choices," and "Choices to Avoid," with recommended servings per week.
In the FY2022-2026 EPA Strategic Plan, the EPA has committed in Goal 5 to "Ensure Clean and
Safe Water for All Communities" and specifically under Objective 5.2 to "Protect and Restore
Waterbodies and Watersheds, to address sources of water pollution and ensure water quality
standards are protective of health and needs of all people and ecosystems." The EPA has
identified several strategies it will undertake to help protect public health and ensure clean and
safe waters that include developing nationally recommended water quality criteria and
addressing contaminants that endanger human health. The EPA's approach to making fish safe
to eat, which is a human health benefit, has been to:
• Work collaboratively with air agencies to maintain and improve the nation's air
quality.
• Encourage development of statewide mercury reduction strategies.
• Reduce air deposition of mercury.
1
-------
• Improve public information and notification offish contamination risks.
The Agency can assess progress towards this goal through the measurement of blood mercury
concentrations among women of childbearing age as reported by the Centers for Disease Control
and Prevention's (CDC) National Health and Nutrition Examination Survey (NHANES). NHANES is
a continuous survey designed to collect data on the health and nutritional status of the U.S.
population. The NHANES reports data on chemicals, or their metabolites as measured in blood
and urine samples collected from a statistically representative sample of the U.S. population. CDC
releases NHANES data every two years and reports environmental exposure results for every
NHANES release in the National Report on Human Exposure to Environmental Chemicals (Center
for Disease Control and Prevention, 2021a, b).
A 2013 study on blood mercury trends in women of childbearing age from NHANES 1999-2010
(Birch et al., 2014; United States Environmental Protection Agency, 2013) found blood mercury
concentrations in NHANES survey release 1999-2000 to be statistically significantly higher than
the mean of the subsequent 10 releases (2001-2010) for both blood THg and blood MeHg.The EPA
reference dose (RfD) for MeHg is 0.1 [jg/kg-day (United States Environmental Protection Agency,
2001). This is equivalent to a blood mercury concentration of 5.8 |jg/L. An RfD is an estimate of the
maximum daily intake that is not likely to cause harmful effects across a lifetime. From 2008 to
2018, the EPA used the percent of women of childbearing age that have blood mercury
concentrations over 5.8 |jg/L as one measure of the progress towards making fish and shellfish
safer to eat. In the study of NHANES 1999-2010, the percentage of women of childbearing age with
blood MeHg >5.8 [jg/L was significantly higher in survey release 1999-2000. The study also found a
significant quadratic trend1 in blood MeHg concentration since 1999-2000. This quadratic trend
indicates decreasing blood MeHg concentrations between NHANES survey release 2001-2002 and
2003-2004, followed by relatively small changes and a slight increase in the survey release 2009-
2010. There was a significant relationship between mercury intake from fish consumption and
blood MeHg, although mercury intake did not fully explain the differences observed across the
survey releases.
A2009 study (Mahaffey et al., 2009) that investigated regional and coastal differences in NHANES
1999-2004 blood mercury data found that elevated blood mercury occurred more commonly in
women living in coastal areas of the United States and that exposure varied regionally with those
1A non-linear trend described by a second-order polynomial function.
2
-------
residing in the Northeast having the highest blood mercury concentrations followed by the South,
West, and Midwest.
This study focuses on NHANES1999-March 2020 data with the goal to investigate national trends
in blood mercury concentrations and fish consumption among women of 16-49 years of age. The
specific objectives are to assess:
1. Trends in blood mercury concentrations over time and by demographic
characteristics.
1. Association between blood mercury concentration and demographic characteristics.
2. Association between estimated usual intake of mercury and blood mercury
concentration.
3. Trends in frequency and amounts offish consumed and the association offish
consumption and mercury intake with demographic characteristics.
4. Geographic distribution of blood MeHg.
3
-------
DATA AND METHODS
2
2.1 NHANES and Methods Overview
NHANES is designed to assess the health and nutritional status of adults and children in the
United States. It is conducted by the National Center for Health Statistics (Center for Disease
Control and Prevention, 2013), part of the CDC that is responsible for producing vital and health
statistics for the United States. NHANES collects health-related data from a nationally
representative sample of about 5,000 non-institutionalized individuals located in 15 counties in the
United States each year and releases the data on two-year cycles. The survey includes interview
and examination components. The interview includes demographic, socioeconomic, dietary, and
health-related questions. The examination consists of medical, dental, and physiological
measurements and laboratory tests of blood and urine. Data from both components of all
NHANES releases 1999-March 2020 were used to investigate the trend of blood mercury
concentration across NHANES releases by demographic categories. This analysis focused on data
from the three most recent NHANES survey releases (2013-2014,2015-2016, and 2017-March 20202)
to assess fish consumption and compare results with the NHANES 1999-2010 study on trends in
fish consumption (Birch et al., 2014; United States Environmental Protection Agency, 2013).
2.2 NHANES Data
The required NHANES data files and variables were identified and downloaded from the NHANES
website. These files were merged to create a dataset customized to the needs of this project. For
each NHANES survey release, the study data include:
• Demographics: Characteristics previously shown to be related to blood mercury
and/or fish consumption (gender, age, race/ethnicity, education, and annual income),
and sampling weights, pseudo-stratum, and pseudo-primary sampling unit (PSU)
variables. The pseudo-stratum and pseudo-PSU variables provide information on how
participants were selected and are needed to calculate standard errors and p-values.
2 Due to the COVID-19 pandemic, field operation and data collection for NHANES 2019-2020 were
suspended in March 2020. The 2019-March 2020 data were combined with the full data set from
NHANES 2017-2018 to create a nationally representative 2017-March 2020 data file.
4
-------
They are modified from the actual N HAN ES strata and PSUs for disclosure control and
are thus prefixed "pseudo."
• Survey weights: Appropriate survey weights of each NHANES release are selected for
this analysis (Table 11.
Table 1. Survey weights and adjustment factor for combining multiple NHANES cycles
Survey release
Survey weight
Adjusted weight factor
1999-2000
WtMec4Yr*2
2/21.2
2001-2002
WtMec4Yr*2
2/21.2
2003-2004
WtMEC2Yr
2/21.2
2005-2006
WtMEC2Yr
2/21.2
2007-2008
WtMEC2Yr
2/21.2
2009-2010
WtMEC2Yr
2/21.2
2011-2012
WtMEC2Yr
2/21.2
2013-2014
WTSH2YR
2/21.2
2015-2016
WTSH2YR
2/21.2
2017-March 2020
WTMECPRP
3.2/21.2
Data processing and analyses were performed using the Statistical Analysis System (SAS) version
9.4 (SAS Institute, 2016). In general, two-year Mobile Examination Center (MEC) exam weights
(WTMEC2YR) were used for this analysis. The two-year MEC exam weights for NHANES 1999-2000
and 2001-2002 were based on a different population and hence not comparable. An adjusted
four-year MEC survey weight (WTMEC4YR) was created to account for the two different reference
populations (Johnson et al., 2013).
In the NHANES 2013-2014and 2015-2016, blood mercury was analyzed in one-half of the 12 years
and older population.3 Special sample weights (WTSH2YR) were created for the subsample to
account for the additional probability of selection into the subsample, as well as the additional
nonresponse to the lab tests results.
• MEC exam weights (WTMECPRP) created for combined NHANES 2017-2018 and 2019-
March 2020 data were used for NHANES 2017-March 2020 data.
3 NHANES collected data for individuals aged 12 and older. This study uses a subset of the data for
women aged 16 to 49.
5
-------
• In addition, the NHANES2017-March 2020 data covered 3.2years compared with 2
years for other N HAN ES data files. Because the period differs from earlier cycles, the
survey weights were adjusted when 2017-March 2020 data files were combined with
other two-year cycles (Akinbami et al., 2022). In this study, data from NHANES cycles
from 1999-2000 through 2017-March 2020 covered 21.2 years of data collection. For
each two-year cycle, the adjusted factor was 2/21.2. And the survey weights for
2017-March 2020 data were adjusted using a factor of 3.2/21.2.
• Laboratory results:
— Blood total mercury (THg) and inorganic mercury (IHg) concentrations.
Hematocrit values, related to blood mercury in that mercury binds to the red
blood cells.
• Body measures: Body weight is related to fish consumption and blood mercury.
• Dietary intake, 24-hour recall: Data necessary to estimate usual intake offish (food
codes, meal name, amount eaten; one record per food item eaten). Usual fish intake is
the long-term average intake of raw finfish and shellfish (from marine, estuarine, and
fresh waters).
Dietary intake, 30 day frequency of consumption: Data used to estimate usual intake offish
(number of times participants reported consuming fish in previous 30 days, calculated from
reports for the following species as collected by NHANES-clams, crabs, crayfish, lobster, mussels,
oysters, scallops, shrimp, other shellfish, other unknown shellfish, breaded fish products, tuna,
bass, catfish, cod, flatfish, haddock, mackerel, perch, pike, pollock, porgy, salmon, sardines, sea
bass, shark, swordfish, trout, walleye, other fish, and other unknown fish).
2.2.1 Blood Mercury Data
The laboratory data files from all survey releases contain measurements of THg and IHg in blood.
Laboratory methods are available on the N HAN ES website. Table 2 presents the total sample size
(N) by survey release, along with the detection limit (DL), the percentage of concentrations
observed below the DL, and the standard error of the percentage for blood THg and IHg.
6
-------
Table 2. Sample size and weighted percent of results below the limit of detection,
Survey release
N
DL
THg
Weighted
percent 20 runs of blood blank)
2.2.2 Dietary Recall Data
NHANES conducts two 24-hour dietary recalls with participants. The first isan in-person interview
and occurs during the examination and the second is a telephone interview 3 to 10 days later.
Participants report every food and drink item they consumed in the 24 hours before the interview,
including the amount. The food items are coded using U.S. Department of Agriculture (USDA)
food codes from the Food and Nutrient Database for Dietary Studies (FNDDS).The FNDDS files
are available from the Agriculture Research Service of the USDA (FNDDS DOWNLOAD
DATABASES: USDAARS1. The recipes for the food codes were searched to find all food codes that
contain finfish or shellfish, including mixed dishes. All records in the 24-hour data file for women
aged 16-49 years that were for fish-containing food codes were extracted. The recipe file and 24-
hour recall data were merged to calculate quantity of raw fish consumed per recipe. A detailed
description of how the fish-containing records were extracted can be found in the 2014 U.S. EPA
report, Estimated Fish Consumption Rates for the U.S. Population and Selected Subpopulations
(NHANES2003-2010) (United States Environmental Protection Agency, 2014).
Additionally, participants are asked to report the number of times they consumed various types of
finfish and shellfish over the past 30 days. Responses to these questions are combined to derive a
7
-------
variable that provides the number of times each respondent consumes any seafood in the past 30
days (frequency of seafood consumption).
2.3 Fish Tissue Mercury Data
To estimate mercury intake, data on mercury concentrations in fish tissue are needed. Mercury
concentration in fish varies greatly among species and within species, with older and larger fish
having higher concentrations (United States Environmental Protection Agency, 1997). We
updated our database offish tissue mercury concentrations used in the report Trends in Blood
Mercury Concentrations and Fish Consumption Among U.S. Women of Childbearina Aae
NHANES. 1999-2010 (United States Environmental Protection Agency, 2013) with data through
2020. This addition included data from the peer reviewed articles (Imanse et a/., 2022; Janssen et
a/., 2021; Janssen et a/., 2019; Malinowski, 2019; Melnyk et a/., 2021; Sackett et a/., 2017; Taylor and
Calabrese, 2018; Whitney, 2021; Wolff et a/., 2016). Fish tissue mercury concentrations for fish
caught between 2011 and 2020 were also extracted from the following databases and reports
provided by federal and state governments:
• 2013-2014 National Rivers and Streams Assessment Fish Tissue Study (United States
Environmental Protection Agency, 2023).
• Assessment of mercury sources in Alaskan lake food webs: U.S. Geological Survey data
release (Lepak, 2022).
• Total Mercury Concentrations in Smallmouth Bass from Chesapeake Bay Tributaries, USA
Dataset, 2013-2017 (Willacker, 2020).
• Fish mercury concentration data and ancillary data for streams and rivers across New York
States (United States), 1969-2016, including environmental characteristics of selected
locations sampled during 2007-16 (Murray, 2020).
• Hg Concentrations of Fish Tissue Samples in the Vicinity of Yellow Pine, Idaho (McGee,
2020).
• Mercury Contaminant Levels in Louisiana Biota (Louisiana Department of Environmental
Quality, 2016-2020).
• Selenium and mercury in the Kootenai River, Montana and Idaho, 2018-2019: U.S.
Geological Survey data release (Mebane, 2019).
• Status and trends of mercury in fish tissue in New Hampshire waterbodies, 1992-2016 (New
Hampshire Department of Environmental Services, 2018).
8
-------
• Measuring Mercury Trends in Freshwater Fish in Washington State (Washington State
Department of Ecology, 2011-2016).
• 2015 Great Lakes Human Health Fish Fillet Tissue Study (United States Environmental
Protection Agency 2015).
Additional data included fish tissue mercury concentrations for the following species that were
used for the fish tissue mercury database developed for this analysis: bass, perch, mackerel, carp,
catfish, haddock, trout, rockfish, flatfish, sea bass, herring, pompano, pike, salmon, porgy, and
whitefish.
To estimate the geometric mean mercury concentration for each fish species, we used the SAS
MIXED procedure and modeled the log-transformed fish tissue mercury concentration by fish
species, treating the data source as a random effect. Some of the data sources reported average
concentrations for multiple fish samples and some sources reported mercury concentrations for
each individual fish sampled. In order to account for this in the model, we included a weighting
factor. The weighting factor allowed the model to account for different variances due to both data
source and number of individual fish samples contributing to each reported value, which
modeled the error variance as a power function of the number of samples averaged to obtain the
reported value. The predicted values were converted to geometric mean fish mercury
concentrations. The average mercury concentration weighted by 30-day consumption frequency
was used for fish not specified in the dietary recall food recipe code. To the extent it could be
tested, there were no consistent time trend in the fish mercury concentration data in the sources
that we used. Table A-l provides the microgram (fjg) of mercury per gram offish by species group
used in the analysis.
2.4 Statistical Methods
The relationship between blood mercury concentrations, mercury intake, time, and demographic
characteristics were assessed using:
1. Summary statistics based on the imputed concentrations of blood MeHg and the
percentage of participants with blood THg and blood MeHg concentrations over 5.8
mq/l.
2. Comparison of imputed concentration of blood MeHg overtime by demographic
characteristics.
9
-------
3. Estimation of usual intake offish (Tooze et al., 2010; Tooze et al., 2006; United States
Environmental Protection Agency, 2014).
4. Regression calibration and linear regression modeling to predict imputed blood
MeHg concentration from age, race, education, income, usual intake of mercury, and
log-transformed hematocrit concentration (Kipnis et al., 2009).
5. Regression modeling to predict amount offish consumption.
6. Logistic regression to predict the probability of reporting any fish consumption in the
previous 30 days.
7. Regression modeling to predict mercury intake.
For the analysis, we imputed blood MeHg concentration from blood total and inorganic
measurements. Blood THg and IHg measurements below the DL were also imputed. Details of
the estimation of usual intake of mercury and the methodology to model MeHg concentrations
are described in Section 2.4.1. Details of the imputation methodology are discussed in Section
2.4.2.
A box plot and table of blood MeHg concentrations were generated to provide sample sizes,
geometric means, 95 percent confidence intervals of geometric means, and percentiles (25th,
75th, and 90th) by survey release. Geometric means of blood MeHg concentrations and their
standard errors were generated by age group, race/ethnicity, income, and education. A test with p
value less than 0.05 is considered significant. These data were generated by averaging the 20
imputed values for each individual, then calculating the statistics from those values utilizing a SAS
software survey procedure to incorporate the uncertainty due to the survey sample design. Data
from all NHANES survey releases were used to investigate the trend of blood MeHg concentration
over time (i.e., 1999-2004,2005-2010,2011-2016, and 2017-March 2020) (Table 4).
The frequency distribution offish consumption, estimated 30-day fish consumption amount,
estimated 30-day mercury intake, and the estimated 30-day mercury intake per unit body weight
were calculated. Detailed tables of these distributions were generated to provide sample sizes,
arithmetic means, percentiles (25th, 50th, 75th, 90th, and 95th) and their 95 percent confidence
intervals, by survey releases, race/ethnicity, age, income, and education. These tables are in the
Appendix (Tables A-4 to A-81. Plots and analytic extracts based on these tables are presented in
the report.
Data processing and analyses were performed using SAS software, 9.4 (SAS Institute, 2016) and
following the NHANES Analytical Guidelines posted on the NHANES website (Akinbami et al.,
10
-------
2022). All analyses were weighted using the statistical weights recommended in the guidelines as
detailed in Section 2.2, and the sampling design variables were used in calculating the variance of
the estimates.
Age was categorized into four groups: 16-19,20-29,30-39, and 40-49 years. Race/ethnicity groups
recorded in N HAN ES consistently across the 10 survey periods include Mexican American, Other
Hispanic, Non-Hispanic White, Non-Hispanic Black, and "Other Race." "Other Race" consists of
Asian, Native American, Pacific and Caribbean Islander, Alaska Native, multiracial, and unknown
race.
Hematocrit was included in the analysis because approximately 80 percent of MeHg binds to red
blood cells (Clarkson and Magos, 2006; Rothenberg et a/., 2015). This variable was log-transformed
for modeling.
Some demographic information (e.g., family and household income; education for youth 6-19
years) were modified or not included in the 2017-March 2020 public-use data file release due to
potential disclosure risks (Akinbami et a/., 2022).
Family and household income is not included in the 2017-March 2020 release. The ratio of family
income to the federal poverty level is included as for previous cycles. The seven income categories
used for the analysis are based on this ratio: less than one times the ratio, one to less than two
times the ratio, two to less than three times the ratio, three to less than four times the ratio, four to
less than five times the ratio, greater than or equal to five times the ratio, and missing values.
Education level for adults aged 20 and over in the 2017-March 2020 release is included. Education
is categorized as less than, equal to, or greater than the median education level for the
participant's age for all previous NHANES cycles, and adults aged 20 and over for NHANES 2017-
March 2020. Education level for children and youth of 6-19 years was not included in the 2017-
March 2020 release and thus for this study, participants aged 16-19 years are categorized as an
unknown education group.
2.4.1 Usual Intake of Fish and Mercury
The National Cancer Institute (NCI) provides SAS macros (titled MIXTRAN and IN DM NT) to
calculate the distribution of usual intake of dietary components (such as fish and mercury from
11
-------
fish) and for calculating the expected mean of transformed usual intake (TUI) that can be used for
calculating the relationship between usual intake and another dependent variable, such as blood
mercury concentrations. Using fish consumption as an example, the MIXTRAN macro fits the
following models in order to predict the distribution of usual intake across the population:
• A logistic mixed model for predicting the probability of consuming fish in any 24-hour
dietary recall.
• A linear mixed model to predict the reported amount offish consumed in dietary recalls
where fish consumption is reported. The fish consumption is modeled on a transformed
scale using a Box-Cox transformation.
• The usual intake is the product of the probability offish consumption and the amount
consumed.
The output from the MIXTRAN macro can be used as input to the INDIVINT macro to calculate the
expected mean fish consumption for each individual or equivalently the regression calibration
estimate offish consumption that can then be used to predict blood mercury concentration in a
linear regression model.
The NCI macros have some limitations when applied to the NHANES fish consumption data,
including:
• With many predictors, the MIXTRAN macro may take a long time to converge to the final
parameter estimates or may fail to converge.
• When assuming the random effects in the logistic and linear mixed models are correlated,
MIXTRAN may fail to converge when using the NHANES survey weights.
• With the NHANES data, the MIXTRAN macro needs to be run multiple times, once for each
replicate weight created to calculate variance of the parameter estimates.
Due to these limitations, Westat created a modification to the NCI method, hereafter referred to
as the EPA method. The EPA method has the following steps:
1. Fit the weighted logistic model using the SAS SURVEYLOGISTIC procedure, saving
the logit transformed predicted probability of fish consumption (call this LogitXBeta).
2. To estimate the variance components, fit a logistic mixed model similar to that used
by MIXTRAN except the only predictor is LogitXBeta and the intercept is set to zero.
3. Determine a Box-Cox transformation that makes the reported distribution of amount
consumed roughly normally distributed.
4. Fit a weighted linear model to predict the amount offish consumed, when
consumed, saving the predicted amount.
12
-------
5. To estimate the variance components, fit a linear mixed model similar to that used by
MIXTRAN, except the only predictor is the predicted amount from the previous
regression and the intercept and slope are set to one and zero, respectively.
6. Assuming random effects from the logistic and linear models are independent, the
output from previous modeling steps are used as inputs to the NCI INDIVINT macro
to calculate the regression calibration estimate for transformed usual fish
consumption. See the discussion below regarding selection of the transformation to
use.
7. Finally, the regression calibration estimate of usual fish consumption is used to
predict blood mercury concentration using weighted regression.
The steps above were repeated for each replicate weight.
Regression calibration assumes that a linear model is used to predict the dependent variable
(blood mercury concentration) from the transformed usual intake. The INDIVINT macro assumes
a Box-Cox transformation of usual intake is used and requires that a Box-Cox lambda parameter
be specified. Several values of lambda were tried to identify a lambda for which the relationship
between the regression calibration estimates of usual intake were most linearly related to blood
THg measurements. A Box-Cox lambda of 0.70 was selected and used for all analyses. Because a
value of 0.70 might not be optimal for all analyses, the final regression model allowed for the slope
above and below the median usual intake to differ. In all cases the slope difference was not
statistically significant. To accommodate this modification, non-linear regression was used to fit
the final model.
2.4.2 Imputation
The MeHg concentration was calculated from the difference between the THg measurements
and the IHg measurement. However, due to measurement errors, that difference can be negative.
In addition, a few of the THg concentrations were less than the DL and not otherwise specified
and many of the IHg measurements were below the DL.
Predictors we re selected by stepwise selection using SAS GLMSELSECT to predict the
log-transformed blood THg concentration from calculated TUI and selected main effects.
Significant (p<0.05) main effects and interactions of (TUI) and main effects were identified. The
selected predictors were then tested for significance using SAS SURVEYREG.The final set of main
effect variables were race/ethnicity, income, education, NHANES survey release, centered age in
decade, centered log-transformed body weight, centered log-transformed hematocrit, and TUI.
13
-------
The final set of two-way interactions were interactions ofTUI with centered log-transformed body
weight, centered log-transformed hematocrit, race/ethnicity, and education.
The following procedure was used to impute the missing concentrations for the non-detects and
adjust for the negative values. A Bayesian model was used to:
• Impute the THg values less than the DLasa function of the regression calibration estimate
of usual intake (using the NHANES analysis weight) and other predictors.
• Impute the IHg values less than the DLas a function of the regression calibration estimate
of usual intake (using the NHANES analysis weight), the THg measurement (detected or
imputed) and other predictors.
• Calculate the preliminary organic mercury concentration as the difference between the
total and inorganic mercury concentrations (using detected or imputed values).
• Adjust the differences to be greater than zero such that the log-transformed differences
have a roughly normal distribution.
The following transformation was used to adjust the smallest differences upward to make all
values greater than zero.
Difference + J Difference2 + 0.04
Organic Hg =
Twenty imputed datasets were created for the analysis. For each replicate weight, the final
regression model (Step 7 in the EPA Method) was fit separately for each of the 20 imputed
datasets.
2.5 Estimation of 30-Day Fish Consumption and Mercury Intake
In order to investigate the relationship between fish consumption and mercury intake with
demographic characteristics, estimates of the amount offish consumed over 30 days were
calculated based on the NHANES 24-hour dietary recall data and the 30-day frequency offish
consumption data. The 24-hour data provided the amount of meal consumed at one time.
Information on the amount and species offish ingredients provided in the 24-hour recall can be
obtained by linking the FNDDS database with food code. Many participants who reported
consuming fish during the previous 30 days did not consume fish in the past 24-hours, so a
corresponding amount and species offish consumed in a meal were not available. Therefore, a
statistical model was used to estimate the amount offish they consumed in a meal to calculate
their estimated 30-day consumption. We followed the method described in the report Trends in
Blood Mercury Concentrations and Fish Consumption Among U.S. Women ofChildbearinaAae
14
-------
NHANES. 1999-2010 (United States Environmental Protection Agency, 2013) to estimate the
amount of 30-day consumption offish. The predicted grams offish consumed in a meal for each
species from the model were multiplied by the reported frequency of consumption of
corresponding species to get the amount of 30-day consumption for each fish species. The sum of
the 31 species-specific fish consumption amounts of each participant was the estimated 30-day
consumption offish for each women aged 16-49 years. The estimated mercury intake was
calculated as the product of species-specific fish tissue mercury concentration and the estimated
amount offish consumed at one time for each species. The participant level 30-day mercury
intake was calculated as the sum of mercury intake by fish species.
Mercury intake per body weight can be explained as the product of four components: 1) frequency
offish consumption; 2) weighted average meal size, weighted by frequency of consumption;
3) weighted average fish tissue mercury concentration, weighted by the quantity offish
consumed; and 4) inverse body weight (United States Environmental Protection Agency, 2013).
Logistic regression was used to model the probability of consuming fish in a 30-day period. For
participants who reported consuming fish, regression analysis was used to investigate the
association between these four components and demographic characteristics (education,
race/ethnicity, income, and age group).
15
-------
RESULTS
3
3.1 Blood MeHg Summary Statistics
This section presents the summary statistics based on the imputed concentrations of blood
MeHg by selected subpopulations.
3.1.1 Time Trends in Blood Mercury Concentrations
Figurel presents the distribution of blood MeHg concentration by N HAN ES survey release using
boxplots. The geometric mean blood MeHg concentration is highest in survey release 1999-2000
then declines to the lowest in 2017-March 2020. The geometric mean blood MeHg in the
2005-2006 release slightly higher than that in the 2003-2004 and 2007-2008 releases, and the
geometric mean blood MeHg in the 2009-2010 release slightly higher than that in the 2007-2008
and 2011-2012 releases. This general declining trend in geometric means is significant (p<0.001)
across the ten survey releases. This same general pattern over time is observed in the 25th and
90th percentiles of blood MeHg concentrations (Table 3). Detailed tabulations of the distribution
of blood MeHg concentrations, including sample size, arithmetic mean, geometric mean,
percentiles (25th, 50th, 75th, 90th, 95th), and their 95 percent confidence intervals, by NHANES
releases, race/ethnicity, age group, income, and education are presented in Table A-2.
16
-------
Figure 1. Distribution of log-transformed blood MeHgfpg/Lj, by NHANES survey releases, women aged
16-49 years
100
10-
tn
X
0.1
cd 0.01 -
0 001
'V
%
%
¦o
"a-
NHANES Data Release
\
Table 3. Distribution of blood MeHg concentration (pq/L), by survey release
Survey release
N
Geometric mean
Percentiles
(95% CI)
25th
75th
90th
All Years
15,236
0.57(0.55,0.60)
0.25
1.21
2.58
1999-2000
1,632
0.93(0.78,1.12)
0.41
1.88
4.56
2001-2002
1,799
0.72(0.65,0.79)
0.34
1.46
2.92
2003-2004
1,615
0.53(0.45,0.61)
0.24
1.21
2.60
2005-2006
1,788
0.58(0.51,0.66)
0.25
1.33
2.71
2007-2008
1,486
0.52(0.46,0.59)
0.25
1.06
2.39
2009-2010
1,780
0.67(0.61,0.73)
0.31
1.38
2.73
2011-2012
1,428
0.52(0.44,0.60)
0.23
0.98
2.32
2013-2014
814
0.50(0.45,0.55)
0.23
0.94
2.21
2015-2016
740
0.50(0.45,0.56)
0.24
0.94
2.19
2017-2020
2,154
0.48(0.43,0.54)
0.18
1.01
1.99
The EPA reference dose (RfD) for MeHg is 0.1 (jg/kg-day (United States Environmental Protection
Agency, 2001). This is equivalent to a blood mercury concentration of 5.8 |jg/L, An RfD is an
estimate of the maximum daily intake that is not likely to cause harmful effects across a lifetime.
The EPA uses the percent of women of childbearing age that have blood mercury concentrations
over 5.8 pg/L as one measure of the progress towards making fish and shellfish safer to eat. The
17
-------
calculated weighted prevalence of both blood MeHg and THg concentrations over 5.8 |jg/L by
survey release are presented in Table 4. The percentages of women of childbearing age with
blood MeHg concentrations above 5.8 |jg/L in 1999-2000 is about 3.5 times that found in 2001-
March 2020, representing a 71 percent decrease. There is a significant difference between the
survey releases for MeHg (Rao-Scott Chi-square p<0.001). No significance difference (p=0.12) is
found between the survey releases after removing the 1999-2000 survey release.
Table 4. Weighted percent of women 16 and 49 years with blood MeHg >5.8 pg/L and blood THg>5.8pg/L, by survey
release
Survey release
N
Blood MeHg
Blood THg
Weighted %
SE
Weighted %
SE
1999-2000
1,632
6.9
1.65
7.3
1.67
2001-2002
1,799
3.5
0.83
3.9
0.88
2003-2004
1,615
1.7
0.69
2.5
0.83
2005-2006
1,788
2.4
0.58
2.7
0.60
2007-2008
1,486
2.3
0.50
2.5
0.55
2009-2010
1,780
2.1
0.37
2.3
0.41
2011-2012
1,428
1.3
0.45
1.8
0.50
2013-2014
814
2.3
0.69
2.7
0.66
2015-2016
740
1.8
0.50
1.8
0.49
2017-2020
2,154
1.2
0.37
1.3
0.37
NOTE: Geometric mean and percentiles were calculated from the mean of imputed values for each respondent.
Decreasing trend over time, p<0.001 for both blood MeHg and THg concentrations based on logistic regression.
3.1.2 Demographic Distributions
Table 5 presents the comparison of blood MeHg concentration over time by demographic
characteristics for women 16-49. NHANES survey periods are grouped as 1999-2004,2005-2010,
2011-2016, and 2017-March 2020. The geometric mean of blood MeHg concentrations decreases
significantly over time in most demographic categories. There is a 31 percent decrease in blood
MeHg concentration from 1999-2004 to 2017-March 2020. Within each time period, people of
"Other Race," 40-49 years of age, with income greater than or equal to 5 times the poverty line
and education level higher than median level of the age group, have higher blood MeHg
concentrations compared to the rest of the groups within each demographic characteristic.
18
-------
Table 5. Comparison of Imputed Blood MeHg concentration (jjg/L) over time, by demographic characteristics, women
aged 76-49 years
Blood mercury
concentrations
1999-2004
2005-2010
2011-2016
2017-March 2020
p-value
Geo.
mean
SE
Geo.
mean
SE
Geo.
mean
SE
Geo.
mean
SE
(F test)
OVERALL
0.70
0.03
0.59
0.02
0.51
0.02
0.48
0.03
<.0001
Race/Ethnicity
Mexican American
0.53
0.03
0.44
0.02
0.39
0.02
0.42
0.03
0.0027
Other Hispanic
0.85
0.07
0.66
0.05
0.59
0.04
0.54
0.05
0.0320
Non-Hispanic White
0.66
0.04
0.55
0.03
0.47
0.02
0.45
0.05
0.0003
Non-Hispanic Black
0.90
0.06
0.65
0.03
0.53
0.04
0.56
0.05
<.0001
Other Race
1.17
0.13
1.25
0.12
0.89
0.07
0.64
0.05
<.0001
Age, Years
16 to 19
0.44
0.03
0.37
0.02
0.31
0.02
0.33
0.03
<0001
20 to 29
0.60
0.03
0.52
0.03
0.47
0.03
0.49
0.06
0.0122
30 to 39
0.80
0.06
0.63
0.04
0.55
0.03
0.51
0.03
<0001
40 to 49
0.83
0.05
0.71
0.03
0.61
0.03
0.52
0.04
<0001
Ratio of family income to poverty guidelines
0 to = 5x poverty line
1.07
0.08
0.94
0.06
0.81
0.06
0.68
0.06
0.0012
Missing/Refused/DK
0.78
0.09
0.63
0.07
0.60
0.04
0.54
0.07
0.2421
Education
Median education forage
0.98
0.07
0.82
0.05
0.72
0.04
0.69
0.07
0.0039
3.2 Blood MeHg Modeling
The parameter estimates and p-values from the multivariable modeling of blood MeHg
concentrations are presented in Table A-3. The parameter estimate provides the direction and
magnitude of the effect the predictor has on log-transformed blood MeHg concentration. Level of
education (p=0.0007) and race/ethnicity (p<0.001) are significantly associated with
log-transformed blood MeHg concentrations. Figure 2 shows the relative blood MeHg
concentrations for different education and racial/ethnic groups represented by multiplicative
difference from the overall MeHg concentration. The red diamond shows the estimate and the
yellow bar shows the 95 percent confidence interval of the estimate. If the 95 percent confidence
interval does not cross 1, then the estimate is statistically significant at p<0.05. Blood MeHg
concentration increases with education level. The difference in blood MeHg concentrations
19
-------
between women with education levels above the median for their age and those at the median
level is more pronounced than the difference between women with education levels at the
median and those below it for their age. There is significant difference (p<0.0001) in blood MeHg
concentrations between the race/ethnicity groups, with non-Hispanic White women of
childbearing age having the highest blood MeHg concentrations and Mexican American having
the lowest.
Figure 2. Relative blood MeHg concentrations with 95 percent confidence intervals, by
demographic characteristics(NHANES 2013-March 2020)
Education..
< Median for age
Median for age -
>Median for age -
Unknown Edu -
Race/Ethnicity --
Mexican American -
Other Hispanic -
Non-Hispanic White -
Non-Hispanic Black-
Other Race -
0.0
Estimate ~ Confidence Interval
0.75
0.65
0.5
1.0 1.5 2.0
Multiplicative Factor
2.5
3.0
TUI of mercury (fjg Hg/day) through fish consumption is one of the significant predictors of blood
MeHg concentration (p<0.001), with higher consumption associated with higher blood MeHg
concentration.
Figure 3 shows the slopes for the relationship between TUI of mercury through fish consumption
and log-transformed MeHg overall and by demographic group. The strength of the relationship
between usual intake of mercury through fish consumption and the blood MeHg concentration
20
-------
increases as the value of the slope factor increases. The following provides a basis for
interoperating theTUI slope parameters. Consider comparing blood MeHg concentrations
between subjects whose usual intake of mercury from fish consumption differs bylO percent
while holding all other factors constant. If all MeHg comes from mercury intake through fish
intake, increasing usual intake of mercury through fish consumption by 10 percent should
increase blood MeHg concentration by 10 percent (TUI slope =0.1). If half of the blood MeHg
concentrations comes from usual intake of mercury through fish consumption and half comes
from other sources, increasing usual intake of mercury through fish consumption by 10 percent
should increase blood MeHg concentrations by 5 percent (TUI = 0.05). The slope relating log-
transformed usual intake of mercury through fish consumption to log-transformed MeHg is an
estimate of the fraction of blood MeHg concentration from usual intake of mercury through fish
consumption. The slope for TUI is a close approximation to the slope of the relationship between
log-transformed usual intake and log-transformed blood MeHg concentration.
All slopes are greater than zero, indicating a positive relationship between TUI of mercury through
fish consumption and blood MeHg concentration. The slope varies significantly based on level of
both education (p=0.01) and race/ethnicity (p<0.001), indicating that blood MeHg concentration
increases at different rates with increasing intake offish mercury by education and race/ethnic
groups. Participants whose education level is higher than the median education level of the
corresponding age group have a higher slope than the rest of the education groups indicating
that blood MeHg concentration increase more to a unit increase in TUI of mercury for those with
greater than median level education compared to the other participants. Non-Hispanic White
participants have the highest slope among all race/ethnic groups and therefore have blood MeHg
concentrations that increase faster as TUI of mercury increases.
21
-------
Figure 3. Slope parameter relating transformed usual fish intake of mercury and
log-transformed blood MeHg concentrations, overall and by demographic group, with 95
percent confidence intervals
£
TO
to
Q_
"C:
£Z
to
S
US
£
r3
to
LL
Overall
Race/Ethnicity
Mexican American
Other Hispanic -
Non-Hispanic White -
Non-Hispanic Black-
Other Race -
Education
< Median for age
Median for age -
>Median for age -
Unknown Edue
-1—
0.0
—I—
0.5
Estimate ~ Confidence Interval
1.57
1.38
1.0 1.5 2.0
TUI slope for predicting blood MeHg Concentration
—i—
2.5
—r
3.0
Due to the model's complexity, the slope parameters presented in Figure 3 may be difficult to
interpret in terms of TUI of mercury (fjg Hg/day). Figure 4 shows the predicted relationship
between blood MeHg concentrations and TUI of mercury by racial/ethnic groups. For
non-Hispanic White participants, blood MeHg concentrations increase at a faster rate with TUI of
mercury compared to the rest of the groups. The slope is similar among participants who identify
as Mexican American, other Hispanic, and non-Hispanic Black.
22
-------
Figure 4. Estimated blood MeHg given usuai intake of fish mercury, by race/ethnicity
~ Histogram of usual intake
Non-Hispanic White
Other Hispanic
Usual intake (main effect)
Mexican American
Non-Hispanic Black
Other Race
Other significant predictors include NHANES survey release (p<0.001), log-transformed body
weight (p=0.002) and its interaction with TUI of mercury (p<0.001), and log-transformed
hematocrit concentration (p=0.005).
3.3 Trends in Fish Consumption
This section presents trends in fish consumption of NHANES 2013-March 2020 and compares
them to trends in fish consumption of NHANES 1999-2010 (United States Environmental
Protection Agency, 2013). NHANES 2011-2012 data was not included in the current study and
therefore not part of this trends analysis.
23
-------
3.3.1 Trends in Frequency of Consumption
Figure 5a presents the weighted percent of women aged 16-49 years in each of the six categories
of reported frequency offish consumption by NHANES survey releases (2013-March 2020).
Detailed tabulations are in Table A-4. There are no significant differences in reported frequency of
consumption between survey releases (Rao-Scott chi-square p-values: p=0.66 for total fish, p=0.56
for finfish, p=0.74for shellfish). Figure 5b displays the same distribution of NHANES 1999-2010
(United States Environmental Protection Agency 2013). While there are statistically significant
differences in consumption frequency between the survey releases 1999-2010, there is not a
consistent trend over time (Rao-Scott chi-square p-values; p=0.03 for total fish, p=0.02 for finfish,
and p=0.16for shellfish). Comparing the frequency of consumption of total fish between the two
studies, there are 7 percent more women reported not consuming fish in the previous 30 days in
2013-2014 than in 2009-2010, and approximate 5 percent decrease in women reported consuming
fish 6 times or more from 2009-2010 (31.8%) to 2013-2014 (26.4%). This indicates a shift of
consuming fish less frequently in women of childbearing age.
The fish consumption frequencies are similar between survey releases for each consumption
category. However, there are some differences in the frequency of consumption by category of
fish. For example, in NHANES 2013-2014, the percent of women who did not consume total fish
(finfish and shellfish) in the previous 30 days (27%) is less than the percent of women who did not
consume finfish (39%) or shellfish only (49%) in the previous 30 days. This is likely due to the fact
that some women only consumed finfish or shellfish. Similarly, the percent of women who
consumed fish a total of six or more times in the previous 30 days (26%) was greater than the
percent of women who consumed finfish (15%) or shellfish (8%) only six or more times. This
difference is a result of participants who may have consumed finfish or shellfish less than six times,
but when combined, they consumed either finfish or shellfish six or more times. These findings
are similar to a previous study of NHANES 1999-2010 data (United States Environmental Protection
Agency, 2013).
24
-------
Figure 5a. Weighted percent of participants by 30-day fish consumption frequency, by NHANES
survey release (2013-March 2020), women aged 16-49 years
100
90
80
70
Sf 60
50
c
0
u
0J 40
CL
30
20
10
0
&
o
fN
O
fN
O
fN
O O
fN fN
in
o
fN
o
fN
K£>
O
fN
O
»
o o o o
o o o o
fN fN fN fN
O
O
O
fN
¦sT KD CO
O O O
o o o
fN fN fN
ch lo r^
o o o
o o o
fN fN fN
Shellfish Only
¦ 6+ times
¦ 4-5 times
~ 3 times
~ 2 times
~ 1 time
~ 0 times
25
-------
Figure 6a presents the frequency of consumption by income, race/ethnicity, education, and age of
NHANES2013-March 2020. There are significant differences in frequency of total fish consumption
between income groups, race/ethnicity, age, and education (Rao-Scott Chi-Square p-values
<0.0001). These demographic characteristics were included in the analysis of relationship between
fish consumption and intake of mercury with demographic factors. Women with higher income
tend to eat fish more frequently. Individuals of "Other Race" eat fish more frequently compared to
Mexican American, non-Hispanic White, non-Hispanic Black, and other Hispanic. These findings
are consistent with a previous study of N HAN ES1999-2010 data (United States Environmental
Protection Agency, 2013) shown in Figure 6b. Older age is associated with increased frequency of
fish consumption in both studies. Women aged 30-39 years consume fish more frequently than
those in other age groups of N HAN ES 2013-March 2020, while women of 40-49 years consume
fish more frequently based on N HAN ES 1999-2010 (United States Environmental Protection
Agency, 2013). Women with higher levels of education are associated with higher frequency offish
consumption based on NHANES2013-March 2020.
There is a trend of decreasing frequency offish consumption among women of "Other Race,"
non-Hispanic White, and ages30-39 and 40-49. For example, comparing the estimates by
race/ethnicity between N HAN ES 1999-2010 and 2013-March 2020, the percent of women of "Other
Race" who reported total fish consumption of 6 or more times decreased from 46.6 percent to
35.4 percent, and the reported frequency of not consuming fish increased from 19.5 percent to
22.3 percent. Similar patterns were also observed in Non-Hispanic White.
26
-------
Figure 6a. Weighted percent of participants by 30-day total fish consumption frequency, by demographic
characteristics, women aged 16-49years, NHANES 2013-March 2020
100
90
80
70
60
£
c 50
>
o o
O. CL
> >
-t-J o
GL
qj
>
O fN
Q_ V
X o
QJ QJ
S 8
Q. Q_
X X X X
m in in
V V V ,|
O O O A
Income
c
-------
Figure 6b. Weighted percent of participants by 30-day total fish consumption frequency, by demographic
characteristics, women aged 16-49 years, NHANES1999 -2010
100%
10%
0%
o
o
o
cT
(N
w
V
8 8
O O
LO
^r
w
V
o
4->
8
o
s*
-uy
Lf)
w
v
o
o
a
in
try
QJ
>
o
ID
c
03
LO
w
o
C
03
8
o
a
<:
o
c
c
o
o
"O
(U
cu
C£
*
"D
-
>-
>¦
>•
or>
4-J
¦w
10
o
o
O
rsl
ro
Income
Race/Ethnicity
Age
3.3.2 Trends in Estimated Amounts Consumed Over the
Previous 30 days
The estimated amounts offish consumed has remained relatively consistent over the NHANES
survey releases from 2013-March 2020. Detailed tabulation of the amounts offish consumed (in
grams [g]), mercury intake (in micrograms [(jg]), and mercury intake per unit body weight (yg
Hg/kg bw) are tabulated in Tables A-5 through A-7 by NHANES releases, and Table A-8 by
race/ethnicity, age, income, and education. While the average amounts of total fish eaten per
participant and NHANES release for NHANES 2013-March 2020 (318.8-335.3 g) are at the higher
end of those found in NHANES 1999-2010 (254.6-322.5 g) (United States Environmental Protection
Agency, 2013), the estimated mercury intake from total fish consumption in NHANES 2013-March
2020 (22.83-25.62 (jg) are lower than those of the previous survey releases (29.33-37.40 (jg),
indicating women probably choose to eat fish with lower mercury concentration. This leads to
28
-------
lower estimates of mercury intake per unit body weight in NHANES 2013-March 2020 (0.32-0.36
[jg Hg/kg bw) compared to those found in NHANES1999-2010 (0.42-0.54 [jg Hg/kg bw).
3.3.3 Association Between Fish Consumption Frequency
and Blood Mercury
Figure 7a presents the distribution of mean blood MeHg concentrations by the 30-day frequency
of total fish consumption and NHANES survey release (2013-March 2020). Detailed tabulations are
in Table A-9. Blood MeHg concentration increases with frequency offish consumption (p<0.0001).
This agrees with previous studies that people who eat fish more frequently tend to have higher
blood mercury concentrations (Birch et al., 2014; Mahaffey et al., 2004; Mahaffey et al., 2009;
United States Environmental Protection Agency 2013). The distribution of blood MeHg
concentrations overtime are not consistent by frequency of consumption groups. The arithmetic
mean of blood MeHg concentration for women who ate fish six or more times decreases from 2.22
(1.66, 2.77) [jg/L in 2013-2014 to 1.66 (1.47,1.85) [jg/L in 2017-March 2020. For women who ate fish two
times, the mean blood MeHg concentration in 2005-2006 (1.04(0.55,1.53) |jg/L) is higher than that
in survey releases 2013-2014 and 2017-March 2020. The blood MeHg concentrations have small
variations for those who do not eat fish or ate fish one, three, four and five times. Based on the
study offish consumption of NHANES 1999-2010 as presented in Figure 7b (United States
Environmental Protection Agency, 2013), there is statistically significant decreasing trend of blood
MeHg concentration over time, indicating that women who consume fish more often may be
shifting to fish with lower concentrations of mercury.
29
-------
Figure 7a. Mean blood MeHg concentrations by reported frequency of total fish consumption in 30 days,
women aged 16-49 years, NHANES 2013-March 2020(with 95% confidence intervals, median, and 90th percentile)
1
¦o
o
o
5
4
3
2
» Arith rn etic M ea n
M median
* 90th pctl
y-error bar=95% CI
&
rt
i
o
KD
o
IX)
o
—l
fN
i—1
iH
fN
t-H
1—1
fN
o
o
O
O
O
O
o
o
O
fN
fN
fN
fN
fN
fN
CM
fN
fN
m
LTl
r-
t—I
O
O
fN
O
fN
O
fN
O
fN
3 TIMES
4-5 TIMES 6+ TIMES
Number of Times Finfish/Shellfish Consumed in Past 30 Days
Figure 7b. Mean blood MeHg concentrations by reported frequency of total fish consumption in 30 days, women aged
16-49years, NHANES 1999-March 2010(with 95% confidence intervals, median, and 90th percentile)
30
-------
3.3.4 Relationship Between Fish Consumption and Intake
of Mercury with Demographic Factors
This section examines the statistical association between fish consumption and demographic
characteristics using the method detailed in Trend in Blood Mercury Concentrations and Fish
Consumption Among U.S. Women of Childbearina Aae NHANES1999-2010 (United States
Environmental Protection Agency, 2013). Estimates of the amount offish consumed in the
previous 30 days, mercury intake, and mercury intake per unit body weight were calculated using
the method described in section 2.5. Logistic regression was applied to model the probability of a
person reporting any fish consumption in the previous 30 days using education, race/ethnicity,
income, and age. For those who reported consumption offish, five regression models were fit to
predict fish consumption and mercury intake variables from demographic characteristics. The five
variables were (1) mercury intake per unit body weight and the four components of this variable,
(2) number of meals in the previous 30 days, (3) amount offish consumed in a meal, (4) the
mercury concentration in the fish consumed calculated as the ratio of mercury intake to fish
consumption in the previous 30 days, and (5) the inverse of body weight. The model results are
presented in Tables A-10 to A-15.
Figure 8 presents the results from the logistic regression models by education groups of NHANES
2013-March 2020. Education is categorized as less than, equal to, or greater than the median
education level for the participant's age for NHANES releases 2013-2014,2015-2016, and 2017-March
2020, and an additional unknown level for participants 16-19years in NHANES 2017-March 2020.
The percentages in parentheses next to the education group, e.g.,
-------
consumption or mercury intake variable. For example, women with greater than median
education level of the participant's age (maroon dot of >Median of age) eat fish more frequently
than typical women (horizontal line). The blue dot is the RR of the inverse of body weight. A RR
less than one indicates higher body weight compared to a typical participant and a RR greater
than one indicates lower body weight than typical.
There are significant differences (p<0.05) for all fish consumption and mercury intake variables by
education except the mercury concentration in fish consumed (p=0.25).The percent of women
who consumed fish in the previous 30 days and mercury intake per unit body weight increase
with increasing known level of education. Women with less than median level education have the
lowest mercury intake per body weight: they generally eat fish less frequently eat a smaller meal
size, and have the highest body weight. The proportion of women with unknown education level
who consumed fish in the previous 30 days is the highest among all groups. They eat fish more
frequently compared to women with less than median level education and median level
education; however, their meal size and mercury concentration in fish consumed are the lowest,
resulting in relative low mercury intake per unit body weight.
Figure 8. Relative ratios and 95 percent confidence intervals from models predicting fish consumption
and mercury intake variables by education, NHANES 2013-March 2020
2.25
2
1.75
1.5
1.25
0.8
0.6
0.4
Fish and Mercury Variables versus Education Level
¦I
J
"ttT I "
II
-85%
-80%
-75% -E
-70% £
-65%
1 1 1 1
< Median for age (30%) Median for age (33%) >Medianforage (32%) Unknown Edu (5%)
Education Level
• % ate fish last 30 days •* meals in 30 days • Meal size
• Hg cone in fish consumed • Inverse body weight • Hg intake per body weight
32
-------
Figure 9a presents the RRs for fish consumption and mercury intake by race/ethnicity of N HANES
2013-March 2020. The p-vaiues testing the overall significance of race/ethnicity in all models is
p<0.0001, except for the model of mercury concentration of the fish consumed (p=0.21) and
mercury intake per unit body weight (p=0.28).The percent of non-Hispanic Black women who
consumed fish in the previous 30 days is higher compared to other racial/ethnic groups while
consumption among non-Hispanic White women is lower, which is consistent with that found in
N HAN ES 1999-2010 study (United States Environmental Protection Agency, 2013) presented in
Figure 9b.
Figure 9a. Relative ratios and 95 percent confidence intervals from models predicting fish consumption and
mercury intake variables by race/ethnicity, NHANES 2013-March 2020
(U
cr
2.25
2
1.75
1.5
1.25
1
-5 0 8
cr
0.6
0.4
e>.
<%>
Fish and Mercury Variables versus Race
%
I-,
r 11*
i
** i
i
{ i*
5
M
i ''
~
\
\
\
r*
%
%
\
85%
%
%
'ft
r/
v
%
Race
fO
"O
Q
CO
(i
80% •-
-C
(/)
-75%
C7)
(T5
O)
-70% ~
-65%
c
O)
o
0>
LL
• % ate fish last 30 days
• Hgconc. in fish consumed
• # meals in 30 days • Meal size
# Inverse body weight • Hg intake per body weight
33
-------
Figure 9b. Relative ratios and 95 percent confidence intervals from models predicting fish consumption and mercury
intake variables by race/ethnicity, NHANES1999-2070
rs
CE
a
09
a.
2,25 s
2
1-75
15
1.25
1
0 8
06
04 -
Or*
Fish and Mercury Variables versus Race
%
\
"4.
%
\
%
%
j 1
*
* T
r
-r- ^? £ T * * iT*:
. 1 *»
ft tjT,
*-
i
%
%
%
-85%
- 80% S
-75%
-70%
-65%
Ou
%
Race
A % ate Ash last 30 days o * meals in 30 days * Meal size
~ hq cone in fi she onsumtd # ifweise body weighi ¦ Hq intake per body w«?>ghi
n
*D
8
JC
to
C2
j|
O
k>
a
In general, non-Hispanic Black women consume the largest meal size, eat fish more frequently,
consume fish with higher mercury concentration, and have higher body weights than a typical
woman of childbearing age, resulting in the highest mercury intake per unit body weight of the
racial/ethnic groups in NHANES 2013-March 2020. In NHANES 1999-2010 (Figure 9b), non-Hispanic
Black women have similar distributions compared to a typical woman of childbearing age except
they consume smaller than typical meal size, resulting in a less than typical mercury intake per
unit body weight.
in NHANES 2013-March 2020, women who identity as "Other Race" eat fish the most frequently,
consume fish with higher mercury concentration, have lower than typical body weight; however,
while they consume the smallest meal size, they have a slightly higher than typical mercury intake
per unit body weight. In NHANES 1999-2010, women of "Other Race" have the same trends in
34
-------
these fish consumption and mercury intake variables. With a close to typical meal size, they have
the largest mercury intake per unit body weight of all racial/ethical groups.
In NHANES 2013-March 2020, non-Hispanic White women generally consume fish the least
frequently and have higher than typical body weight; however, they consume larger meals,
consume fish with higher than typical concentration of mercury, resulting in a close to typical
mercury intake per unit body weight. In N HANES1999-2010, the fish consumption and mercury
intake variables of non-Hispanic White women follow the same trends except they consumed
smaller than typical meal size and yield a less than typical mercury intake per unit body weight.
In NHANES 2013-March 2020, Mexican American women consume larger meal sizes; however,
they eat less frequently and consumed fish with lower concentration of mercury, resulting in the
lowest mercury intake per unit body weight. Other Hispanic women consume fish less frequently,
eat smaller meal size, consume fish with lower contraction of mercury, and have lower body
weight than a typical woman of childbearing age, resulting in a close to typical mercury intake per
unit body weight. The trends of the fish consumption and mercury intake variables compared to
values of typical women of childbearing age are similar in Mexican American and other-Hispanic
participants of N HAN ES 1999-2010 (Figure 9b).
Figure 10 presents the RRs for fish consumption and mercury intake variables by income of
NHANES 2013-March 2020 survey releases. There are significant differences (p=0.0004) for the
proportion of women who consumed fish in the previous 30 days by income. Higher income is
associated with higher proportion of women who consumed fish in the previous 30 days in
general. Women with income equal to or greater than five times the poverty line consume fish
the most frequently, consume larger meal sizes and fish with higher concentration of mercury,
and have less than typical body weight, resulting in the highest mercury intake per unit body
weight. In the 2013 study of trends in fish consumption among women of childbearing age using
NHANES 1999-2010 data (United States Environmental Protection Agency, 2013), income is
categorized with different survey variables and therefore not comparable to this study.
35
-------
Figure 10. Relative ratios and 95 percent confidence intervals from models predicting fish consumption and
mercury intake variables by income, NHANES 2013-March 2020
Fish and Mercury Variables versus Income
Household Income
• % ate fish last 30 days • # meals in 30 days • Meal see
~ Hg cone in flsh consumed • Inverse body weight • Hg intake per body weight
Figure 11a presents the RRs for fish consumption and mercury intake by age group. The p-values
testing the overall significance of age group in all models are less than 0.05, except for the models
of mercury concentration of the fish consumed (p=0.16) and amount of meal consumed (p=0.29).
The proportion of women who consumed fish in the previous 30 days increases with age. As
observed in a previous study on trends in blood mercury concentrations and fish consumption
among U.S. women of childbearing Age NHANES 1999-2010 (Birch et al., 2014; United States
Environmental Protection Agency 2013) and presented in Figure lib, women 16-19 years eat fish
the least frequently, consume the smallest meal size, have the lowest body weight, and consume
fish with the lowest concentration of mercury resulting in the lowest mercury intake per unit
body weight. The value of mercury intake per unit body weight increases with age to the highest
level in women 30-39 years and then drops in women 40-49 years to a level comparable to
women 20-39years. Based on NHANES 1999-2010 study (United States Environmental Protection
Agency 2013) women of 40-49 years have the highest intakes of mercury per unit body weight.
36
-------
Figure 11a. Relative ratios arid 95 percent confidence intervals from models predicting fish consumption and
mercury intake variables by age group, NHANES 2013-March 2020
Fish and Mercury Variables versus Age Group
ro
tr
£
2®
cr
2 25
2
1 75 ¦
1.5-
1.25-
TI
Tf"
08-
06-
0 4 -
kl
16 to 19 (12%) 20 to 29 (30%) 30 to 39 (29%)
Age Group
iif "
85%
80%
5
«
o>
75% ~
- 70% £
CL
65%
40 to 49 (29%)
• % ate fish last 30 days ~ * meats in 30 days • Meal see
• Hg cone in flsh consumed • Inverse body weight • Hg Intake per body weight
37
-------
Figure lib. Relative ratios and 95 percent confidence intervals from models predicting fish consumption and mercury
intake variables by age group, NHANES1999 - 2010
o
*55
a
4)
£
•11
a:
225
2
1.75
1.5
1 25
0 8 -
0 6 -
Fish and Mercury Variables versus Age Group
!!
J
{
1
-85%
-80%
¦u
s
tA
in
MS
CT)
'75% *
70%
B
D
£L
fi5%
0 4 -
16 to 19 (11%) 20 to 29 (28%) 3010 39 (30%)
Age Group
40 to 49 (32%)
* % ate fish last 30 days o * meals in 30 days • Meal size
~ Hgconi infish consum* 1 • inverse body weight ¦ Hg intake per body weight
38
-------
DISCUSSION and CONCLUSIONS
Trends in geometric mean blood MeHg concentrations over time and bv demographic
characteristics
The results of this study indicate that there is a significant (p<0.0001. Table A-3) decreasing trend of
blood MeHg concentrations overtime across the ten N HAN ES survey releases I see Figure 1 and
Table 31 after controlling for demographic characteristics. The geometric mean blood MeHg
concentration is highest in NHANES survey release 1999-2000, then declines to the lowest in
N HAN ES survey release 2017-March 2020. Additionally the geometric mean blood MeHg
concentration in the 2005-2006 release is slightly higher than that in the 2003-2004 and
2007-2008 releases, while the geometric mean blood MeHg in the 2009-2010 release is slightly
higher than those in the 2007-2008 and 2011-2012 releases. This does not appear to be a
meaningful increase and is likely due to the fluctuations in the data. The geometric mean blood
MeHg concentration in NHANES 1999-2000 is 1.94 times higher than the geometric mean in
NHANES 2017-March 2020 data, representing a 48 percent decrease between NHANES 1999-2000
and 2017-March 2020. Similar decreasing trends in blood MeHg concentration over time are found
in NHANES 1999-2010 (Birch et a/., 2014; United States Environmental Protection Agency, 2013).
The studies using NHANES 1999-2010 found that the linear time trend in the mean of blood MeHg
concentrations was statistically significant (p=0.006), but there was no significant trend from 2001
to 2010 (p=0.74). With additional NHANES survey releases in this study, excluding survey release
1999-2000 does not change the significance of the linear trend from 2001-March 2020.
The percentages of women of childbearing age with blood MeHg concentrations over 5.8 |jg/L in
1999-2000 is about 3.5 times that found in 2001-March 2020, representing a 71 percent decrease.
There is a significant difference between the survey releases for blood MeHg concentrations
(Rao-Scott Chi-square p<0.001), with 1999-2000 having approximately three times the amount of
women with levels over 5.8 [jg/L compared to the other NHANES releases. However, no
significance difference (p=0.12) is found between the survey releases after removing survey release
NHANES 1999-2000. Similar patterns were found in previous studies of NHANES 1999-2010 (Birch
et a/., 2014; United States Environmental Protection Agency, 2013).
There are significantly (p<0.05) decreasing trends of blood MeHg concentrations over time by
most demographic characteristics across these NHANES survey releases 1999-2004,2005-2010,
39
-------
2011-2016, and 2017-March 2020. The geometric mean blood MeHg concentration in women of
"Other Race" in 1999-2004 is 1.82 times higher than the geometric mean for the same
demographic in 2017-March 2020, representing a 45 percent decrease between N HAN ES 1999-
2004 and 2017-March 2020. Within each of these survey release periods, higher blood MeHg
concentrations are observed with increasing age, ratio of family income to poverty, education
level, and among participants who reported their race as "Other Race."
Association of blood MeHg concentration with intake and demographic characteristics
A multivariable model is used to investigate the association between MeHg concentration and
TUI through fish consumption and other demographic characteristics. The model found the
following factors to be significantly associated with blood MeHg are (1) TUI through fish
consumption (p<0.0001), (2) education (p=0.0007), (3) race/ethnicity (p<0.0001), (4) NHANES survey
release (p<0.0001), (5) log-transformed hematocrit (p=0.005), and (6) log-transformed body weight
(p=0.002). In addition, the rate of increase in blood MeHg concentration due to usual intake of
mercury varies by education, race/ethnicity, and log-transformed body weight. Household income
is marginally significant at the 5 percent level (p=0.054) in predicting blood MeHg concentrations.
Geographic differences in blood mercury
The geographic findings of this study (Appendix B) are similar to what was found in previous
studies (Cusack et a/., 2017; Mahaffey et a/., 2009). There are geographic differences in blood MeHg
concentrations with higher levels among residents of coastal counties compared to non-coastal
counties. Residents of the Northeast region have the highest blood MeHg concentrations of the
four regions and residents of the Midwest have the lowest blood MeHg concentrations.
Trends in frequency offish consumption and the association offish consumption and
mercury intake with demographic characteristics
One limitation of this study is that the analysis offish consumption over the previous 30 days used
data from N HAN ES 2013-March 2020, and results were compared toa previous study using
N HAN ES 1999-2010 (United States Environmental Protection Agency, 2013). N HAN ES 2011-2012
was not published during the previous study and was not included in the current study, leaving a
gap between the two studies. This gap limits the ability to investigate the trend in fish
consumption from 1999 to March 2020.
Blood MeHg concentrations are positively associated with the frequency offish consumption in
both N HAN ES 2013-March 2020 and N HAN ES 1999-2010. There are no significant differences in the
40
-------
reported frequency of consumption offish over the previous 30 days across NHANES survey
releases 2013-March 2020 while significant differences of reported frequency of consumption are
found across the six NHANES survey releases between 1999 and 2010 (United States
Environmental Protection Agency 2013). The association between the frequency offish
consumption and demographic characteristics are consistent between NHANES 1999-2000 and
NHANES 2013-March 2020, respectively. Women of older age and with higher income tend to eat
fish more frequently. Individuals of "Other Race" eat fish more frequently compared to Mexican
American, non-Hispanic White, non-Hispanic Black, and other Hispanic. Women with higher
education levels are found to eat fish more frequently.
The estimated amounts of total fish eaten over the previous 30 days in NHANES 2013-March 2020
are at the higher end of those found in NHANES 1999-2010. The estimated mercury intake from
total fish consumption over the previous 30 days in NHANES 2013-March 2020 are lower than
those found in NHANES 1999-2010. This suggests that women of childbearing age are probably
choosing to eat fish with lower mercury concentration leading to lower estimates of mercury
intake per unit body weight in NHANES 2013-March 2020 compared to those found in NHANES
1999-2010.
In addition to continuing to monitor the time trend in blood MeHg and fish consumption, a future
study on the geographic distribution of blood MeHg using NHANES releases of 2013-2014 and
later would be a valuable supplement to this study, helping to investigate trends in geographic
differences in blood MeHg.
41
-------
QUALITY CONTROL/QUALITY
ASSURANCE
5
This section details the steps that were taken to ensure the quality of the results.
The fish tissue mercury concentration database developed for previous EPA studies on trends in
blood mercury concentrations and fish consumption was updated with data on fish samples
collected from 2013 to 2020. Data were downloaded when they were available online. Other data
extracted from peer reviewed journal articles and reports were checked by a second individual to
ensure all information were correct.
The N HAN ES 2003-2010 24-hour recall data were processed to extract all reports offish
consumption for the 2014 U.S. EPA report (United States Environmental Protection Agency 2014),
Estimated Fish Consumption Rates for the U.S. Population and Selected Subpopulations
(NHANES2003-2010). At that time, the processing was done independently by two individuals and
results were compared. A final program with macros applicable to all NHANES releases in general
was created to process NHANES 24-hour dietary recall data. In the study of the geographic
distributions of blood mercury concentration (Appendix B) of NHANES 1999-2012, additions were
made to that code to include processing of the NHANES 1999-2000,2001-2002, and 2011-2012 data,
based on the code for the NHANES 2003-2010 data. The current analysis is built upon these
previous analyses with the addition of NHANES 2013-2014,2015-2016, and 2017-March 2020 data.
This analysis utilized the software created for the estimation of usual fish consumption rates, the
EPA Method, developed for Estimated Fish Consumption Rates for the U.S. Population and
Selected Subpopulations (NHANES 2003-2010) (United States Environmental Protection Agency,
2014). This software has previously undergone quality checks. This analysis also utilized the
INDIVINT macro, which is one of three NCI Method programs. It is available from NCI.
The imputation method was evaluated using the NHANES 2011-2012 data on methyl and ethyl
mercury.
42
-------
The models used a large number of potential predictors. These were very useful in avoiding
misleading results that may be found in simpler models that do not account for these
interrelationships.
43
-------
REFERENCES
6
Akinbami, L. J.; Chen, T. C.; Davy, 0.; Ogden, C. L.; Fink, S.; Clark, J.; Riddles, M. K.; Mohadjer, L. K.(2022) National Health and
Nutrition Examination Survey, 2017-March 2020 Prepandemic File: Sample Design, Estimation, and Analytic Guidelines.
Vital Health Stat 1,1-36. https://dx.doi.Org/10.15620/cdc:115454.
Birch, R. J.; Bigler, J.; Rogers, J. W.; Zhuang, Y.; Clickner, R. P. (2014) Trends in blood mercury concentrations and fish
consumption among U.S. women of reproductive age, NHANES, 1999-2010. Environ Res, 133,431-438.
httPs://doi.org/10.1016/i.envres.2014.02.001.
Bjornberg, K. A.; Vahter, M.; Petersson-Grawe, K.; Glynn, A.; Cnattingius, S.; Darnerud, P. 0.; Atuma, S.; Aune, M.; Becker, W.;
Berglund, M. (2003) Methyl mercury and inorganic mercury in Swedish pregnant women and in cord blood: influence of
fish consumption. Environ Health Perspect, 111,637-641. https://doi.org/10.1289/ehp.111-1241457.
Bramante, C. T.; Spiller, P.; Landa, M. (2018) Fish Consumption During Pregnancy: An Opportunity, Not a Risk. JAMA Pediatr, 172,
801-802. https://doi.org/10.1001/iamapediatrics.2018.1619.
Caldwell KL, Mortensen ME, Jones RL, Caudill SP, Osterloh JD. Total blood mercury concentrations in the U.S. population: 1999-
2006. Int J Hyg Environ Health. 2009 Nov:212(6):588-98. https://doi.Org/10.1016/i.iiheh.2009.04.004.
Centerfor Disease Control and Prevention. (2021a) Fourth national report on human exposure to environmental chemicals.
Updated tables, March 2021: volume one, NHANES 1999-2010 in: National Centerfor Environmental Health. Division of
Laboratory, S.; National Centerfor Health, S.; National, H.; Nutrition Examination, S. (Eds.), Atlanta, GA.
Centerfor Disease Control and Prevention. (2021b) Fourth national report on human exposure to environmental chemicals.
Updated tables, March 2021: volume two, NHANES 2011-2016 in: National Centerfor Environmental Health. Division of
Laboratory, S.; National Centerfor Health, S.; National, H.; Nutrition Examination, S. (Eds.). Centerfor Disease Control
and Prevention, Atlanta, GA.
Centerfor Disease Control and Prevention, N. C. f. H. S. N.(2013)Aboutthe National Health and Nutrition Examination Survey.
U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, Hyattsville, MD.
Clarkson, T. W.; Magos, L. (2006) The toxicology of mercury and its chemical compounds. Crit Rev Toxicol, 36,609-662.
https://doi.org/10.1080/1040844060Q845619
Cusack, L. K.; Smit, E.; Kile, M. L.; Harding, A. K. (2017) Regional and temporal trends in blood mercury concentrations and fish
consumption in women of child bearing age in the United States using NHANES data from 1999-2010. Environ Health, 16,
10. https://doi.org/10.1186/s12940-017-0218-4.
EPA_FDA. (2021) Advice about Eating Fish and Shellfish in: United States Environmental Protection Agency and United States
Food and Drug Administration (Ed.). Available at: https://www.epa.gov/choose-fish-and-shellfish-wiselv/epa-fda-
advice-about-eating-fish-and-shellfish
Golding, J.; Gregory, S.; Iles-Caven, Y.; Hibbeln, J.; Emond, A.; Taylor, C. M.(2016)Associations between prenatal mercury
exposure and early child development in the ALSPAC study. Neurotoxicology, 53,215-222.
https://doi.org/10.1016%2Fi.neuro.2016.02.0Q6.
Imanse, S. M.; Anchor, C. L.; Anchor, G. C.; Landolfi, J. A.; Kinsel, M. J.; Levengood, J. M.; Delaney, M. A.; Terio, K. A.(2022)
Pathologic impacts of contaminants in freshwater fish of Cook County IL. Aquat Toxicol, 242,106043.
httPs://doi.org/10.1016/i.aguatox.2021.106043.
Janssen, S. E.; Hoffman, J. C.; Lepak, R. F.; Krabbenhoft, D. P.; Walters, D.; Eagles-Smith, C. A.; Peterson, G.; Ogorek, J. M.;
DeWild, J. F.; Cotter, A.; Pearson, M.; Tate, M. T.; Yeardley, R. B., Jr.; Mills, M. A. (2021) Examining historical mercury
44
-------
sources in the Saint Louis River estuary: How legacy contamination influences biological mercury levels in Great Lakes
coastal regions. Sci Total Environ, 779,146284. https://doi.Org/10.1016/i.scitotenv.2021.146284.
Janssen, S. E.; Riva-Murray, K.; DeWild, J. F.; Ogorek, J. M.; Tate, M. T.; Van Metre, P. C.; Krabbenhoft, D. P.; Coles, J. F.(2019)
Chemical and Physical Controls on Mercury Source Signatures in Stream Fish from the Northeastern United States.
Environ Sci Technol, 53,10110-10119. https://doi.org/10.1021/acs.est.9b05594.
Johnson, C. L.; Paulose-Ram, R.; Ogden, C. L.; Carroll, M. D.; Kruszon-Moran, D.; Dohrmann, S. M.; Curtin, L. R.(2015) National
health and nutrition examination survey: analytic guidelines, 1999-2010. Vital Health Stat 2,1-24.
https://www.cdc.gov/nchs/data/series/sr_02/sr02_161.pdf
Kipnis, V.; Midthune, D.; Buckman, D. W.; Dodd, K. W.; Guenther, P. M.; Krebs-Smith, S. M.; Subar, A. F.; Tooze, J. A.; Carroll, R. J.;
Freedman, L. S. (2009) Modeling data with excess zeros and measurement error: application to evaluating relationships
between episodically consumed foods and health outcomes. Biometrics, 65,1005-1010. https://doi.org/10.1111/i.1541-
0420.2009.01225.x.
Lepak, R. F., Bartz, K.K., Ogorek, J.M., Tate, M.T., DeWild, J.F., and Janssen, S.E. (2022) Assessment of mercury sources in
Alaskan lake food webs: U.S. Geological Survey data release, in: United States Geological Survey (Ed.).
Louisiana Department of Environmental Quality. (2016-2020) Mercury Contaminant Levels in Louisiana Biota in: Quality, L. D. o. E.
(Ed.).
Mahaffey, K. R.; Clickner, R. P.; Jeffries, R. A. (2009) Adult women's blood mercury concentrations vary regionally in the United
States: association with patterns offish consumption (NHANES1999-2004). Environ Health Perspect, 117,47-55.
https://doi.org/10.1289%2Fehp.11674.
Mahaffey, K. R.; Clickner, R. P.; Bodurow, C. C. (2004) Blood organic mercury and dietary mercury intake: National Health and
Nutrition Examination Survey, 1999 and 2000. Environ Health Perspect, 112,562-570.
https://doi.org/10.1289%2Fehp.6587.
Malinowski, C. R. (2019) High mercury concentrations in Atlantic Goliath Grouper: Spatial analysis of a vulnerable species. Mar
Pollut Bull, 145, 81-91. https://doi.Org/10.1016/i.marpolbul.2019.04.006.
McGee, B. N. R., D.L.; Pribil, M. (2020) Hg Concentrations of Fish Tissue Samples in the Vicinity of Yellow Pine, Idaho in: United
States Geological Survey (Ed.). Available at https://www.usgs.gov/data/ho-concentrations-fish-tissue-samples-
vicinitv-vellow-pine-idaho.
Mebane, C. A., and Schmidt, C.G. (2019) Selenium and mercury in the Kootenai River, Montana and Idaho, 2018-2019: U.S.
Geological Survey data release in: United States Geological Survey (Ed.). Available at
https://www.usgs.gov/data/selenium-and-mercurv-kootenai-river-montana-and-idaho-2018-2019.
Melnyk, L. J.; Lin, J.; Kusnierz, D. H.; Pugh, K.; Durant, J. T.; Suarez-Soto, R. J.; Venkatapathy, R.; Sundaravadivelu, D.; Morris, A.;
Lazorchak, J. M.; Perlman, G.; Stover, M. A. (2021) Risks from mercury in anadromousfish collected from Penobscot
River, Maine. Sci Total Environ, 781,146691. https://doi.Org/10.1016/i.scitotenv.2021.146691.
Mergler, D.; Anderson, H. A.; Chan, L. H.; Mahaffey, K. R.; Murray, M.; Sakamoto, M.; Stern, A. H.; Panel on Health, R.; Toxicological
Effects of, M. (2007) Methylmercury exposure and health effects in humans: a worldwide concern. Ambio, 56,5-11.
https://doi.org/10.1579/0044-7447(2007)56[ 5:meaheil2.0.co:2.
Murray, K. R.; Cleckner, L.B.; Razavi, N.R.; Richter, W.(2020) Fish mercury concentration data and ancillary data for streams and
rivers across New York States (United States), 1969-2016, including environmental characteristics of selected locations
sampled during 2007-16 in: United States Geological SurveyfEd.). Available at
https://data.usgs.gov/datacatalog/data/USGS:5e1f565ce4b0ecf25c61ed64.
National Research Council. (2000) Toxicological Effects of Methylmercury. The National Academies Press: Washington, DC.
https://doi.org/10.17226/9899.
New Hampshire Department of Environmental Services. (2018) Status and trends of mercury in fish tissue in New Hampshire
waterbodies, 1992-2016. https://www.des.nh.goV/sites/g/files/ehbemt541/files/documents/2020-01/r-wd-17-22.pdf
45
-------
Park, S.; Lee, B. K. (2013) Strong positive associations between seafood, vegetables, and alcohol with blood and urinary arsenic
levels in the Korean adult population. Arch Environ Contam Toxicol, 64 (1), 167-170. https://doi.oro/10.1007/s00244-012-
9808-x
Rice, G.; Swartout, J.; Mahaffey, K.; Schoeny, R. (2000) Derivation of U.S. EPA's oral Reference Dose(RfD)for methylmercury.
Drug Chem Toxicol, 23,41-54. https://doi.org/10.1081/dct-100100101.
Rothenberg, S. E.; Korrick, S. A.; Fayad, R. (2015)The influence of obesity on blood mercury levels for U.S. non-pregnant adults
and children: NHANES 2007-2010. Environ Res, 138,173-180. https://doi.Org/10.1016/i.envres.2015.01.018.
Sackett, D. K.; Drazen, J. C.; Popp, B. N.; Choy, C. A.; Blum, J. D.; Johnson, M. W. (2017)Carbon, Nitrogen, and Mercury Isotope
Evidence for the Biogeochemical History of Mercury in Hawaiian Marine Bottomfish. Environ Sci Technol, 51,13976-
13984. https://doi.org/10.1021/acs.est.7b04893.
Sanzo, J. M.; Dorronsoro, M.; Amiano, P.; Amurrio, A.; Aguinagalde, F. X.; Azpiri, M. A.; Spain, E. G. o. (2001) Estimation and
validation of mercury intake associated with fish consumption in an EPIC cohort of Spain. Public Health Nutr, 4,981-988.
SAS Institute. (2016) Statistical Analysis System, 9.4 ed. https://doi.ora/10.1079/phn2001170.
Stratakis, N.; Conti, D. V.; Borras, E.; Sabido, E.; Roumeliotaki, T.; Papadopoulou, E.; Agier, L.; Basagana, X.; Bustamante, M.;
Casas, M.; Farzan, S. F.; Fossati, S.; Gonzalez, J. R.; Grazuleviciene, R.; Heude, B.; Maitre, L.; McEachan, R. R. C.;
Theologidis, I.; Urquiza, J.; Vafeiadi, M.; West, J.; Wright, J.; McConnell, R.; Brantsaeter, A. L.; Meltzer, H. M.; Vrijheid,
M.; Chatzi, L.(2020)Association of Fish Consumption and Mercury Exposure During Pregnancy With Metabolic Health
and Inflammatory Biomarkers in Children. JAMA NetwOpen, 3, e201007.
https://doi.org/10.1001%2Fiamanetworkopen.2020.1007.
Svensson, B. G.; Schutz, A.; Nilsson, A.; Akesson, I.; Akesson, B.; Skerfving, S. (1992) Fish as a source of exposure to mercury and
selenium. Sci Total Environ, 126,61-74. https://doi.org/10.1016/0048-9697(92)90484-a.
Taylor, D. L.; Calabrese, N. M. (2018) Mercury content of bluecrabs(Callinectes sapidus)from southern New England coastal
habitats: Contamination in an emergent fishery and risks to human consumers. MarPollut Bull, 126,166-178.
https://doi.Org/10.1016/i.marpolbul.2017.10.089.
Taylor, C.; Golding, J.; Emond, A. M. (2016) Blood mercury levels and fish consumption in pregnancy: Risks and benefits for birth
outcomes in a prospective observational birth cohort. International Journal of Hygiene and Environmental Health, 219,
513-520. https://doi.org/10.1016%2Fi.iiheh.2016.05.0Q4.
Tooze, J. A.; Kipnis, V.; Buckman, D. W.; Carroll, R. J.; Freedman, L. S.; Guenther, P. M.; Krebs-Smith, S. M.; Subar, A. F.; Dodd, K.
W. (2010) A mixed-effects model approach for estimating the distribution of usual intake of nutrients: the NCI method.
Stat Med, 29,2857-2868. https://doi.org/10.1002/sim.4063.
Tooze, J. A.; Midthune, D.; Dodd, K. W.; Freedman, L. S.; Krebs-Smith, S. M.; Subar, A. F.; Guenther, P. M.; Carroll, R. J.; Kipnis, V.
(2006) A new statistical method for estimating the usual intake of episodically consumed foods with application to their
distribution. J Am Diet Assoc, 106,1575-1587. https://doi.Org/10.1016/i.iada.2006.07.003.
United States Environmental Protection Agency. (2023)2013-2014 National Rivers and Streams Assessment Fish Tissue Study in:
United States Environmental Protection Agency (Ed.). Available at https://www.epa.gov/choose-fish-and-shellfish-
wiselv/2013-2014-national-rivers-and-streams-assessment-fish-tissue-studv.
United States Environmental Protection Agency. (2015)2015 Great Lakes Human Health Fish Fillet Tissue Study in: United States
Environmental Protection Agency (Ed.). Available at https://www.epa.gov/choose-fish-and-shellfish-wiselv/2015-
great-lakes-human-health-fish-fillet-tissue-studv.
United States Environmental Protection Agency. (2014) Estimated Fish Consumption Rates for the U.S. Population and Selected
SubpopulationsfNHANES2003-2010). Report No. EPA-820-R-14-002; United States Environmental Protection Agency,
Office of Science and Technology, Office of Water. Available at https://www.epa.gov/sites/default/files/2015-
01/documents/fish-consumption-rates-2014.pdf.
46
-------
United States Environmental Protection Agency. (2013) Trends in Blood Mercury Concentrations and Fish Consumption Among U.S.
Women ofChildbearingAge, NHANES1999-2010. Report No. EPA-823-R-13-002; U.S. Environmental Protection Agency:
Washington D.C. Available at https://www.epa.aov/sites/default/files/2018-11/documents/trends-blood-mercurv-
concentrations-report.pdf.
United States Environmental Protection Agency. (2001) Integrated Risk Information System Chemical Assessment Summary:
Methylmercury(MeHg)CASRN 22967-92-6. United States Environmental Protection Agency, National Centerfor
Environmental Assessment. Available at https://iris.epa.gov/Chemicall_andino/&substance_nmbr^73.
United States Environmental Protection Agency. (1997) Mercury Study Report to Congress, Volume IV: An Assessment of Exposure
to Mercury in the United States, Report No. EPA/452-R-97-006; USEPA Office of Air Quality Planning & Standards and
Office of Research and Development.: Washington, D.C.
Washington State Department of Ecology. (2011-2016) Measuring Mercury Trends in Freshwater Fish in Washington State in:
Ecology, W.S. D. o. (Ed.).
Whitney, M. C. (2021) Metal accumulation in Lake Michigan prey fish: Influence of ontogeny, trophic position, and habitat. Journal
of Great Lakes research, v. 47, pp. 1746-1755-2021 v.1747 no. 1746. https://doi.org/10.1016/i.ialr.2021.08.019.
Willacker, J. J., Eagles-Smith, C.A., Blazer, V.S. (2020) Total Mercury Concentrations in Smallmouth Bass from Chesapeake Bay
Tributaries, USA Dataset, 2013-2017 in: United States Geological Survey (Ed.). Available at
https://data.usgs.gov/datacataloo/data/USGS:5d7ff0f3e4b0c4f70d04be43.
Wolff, S.; Brown, G.; Chen, J.; Meals, K.; Thornton, C.; Brewer, S.; Cizdziel, J. V.; Willett, K. L. (2016) Mercury concentrations in fish
from three major lakes in north Mississippi: Spatial and temporal differences and human health risk assessment. J
Toxicol Environ Health A, 79,894-904. https://doi.org/10.1080/15287394.2016.1194792.
47
-------
APPENDIX A
Extended Data Tables
-------
Table A-1 Fish tissue mercury concentrations used in analysis, by species
Species group
ygHg/g
fish
Species group
vgHg/g
fish
Species group
vgHg/g
fish
Swordfish
0.739
Snapper
0.106
Herring/Shad
0.042
Barracuda
0.483
Catfish
0.090
Smelt
0.036
Shark
0.410
Fish, not specified
0.083
Salmon
0.036
Ray
0.410
Haddock
0.066
Mullet
0.032
Mackerel
0.311
Cod
0.065
'Shellfish, not
specified
0.029
Tuna (Fresh/Frozen)
0.281
Crab
0.064
Octopus
0.017
Pike
0.229
Croaker
0.058
Abalone
0.016
Pompano/Mahi Mahi
0.190
Trout
0.056
Sardine
0.016
Perch/Bass/Walleye/
Bluegill
0.179
Whitefish
0.054
Anchovy
0.016
Lobster
0.173
Crayfish
0.053
Mussel
0.015
Halibut
0.170
Whiting
0.052
Scallop
0.015
Rockfish/Redfish/
Orange Roughy
0.169
Flatfish
0.050
Shrimp
0.014
Sea Bass
0.165
Squid
0.049
Oyster
0.014
Eel
0.152
Breaded Fish
Products
0.047
Clam
0.011
'Tuna
0.123
Conch
0.047
Pollock
0.011
Sturgeon
0.122
Snail
0.047
Jellyfish
0.010
Tuna(Canned)
0.113
20ther shellfish
0.047
Tilapia
0.010
Carp/Sucker
0.109
Porgy/Scup
0.046
Caviar/Roe
0.000
'Tuna, weighted mean of Tuna (Fresh/Frozen) and Tuna (Canned).
20ther shellfish, weighted mean ofabalone, conch, jellyfish, octopus, squid, and snail.
3Shellfish, NS (non-specified), weighted mean of abalone, clam, conch, crayfish, jellyfish, lobster, mussel, octopus,
oyster, scallop, shrimp, squid, crab, and snail.
A-l
-------
Table A-2. Distribution of blood MeHg concentrations ((jg/L), by NHANES survey releases, age, income, and race/ethnicity, women aged 16-49 years, NHANES
2013-March 2020
N
Arithmetic
Geometric mean
Selected percentiles (95% CI)
mean
(95% CI)
25th
50th
75th
90th
95th
(95% CI)
NHANES Survey Release
1999-2000
1632
1.83(1.43,2.22)
0.93(0.78,1.12)
0.41(0.31, 0.51)
0.89(0.68,1.09)
1.88(1.16,2.60)
4.56(3.25,5.86)
6.82(3.57,10.07)
2001-2002
1799
1.30(1.13,1.47)
0.72(0.65,0.79)
0.34(0.29,0.38)
0.70(0.60, 0.79)
1.46(1.23,1.70)
2.92(2.56,3.28)
4.42(3.44,5.39)
2003-2004
1615
1.06(0.91,1.22)
0.53(0.45,0.61)
0.24(0.19,0.29)
0.52(0.40, 0.64)
1.21(0.98,1.44)
2.60(2.00,3.19)
3.77(2.87,4.67)
2005-2006
1788
1.13(0.99,1.27)
0.58(0.51, 0.66)
0.25(0.18, 0.32)
0.63(0.52,0.73)
1.33(1.07,1.59)
2.71(2.19,3.23)
3.99(2.93,5.05)
2007-2008
1486
0.99(0.84,1.13)
0.52(0.46, 0.59)
0.25(0.20,0.29)
0.51(0.43,0.59)
1.06(0.82,1.30)
2.39(1.81,2.97)
3.46(2.70,4.23)
2009-2010
1780
1.18(1.08,1.29)
0.67(0.61, 0.73)
0.31(0.26,0.37)
0.64(0.53, 0.74)
1.38(1.15,1.61)
2.73(2.41,3.04)
4.03(3.50,4.55)
2011-2012
1428
0.93(0.75,1.11)
0.52(0.44, 0.60)
0.23(0.20, 0.26)
0.45(0.37,0.54)
0.98(0.65,1.32)
2.32(1.66,2.99)
3.31(2.22,4.41)
2013-2014
814
0.98(0.84,1.12)
0.50(0.45,0.55)
0.23(0.18,0.28)
0.43(0.37, 0.49)
0.94(0.78,1.11)
2.21(1.62,2.80)
3.66(3.13,4.20)
2015-2016
740
0.90(0.80,1.01)
0.50(0.45,0.56)
0.24(0.21, 0.27)
0.44(0.37, 0.51)
0.94(0.77,1.10)
2.19(1.71,2.68)
3.38(2.79,3.96)
2017-2020
2154
0.87(0.75, 0.98)
0.48(0.43, 0.54)
0.18(0.15, 0.21)
0.43(0.36, 0.50)
1.01(0.79,1.22)
1.99(1.54,2.44)
3.06(2.29,3.83)
Race/Ethnicity
Mexican American
3396
0.69(0.65,0.73)
0.44(0.42,0.46)
0.24(0.23, 0.26)
0.42(0.39,0.45)
0.80(0.74, 0.85)
1.44(1.33,1.56)
2.09(1.77,2.41)
Other Hispanic
1259
1.16(1.00,1.31)
0.66(0.62, 0.71)
0.31(0.28,0.34)
0.65(0.59,0.71)
1.39(1.24,1.54)
2.60(2.27,2.92)
3.68(3.02,4.34)
Non-Hispanic White
5725
1.05(0.98,1.11)
0.54(0.51, 0.57)
0.23(0.22,0.25)
0.51(0.47,0.54)
1.17(1.09,1.24)
2.54(2.27,2.82)
3.84(3.52,4.17)
Non-Hispanic Black
3497
1.13(1.05,1.21)
0.66(0.62, 0.70)
0.32(0.30,0.34)
0.64(0.60,0.69)
1.27(1.16,1.37)
2.38(2.10,2.66)
3.61(3.19,4.03)
Other Race
1359
2.01(1.85,2.17)
0.95(0.88,1.03)
0.34(0.29,0.38)
0.94(0.82,1.06)
2.50(2.18,2.83)
5.27(4.64,5.90)
7.64(6.26,9.01)
Age, Years
16 to 19
3349
0.63(0.59,0.68)
0.36(0.35, 0.38)
0.16(0.15,0.17)
0.33(0.30,0.35)
0.69(0.63, 0.75)
1.46(1.32,1.60)
2.10(1.94,2.26)
20 to 29
4111
0.97(0.91,1.03)
0.52(0.49,0.55)
0.23(0.20, 0.25)
0.49(0.45,0.53)
1.12(1.01,1.22)
2.34(2.14,2.55)
3.43(3.01,3.84)
30 to 39
3943
1.22(1.13,1.32)
0.63(0.60, 0.66)
0.28(0.26,0.29)
0.60(0.56,0.65)
1.34(1.22,1.45)
2.92(2.54,3.29)
4.30(3.79,4.81)
40 to 49
3833
1.27(1.19,1.34)
0.68(0.65, 0.72)
0.32(0.30,0.34)
0.65(0.60,0.69)
1.39(1.28,1.51)
3.03(2.75,3.30)
4.31(4.00,4.62)
Annual Income
Oto <1x poverty line
3814
0.79(0.73,0.85)
0.44(0.42, 0.47)
0.22(0.20, 0.24)
0.41(0.38,0.45)
0.84(0.78, 0.90)
1.67(1.49,1.84)
2.74(2.41,3.07)
A-2
-------
N
Arithmetic
Geometric mean
Selected percentiles (95% CI)
mean
(95% CI)
25th
50th
75th
90th
95th
(95% CI)
Ix to <2x poverty
3556
0.83(0.78, 0.88)
0.46(0.44,0.48)
0.22(0.20, 0.24)
0.43(0.40, 0.46)
0.90(0.83,0.97)
1.76(1.59,1.94)
2.74(2.40,3.07)
2x to <3x poverty line
1995
1.00(0.92,1.08)
0.53(0.50, 0.57)
0.25(0.23,0.27)
0.50(0.46,0.54)
1.11(0.98,1.24)
2.40(2.17,2.63)
3.31(2.78,3.85)
3x to <4x poverty line
1580
1.06(0.98,1.14)
0.57(0.53,0.61)
0.25(0.21,0.28)
0.56(0.50, 0.62)
1.17(1.05,1.28)
2.54(2.18,2.90)
3.74(3.33,4.15)
4x to <5x poverty line
1102
1.15(1.04,1.27)
0.62(0.57,0.66)
0.29(0.26, 0.32)
0.59(0.51,0.68)
1.25(1.11,1.38)
2.71(2.19,3.23)
4.36(3.35,5.38)
>=5x poverty line
2051
1.66(1.55,1.78)
0.89(0.83, 0.94)
0.39(0.34,0.43)
0.92(0.84,1.01)
2.02(1.80,2.24)
3.98(3.66,4.30)
5.90(5.03,6.77)
Missing/Refused/DK
1138
1.25(1.08,1.42)
0.64(0.58,0.71)
0.27(0.24, 0.31)
0.57(0.50, 0.65)
1.47(1.14,1.79)
3.01(2.20,3.83)
4.51(3.28,5.74)
Education
Median education forage
3411
1.52(1.42,1.62)
0.80(0.75, 0.85)
0.35(0.31,0.38)
0.79(0.72, 0.87)
1.82(1.63,2.01)
3.78(3.53,4.04)
5.33(4.69,5.97)
Unknown education
373
0.51(0.42, 0.61)
0.33(0.29, 0.38)
0.16(0.15,0.17)
0.27(0.21,0.34)
0.58(0.45, 0.72)
1.10(0.65,1.55)
1.61 (, f
a. Missing confidence interval because of stratum with single sampling unit.
A-3
-------
Table A-J. Regression parameter estimates and p-valuesfrom models predicting log-transformed blood MeHg
concentrations
Dependent variable
Log-transformed MeHg
Parameter
Estimate
LCL
UCL
tValue
Probt
fValue
ProbF
Intercept
0.8410
0.5293
1.1528
5.3297
0.0000
Factors affecting the intercept
Education
5.6554
0.0007
< Median forage
-0.2932
-0.5570
-0.0293
-2.1952
0.0297
>Median forage
0.2599
-0.0359
0.5558
1.7356
0.0846
Unknown education
0.0788
-0.5782
0.7357
0.2369
0.8131
Median forage
-0.0455
-0.2813
0.1903
-0.3814
0.7034
Annual Income
2.0628
0.0541
Oto =5x poverty line
0.1345
0.0076
0.2614
2.0942
0.0379
Missing/Refused/DK
-0.0022
-0.1654
0.1610
-0.0262
0.9791
Race/Ethnicity
6.6549
2.37E-05
Mexican American
-0.4295
-0.6110
-0.2480
-4.6743
0.0000
Other Hispanic
0.1070
-0.1179
0.3318
0.9399
0.3488
Non-Hispanic White
0.4437
0.1775
0.7099
3.2926
0.0012
Non-Hispanic Black
-0.0661
-0.2674
0.1352
-0.6487
0.5175
Other Race
-0.0551
-0.3539
0.2437
-0.3642
0.7162
NHANES Survey release
6.0979
1.28E-08
1999-2000
0.4773
0.3125
0.6422
5.7200
0.0000
2001-2002
0.0918
-0.0898
0.2734
0.9991
0.3193
2003-2004
-0.1155
-0.2421
0.0111
-1.8028
0.0734
2005-2006
-0.0901
-0.2296
0.0495
-1.2754
0.2041
2007-2008
0.0132
-0.1029
0.1292
0.2243
0.8228
2009-2010
0.1350
0.0166
0.2534
2.2534
0.0257
2011-2012
-0.1197
-0.2953
0.0559
-1.3462
0.1802
2013-2014
-0.1850
-0.2957
-0.0742
-3.3010
0.0012
2015-2016
-0.0640
-0.2403
0.1123
-0.7169
0.4745
2017-2020
-0.1431
-0.2852
-0.0011
-1.9909
0.0483
Age, decade (centered)
0.0212
-0.0290
0.0714
0.8328
0.4063
Log-transformed body weight
(centered)
-0.7472
-1.2101
-0.2843
-3.1887
0.0017
Log-transformed hematocrit
(centered)
1.8543
0.5673
3.1413
2.8462
0.0050
A-4
-------
Dependent variable
Log-transformed MeHg
Age, decade (centered^log-
transformed body weight (centered)
-0.0769
-0.2327
0.0789
-0.9753
0.3310
Factors predicting the slope for transformed usual fish intake
Transformed usual fish intake
1.5657
1.3056
1.8258
11.8926
0.0000
Transformed usual fish intake*log-
transformed body weight (centered)
-0.6468
-1.0244
-0.2691
-3.3839
0.0009
Transformed usual fish intake*log-
trans formed hematocrit (centered)
0.8567
-0.2088
1.9222
1.5885
0.1143
Transformed usual fish
intake*Race/Ethnicity
8.4232
8.63E-07
Transformed usual fish intake*Mexican
American
-0.2898
-0.4434
-0.1361
-3.7279
0.0003
Transformed usual fish
intake*Other Hispanic
-0.0686
-0.2514
0.1141
-0.7422
0.4591
Transformed usual fish
intake*Non-Hispanic White
0.5229
0.3130
0.7328
4.9212
0.0000
Transformed usual fish
intake*Non-Hispanic Black
-0.1758
-0.3507
-0.0010
-1.9865
0.0488
Transformed usual fish
intake*Other Race
0.0113
-0.2342
0.2569
0.0911
0.9276
Transformed usual fish
intake*Education
3.7225
0.0109
Transformed usual fish intake*< Median
forage
-0.1855
-0.3953
0.0243
-1.7470
0.0827
Transformed usual fish intake*>Median
forage
0.1903
-0.0485
0.4290
1.5741
0.1175
Transformed usual fish
intake*Unknown Edu
0.0261
-0.5024
0.5547
0.0977
0.9223
Transformed usual fish intake*Median
forage
-0.0309
-0.2208
0.1589
-0.3221
0.7479
TUIV
-0.0177
-0.1347
0.0993
-0.2986
0.7657
Note: p-values in bold areforthe F-Test for differences across categories in categorical predictors.
A-5
-------
Table A-4. Weighted percentages and their standard errors for categorized reports of 30-day frequency of consumption offish, by NHANES survey releases,
income, race/ethnicity, age, and education, women aged 16-49 years, NHANES 2013-March 2020
Parameter
N
0 times
Itime
Percent (standard error)
2 times 3 times
4-5 times
6+times
NHANES Survey release
Total Fish
2013-2014
814
27.2 (2.4)
16.0(2.0)
11.6(1.2)
8.9(1.1)
9.8(1.4)
26.4(2.2)
2015-2016
740
26.5(2.1)
15.2(1.3)
10.1(2.0)
7.8(1.1)
14.9(1.4)
25.5(2.5)
2017-March 2020
2,154
25.5(1.6)
13.8(1.1)
12.4(1.2)
9.0(0.9)
13.3(1.1)
26.1(2.1)
Finfish Only
2013-2014
814
39.3(2.6)
19.0(2.2)
11.2(1.2)
7.4(1.1)
8.3(1.4)
14.7(1.4)
2015-2016
740
38.0(1.9)
17.1(1.7)
12.6(1.6)
8.7(1.4)
10.9(1.7)
12.8(1.5)
2017-March 2020
2,154
37.8(1.3)
15.4(1.2)
14.6(1.0)
8.7(0.6)
10.2(1.1)
13.3(0.9)
Shellfish Only
2013-2014
814
49.3(2.6)
16.0(1.2)
13.3(1.0)
6.1(1.1)
6.9(1.3)
8.4(1.6)
2015-2016
740
49.1(4.0)
17.9(2.2)
12.3(1.4)
5.6(1.2)
6.2(0.9)
9.0(1.3)
2017-March 2020
2,154
47.6(3.1)
17.7(1.1)
11.2(0.9)
8.2(1.2)
8.1(1.4)
7.3(1.0)
Income
Total Fish
Oto <1x poverty line
876
33.5(1.8)
16.1(1.4)
11.3(1.4)
11.1(1.5)
11.1(1.4)
16.9(1.9)
1xto <2x poverty line
856
30.8(2.1)
15.9(1.9)
9.6(1.4)
9.5(1.3)
11.2(1.3)
23.1(2.0)
2xto <3x poverty line
516
27.0(2.6)
17.3(2.3)
11.3(1.8)
11.2(1.8)
10.9(1.8)
22.3(2.1)
3x to <4x poverty line
368
25.2(3.9)
14.0(2.4)
12.9(2.3)
6.0(1.3)
15.2(2.2)
26.8(4.1)
4xto <5x poverty line
251
20.0(3.1)
13.1(2.5)
15.3(3.3)
8.8(2.3)
10.5(2.3)
32.2(4.4)
>= 5x poverty line
498
21.0(2.5)
11.7(2.1)
10.1(1.6)
5.8(1.4)
15.1(1.9)
36.4(2.9)
Missing/Refused/DK
343
19.5(2.5)
15.9(2.5)
13.9(2.9)
7.3(1.9)
16.8(3.3)
26.6(3.6)
Finfish Only
Oto <1x poverty line
876
45.5(1.9)
17.6(1.6)
13.1(1.4)
8.0(1.3)
5.6(0.8)
10.2(1.4)
1xto <2x poverty line
856
42.1(2.2)
17.0(1.8)
11.9(1.7)
8.7(1.0)
9.4(1.2)
10.9(1.4)
2xto <3x poverty line
516
40.1(2.7)
20.8(2.9)
13.7(1.9)
4.9(0.8)
8.5(1.5)
12.0(1.8)
3x to <4x poverty line
368
37.0(3.9)
16.1(2.6)
11.1(1.9)
9.7(2.1)
11.8(2.5)
14.3(3.0)
4xto <5x poverty line
251
32.0(3.8)
15.1(3.2)
15.5(3.4)
8.7(1.9)
11.1(3.0)
17.5(3.2)
A-6
-------
Parameter
N 0 times
>=5x poverty line 498 31.9(2.7)
Missing/Refused/DK 343 34.3(3.2)
Shellfish Only
Oto <1x poverty line
876
56.5(2.4)
1xto <2x poverty line
856
54.2(2.7)
2xto <3x poverty line
516
49.4(3.1)
3x to <4x poverty line
368
48.4(4.5)
4xto <5x poverty line
251
44.0(4.3)
>= 5x poverty line
498
39.6(3.5)
Missing/Refused/DK
343
41.4(3.9)
Race/Ethnicity
Total Fish
Mexican American
624
23.4(1.8)
Other Hispanic
380
21.2(2.3)
Non-Hispanic White
1,199
30.0(1.7)
Non-Hispanic Black
877
18.4(2.1)
Other Race
628
22.3(2.1)
Finfish Only
Mexican American
624
40.5(2.2)
Other Hispanic
380
34.5(2.3)
Non-Hispanic White
1,199
41.4(1.5)
Non-Hispanic Black
877
31.8(2.2)
Other Race
628
29.2(2.2)
Shellfish Only
Mexican American 624 39.8(2.2)
Other Hispanic 380 43.9(2.7)
Non-Hispanic White 1,199 53.1(2.9)
Non-Hispanic Black 877 42.5(2.7)
Other Race 628 43.6(2.9)
Percent (standard error)
7 time 2 times 3 times 4-5 times 6+times
14.1(2.1) 12.1(2.0) 10.6(1.6) 12.4(2.1) 18.9(1.8)
18.3(2.3) 17.4(3.2) 6.3(1.5) 12.6(2.5) 11.0(2.3)
16.6(1.3)
9.2(1.2)
7.0(1.1)
6.0(1.0)
4.6(0.8)
18.2(1.8)
10.1(1.2)
5.0(0.9)
5.9(0.8)
6.6(1.3)
16.5(2.6)
14.0(2.1)
6.6(1.5)
6.2(1.3)
7.2(1.4)
17.5(3.0)
15.8(2.5)
3.2(0.9)
7.9(2.1)
7.2(2.1)
19.5(3.6)
9.9(1.9)
8.5(2.7)
9.4(2.2)
8.6(2.3)
16.0(2.3)
13.8(2.0)
7.4(1.3)
9.8(2.3)
13.4(1.9)
18.1(2.6)
12.5(2.4)
14.4(2.9)
4.7(1.1)
8.8(2.1)
16.9(1.7)
14.6(1.5)
10.8(1.4)
14.5(1.5)
19.9(2.1)
17.8(2.1)
12.4(2.9)
9.6(1.4)
12.7(2.8)
26.3(2.9)
15.4(1.2)
10.4(1.3)
7.5(0.9)
11.8(1.3)
24.8(1.7)
12.5(1.1)
13.4(1.3)
10.7(1.2)
15.6(1.4)
29.4(2.5)
10.1(1.5)
11.0(1.3)
9.2(1.6)
12.1(2.1)
35.4(2.7)
20.2(1.7)
14.5(1.3)
8.9(1.3)
6.0(1.2)
9.9(1.9)
20.5(1.9)
13.8(2.0)
7.8(1.7)
8.9(1.3)
14.5(2.0)
16.0(1.3)
12.2(1.1)
7.2(0.8)
10.5(1.1)
12.7(1.0)
16.7(1.8)
16.7(1.4)
9.8(1.4)
10.1(1.2)
14.8(2.0)
15.7(1.5)
10.9(1.3)
12.6(2.2)
11.2(1.6)
20.3(2.2)
24.5(1.6) 15.4(1.4) 8.3(1.3) 7.7(1.0) 4.2(1.0)
21.6(2.3) 12.0(1.7) 6.7(1.3) 7.1(1.5) 8.6(2.2)
16.2(1.4) 10.7(1.0) 6.3(1.1) 6.9(1.2) 6.7(0.8)
13.6(1.3) 15.8(1.4) 8.3(1.1) 8.6(1.2) 11.3(1.5)
16.0(1.9) 11.4(1.6) 6.5(1.2) 6.5(0.9) 15.9(1.9)
A-7
-------
Parameter
Age
Total Fish
Finfish Only
Shellfish Only
Education
Total Fish
Finfish Only
Shellfish
16 to 19 years 654
20 to 29 years 963
30 to 39 years 1,056
40 to 49 years 1,035
16 to 19 years 654
20 to 29 years 963
30 to 39 years 1,056
40 to 49 years 1,035
16 to 19 years 654
20 to 29 years 963
30 to 39 years 1,056
40 to 49 years 1,035
< Median forage 1,175
Median for age 1,256
>Medianforage 904
Unknown education 373
< Median forage 1,175
Median for age 1,256
>Medianforage 904
Unknown education 373
< Median forage 1,175
Median for age 1,256
0 times
41.1(2.6)
29.2(1.8)
24.4(1.7)
19.3(1.7)
58.2(2.6)
42.1(1.8)
33.9(1.8)
30.8(2.1)
59.4(2.6)
48.1(2.6)
46.3(2.5)
46.8(2.7)
33.8(2.0)
24.3(1.7)
19.9(2.0)
35.6(2.7)
46.4(2.0)
36.5(1.7)
30.0(2.1)
55.2(3.6)
55.0(2.5)
49.7(2.6)
Percent (standard error)
7 time 2 times 3 times 4-5 times 6+times
18.3(2.6)
14.5(1.4)
13.4(1.2)
15.2(1.7)
19.6(2.1)
14.9(1.6)
16.1(1.4)
18.7(1.9)
15.9(1.9)
17.3(1.8)
15.9(1.2)
19.2(1.6)
11.2(1.5)
9.8(1.1)
10.8(1.5)
14.1(1.5)
8.3(1.2)
13.9(1.7)
14.0(1.1)
13.0(1.4)
8.5(1.0)
11.5(1.3)
11.3(1.2)
15.0(1.3)
8.6(1.2)
8.7(1.2)
8.6(1.0)
8.8(1.0)
7.5(1.5)
6.2(1.0)
9.5(1.7)
9.7(1.3)
7.2(1.3)
5.8(1.0)
8.3(1.2)
6.3(1.1)
10.3(2.0)
12.0(1.3)
12.2(1.1)
15.0(1.4)
3.7(0.9)
9.1(1.3)
11.9(1.5)
11.1(1.5)
5.5(1.4)
8.6(1.2)
8.3(1.3)
5.4(1.0)
10.5(1.7)
25.8(1.8)
30.6(1.9)
27.7(2.0)
2.6(0.6)
13.8(1.3)
14.5(1.4)
16.7(1.5)
3.6(0.9)
8.7(1.2)
9.9(1.2)
7.3(1.2)
16.9(1.6)
16.4(1.3)
10.8(1.3)
17.6(3.3)
19.1(1.7)
18.1(1.4)
13.3(1.5)
18.8(2.9)
17.7(1.4)
17.7(1.5)
12.3(1.3)
11.9(1.2)
10.7(1.4)
8.9(1.6)
13.8(1.3)
13.3(1.2)
13.0(1.4)
7.4(1.6)
12.1(1.3)
11.0(0.9)
9.1(1.0)
8.7(1.0)
8.0(1.2)
9.9(1.7)
6.7(0.9)
7.8(1.0)
9.9(1.1)
11.7(2.5)
5.8(0.9)
6.9(0.8)
10.9(1.3)
13.8(1.3)
12.9(1.4)
15.1(3.6)
6.8(1.0)
10.4(1.0)
13.2(1.6)
3.3(1.1)
4.6(0.9)
6.6(0.7)
16.9(1.5)
24.9(1.6)
37.7(2.6)
13.0(2.5)
7.3(0.9)
14.0(1.2)
20.5(1.7)
3.5(1.0)
4.8(0.8)
8.0(1.1)
A-8
-------
Parameter
N
0 times
Itime
Percent (standard error)
2 times 3 times
4-5 times
6+times
>Medianforage
Unknown education
904
373
40.6(2.8)
51.6(3.3)
16.2(1.7)
18.6(2.2)
13.8(1.4) 7.4(1.4)
8.6(1.4) 9.0(1.8)
10.3(1.7)
6.8(2.1)
11.7(1.4)
5.4(1.8)
A-9
-------
Table A-5. Estimated amount offish consumed (g) in last 30 days, by NHANES survey releases, women aged 16-49 years, NHANES 2013-March 2020
N Arithmetic mean Selected percentiles (95% CI)
(95% Cl) 25th 50th 75th 90th 95th
Estimated amount of fish consumed (g)in last 30 days
Shellfish
2013-2014 814 97.6(69.2,126.0)
2015-2016 740 91.3(77.8,104.7)
2017-March 2020 2154 93.0(79.8,106.1)
0.0(-20.7,20.7) 28.1(7.3,48.9) 107.5(74.0,141.1) 266.7(186.9,346.5) 407.4(285.6,529.1)
0.0(-21.2,21.2) 28.6(7.4,49.8) 99.3(70.4,128.2) 271.0(189.8,352.3) 417.4(354.2,480.7)
0.0(-20.9,20.9) 37.9(16.7,59.0) 104.6(76.3,132.9 ) 249.3(201.1,297.5 ) 393.7(326.9,460.4)
A-10
-------
Table A-6. Estimated mercury intake (|jg) in last 30 days, by NHANES survey releases, women aged 16-49 years, NHANES 2013-March 2020
N
Arithmetic mean
(95% CI)
25th
50th
Selected percentiles (95% CI)
75th 90th
95th
Estimated mercury intake (fig) in last 30 days
Shellfish
2013-2014
814
3.65(2.30,4.99)
0.00(-0.29,0.29)
0.34(0.05,0.63)
2.71(1.02,4.39)
10.01(4.57,15.46)
17.51(13.26,21.77)
2015-2016
740
3.37(2.75,3.99)
0.00(-0.29,0.29)
0.36(0.07,0.65)
2.05(0.56,3.55)
11.87(7.72,16.02)
19.47(13.46,25.48)
2017-March 2020
2154
3.22(2.74,3.70)
0.00(-0.29,0.29)
0.52(0.23,0.81)
2.10(1.02,3.18)
9.63(6.69,12.56)
15.84(13.93,17.75)
Finfish
2013-2014
814
21.97(19.47,24.47)
0.00(-1.82,1.82)
8.03(5.29,10.78)
26.25(22.40,30.09)
57.05(46.69,67.41)
91.87(68.70,115.04)
2015-2016
740
19.46(16.42,22.50)
0.00(-1.76,1.76)
8.55(6.99,10.12)
23.04(17.89,28.20)
47.94(35.97,59.90)
81.46(50.33,112.59)
2017-March 2020
2154
20.51(18.37,22.65)
0.00(-1.70,1.70)
8.02(6.85,9.19)
23.63(20.57,26.69)
54.35(46.25,62.44)
85.81(69.88,101.74)
Total Fish
2013-2014
814
25.62(22.16,29.08)
0.00(-0.57,0.57)
9.39(6.48,12.30)
30.93(26.22,35.64)
66.46(51.48,81.44)
104.26(75.57,132.94)
2015-2016
740
22.83(19.55,26.12)
0.00(-0.30,0.30)
9.91(7.34,12.47)
29.16(24.93,33.39)
55.94(39.64,72.23)
84.27(51.47,117.07)
2017-March 2020
2154
23.73(21.24,26.21)
0.00(-0.29,0.29)
9.55(7.94,11.16)
28.59(25.14,32.03)
62.66(55.21,70.11)
91.02(75.33,106.72)
A-ll
-------
Table A- 7. Estimated mercury intake per unit body weight (pg Hg/kg bw)in last 30 days, by NHANES survey releases, women aged 16-49 years, NHANES 2013-March 2020
N
Arithmetic Mean
(95% CI)
25th
50th
Selected percentiles (95% CI)
75th
90th
95th
Estimated mercury intake per unit body weight (fig Hg/kg bw) in last 30 days
Shellfish
2013-2014
814
0.05(0.03,0.06)
0.00(0.00,0.00)
0.00(0.00,0.01)
0.03(0.01,0.05)
0.13(0.07,0.19)
0.24(0.16,0.31)
2015-2016
740
0.05(0.04,0.06)
0.00(0.00,0.00)
0.00(0.00,0.01)
0.03(0.02,0.05)
0.15(0.10,0.20)
0.26(0.21,0.32)
2017-March 2020
2154
0.04(0.04,0.05)
0.00(0.00,0.00)
0.01(0.00,0.01)
0.03(0.02,0.05)
0.12(0.09,0.15)
0.22(0.19,0.25)
Finfish
2013-2014
814
0.31(0.28,0.35)
0.00(-0.03,0.03)
0.09(0.05,0.13)
0.36(0.30,0.42)
0.84(0.60,1.08)
1.30(0.98,1.62)
2015-2016
740
0.27(0.23,0.31)
0.00(-0.03,0.03)
0.10(0.07,0.12)
0.32(0.26,0.38)
0.72(0.53,0.91)
1.12(0.78,1.45)
2017-March 2020
2154
0.28(0.25,0.31)
0.00(-0.02,0.02)
0.10(0.08,0.12)
0.32(0.27,0.37)
0.74(0.60,0.87)
1.15(0.90,1.41)
Total Fish
2013-2014
814
0.36(0.31,0.41)
0.00(-0.01,0.01)
0.13(0.08,0.17)
0.44(0.35,0.52)
0.93(0.69,1.17)
1.39(0.92,1.87)
2015-2016
740
0.32(0.27,0.37)
0.00(0.00,0.00)
0.13(0.10,0.17)
0.39(0.32,0.46)
0.82(0.60,1.04)
1.34(0.92,1.76)
2017-March 2020
2154
0.32(0.29,0.36)
0.00(0.00,0.00)
0.13(0.10,0.16)
0.38(0.34,0.43)
0.81(0.71,0.92)
1.30(1.09,1.51)
A-12
-------
Table A-8. Estimated amounts offish consumed (g), mercury intake ((jg), and mercury intake per unit body weight ((jg/kg), by race/ethnicity, age, income, and
education, women aged 16-49 years, NHANES 2013-March 2020
Parameter
N
Arithmetic mean
(95% CI)
25th
50th
75th
90th
95th
Estimated amount of fish consumed (g)
Race/Ethnicity
Mexican American
624
300.4(250.9,349.9
37.6(17.7,57.5)
149.3(126.3,172.3)
373.1(308.8,437.3)
706.0(549.8,862.3)
1013.3(709.0,1317.5)
Other Hispanic
380
329.6(267.9,391.3)
32.6(14.7,50.5)
155.2(112.3,198.2)
424.1(322.0,526.3
769.2(544.4,994.0)
1331.9(678.9,1984.8)
Non-HispanicWhite
1199
305.4(278.1,332.7)
0.0(-20.5,20.5)
143.4(109.4,177.4)
412.7(352.5,473.0)
791.8(715.6,868.0)
1151.5(972.1,1330.9)
Non-Hispanic Black
877
443.2(375.9,510.5)
71.2(38.7,103.6)
232.0(196.2,267.8
510.9(424.2,597.7)
1057.1(764.7,1349.5)
1600.3(1048.8,2151.8)
Other Race
628
324.7(288.3,361.1)
38.8(14.8,62.7)
174.7(130.6,218.7)
449.0(387.1,510.9)
871.1(694.4,1047.8)
1170.1(973.9,1366.4)
Age, Years
16 to 19
654
142.0(122.8,161.1)
0.0(-21.2,21.2)
59.5(42.8,76.2)
200.6(157.4,243.9)
384.4(295.2,473.6)
594.0(496.3,691.8)
20 to 29
963
319.4(286.1,352.7)
0.0(-21.4,21.4)
147.8(118.9,176.7)
436.9(374.3,499.4)
843.9(711.4,976.3)
1146.5(977.5,1315.5)
30 to 39
1056
374.6(336.3,413.0)
31.4(9.0,53.8)
186.9(138.7,235.0)
503.0(430.1,575.9)
959.7(800.7,1118.7)
1300.1(1052.0,1548.1)
40 to 49
1035
358.5(326.7,390.3)
58.9(44.0,73.8)
195.5(158.4,232.5)
464.5(400.9,528.1)
857.6(800.3,914.9)
1391.3(1106.7,1675.9)
Annual Income
Oto <1x poverty line
876
265.9(223.9,307.9)
0.0(-21.1,21.1)
105.3(78.4,132.2)
317.7(254.0,381.4)
689.2(556.7,821.7)
993.5(738.6,1248.3)
1xto <2x poverty
856
277.0(243.7,310.3)
0.0(-21.4,21.4)
130.8(97.5,164.2)
389.2(339.6,438.8)
734.6(618.8,850.3)
1039.6(784.4,1294.9)
2x to <3x poverty line
516
295.4(250.0,340.8)
0.0(-29.0,29.0)
136.4(101.2,171.7)
369.1(280.8,457.5)
762.5(547.6,977.5)
1184.7(813.2,1556.1)
3x to <4x poverty line
368
317.8(256.3,379.2)
0.0(-33.6,33.6)
161.4(97.9,224.9)
463.5(314.6,612.4)
820.2(631.9,1008.5)
1104.4(717.8,1491.0)
4xto <5x poverty line
251
396.0(326.0,466.0)
45.1(1.2,89.0)
199.6(127.5,271.8)
531.1(390.1,672.0)
968.8(709.1,1228.5)
1373.9(585.9,2161.8)
>= 5x poverty line
498
426.3(382.0,470.6)
44.7(8.1,81.4)
279.3(209.4,349.2)
591.6(509.8,673.3)
1085.6(954.2,1217.0)
1349.8(1175.0,1524.6)
Missing/Refused/DK
343
331.6(287.7,375.6)
42.4(14.2,70.7)
192.6(134.4,250.7)
380.8(316.6,444.9)
773.9(631.5,916.3)
1098.7(566.2,1631.3)
Education
Median education forage
904
425.4(387.7,463.2)
64.8(41.9,87.8)
273.8(214.3,333.3)
626.2(561.2,691.3)
1024.1(904.8,1143.4)
1376.7(1180.9,1572.4)
Unknown education
373
170.2(137.0,203.4)
0.0(-20.9,20.9)
76.4(38.2,114.6)
254.2(182.6,325.7)
437.8(307.4,568.1)
656.0(516.1,796.0)
Estimated mercury intake (fig)
Race/Ethnicity
Mexican American
624
22.30(17.24,27.37)
0.43(0.19,0.68)
9.39(6.61,12.18)
25.28(19.43,31.12)
56.13(47.26,65.01)
82.21(43.82,120.60)
Other Hispanic
380
24.49(19.88,29.10)
0.41(0.17,0.65)
10.39(7.56,13.23)
28.28(21.98,34.57)
63.97(52.32,75.62)
83.77(50.14,117.41)
Non-HispanicWhite
1199
22.83(20.90,24.75)
0.00(-0.27,0.27)
9.11(7.29,10.93)
28.19(24.99,31.38)
60.57(52.58,68.57)
95.47(78.19,112.75)
Non-Hispanic Black
877
32.03(27.29,36.77)
1.12(-0.29,2.52)
15.38(11.82,18.95)
36.45(31.17,41.73)
72.63(56.01,89.26)
123.50(87.85,159.15)
Other Race
628
22.15(19.27,25.04)
0.86(-0.35,2.07)
9.38(7.02,11.73)
29.42(24.25,34.59)
62.97(49.79,76.15)
89.73(67.39,112.06)
Age, Years
16 to 19
654
9.12(7.94,10.31)
0.00(-0.30,0.30)
1.91(0.51,3.31)
12.41(9.90,14.91)
26.18(22.06,30.30)
35.65(25.90,45.40)
A-13
-------
Parameter N Arithmetic mean 25th
(95% CI)
20 to 29 963 22.18(19.86,24.51) 0.00(-0.30,0.30)
30 to 39 1056 26.69(23.82,29.56) 0.40(-0.17,0.97)
40 to 49 1035 29.16(26.05,32.26) 0.94(0.14,1.74)
Annual Income
0 to <1x poverty line 876 19.86(16.85,22.86) 0.00(-0.30,0.30)
1x to <2x poverty 856 20.93(18.40,23.46) 0.00(-0.30,0.30)
2xto<3x poverty line 516 20.88(17.23,24.53) 0.00(-0.63,0.63)
3x to <4x poverty line 368 24.51(18.35,30.66) 0.00(-0.78,0.78)
4xto<5x poverty line 251 26.63(21.46,31.81) 0.73(-0.87,2.32)
>=5x poverty line 498 31.51(27.44,35.58) 0.90(-0.70,2.51)
Missing/Refused/DK 343 24.78(19.75,29.81) 0.59(-0.01,1.19)
Education
Median education forage 904 31.45(28.30,34.61) 1.25(-0.09,2.60)
Unknown education 373 10.63(8.46,12.80) 0.00(-0.30,0.30)
Estimated mercury intake per unit body weight (fig Hg/kg bw)
Race/Ethnicity
Mexican American
624
0.32(0.24,0.41)
0.01(0.00,0.01)
Other Hispanic
380
0.37(0.29,0.44)
0.01(0.00,0.01)
Non-HispanicWhite
1199
0.31(0.29,0.34)
0.00(0.00,0.00)
Non-Hispanic Black
877
0.38(0.33,0.43)
0.02(-0.01,0.04)
Other Race
628
0.36(0.31,0.41)
O.OK-0.01,0.03)
Age, Years
16 to 19 654 0.14(0.12,0.16) 0.00(-0.01,0.01)
20 to 29 963 0.32(0.29,0.35) 0.00(0.00,0.00)
30 to 39 1056 0.36(0.31,0.40) 0.00(0.00,0.01)
40 to 49 1035 0.39(0.35,0.44) 0.01(0.00,0.02)
Annual Income
Oto <1x poverty line
876
0.26(0.23,0.30)
0.00(0.00,0.00)
1xto <2x poverty
856
0.28(0.25,0.31)
0.00(0.00,0.00)
2x to <3x poverty line
516
0.28(0.23,0.34)
0.00(-0.01,0.01)
3x to <4x poverty line
368
0.33(0.25,0.41)
0.00(-0.01,0.01)
4xto <5x poverty line
251
0.38(0.30,0.46)
O.OK-0.01,0.03)
>= 5x poverty line
498
0.46(0.39,0.52)
O.OK-0.01,0.04)
Missing/Refused/DK
343
0.36(0.28,0.43)
O.OK-0.01,0.03)
A-14
50th
75th
90th
95th
9.14(7.50,10.78)
12.20(9.44,14.95)
12.23(10.29,14.16)
5.09(2.39,7.80)
9.13(6.45,11.80)
9.04(6.82,11.27)
11.17(6.59,15.74)
12.32(8.30,16.34)
16.96(12.44,21.49)
9.77(7.62,11.93)
5.29(2.57,8.01)
10.12(8.25,11.99)
16.17(12.27,20.06)
2.60(0.78,4.42)
27.23(23.90,30.56)
34.90(30.45,39.35)
33.68(28.68,38.68)
21.05(17.57,24.52)
29.47(24.45,34.49)
25.22(20.33,30.11)
29.99(20.99,38.99)
33.69(25.86,41.52)
38.01(32.23,43.79)
24.10(16.46,31.73)
19.82(16.25,23.39)
29.58(26.50,32.65)
38.70(34.86,42.54)
13.30(8.43,18.18)
63.87(52.50,75.23)
62.99(54.08,71.90)
73.81(59.19,88.42)
53.88(43.83,63.92)
57.89(47.15,68.63)
50.15(34.18,66.11)
55.44(31.80,79.08)
68.71(31.54,105.89)
80.89(63.29,98.49)
63.60(51.90,75.31)
48.82(39.91,57.72)
61.80(52.37,71.23)
76.40(62.63,90.17)
28.74(24.80,32.68)
88.93(71.51,106.35)
97.75(70.58,124.91)
128.91(98.42,159.40)
77.27(56.77,97.77)
81.20(66.02,96.37)
89.64(54.93,124.34)
90.28(33.13,147.43)
113.20(55.38,171.01)
113.00(82.33,143.68)
121.19(70.30,172.08)
72.54(61.62,83.45)
101.15(72.12,130.19)
129.84(100.67,159.02)
44.81(27.84,61.78)
0.12(0.09,0.16)
0.14(0.10,0.19)
0.12(0.09,0.14)
0.18(0.14,0.22)
0.13(0.09,0.17)
0.02(0.00,0.04)
0.13(0.10,0.16)
0.15(0.11,0.19)
0.16(0.13,0.19)
0.07(0.05,0.09)
0.11(0.07,0.14)
0.11(0.08,0.13)
0.13(0.07,0.20)
0.17(0.10,0.24)
0.21(0.15,0.26)
0.14(0.11,0.17)
0.34(0.27,0.42)
0.39(0.29,0.49)
0.40(0.36,0.44)
0.44(0.40,0.49)
0.42(0.35,0.49)
0.19(0.15,0.22)
0.37(0.31,0.43)
0.45(0.40,0.49)
0.44(0.38,0.50)
0.29(0.24,0.35)
0.36(0.30,0.42)
0.32(0.25,0.40)
0.41(0.32,0.49)
0.45(0.36,0.54)
0.52(0.42,0.62)
0.32(0.24,0.41)
0.79(0.66,0.92)
0.95(0.70,1.19)
0.79(0.66,0.92)
0.96(0.68,1.24)
1.01(0.75,1.27)
0.42(0.36,0.48)
0.88(0.69,1.07)
0.88(0.75,1.02)
0.98(0.78,1.18)
0.70(0.56,0.85)
0.80(0.63,0.97)
0.77(0.55,0.99)
0.74(0.47,1.01)
0.94(0.48,1.39)
1.19(0.98,1.39)
0.79(0.40,1.17)
1.20(0.75,1.66)
1.39(1.12,1.66)
1.26(0.97,1.55)
1.42(1.13,1.72)
1.41(1.16,1.67)
0.63(0.53,0.72)
1.37(1.17,1.57)
1.32(0.96,1.67)
1.71(1.19,2.23)
1.06(0.77,1.34)
1.27(1.08,1.46)
1.12(0.72,1.52)
1.17(0.39,1.95)
1.56(0.75,2.37)
1.77(1.37,2.17)
1.52(0.83,2.20)
-------
Parameter
N
Arithmetic mean
(95% CI)
25th
50th
75th
90th
95th
Education
Median education forage
904
0.45(0.41,0.49)
0.02(0.00,0.04)
0.22(0.17,0.27)
0.53(0.46,0.59)
1.14(0.95,1.34)
1.90(1.47,2.33)
Unknown education
373
0.17(0.13,0.20)
0.00(0.00,0.00)
0.04(0.01,0.07)
0.20(0.11,0.30)
0.44(0.32,0.55)
0.67(0.46,0.88)
A-15
-------
Table A-9. Blood MeHg concentrations (pg/L), by frequency of consuming fish, by NHANES survey releases, women aged 16-49 years, NHANES 2013-March 2020
SURVEY RELEASE
TIMES EATEN IN
30 DAYS
N
ARITH. MEAN
(95% CI)
SELECTED PERCENTILES (95% CI)
25th
50th
75th
90th
95th
2013-2014
0
201
0.31(0.25,0.37)
0.14(0.13,0.14)
0.20(0.15,0.24)
0.30(0.20,0.40)
0.53(0.04,1.02)
0.92(0.19,1.66)
1
137
0.61(0.40,0.82)
0.16(0.12,0.21)
0.32(0.27,0.37)
0.57(0.38,0.77)
1.09(0.69,1.49)
1 •47(.,.)a
2
111
0.63(0.54,0.73)
0.27(0.19,0.35)
0.42(0.29,0.54)
0.72(0.46,0.99)
1.31(0.90,1.73)
1.60(0.67,2.53)
3
77
0.62(0.54,0.70)
0.32(0.26,0.38)
0.56(0.45,0.66)
0.84(0.66,1.02)
1.08(0.48,1.69)
i.igf.,.)3
4-5
79
0.86(0.76,0.97)
0.37(0.26,0.48)
0.52(0.45,0.59)
0.90(0.71,1.09)
1.66(0.71,2.62)
3.23(2.21,4.24)
6 and up
209
2.22(1.66,2.77)
0.59(0.47,0.72)
1.24(0.96,1.52)
2.64(1.81,3.47)
4.40(2.84,5.96)
7.84(2.69,12.99)
2015-2016
0
197
0.31(0.26,0.36)
0.15(0.14,0.15)
0.21(0.16,0.26)
0.31(0.27,0.35)
0.49(0.37,0.60)
0.76(0.49,1.03)
1
106
0.42(0.31,0.52)
0.17(0.13,0.21)
0.30(0.27,0.33)
0.46(0.34,0.57)
0.61(0.47,0.76)
0.78 (-0.18,1.74)
2
80
1.04(0.55,1.53)
0.27(0.22,0.32)
0.49(0.35,0.63)
0.89(0.39,1.39)
2.04 (-0.99,5.07)
4.01(.,.)a
3
69
0.71(0.52,0.90)
0.35(0.22,0.48)
0.51(0.30,0.71)
0.87(0.43,1.30)
i.ogf.,.)3
1.42 („.)a
4-5
100
1.00(0.79,1.21)
0.45(0.37,0.54)
0.68(0.48,0.88)
1.11(0.69,1.52)
2.01(0.64,3.39)
3.01(1.36,4.67)
6 and up
188
1.77(1.51,2.02)
0.59(0.39,0.78)
1.11(0.90,1.33)
2.42(1.75,3.09)
3.81(2.69,4.93)
5.38(3.47,7.29)
2017-MARCH 2020
0
554
0.34(0.29,0.40)
0.15(0.14,0.15)
0.16(0.15,0.17)
0.34(0.24,0.43)
0.61(0.42,0.80)
1.00(0.63,1.38)
1
290
0.48(0.38,0.58)
0.16(0.15,0.17)
0.28(0.24,0.32)
0.53(0.31,0.75)
1.31(0.74,1.89)
1.54(1.14,1.94)
2
252
0.55(0.46,0.64)
0.21(0.16,0.25)
0.43(0.36,0.51)
0.64(0.44,0.84)
1.11(0.66,1.57)
1.34(0.69,1.99)
3
211
0.78(0.65,0.91)
0.25(0.18,0.32)
0.39(0.26,0.53)
0.95(0.70,1.20)
1.67(1.31,2.03)
2.25(1.83,2.67)
4-5
282
1.06(0.83,1.29)
0.33(0.25,0.40)
0.62(0.47,0.78)
1.16(0.72,1.61)
2.73(1.90,3.56)
3.44(1.67,5.22)
6 and up
565
1.66(1.47,1.85)
0.61(0.51,0.70)
1.08(0.88,1.28)
2.01(1.51,2.50)
3.68(2.84,4.52)
4.65(3.24,6.06)
a Missing confidence interval because of stratum with single sampling unit.
A-16
-------
Table A-10. Parameter estimates and odds ratio from the logistic model predicting the probability of reporting any
Parameter
Std. error
p-Value
Odds ratio
Intercept
1.3154
0.06
<0.0001
Income, Overall
0.0004
Oto =5x poverty line
0.1546
0.15
0.3051
1.17
Missing/Refused/DK
0.3025
0.15
0.0430
1.35
Race, Overall
<0.0001
Mexican American
0.1636
0.09
0.0733
1.18
Non-Hispanic Black
0.3179
0.12
0.0081
1.37
Non-Hispanic White
-0.5135
0.08
<0.0001
0.60
Other Hispanic
0.1544
0.13
0.2407
1.17
Other Race
-0.1225
0.12
0.3129
0.88
Education, Overall
<0.0001
< Median forage
-0.4417
0.10
<0.0001
0.64
Median forage
0.0740
0.09
0.4152
1.08
>Median forage
0.1674
0.11
0.1204
1.18
Unknown education
0.2003
0.15
0.1773
1.22
Age, Overall
<0.0001
16 to 19
-0.6722
0.13
<0.0001
0.51
20 to 29
-0.0005
0.09
0.9953
1.00
30 to 39
0.1792
0.08
0.0182
1.20
40 to 49
0.4935
0.08
<0.0001
1.64
A-17
-------
Table A-11. Parameter estimates and relative ratios from the model predicting the log-transformed frequency of fish
consumption in the previous 30 days (times), NHANES 2013-March 2020
Parameter
Std. error
p-Value
Odds ratio
Intercept
1.3014
0.04
<0.0001
Income, Overall
0.1650
Oto =5x poverty line
0.1556
0.06
0.0125
1.17
Missing/Refused/DK
-0.0023
0.06
0.9694
1.00
Race, Overall
<0.0001
Mexican American
-0.0966
0.05
0.0810
0.91
Non-Hispanic Black
0.0889
0.04
0.0344
1.09
Non-Hispanic White
-0.1258
0.04
0.0020
0.88
Other Hispanic
-0.0401
0.06
0.5124
0.96
Other Race
0.1737
0.05
0.0008
1.19
Education, Overall
<0.0001
< Median forage
-0.1865
0.04
<0.0001
0.83
Median forage
-0.0414
0.04
0.3249
0.96
>Median forage
0.1342
0.05
0.0146
1.14
Unknown education
0.0937
0.09
0.3092
1.10
Age, Overall
<0.0001
16 to 19
-0.3314
0.06
<0.0001
0.72
20 to 29
0.1203
0.04
0.0044
1.13
30 to 39
0.1582
0.04
0.0002
1.17
40 to 49
0.0529
0.04
0.1498
1.05
A-18
-------
Table A-12. Parameter estimates and relative ratios from the model predicting the log-transformed amount offish
consumed in a meal (meal size)(q), NHANES 2013-March 2020
Parameter
Std. error
p-Value
Odds ratio
Intercept
4.1932
0.01
<0.0001
Income, Overall
0.4531
Oto =5x poverty line
0.0166
0.02
0.3624
1.02
Missing/Refused/DK
-0.0467
0.02
0.0670
0.95
Race, Overall
<0.0001
Mexican American
0.0881
0.02
<0.0001
1.09
Non-Hispanic Black
0.1170
0.01
<0.0001
1.12
Non-Hispanic White
0.0923
0.01
<0.0001
1.10
Other Hispanic
-0.0247
0.02
0.1802
0.98
Other Race
-0.2727
0.01
<0.0001
0.76
Education, Overall
0.0155
< Median forage
-0.0096
0.01
0.5160
0.99
Median forage
0.0208
0.01
0.1598
1.02
>Median forage
0.0450
0.02
0.0080
1.05
Unknown education
-0.0562
0.03
0.0958
0.95
Age, Overall
0.2898
16 to 19
-0.0482
0.03
0.1143
0.95
20 to 29
0.0050
0.01
0.7275
1.00
30 to 39
0.0270
0.01
0.0628
1.03
40 to 49
0.0162
0.01
0.2479
1.02
A-19
-------
Table A-13. Parameter estimates and relative ratios from the model predicting the log-transformed mercury concentration
Parameter
Std. error
p-Value
Odds ratio
Intercept
-2.9757
0.03
<0.0001
Income, Overall
0.4026
Oto =5x poverty line
0.0178
0.05
0.7121
1.02
Missing/Refused/DK
-0.0609
0.06
0.3112
0.94
Race, Overall
0.2109
Mexican American
-0.0755
0.04
0.0510
0.93
Non-Hispanic Black
0.0201
0.03
0.5046
1.02
Non-Hispanic White
0.0447
0.03
0.0891
1.05
Other Hispanic
-0.0176
0.03
0.6127
0.98
Other Race
0.0283
0.03
0.3708
1.03
Education, Overall
0.2515
< Median forage
0.0432
0.04
0.2366
1.04
Median forage
0.0703
0.04
0.0508
1.07
>Median forage
0.0273
0.05
0.5543
1.03
Unknown education
-0.1408
0.09
0.1102
0.87
Age, Overall
0.1633
16 to 19
-0.0711
0.06
0.2513
0.93
20 to 29
-0.0322
0.04
0.3688
0.97
30 to 39
0.0428
0.03
0.1317
1.04
40 to 49
0.0605
0.04
0.0953
1.06
A-20
-------
Table A-14. Parameter estimates and relative ratios from the model predicting the log-transformed inverse of body
Parameter
Std. error
p-Value
Odds ratio
Intercept
-4.2811
0.01
<0.0001
Income, Overall
0.4767
Oto =5x poverty line
0.0217
0.01
0.1222
1.02
Missing/Refused/DK
0.0282
0.02
0.1238
1.03
Race, Overall
<0.0001
Mexican American
0.0009
0.01
0.9383
1.00
Non-Hispanic Black
-0.1202
0.01
<0.0001
0.89
Non-Hispanic White
-0.0249
0.01
0.0151
0.98
Other Hispanic
0.0675
0.01
<0.0001
1.07
Other Race
0.0767
0.02
<0.0001
1.08
Education, Overall
0.0016
< Median forage
-0.0218
0.02
0.1645
0.98
Median forage
-0.0186
0.01
0.1388
0.98
>Median forage
0.0465
0.01
0.0011
1.05
Unknown education
-0.0061
0.03
0.8079
0.99
Age, Overall
<0.0001
16 to 19
0.0934
0.02
0.0004
1.10
20 to 29
0.0236
0.01
0.1029
1.02
30 to 39
-0.0563
0.01
<0.0001
0.95
40 to 49
-0.0607
0.01
<0.0001
0.94
A-21
-------
Table A-15. Parameter estimates and relative ratios from the model predicting the log-transformed mercury intake per unit body
Parameter
Std. error
p-Value
Odds ratio
Intercept
-1.7622
0.06
<0.0001
Income, Overall
0.5203
Oto =5x poverty line
0.2118
0.10
0.0435
1.24
Missing/Refused/DK
-0.0817
0.10
0.4038
0.92
Race, Overall
0.2830
Mexican American
-0.0832
0.08
0.3079
0.92
Non-Hispanic Black
0.1058
0.05
0.0363
1.11
Non-Hispanic White
-0.0137
0.05
0.7691
0.99
Other Hispanic
-0.0150
0.09
0.8678
0.99
Other Race
0.0060
0.07
0.9303
1.01
Education, Overall
0.0003
Median forage
0.2530
0.08
0.0033
1.29
Unknown education
-0.1094
0.14
0.4509
0.90
Age, Overall
0.0035
16 to 19
-0.3573
0.10
0.0010
0.70
20 to 29
0.1166
0.06
0.0404
1.12
30 to 39
0.1717
0.06
0.0057
1.19
40 to 49
0.0689
0.06
0.2890
1.07
A-22
-------
APPENDIX B
Geographic Distribution of Blood MeHg
in the General U.S. Population Using
NHANES 2009-2012
-------
Bl. Introduction
This analysis documents a study of the 1999-2012 National Health and Nutrition Examination
Survey (NHANES) blood mercury and fish consumption data from the general U.S. population of
women 16 to 49 years. Trends over time in blood methylmercury (MeHg) concentrations and
demographic characteristics that are related to blood mercury concentrations have been
updated and addressed with additional NHANES data 2013-March 2020 in the main report.
Information presented in this appendix summarizes the analysis of the geographic distribution of
blood mercury using NHANES 2009-2012 to reflect levels of recent years at the time of analysis
and the blood MeHg modeling using NHANES 1999-2012.
B2. Methods
This geographic analysis applied the same methods detailed in sections 2.1 to 2.4 in the main
report. In addition, serum cotinine concentration from laboratory tests and alcohol consumption
data collected from questionnaires were included in this analysis. Serum cotinine served as a
biomarker for assessing smoking status. The serum cotinine concentrations were
log-transformed. Alcohol consumption has been shown to be related to blood mercury levels
(Park and Lee, 2013). Alcohol consumption data are available for participants age 20 years and
older. The analysis variable used in the model has four levels - greater than or equal to 12 drinks in
past year, less than 12 drinks in past year, participants less than 20 years old, and participants with
missing data for this variable.
Household annual income is derived using two sets of questions, the first for less than or greater
than $20K and the second with a more detailed breakdown. The seven income categories used
for the analysis are: less than $20K, $20K to $45K, greater than $45K to $75K, greater than $75K,
greater than or equal to $20K (if second, more detailed question not answered), "refused" or "don't
know," and missing information.
U.S. Census Region and coastal status data were obtained from the NCHS Research Data Center
(http://www.cdc.aov/rdc/1 through a proposal process for access to non-publicly available NHANES
data. NCHS analysts derived the census region variable from the state of residence of the
participant and the coastal status variable from the county of residence. If a county was adjacent
to the Atlantic Ocean, Pacific Ocean, Gulf of Mexico, or one of the Great Lakes, it was considered
coastal. All other counties were considered non-coastal. A list of coastal counties based on
B-l
-------
methods detailed in Mahaffey et al 2009, regardless of the geographic association of NHANES
participants can be found in Table B-l. All analyses utilizing these restricted access variables were
conducted at NCHS Research Data Center in Hyattsville, MD.
B-2
-------
Table B-1 List of coastal counties
State Coastal Counties State Coastal Counties State Coastal Counties
Alabama
Baldwin County
Flor
da
Baker County
Florida
St. Johns County
Alabama
Mobile County
Flor
da
Bay County
Florida
St. Lucie County
Alaska
Aleutians East Borough
Flor
da
Bradford County
Florida
Sumter County
Alaska
Aleutians West
Flor
da
Brevard County
Florida
Suwannee County
Alaska
Anchorage Borough
Flor
da
Broward County
Florida
Taylor County
Alaska
Bethel
Flor
da
Calhoun County
Florida
Union County
Alaska
Bristol Bay Borough
Flor
da
Charlotte County
Florida
Volusia County
Alaska
City & Borough of Juneau
Flor
da
Citrus County
Florida
Wakulla County
Alaska
City & Borough of Sitka
Flor
da
Clay County
Florida
Walton County
Alaska
Dillingham
Flor
da
Collier County
Florida
Washington County
Alaska
Haines Borough
Flor
da
Columbia County
Georgia
Bryan County
Alaska
Kenai Peninsula Borough
Flor
da
DeSoto County
Georgia
Camden County
Alaska
Ketchikan Gateway Borough
Flor
da
Dixie County
Georgia
Chatham County
Alaska
Kodiak Island Borough
Flor
da
Duval County/City of Jacksonville
Georgia
Glynn County
Alaska
Lake And Peninsula Borough
Flor
da
Escambia County
Georgia
Liberty County
Alaska
Nome
Flor
da
Flagler County
Georgia
Mcintosh County
Alaska
North Slope Borough
Flor
da
Franklin County
Hawaii
Hawaii County
Alaska
Northwest Arctic Borough
Flor
da
Gadsden County
Hawaii
Honolulu City and County
Alaska
Prince of Wales-Outer Ketchikan
Flor
da
Gilchrist County
Hawaii
Kalawao
Alaska
Skagway-Hoonah-Angoon
Flor
da
Glades County
Hawaii
Kauai County
Alaska
Valdez-Cordon
Flor
da
Gulf County
Hawaii
Maui County
Alaska
Wade Hampton
Flor
da
Hamilton County
Illinois
Cook
Alaska
Wrangell-Petersburg
Flor
da
Hardee County
Illinois
DuPage
Alaska
Yakutat
Flor
da
Hendry County
Illinois
Kane
California
Alameda County
Flor
da
Hernando County
Illinois
Lake
California
Contra Costa County
Flor
da
Highlands County
Illinois
McHenry
California
Del Norte County
Flor
da
Hillsborough County
Illinois
Will
California
Humboldt County
Flor
da
Holmes County
Indiana
Lake
California
Los Angeles County
Flor
da
Indian River County
Indiana
LaPorte
California
Marin County
Flor
da
Jackson County
Indiana
Porter
California
Mendocino County
Flor
da
Jefferson County
Louisiana
Assumption Parish
California
Monterey County
Flor
da
Lafayette County
Louisiana
Cameron Parish
California
Napa County
Flor
da
Lake County
Louisiana
Iberia Parish
California
Orange County
Flor
da
Lee County
Louisiana
Jefferson Parish
California
San Diego County
Flor
da
Leon County
Louisiana
Lafayette Consolidated Government
California
San Francisco City & County
Flor
da
Le\ty County
Louisiana
Lafourche Parish
California
San Luis Obispo County
Flor
da
Liberty County
Louisiana
Livingston Parish
California
San Mateo County
Flor
da
Madison County
Louisiana
Orleans Parish
California
Santa Barbara County
Flor
da
Manatee County
Louisiana
Plaquemines Parish
California
Santa Clara County
Flor
da
Marion County
Louisiana
St. Bernard Parish
California
Santa Cruz County
Flor
da
Martin County
Louisiana
St. Charles Parish
California
Solano County
Flor
da
Miami-Dade County
Louisiana
St. James Parish
California
Sonoma County
Flor
da
Monroe County
Louisiana
St. John The Baptist Parish
California
Ventura County
Flor
da
Nassau County
Louisiana
St. Mary Parish
Connecticut
Fairfield County
Flor
da
Okaloosa County
Louisiana
St. Tammany Parish
Connecticut
Hartford County
Flor
da
Okeechobee County
Louisiana
Tangipahoa Parish
Connecticut
Middlesex County
Flor
da
Orange County
Louisiana
Terrebonne Parish
Connecticut
New Ha\en County
Flor
da
Osceola County
Louisiana
Vermilion Parish
Connecticut
New London County
Flor
da
Palm Beach County
Maine
Androscoggin County
Connecticut
Tolland County
Flor
da
Pasco County
Maine
Cumberland County
Connecticut
Windham County
Flor
da
Pinellas County
Maine
Hancock County
Delaware
Kent County
Flor
da
Polk County
Maine
Kennebec County
Delaware
New Castle County
Flor
da
Putnam County
Maine
Knox County
Delaware
Sussex County
Flor
da
Santa Rosa County
Maine
Lincoln County
District Of Columbia
District Of Columbia
Flor
da
Sarasota County
Maine
Sagadahoc County
Florida
Alachua County
Flor
da
Seminole County
Maine
Waldo County
B-3
-------
State
Coastal Counties
State
Coastal Counties
State
Coastal Counties
Maine
Washington County
Michigan
Luce
New York
Onondaga
Maine
York County
Michigan
Mackinac
New York
Ontario
Maryland
Anne Arundel County
Michigan
Macomb
New York
Orleans
Maryland
Baltimore City
Michigan
Manistee
New York
Oswego
Maryland
Baltimore County
Michigan
Marquette
New York
Seneca
Maryland
Calwrt County
Michigan
Mason
New York
Wayne
Maryland
Caroline County
Michigan
Menominee
New York
Wyoming
Maryland
Cecil County
Michigan
Midland
New York
Bronx County
Maryland
Charles County
Michigan
Monroe
New York
Kings County (Brooklyn)
Maryland
Dorchester County
Michigan
Muskegon
New York
Nassau County
Maryland
Harford County
Michigan
Oakland
New York
New York County (Manhattan)
Maryland
Howard County
Michigan
Oceana
New York
Queens County
Maryland
Kent County
Michigan
Ontonagon
New York
Richmond County (Staten Island)
Maryland
Montgomery County
Michigan
Ottawa
New York
Rockland County
Maryland
Prince George's County
Michigan
Presque Isle
New York
Suffolk County
Maryland
Queen Anne's County
Michigan
Saginaw
New York
Westchester County
Maryland
Somerset County
Michigan
Sanilac
North Caro
ina
Beaufort County
Maryland
St. Mary's County
Michigan
Schoolcraft
North Caro
ina
Bertie County
Maryland
Talbot County
Michigan
St. Clair
North Caro
ina
Brunswick County
Maryland
Wicomico County
Michigan
Tuscola
North Caro
ina
Camden County
Maryland
Worcester County
Michigan
Van Buren
North Caro
ina
Carteret County
Massachusetts
Barnstable County
Michigan
Washtenaw
North Caro
ina
Chowan County
Massachusetts
Bristol County
Michigan
Wayne
North Caro
ina
Crawn County
Massachusetts
Dukes County
Minnesota
Carlton
North Caro
ina
Currituck County
Massachusetts
Essex County
Minnesota
Cook
North Caro
ina
Dare County
Massachusetts
Middlesex County
Minnesota
Lake
North Caro
ina
Hyde County
Massachusetts
Nantucket County
Minnesota
St. Louis
North Caro
ina
Jones County
Massachusetts
Norfolk County
Mississippi
Hancock County
North Caro
ina
New Hanowr County
Massachusetts
Plymouth County
Mississippi
Harrison County
North Caro
ina
Onslow County
Massachusetts
Suffolk County
Mississippi
Jackson County
North Caro
ina
Pamlico County
Michigan
Alcona
New Hampshire
Rockingham County
North Caro
ina
Pasquotank County
Michigan
Alger
New Hampshire
Strafford County
North Caro
ina
Pender County
Michigan
Allegan
New Jersey
Atlantic County
North Caro
ina
Perquimans County
Michigan
Alpena
New Jersey
Bergen County
North Caro
ina
Tyrrell County
Michigan
Antrim
New Jersey
Burlington County
North Caro
ina
Washington County
Michigan
Arenac
New Jersey
Camden County
Ohio
Ashtabula
Michigan
Baraga
New Jersey
Cape May County
Ohio
Cuyahoga
Michigan
Bay
New Jersey
Cumberland County
Ohio
Erie
Michigan
Benzie
New Jersey
Essex County
Ohio
Geauga
Michigan
Berrien
New Jersey
Glouchester County
Ohio
Huron
Michigan
Charlewix
New Jersey
Hudson County
Ohio
Lake
Michigan
Cheboygan
New Jersey
Middlesex County
Ohio
Lorain
Michigan
Chippewa
New Jersey
Monmouth County
Ohio
Lucas
Michigan
Delta
New Jersey
Ocean County
Ohio
Medina
Michigan
Emmet
New Jersey
Passaic County
Ohio
Ottawa
Michigan
Genesee
New Jersey
Salem County
Ohio
Sandusky
Michigan
Gladwin
New Jersey
Union County
Ohio
Seneca
Michigan
Gogebic
New York
Cattaraugus
Ohio
Summit
Michigan
Grand Traverse
New York
Cayuga
Ohio
Wood
Michigan
Houghton
New York
Chautauqua
Oregon
Clatsop County
Michigan
Huron
New York
Erie
Oregon
Columbia County
Michigan
Iosco
New York
Genesee
Oregon
Coos County
Michigan
Kalkaska
New York
Jefferson
Oregon
Curry County
Michigan
Keweenaw
New York
Livingston
Oregon
Douglas County
Michigan
Lapeer
New York
Monroe
Oregon
Lane County
Michigan
Leelanau
New York
Niagara
Oregon
Lincoln County
B-4
-------
State
Coastal Counties State
Coastal Counties
Oregon
Multnomah County
Virginia
Manassas City
Oregon
Tillamook County
Virginia
Manassas Park City
Oregon
Washington County
Virginia
Matthews County
Pennsylvania
Crawford
Virginia
Middlesex County
Pennsylvania
Erie
Virginia
New Kent County
Pennsylvania
Delaware County
Virginia
Newport News City
Pennsylvania
Montgomery County
Virginia
Norfolk City
Pennsylvania
Philadelphia County
Virginia
Northampton County
Rhode Island
Bristol County
Virginia
Northumberland County
Rhode Island
Kent County
Virginia
Poquoson City
Rhode Island
Newport County
Virginia
Portsmouth City
Rhode Island
Providence County
Virginia
Prince William County
Rhode Island
Washington County
Virginia
Richmond City
South Carolina
Beaufort County
Virginia
Richmond County
South Carolina
Berkeley County
Virginia
Stafford County
South Carolina
Charleston County
Virginia
Suffolk City
South Carolina
Colleton County
Virginia
Surry County
South Carolina
Georgetown County
Virginia
Virginia Beach City
South Carolina
Horry County
Virginia
Westmoreland County
South Carolina
Jasper County
Virginia
Williamsburg City
Texas
Aransas County
Virginia
York County
Texas
Brazoria County
Washington
Clallam County
Texas
Calhoun County
Washington
Clark County
Texas
Cameron County
Washington
Cowlitz County
Texas
Chambers County
Washington
Grays Harbor County
Texas
Galwston County
Washington
Island County
Texas
Harris County
Washington
Jefferson County
Texas
Jackson County
Washington
King County
Texas
Jefferson County
Washington
Kitsap County
Texas
Kenedy County
Washington
Mason County
Texas
Kleberg County
Washington
Pacific County
Texas
Matagorda County
Washington
Pierce County
Texas
Nueces County
Washington
San Juan County
Texas
Orange County
Washington
Skagit County
Texas
Refugio County
Washington
Snohomish County
Texas
San Patricio County
Washington
Thurston County
Texas
Victoria County
Washington
Wahkiakum County
Texas
Willacy County
Washington
Whatcom County
Virginia
Accomack County
Wisconsin
Ashland
Virginia
Alexandria City
Wisconsin
Bayfield
Virginia
Arlington County
Wisconsin
Brown
Virginia
Charles City County
Wisconsin
Calumet
Virginia
Chesapeake City
Wisconsin
Door
Virginia
Clifton Forge City
Wisconsin
Douglas
Virginia
Essex County
Wisconsin
Iron
Virginia
Fairfax City
Wisconsin
Kenosha
Virginia
Fairfax County
Wisconsin
Kewaunee
Virginia
Falls Church City
Wisconsin
Manitowoc
Virginia
Gloucester County
Wisconsin
Marinette
Virginia
Hampton City
Wisconsin
Milwaukee
Virginia
Henrico County
Wisconsin
Oconto
Virginia
Isle of Wight County
Wisconsin
Ozaukee
Virginia
James City County
Wisconsin
Racine
Virginia
King and Queen County
Wisconsin
Sheboygan
Virginia
King George County
Wisconsin
Washington
Virginia
Lancaster County
Wisconsin
Waukesha
B-5
-------
B3. Results
B3.1 Blood MeHg Summary Statistics of Geographic
Distributions
Figure B-l presents the geometric mean and 95 percent confidence interval of blood MeHg
concentration by U.S. census region and coastal status for women 16 to 49 years during the survey
period 2009-2012. Women in the Northeast region have the highest geometric mean
concentrations while those in the Midwest have the lowest and those in coastal areas have higher
geometric mean concentrations than those residing in non-coastal areas. Similar patterns are
observed in the percentiles (Table B-2).
Figure B-1 Geometric Mean and 95 percent CI of MeHg by geography (NHANES 2009-2012)
^Geo. Mean 95% CI
Region
Midwest
South
West
Northeast
0.1
MeHg, Mg/L (log scale)
~ 0.41
~ 0.50
0.69
~
1.08
Coastal Status
Noncoastal
Coastal
+ 0.45
~ 0.79
B-6
-------
Geographic locations
N
Geometric, mean
Percentiles
(95% CI)
25th
75th
90th
Northeast
490
1.08(0.87,1.33)
0.49
2.57
3.90
Midwest
678
0.41(0.36,0.47)
0.21
0.72
1.46
West
841
0.69(0.54,0.90)
0.31
1.51
2.71
South
1164
0.50(0.44,0.58)
0.24
1.00
1.94
Noncoastal
1593
0.45(0.41,0.51)
0.22
0.84
Ml
Coastal
1580
0.79(0.71,0.88)
0.35
1.71
3.25
Note: Geometric mean and percentiles were calculated from the mean of 20 imputed values for each respondent.
B3.2 Blood MeHg Modeling
The results of the multivariable modeling of blood MeHg concentrations are described in this
section. Transformed usual intake of mercury (TUI) through fish consumption (|jg Hg/day) was the
most highly significant predictor of blood MeHg concentration, with higher consumption
associated with higher blood MeHg concentration. Additionally the interaction between
race/ethnicity and TUI is significant, indicating that blood MeHg concentration increases at
different rates with increasing intake offish mercury by racial/ethnic group. Other significant
predictors include the geographic variables, education, and survey release. Figure B-2 presents
the multiplicative change in blood MeHg concentration with 95 percent confidence intervals, for
region, coastal status, and demographic characteristics. Coastal status and region are associated
with blood MeHg concentration (p<0.0001 and p=0.02, respectively), with coastal residence and
residence in the Northeast associated with higher blood MeHg concentrations. Residence in the
Midwest is associated with lower blood MeHg concentration. Education is positively associated
with blood MeHg concentrations (p=0.008). NHANES survey release is also associated with blood
MeHg concentration (p<0.0001). The earliest study year, N HAN ES 1999-2000, has the highest
concentrations and NHANES 2003-2004 and 2011-2012 have the lowest concentrations.
B-7
-------
Figure B-2. Multiplicative change for statistically significant predictors of blood MeHg concentration (NHANES
1999-2012)
^ Estimate 95% CI Multiplicative Change (log-scale)
0.5 1 2
Coastal Status
Coastal ^1.15
Noncoastal A 0.87
Region
Northeast ^ 114
Midwest ^ 0.65
West | ~ 1.07
South ^ 0.97
Education
< Median for age + 0.69
= Median for age ^ 1.03
> Median for age ^ 1.09
Survey Year
1999-2000 + 1.51
2001-2002 + 1.09
2003-2004 ^ 0.81
2005-2006 + 0.66
2007-2008 ^ 0.97
2009-2010 | + 107
2011-2012 ~ 0.83
Of the remaining predictors in the model, only hematocrit, had a significant association with
blood MeHg concentration (p<0.0001), with increasing hematocrit associated with increasing
blood MeHg concentration. Table B-3 provides the full model results.
B-8
-------
Table B-3. Modeling results predicting blood MeHg concentrations
Parameter
Estimate
LCL
UCL
tValue
Probt
fValue
ProbF
Intercept
0.6604
0.4314
0.8893
5.72
<0.0001
TUI
1.3939
1.2102
1.5775
15.06
<0.0001
TUIV
-0.0443
-0.1848
0.0962
-0.63
0.5332
TUI x log-transformed body weight (centered)
-0.3744
-0.7538
0.0051
-1.96
0.0533
TUI x Race/Ethnicity
9.06
<0.0001
TUI x Mexican American
-0.2616
-0.4405
-0.0828
-2.90
0.0046
TUIx non-Hispanic Black
-0.1604
-0.3479
0.0271
-1.70
0.0929
TUI x non-Hispanic White
0.5524
0.3438
0.7610
5.25
<0.0001
TUIx other Hispanic
-0.0502
-0.2981
0.1978
-0.40
0.6891
TUIx other race
-0.0802
-0.3524
0.1920
-0.58
0.5603
Age, decade (centered)
0.0101
-0.0552
0.0754
0.31
0.7602
Log-transformed body weight (centered)
-0.5116
-1.0021
-0.0210
-2.07
0.0414
Log-transformed body weight2 (centered)
-0.2173
-0.6731
0.2384
-0.95
0.3466
Coastal Status
23.70
<0.0001
Coastal
0.1395
0.0826
0.1963
4.87
<0.0001
Non-coastal
-0.1395
-0.1963
-0.0826
-4.87
<0.0001
Region
3.36
0.0222
Northeast
0.1305
0.0090
0.2521
2.13
0.0358
Midwest
-0.1677
-0.2856
-0.0498
-2.82
0.0058
West
0.0676
-0.0363
0.1716
1.29
0.2001
South
-0.0304
-0.1111
0.0502
-0.75
0.4558
Alcohol Consumption
2.58
0.0581
<12 drinks past year
0.1058
0.0248
0.1867
2.59
0.0111
>12 drinks past year
0.0264
-0.0896
0.1424
0.45
0.6527
<20 years old
-0.0958
-0.2253
0.0338
-1.47
0.1458
Refused/Don't know/Missing
-0.0364
-0.2001
0.1273
-0.44
0.6602
Household Income
1.64
0.1458
<$20K
-0.1030
-0.2217
0.0157
-1.72
0.0885
$20K to <$45K
-0.0348
-0.1481
0.0784
-0.61
0.5429
$45K to <$75K
0.0932
-0.0214
0.2078
1.61
0.1102
$75K and over
0.1173
-0.0281
0.2626
1.60
0.1130
$20Kand over
0.1187
-0.1362
0.3737
0.92
0.3579
Refused/Don't know
0.0159
-0.3140
0.3458
0.10
0.9239
Missing
-0.2072
-0.4997
0.0853
-1.41
0.1632
Education
5.10
0.0079
Median forage
0.0834
0.0037
0.1632
2.07
0.0408
B-9
-------
Parameter
Estimate
LCL
UCL
tVatue
Probt
fVatue
ProbF
Race/Ethnicity
4.99
0.0011
Mexican American
-0.3862
-0.6024
-0.1700
-3.54
0.0006
Non-Hispanic Black
-0.0117
-0.2302
0.2067
-0.11
0.9155
Non-Hispanic White
0.4682
0.1973
0.7392
3.43
0.0009
Other Hispanic
0.0392
-0.2748
0.3532
0.25
0.8048
Other race
-0.1095
-0.4383
0.2193
-0.66
0.5105
NHANES Survey Year
9.34
<0.0001
1999-2000
0.4108
0.2809
0.5406
6.28
<0.0001
2001-2002
0.0822
-0.0725
0.2369
1.05
0.2947
2003-2004
-0.2095
-0.3164
-0.1026
-3.89
0.0002
2005-2006
-0.1295
-0.2556
-0.0034
-2.04
0.0444
2007-2008
-0.0327
-0.1316
0.0663
-0.66
0.5140
2009-2010
0.0667
-0.0392
0.1726
1.25
0.2144
2011-2012
-0.1880
-0.3420
-0.0339
-2.42
0.0174
Log -transformed hematocrit (centered)
1.4304
0.8855
1.9752
5.21
<0.0001
Log -transformed cotinine(centered)
0.0298
-0.0155
0.0751
1.31
0.1950
Log -transformed cotinine2 (centered)
-0.0251
-0.0703
0.0201
-1.10
0.2731
Log -transformed cotinine5(centered)
0.0164
-0.0435
0.0763
0.54
0.5880
B-10
------- |