Example Exposure Investigation Report

Westfield
Hampden County I Massachusetts
INFORMATION TO PROTECT OUR COMMUNITIES

Per- and Polyfluoroalkyl
Substances (PFAS)
Exposure Assessment

REPORT

National Center
for Environmental Health
Agency for Toxic Substances
and Disease Registry
11/18/2021
11
/ 18/ 2021

Table of Contents
Table of Contents ................................................................................................. i
Abbreviations .................................................................................................... iii
Executive Summary ............................................................................................ 1
Background and Purpose .................................................................................................................. 1
Exposure Assessment Activities ........................................................................................................ 2
Westfield Community-Wide Findings................................................................................................ 3
Limitations ....................................................................................................................................... 5
Recommendations ........................................................................................................................... 6
For More Information....................................................................................................................... 7

Background and Purpose .................................................................................... 1
What Are PFAS? ............................................................................................................................... 1
Why Westfield? ................................................................................................................................ 3

Methods ............................................................................................................. 4
Sampling Frame................................................................................................................................ 4
Participant Eligibility ........................................................................................................................ 6
Participant Recruitment ................................................................................................................... 6
Data Collection and Analysis ............................................................................................................ 6

Results .............................................................................................................. 13
Profile of Westfield EA Participants ................................................................................................ 13
Comparison of Westfield EA Participants’ Demographics to Sampling Frame Demographics ........... 15
PFAS in Blood ................................................................................................................................. 16
PFAS in Urine.................................................................................................................................. 33
PFAS in Tap Water .......................................................................................................................... 34
PFAS in Household Dust ................................................................................................................. 34

Discussion ......................................................................................................... 36
Generalizability of Westfield EA Community Statistics .................................................................... 37
Relationships Between Demographics and PFAS Blood Levels ........................................................ 38
Significance of Drinking Water Exposures ....................................................................................... 39
Other Exposure Characteristics ....................................................................................................... 40

Westfield Community-Wide Findings ................................................................ 41
Limitations ..................................................................................................................................... 43
Recommendations ......................................................................................................................... 44
For More Information..................................................................................................................... 45

References ........................................................................................................ 45
Appendix A: Additional Tables
Appendix B: Additional Background Statistics
Appendix C: PFAS Blood Levels by Demographics and Exposure Characteristics
i

Tables
Table 1. Summary of recruitment and data collection efforts ..................................................................... 9
Table 2. List of PFAS analyzed in blood, urine, tap water, and dust ........................................................... 10
Table 3. Characteristics of Westfield EA participants ................................................................................. 14
Table 4. Demographic comparison of EA participants and the sampling frame population ...................... 16
Table 5. Community statistics for PFAS in blood in micrograms per liter .................................................. 17
Table 6. Geometric means for PFAS in blood in micrograms per liter, unadjusted and age-adjusted to the
sampling frame ........................................................................................................................................... 18
Table 7. Comparison of PFAS blood geometric means (GMs) and 95th percentiles in Westfield,
Massachusetts, with the U.S. population (NHANES 2015–2016) in micrograms per liter ......................... 20
Table 8. Pearson correlation coefficients between PFAS in blood (log10) .................................................. 21
Table 9. Summary of statistically significant variables (p<0.05) in multivariate regression models .......... 22
Table 10. Community statistics for PFAS in urine reported in micrograms per liter .................................. 34
Table 11. Summary statistics for dust samples collected in Westfield ....................................................... 35

Figures
Figure 1. Sampling frame for Hampden County Exposure Assessment ....................................................... 5
Figure 2. Distribution of PFAS blood levels (log scale) ................................................................................ 17
Figure 3. EA PFAS blood levels compared to national averages ................................................................. 20
Figure 4. PFAS blood levels in adults and children (log scale) .................................................................... 24
Figure 5. PFAS blood levels in adults by sex (log scale) .............................................................................. 25
Figure 6. PFAS blood levels in adults by race and ethnicity (log scale) ....................................................... 26
Figure 7. PFAS blood levels in adults by length of residence in sampling frame (log scale) ....................... 28
Figure 8. PFAS blood levels in adults by blood-donation frequency (log scale) ......................................... 29
Figure 9. PFAS blood levels in adults by frequency of stain-resistant product use (log scale) ................... 30
Figure 10. PFAS blood levels in adults by frequency of consumption of local fruits and vegetables (log
scale) ........................................................................................................................................................... 31
Figure 11. Blood PFAS levels in females by breastfeeding (yes/no) (log scale) .......................................... 32

About ATSDR
The Agency for Toxic Substance and Disease Registry (ATSDR) is a federal public
health agency of the U.S. Department of Health and Human Services (HHS).
ATSDR works with other agencies and state, tribal and local governments to
protect communities from harmful health effects related to exposure to natural
and manmade hazardous substances. For more information about ATSDR, visit
https://www.atsdr.cdc.gov/.

ii

Abbreviations
9Cl-PF3ONS
11Cl-PF3OUdS
AFFF
ATSDR
CDC
DONA
EA
EPA
EtFOSAA
FOD
FtS 4:2
FtS 6:2
FtS 8:2
HFPO-DA (GenX)
LOD
MassDEP
MeFOSAA
µg/L
ng/g
NHANES
N-EtFOSA
N-EtFOSE
N-MeFOSA
N-MeFOSE
n-PFOA
n-PFOS
PFAS
PFBA
PFBS
PFDA
PFDoA
PFDS
PFDoS
PFHpA
PFHpS
PFHxA
PFHxS
PFNA
PFNS
PFOA

9-chlorohexadecafluoro-3-oxanone-1-sulfonic acid
11-chloroeicosafluoro-3-oxaundecane-1-sulfonic acid
aqueous film forming foam, also known as “A triple F”
Agency for Toxic Substances and Disease Registry
Centers for Disease Control and Prevention
4,8-dioxa-3H-perfluorononanoic acid
exposure assessment
U.S. Environmental Protection Agency
N-ethyl perfluorooctanesulfonamidoacetic acid
frequency of detection
fluorotelomer sulfonic acid 4:2
fluorotelomer sulfonic acid 6:2
fluorotelomer sulfonic acid 8:2
hexafluoropropylene oxide dimer acid
limit of detection
Massachusetts Department of Environmental Protection
N-methyl perfluorooctanesulfonamidoacetic acid
micrograms per liter (same as parts per billion or 1,000 parts per trillion)
nanograms per gram (same as parts per billion or micrograms per kilogram)
National Health and Nutrition Examination Survey
N-ethyl perfluorooctanesulfonamide
N-ethyl perfluorooctanesulfonamidoethanol
N-methyl perfluorooctanesulfonamide
N-methyl perfluorooctanesulfonamidoethanol
linear isomer of PFOA
linear isomer of PFOS
per- and polyfluoroalkyl substances
perfluorobutanoic acid
perfluorobutane sulfonic acid
perfluorodecanoic acid
perfluorododecanoic acid
perfluorodecane sulfonic acid
perfluorododecanesulfonate
perfluoroheptanoic acid
perfluoroheptane sulfonic acid
perfluorohexanoic acid
perfluorohexane sulfonic acid
perfluorononanoic acid
perfluorononane sulfonic acid
perfluorooctanoic acid

iii

PFOS
PFOSA
PFPeA
PFPeS
PFTA
PFTrA
PFUnA
ppt
Sb-PFOA
Sm-PFOS
UCMR 3

perfluorooctane sulfonic acid
perfluorooctanesulfonamide
perfluoropentanoic acid
perfluoropentane sulfonic acid
perfluorotetradecanoic acid
perfluorotridecanoic acid
perfluoroundecanoic acid
parts per trillion (same as 1 nanogram per liter)
branched isomers of PFOA
branched isomers of PFOS
Third Unregulated Contaminant Monitoring Rule

iv

Executive Summary
Background and Purpose
PFAS (or per- and polyfluoroalkyl substances) are a family of synthetic chemicals that have been used in
industry and consumer products since the 1950s. There are thousands of different PFAS. This
assessment discusses some of the most commonly studied PFAS, including perfluorooctanoic acid
(PFOA), perfluorooctane sulfonic acid (PFOS), perfluorohexane sulfonic acid (PFHxS), perfluorononanoic
acid (PFNA), perfluorodecanoic acid (PFDA), perfluoroundecanoic acid (PFUnA), and N-methyl
perfluorooctanesulfonamidoacetic acid (MeFOSAA).
PFAS do not occur naturally but are widespread in the environment. They have been found in soil,
water, air, and animal and plant life. Most PFAS (including PFOA, PFOS, PFHxS, and PFNA) are either very
resistant to breaking down or degrade into other PFAS that do not degrade further. Certain PFAS will
therefore remain in the environment indefinitely. Major exposure routes for PFAS include drinking
contaminated water and eating contaminated food, but exposure can also occur through other routes
(i.e., ingestion of contaminated dust). Once PFAS enter people’s bodies, some of them (including PFOA,
PFOS, PFHxS, and PFNA) can remain in the blood for long periods. Most people in the United States have
been exposed to PFAS. At least one PFAS was detected in more than 99% of NHANES samples collected
for the 1999-2000 survey cycle.
The Centers for Disease Control and Prevention (CDC) and the Agency for Toxic Substances and Disease
Registry (ATSDR) are conducting exposure assessments (EAs) in communities that were known to have
PFAS in their drinking water and are near current or former military bases. This report shares results
from the City of Westfield in Hampden County, Massachusetts, near Barnes Air National Guard Base
(the Base). When all EAs are complete, ATSDR will prepare a report analyzing the results across all sites.
Possibly as early as the 1970s, the Base used aqueous film forming foam (AFFF) containing PFAS for its
firefighter training. Over time, the PFAS from the AFFF entered the ground, moved into the groundwater
to offsite locations, and affected nearby municipal wells. By January 2016, the Westfield authorities had
removed from service the two water wells with the highest levels of PFAS contamination pending a new
water treatment system. Based on the information ATSDR has reviewed, the public drinking water
supply in Westfield currently meets the U.S. Environmental Protection Agency’s (EPA) 2016 health
advisory (HA) and state public health guidelines for PFAS in drinking water.1 At this time, ATSDR does not
recommend community members who get drinking water from the City of Westfield’s public water
system use alternative sources of water.
This EA assessed PFAS levels in the blood and urine of Westfield residents and compared them to PFAS
levels in a nationally representative sample. EA participants were recruited from the part of Westfield
that lies north of the Westfield River, where the highest PFAS contamination levels in tap water likely
occurred. Tap water and indoor dust samples from a subset of households were also analyzed for PFAS.
These EA results will help participants and their communities better understand their PFAS exposure,
explain what they can do to protect themselves from exposures, and inform public health efforts related
to protecting communities from sources of PFAS other than contaminated drinking water supplies.

1

ATSDR compared PFAS levels in drinking water to public health guidelines in place at the time of data collection.
The Massachusetts Department of Environmental Protection (MassDEP) published drinking water standards for
PFAS in October 2020.

ES-1

ATSDR will use the data collected from this and other EAs to help inform future studies of PFAS water
contamination.

Exposure Assessment Activities
ATSDR invited a randomly selected sample of Westfield households to participate in this EA. To be
eligible to participate, household residents must have (1) received their drinking water from the
Westfield Water Department, (2) lived north of the Westfield River for at least 1 year before January 20,
2016, (3) been greater than three years old at the time of sample collection (these residents have the
greatest likelihood of past exposures to PFAS via the city’s drinking water supply), and (4) not be anemic
or have a bleeding disorder that would prevent giving a blood sample. Results from randomly selected
households allow ATSDR to estimate exposure for all community members, even those who were not
tested.
In September 2019, 459 eligible people from 247 households participated in the EA sample collection
event. ATSDR performed the following tasks:
•

administered exposure history questionnaires to all participants

•

collected blood and urine samples from most participants

•

collected tap water and dust samples from the homes of 17 randomly selected participants

•

tested 7 PFAS in blood, 14 in urine, 18 in water, and 33 in dust2

•

measured PFHxS, PFOS, PFOA, PFNA, PFDA, and PFUnA across all media, and

•

mailed individual biological and environmental results to participants in May 2020

This report summarizes community PFAS blood levels, measured in serum, for the group of Westfield
residents. In this report, when we write blood levels of PFAS, we are referring to the measurement of
PFAS in the serum fraction of the blood. This report also summarizes urine sample results from a subset
of participants and presents results from the dust and tap water samples. Finally, the relationships
between blood results and the environmental sampling data are explored. The Westfield blood and
urine results are compared to a nationally representative sample of the US population. Specifically,
ATSDR compared Westfield data to those collected by CDC as part of its National Health and Nutrition
Examination Survey (NHANES). The NHANES survey collects blood and urine samples and tests them for
chemicals, including PFAS, from a representative sample of the civilian non-institutionalized U.S.
population. PFAS levels reported by NHANES are also shown by age, race/ethnicity, sex, number of years
living in the community, drinking water consumption patterns, and other exposure parameters.
The samples were collected and analyzed in strict accordance with ATSDR’s Exposure Assessment
Protocol: Biological and Environmental Sampling of PFAS (EA protocol) to ensure their quality. This EA
was designed to estimate geometric mean concentrations of PFOS in blood for the sampling frame (area
north of the Westfield River served by the City of Westfield’s municipal water supply) population, with a
precision goal of at least 15%. The precision is a measure of how wide the confidence interval is around
the estimated geometric mean. ATSDR met this goal as precision for this EA ranged from approximately
8% to 14%, depending on the individual PFAS. ATSDR also calculated geometric means that were
2

The laboratory reports branched and linear isomers of PFOA and PFOS in blood and urine. ATSDR reports on the
sum of the individual isomer concentrations of PFOA and PFOS.

ES-2

adjusted to the age distribution of the sampling frame population to correct for participation bias and to
provide an estimate that is more generalizable to the sampling frame community. ATSDR also calculated
geometric means that were adjusted to the national age distribution for comparison with the 2015–
2016 NHANES survey. To assess possible relationships between blood levels and various demographic
and exposure variables, ATSDR used statistical models. Univariate statistics were used to evaluate one
variable at a time, mostly as a tool to examine the data broadly and find patterns that existed within the
data. Multivariate statistics and regression modeling were used to account for multiple variables
simultaneous to control for potential confounding factors.

Westfield Community-Wide Findings
Finding 1. Blood levels of PFHxS, PFOS, and PFOA in the Westfield community are higher than
national levels.
Geometric means (i.e., averages) for PFHxS, PFOS, and PFOA blood levels were statistically higher
(p<0.05) in Westfield participants when compared to CDC’s NHANES (2015–2016) data, which was
limited to people over 12 years old. The statistically higher blood PFAS levels were observed for both
unadjusted geometric means for all EA participants and geometric means adjusted to the age
distribution of the U.S. population from NHANES 2015–2016.
Of the PFAS analyzed in blood, PFHxS had the largest elevations when compared to national levels. The
age-adjusted geometric mean blood PFHxS level among all Westfield EA participants was 3.4 times
higher than the national average. Blood PFHxS levels were above the national geometric mean for 92%
of the Westfield EA participants and above the NHANES 95th percentile for 46%. The age-adjusted
geometric mean blood PFOS and PFOA levels among Westfield EA participants were 1.1 times higher
than the national level.
Other PFAS measured in this EA (PFNA and PFDA) were not higher than the national average. PFUnA and
MeFOSAA were detected in fewer than 60% of the EA participant samples; due to the large percent of
samples below the limit of detection, geometric means were not calculated.

Finding 2. Elevated blood levels of PFHxS, PFOS, and PFOA may be associated with past
drinking water contamination.
The three PFAS (PFHxS, PFOS, and PFOA) with statistically elevated blood levels compared to national
geometric means were detected in Westfield’s drinking water as early as 2013. It is likely that
contamination began earlier, but no data are available before 2013. The maximum concentrations
observed in active drinking water wells in Westfield were 170 parts per trillion (ppt) for PFHxS, 160 ppt
for PFOS, and 43 ppt for PFOA in 2013. In 2016, Westfield reduced concentrations of PFAS below U.S.
EPA health advisory levels (70 ppt for PFOA and PFOS combined). Before 2016, PFAS-containing AFFF
were primarily formulated with PFOS, but also contained various PFAS precursors that could break down
into other PFAS, such as PFHxS, which could explain the elevated blood PFHxS levels. PFHxS, PFOS, and
PFOA have long biological half-lives (2.1 to 35 years). There were 3 years and 8 months between the
reduction of exposure via contaminated drinking water and collection of biological samples during the
EA. Because of the long half-lives of PFHxS, PFOS, and PFOA, past drinking water exposures may have
contributed to the EA participants’ blood levels.
PFHxS, PFOS, and PFOA were highly correlated in Westfield residents’ blood (Pearson correlation
coefficient, r, between 0.83 and 0.87). This means that typically, residents who had elevated blood

ES-3

PFHxS levels also had elevated blood PFOS and blood PFOA levels. This correlation suggests a common
exposure source, such as the Westfield public water supply, though other sources of exposure may also
have contributed to the observed blood levels.
Additional observations support the finding that past exposure to contaminated drinking water may
have contributed to the elevated blood levels.
•

First, in univariate models, a consistent and statistically significant predictor of participant blood
levels for PFHxS, PFOS, and PFOA was how long the resident had lived in Westfield before
January 2016. Those who lived in the area longest likely drank, in total, more contaminated
water. This relationship remained significant in a multivariate model for PFHxS.

•

Second, in multivariate models, PFHxS and PFOA blood levels in adults statistically increased
with the amount of tap water those adults reported drinking.

Multivariate models conducted separately for males and females suggest that these relationships
(between blood levels and residency duration/tap water consumption) were primarily observed in male
participants.

Finding 3. Age, sex, breastfeeding, use of stain-resistant products, and blood donation were
associated with some PFAS blood levels.
PFAS blood levels varied with different demographic and exposure characteristics of the participant
population. The following statistically significant relationships in the Westfield EA data set were
observed in adult participants (and are consistent with those reported in other non-ATSDR PFAS
studies):
•

Blood levels of PFHxS, PFOS, and PFOA were higher in older participants, and the size of the
effect varied by sex. In males, blood levels for these compounds increased by 0.5% to 1.2% for
every year of participant age. In females, blood levels for these compounds increased by 1.8% to
3.5% for every year of participant age.

•

Males had higher blood levels of PFHxS, PFOS, and PFOA than females. The difference between
males and females was larger in younger people. For example, 30-year-old males had higher
blood PFHxS, PFOS, and PFOA levels than 30-year-old females by 107%, 109%, and 57%,
respectively. For 50-year-old males, this difference was reduced to 31% for PFHxS, 58% for
PFOS, and 19% for PFOA compared to 50-year-old females.

•

Females who breastfed had lower blood levels of PFHxS, PFOS, and PFOA than females who did
not, and this effect was larger in younger women. For example, 30-year-old females who
breastfed had lower blood PFHxS, PFOS, and PFOA levels than 30-year-old females who had
never breastfed by 59%, 36%, and 43%, respectively. For 50-year-old females who had
breastfed, this difference was reduced to 39% for PFHxS, 19% for PFOS, and 21% for PFOA
compared to 50-year-old females who had never breastfed.

•

Only 49 participants reported ever using stain resistant products, and most of these reported
their frequency of use as “rarely.” Participants who reported ever using stain-resistant products
had 44% higher blood levels of PFHxS than those who never reported using these products.
Because of the small sample size for people who ever used stain resistant products, these
results should be interpreted with caution.

ES-4

•

Only 35 participants reported donating blood at least once or more a year. Participants who
reported donating blood at least once or more a year had 34% lower blood levels of PFHxS and
24% lower blood levels of PFOS than participants who never reported donating blood. Because
of the small sample size for people who reported donating blood once or more a year, these
results should be interpreted with caution.

A few associations were observed in children (<18 years), though many variables could not be examined
because of the small number of child participants (n=49). Because of the small sample size, results
should be interpreted with caution. Specifically, blood levels of PFOA decreased with age, and children
who were breastfed had higher blood levels of PFHxS and PFOA compared to non-breastfed children.
Infants born to mothers exposed to PFAS can be exposed in utero and while breastfeeding. However,
based on current science, the benefits of breastfeeding outweigh the risks for infants exposed to PFAS in
breast milk. The final aggregate report on all EA sites will include a more robust analysis of children.

Finding 4. Only one PFAS was detected in urine and at low concentrations.
ATSDR analyzed 47 (10%) of the urine samples collected. Only perfluorobutanoic acid
(PFBA) was detected; it was detected in 59.6% of the 47 samples that were analyzed. ATSDR did not
analyze all participants’ urine samples because none of the species were detected in more than 60% of
the samples analyzed.

Finding 5. All Westfield tap water samples collected during the EA in 2019 met the EPA’s HA
and Massachusetts Department of Environmental Protection’s (MassDEP) public health
guidelines for PFAS in drinking water.
This is based on 16 unfiltered and 8 filtered tap water samples collected in 17 households during the EA.
These results are consistent with recent data collected by the City of Westfield.

Finding 6. Patterns and levels of dust contamination measured in participating EA households
are comparable to those reported in selected U.S. studies.
Among the PFAS detected most frequently in household dust samples, PFOA and PFOS were measured
at the highest concentrations. No nationally representative comparison values are available, but
geometric mean and median concentrations for PFAS measured in dust collected in the small subset of
participating households (n=17) were within the range of levels reported in a few published studies of
other U.S. communities. None of the PFAS measured in this EA’s household dust samples were
statistically correlated with the same PFAS measured in participants’ blood. The final aggregate report
on all EA sites will likely include a more robust comparison of PFAS measured in dust and blood.

Limitations
There are several limitations associated with this assessment.
•

The random sampling recruitment method used for this EA was designed to measure blood PFAS
concentrations that were generalizable to all Westfield residents who lived north of the
Westfield River and were connected to the municipal drinking water supply. However, the EA
participant sample may not be fully representative of the community. Only 18% of the invited
households from the random sample participated in the EA sample collection event, and
participant characteristics were different than those of the area’s overall population.
Participants were older and less likely to be Hispanic or Latino. ATSDR addressed some of these

ES-5

concerns by calculating geometric mean estimates that were adjusted to the age distribution of
the community.
•

Blood, urine, and environmental PFAS concentrations may improve the understanding of
exposure in this community but will not provide discrete information about all sources of
exposure. Additionally, identifying every source of exposure is not possible.

•

Multivariate regression models did not explain a large portion of the variability in participants’
blood PFAS levels (R-squared or R2, a measure of model goodness-of-fit, ranged between 0.14
and 0.23, in the “all adult” models). This means that other factors not identified could influence
the relationships reported in this assessment (see “Statistical Analysis” section for details).

•

This study did not directly assess tap water consumption prior to the reduction of PFAS from the
municipal water system.

•

This EA was not designed to investigate health outcomes. Without additional information about
exposure response relationships, the results of this EA cannot be used to assess current or past
health problems or predict the future occurrence of disease.

•

The dust results are exploratory and should be interpreted with caution. They are based on a
limited set of samples, and in some cases those samples are based on a small sample mass.

Recommendations
This PFAS EA has demonstrated that past exposures to PFAS in drinking water have impacted the levels
of PFAS in people’s bodies. These PFAS are eliminated from the body over a long period of time. This
allowed ATSDR to measure PFAS even though exposures through drinking water were mitigated, or
lowered, years ago.
Although the exposure contribution from PFAS in drinking water in Westfield has been mitigated, there
are actions community members and city officials can take to further reduce exposures to PFAS and
protect public health.
Based on the PFAS drinking water test results from the City of Westfield’s municipal water system,
ATSDR does not recommend an alternate source of drinking water at this time.
1. What the City of Westfield can/should do:
a. Operators of the municipal water system should continue to monitor concentrations of PFAS
in drinking water delivered to the Westfield community to ensure concentrations of PFAS
remain below the EPA’s HA and MassDEP’s guidelines for specific PFAS in drinking water.
Results of PFAS monitoring should be shared with community members through appropriate
communication channels (Consumer Confidence Reports,
https://www.cityofwestfield.org/236/Water-Quality-Reports).
b. All treatment systems to remove PFAS from the municipal drinking water in Westfield
should be maintained appropriately to ensure PFAS concentrations remain below the EPA’s
HA and MassDEP’s guidelines for specific PFAS in drinking water.
2. What community members can/should do:
a. Become familiar with Consumer Confidence Reports
(https://www.cityofwestfield.org/236/Water-Quality-Reports) for information on the City of
Westfield’s water quality.

ES-6

b. Private well owners living in the area affected by PFAS should consider having their wells
tested for PFAS if testing has not been conducted before. To learn more about testing wells
for PFAS visit https://www.mass.gov/info-details/per-and-polyfluoroalkyl-substances-pfasin-private-well-drinking-water-supplies-faq. To learn more about previous testing for PFAS in
private wells in Westfield
visit http://eeaonline.eea.state.ma.us/EEA/fileviewer/Rtn.aspx?rtn=1-0020093.
c. Nursing mothers should continue breastfeeding. Based on current science, the known
benefits of breastfeeding outweigh the risks for infants exposed to PFAS in breast milk.
d. When possible, eliminate or decrease potential exposure to PFAS in consumer products
such as stain-resistant products and food packaging materials. To learn more visit
https://www.fda.gov/food/chemical-contaminants-food/questions-and-answers-pfas-food
e. Pay attention to advisories about food consumption, such as local fish advisories.
f. Discuss any health concerns or symptoms with your health care provider. Share results of
PFAS blood testing with your health care provider and make them aware of ATSDR
resources for clinicians (https://www.atsdr.cdc.gov/pfas/resources/info-for-healthprofessionals.html). Follow the advice of your health care provider and the
recommendations for checkups, vaccinations, and health screening tests.
g. Follow the advice of your child’s health care provider and the recommendations for well
child checkups, vaccinations, and health screening tests. Consult
https://health.gov/myhealthfinder to help identify those vaccinations and tests.
h. For additional information about environmental exposures and children’s health, contact
the Pediatric Environmental Health Specialty Units, a nationwide network of experts in
reproductive and children’s environmental health, https://www.pehsu.net/.
PFAS found in a person’s blood or urine means that exposure has occurred. The presence of PFAS in
blood or urine does not tell us how, where, when, or for how long a person was exposed to PFAS.
Exposure to PFAS does not mean adverse health effects will result, either now or in the future.

For More Information
If you have questions or comments or want more information on the Hampden County (Westfield) EA
site, call 800-CDC-INFO or email [email protected]. For more information on the work CDC/ATSDR is doing
to address PFAS exposure, visit ATSDR’s PFAS website, https://www.atsdr.cdc.gov/pfas/. For other EA or
PFAS-related questions, email [email protected].

ES-7

Background and Purpose
The Centers for Disease Control and Prevention (CDC) and the
Agency for Toxic Substances and Disease Registry (ATSDR) are
conducting exposure assessments (EAs) in communities near
current or former military bases that are known to have had
per- and polyfluoroalkyl substances (PFAS) in their drinking
water. One of these communities is the City of Westfield in
Hampden County, Massachusetts. This report summarizes the
findings of the Westfield EA. When all EAs are complete,
ATSDR will prepare an aggregate report analyzing the results
across all sites.

The PFAS exposure assessment in
Westfield focused on a specific
geographic area north of the
Westfield River where the highest
levels of PFAS in tap water likely
occurred. For purposes of this
report, we use the terms Westfield
EA and Hampden County EA
interchangeably to describe the
exposure assessment conducted in
this area. For more information
and a map of the area see the
“Methods” section of the report.

The EA involved collecting responses to exposure history
questionnaires, biological samples (blood and urine), and
environmental samples (tap water and household dust).
ATSDR collected biological samples and administered
questionnaires at the Westwood Building at 94 North Elm Street in Westfield between September 4 and
September 17, 2019. During the same time frame, ATSDR also took water and dust samples in a subset
of randomly chosen participant homes.
The results of the EA
•
•
•
•
•
•

tell us the amount of PFAS in the blood of individual participants and the Westfield community
and how these levels compare to the general U.S. population,
tell us the amount of PFAS in the urine of a subset of individual participants and the Westfield
community and how these levels compare to the general U.S. population,
provide a better understanding of environmental factors that may affect PFAS exposure,
provide information that may be used to stop or reduce PFAS exposure,
produce information that public health professionals can use to help communities affected by
PFAS, and
inform future studies looking at the effect of PFAS exposure on human health.

The EA does not look at what types of health problems are associated with exposure and is not meant to
determine if PFAS levels in blood or urine are risk factors for illness now or later in life. Additionally, the
EA does not tell us exactly how or where people were exposed or when or how long PFAS exposure
lasted.
ATSDR’s Exposure Assessment Protocol: Biological and Environmental Sampling of PFAS, termed the
PFAS EA Protocol [ATSDR 2019a], provides additional background, describes the criteria for selecting
communities for the EAs, and highlights the procedures ATSDR used in conducting the EAs.

What Are PFAS?
Human exposure to PFAS is a growing environmental health concern. PFAS are synthetic chemicals used
in many industries and consumer products since the 1950s. They have been used in nonstick cookware;
water-repellent clothing; stain-resistant fabrics and carpets; cosmetics; firefighting foams; and products
that resist grease, water, and oil. Exposure to PFAS has been associated with increased cholesterol,

1

decreased vaccine response in children, changes in liver enzymes, small decreases in infant birth
weights, increased risk of high e reason for this discrepancy is blood pressure or pre-eclampsia in
pregnant women, and increased risk of kidney and testicular cancer. [ATSDR 2021; Buck et al. 2011;
Gluge et al. 2020; Wang et al. 2017]
There are thousands of different PFAS. This assessment discusses some of the most commonly studied
PFAS, which include perfluorooctanoic acid (PFOA), perfluorooctane sulfonic acid (PFOS),
perfluorohexane sulfonic acid (PFHxS), perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDA),
and perfluoroundecanoic acid (PFUnA). The manufacture and import of PFOA, precursor chemicals that
can break down to PFOA, and related higher homologue chemicals, have been mostly phased out in the
United States. However, existing stocks of PFOA might still be used and there might be PFOA in some
imported articles. PFOS manufacture in the United States has not been reported to the EPA since 2002,
however, there are some limited ongoing uses of PFOS. These PFAS with long perfluoroalkyl chains are
no longer produced in the United States because of concerns over long biological half-lives. Other
countries may still manufacture and use them, but U.S. manufacturers have replaced these compounds
with shorter chained PFAS which typically have shorter biological half-lives. Some of the PFAS discussed
in this report, such as N-methyl perfluorooctanesulfonamidoacetic acid (MeFOSAA), are considered
precursors that can degrade in the environment or in people to other PFAS [ATSDR 2021; Gluge et al.
2020; Wang et al. 2017].
PFAS do not occur naturally but are widespread in the environment. PFAS can be released into the
environment during their production, use, or disposal. PFAS have been found in air, water, soil,
sediment, animal and plant life, and air. Most PFAS (including PFOA, PFOS, PFHxS, and PFNA) are either
very resistant to breaking down or degrade into other PFAS that do not degrade further. Certain PFAS
will therefore remain in the environment indefinitely. Most people in the United States have been
exposed to PFAS. At least one PFAS was detected in more than 99% of NHANES samples collected for
the 1999-2000 survey cycle [Calafat et al. 2007a]. Exposure can occur via contaminated drinking water
for which ingestion is believed to be the primary exposure route. Studies have shown that showering,
bathing, and swimming in water containing PFAS at levels seen in Westfield are not expected to be an
important contributor to PFAS exposure relative to the contribution from drinking water. [Sunderland
2019]
ATSDR’s PFAS EAs focused on communities with known exposures via contaminated drinking water.
However, residents may have had additional exposures to PFAS, such as from the following [Sunderland
2019]:
•
•
•
•
•
•
•
•
•
•

eating food packaged in materials containing PFAS (e.g., popcorn bags, fast food containers,
pizza boxes)
eating fish or shellfish caught in PFAS-contaminated waters
using consumer products such as stain-resistant carpeting and water-repellent clothing
eating garden vegetables grown with PFAS-contaminated water or in PFAS-contaminated soil
accidentally swallowing PFAS-contaminated soil
drinking infant formula mixed with PFAS-contaminated water
consuming breastmilk from women exposed to PFAS
gestational exposure to PFAS
working in industries that manufacture, process, or use products containing PFAS
background exposure to PFAS due to their ubiquitous nature

2

ATSDR asked study participants about these types of activities to evaluate whether these exposures
might influence PFAS levels in the EA communities.
After PFAS enter the human body, some PFAS can remain there for a long time. Some studies estimate
the half-life of PFHxS is between 4.7 and 35 years [ATSDR 2021]. Half-life estimates range from 3.3 to 27
years for PFOS and from 2.1 to 10.1 years for PFOA [ATSDR 2021].
The body of science about PFAS exposure and health effects is growing rapidly. Some, but not all,
scientific studies have shown that exposure to certain PFAS has been linked to harmful health effects.
While this EA does not examine specific health outcomes associated with PFAS exposure, EA findings
might help inform future studies on how PFAS exposure affects human health.

Why Westfield?
Westfield was one of several sites with identified PFAS drinking water contamination from use of
products such as aqueous film forming foam (AFFF). When selecting EA sites, ATSDR considered the
extent of PFOA and PFOS contamination in drinking water supplies, the duration over which exposure
may have occurred, and the number of potentially affected residents.3
PFAS and precursors that degrade to other PFAS measured in this EA were used in historical AFFF
formulations. Two types of PFAS-containing AFFF were manufactured before 2016 [ITRC 2020]. Both
formulations contained PFAS or PFAS precursors, the use of which resulted in the release of PFOS,
PFHxS, PFOA, and PFHxA into the environment. Possibly as early as the 1970s, the Barnes Air National
Guard Base used AFFF containing PFAS for its firefighter training. Over time, the PFAS from the AFFF
moved off site in groundwater and contaminated nearby municipal wells.
The date when PFAS first entered Westfield’s public water system is not known. These substances were
first detected in the city’s water in 2013, through testing conducted for the U.S. Environmental
Protection Agency’s (EPA’s) Third Unregulated Contaminant Monitoring Rule (UCMR 3) [EPA 2017]. The
rule required testing for six PFAS. At that time, drinking water provided by the Westfield Water
Department came from eight groundwater wells, two surface water reservoirs, and a connection with
the City of Springfield. UCMR 3 testing indicated two of Westfield’s four drinking water wells north of
the Westfield River (Wells #7 and #8) were contaminated with PFAS. Wells #7 and #8 were constructed
in 1978, but the date when they were first contaminated with PFAS is unknown. The highest sampling
result from an active well was 203 parts per trillion (ppt) for the sum of PFOA (43 ppt) and PFOS (160
ppt) in Well #7. PFHxS was also detected in this well at a concentration of 170 ppt. PFAS were not
detected in the city’s other drinking water sources including the wells located south of the river and the
surface water reservoirs.
The levels measured during UCMR 3 were not above EPA’s provisional health advisory, which at the time
was 400 ppt for PFOA and 200 ppt for PFOS. However, when EPA issued a lifetime health advisory for
the sum of PFOA and PFOS levels in drinking water (70 ppt) in 2016, the 2013 contamination levels were

PFHxS data were not available for all sites evaluated so were not considered in the site selection process even though
water contaminated by AFFF often has higher concentrations of PFHxS than PFOA or PFOS.
3

3

above this health advisory. To reduce concentrations of PFOA and PFOS in drinking water, the Westfield
Water Department removed the two contaminated wells (#7 and #8) from service. One (#7) was taken
offline in December 2015, and the second (#8) was taken offline in January 2016. The information ATSDR
obtained from City of Westfield indicates that PFOA and PFOS concentrations in drinking water were
below 70 ppt by January 20, 2016.
In addition to UCMR 3 testing, the Westfield Water Department conducted additional testing in the
summer of 2016, which showed PFAS detections in the two other wells north of the Westfield River (#1
and #2). PFOA+PFOS were detected at 54 ppt in Well #1 and 82 ppt in Well #2. Based on these
preliminary results, the Westfield Water Department issued a health advisory to its customers on
September 16, 2016. However, confirmatory sampling of these two wells on September 19, 2016,
showed considerably lower levels of PFAS contamination. At Well #1, PFOA, PFOS, and PFHxS were
measured at concentrations of 8.3 ppt, 25 ppt, and 45 ppt, respectively; at Well #2, PFOA, PFOS, and
PFHxS concentrations were 3.2 ppt, 6.8 ppt, and 17 ppt, respectively. Based on these results, the
Westfield Water Department issued a health advisory to its customers on March 31, 2017, indicating
that the final sampling results for Wells #1 and #2 were below the EPA health advisory and therefore not
a health concern. In 2017, the Westfield Water Department found that Well #2 exceeded the
Massachusetts Department of Environmental Protection’s (MassDEP) proposed action level at that time
of 70 ppt for total PFOA, PFOS, PFNA, PFHxS, and PFHpA [MassDEP 2018].4 Well #2 was taken offline
until a temporary treatment system was installed. The Westfield Water Department conducted multiple
additional rounds of testing of Wells #1 and #2 between 2017 and 2019 across these sampling events.
The highest PFOA+PFOS concentration was 33.3 ppt (i.e., the result of the Well #1 sample from
September 19, 2016).
The Westfield Water Department installed a temporary treatment system on the contaminated wells
that remains active. This system ensures that PFAS levels in the treated water are below detection
limits. The Westfield Water Department is in the process of installing permanent treatment on its four
affected wells. The Westfield Water Department continues to test its water sources and pursue system
improvements to address PFAS contamination.
The information available to ATSDR indicates that in 2019, the city’s drinking water met the EPA’s HA
and MassDEP public health guidelines for PFAS in drinking water.

Methods
ATSDR’s PFAS EA Protocol [ATSDR 2019a] details the approaches used to recruit participants, collect
samples, administer exposure history questionnaires, and evaluate data. This section briefly describes
how those methods were applied to the Westfield EA.

Sampling Frame
This EA focused on a specific geographic area, called the sampling frame or sampling area. The sampling
frame for this EA was the part of Westfield that lies north of the Westfield River, where the highest PFAS
contamination levels in tap water likely occurred (see Figure 1). Based on a review of Westfield land

4

ATSDR compared PFAS levels in drinking water to public health guidelines in place at the time of data collection.
Subsequent to data collection, in October 2020, the MassDEP published public drinking water standards for PFAS
of 20 ppt for total PFOA, PFOS,PFNA, PFHxS, PFHpA, and PFDA (“PFAS6”).

4

parcel data, ATSDR determined that 4,776 households in the sampling frame were connected to the
city’s water supply. These households formed the sampling frame from which households were
randomly selected for recruitment. Households with private wells were not eligible for participation.
Private well owners living in the area affected by PFAS should consider having their wells tested for PFAS
if testing has not been conducted before. To learn more about testing wells for PFAS visit
https://www.mass.gov/info-details/per-and-polyfluoroalkyl-substances-pfas-in-private-well-drinkingwater-supplies-faq. To learn more about previous testing for PFAS in private wells in Westfield
visit http://eeaonline.eea.state.ma.us/EEA/fileviewer/Rtn.aspx?rtn=1-0020093.

Figure 1. Sampling frame for Hampden County Exposure Assessment

Sampling Frame for Hampden County Site
Westfield, Hampden County, MA

-:

.

I

!"'

I

"

I,,.

91

!:

'O

~

~

sooe"!.t

(§1

:<

I u,

&ij

f

i

;;

. ! r.f

/

'O

~

111

J

J
p,O\pect~"e

w

\
'%,,

l

. .. !

'->'
fl,~~

0

.-·... . ,, \,,. ;; ,,•i
~

~fl,

,p
\

•;

. ,,,

,=tc
✓-""'

'J".,

'O

i
i

J

£ "'!:

",Ii

J

'i j°<,

ME

VT

,:

-f.'t'~.,

NH

57

NY

D
feeding HIiis Rd

.,,

j

189

MA

CT

sovthw

IURI

PA
NJ

NY

-~

■ Exposure Assessment Sampling Area
Westfield-Barnes Regional Airport

0

□ City of Westfield
■ Barnes Air National Guard Base

0.5

MIies
------------■

Data Sour=60% of samples had detections.

Statistical Terms
Geometric mean: The geometric mean is
a type of average and provides an
estimate of the central point of a set of
numbers. It is often used for
environmental data that exhibit a skewed
distribution (e.g., a data set with several
values that are much higher than the rest
of the results). The geometric mean is less
influenced by high values than an
arithmetic mean.
Percentiles (25th, 50th, 75th, 90th, 95th):
A percentile provides additional
information about the distribution of a
data set and represents the value below
which a certain percentage of the data
fall. For example, a 95th percentile of 25
micrograms per liter (µg/L) indicates that
95% of results fall below this
concentration.
Confidence intervals: A confidence
interval provides information about the
reliability of a statistic. In this EA, ATSDR
estimated geometric means for the PFAS
blood levels measured among study
participants. The 95% confidence interval
around the geometric mean represents
the range within which the true
population mean is expected to lie. More
specifically, if we hypothetically repeated
the study 100 times, 95 times out of 100
the mean of the sampling frame
population would fall within this range.
Precision: Precision provides information
on the reproducibility of a study and is
associated with sample size. The larger
the sample size the higher the precision.
In the context of this EA, precision was
estimated based on the width of
confidence intervals around the
geometric mean. A wide confidence
interval indicates low precision while a
narrow confidence interval suggests high
precision.

11

Geometric means were calculated as the measures of central tendency because of the lognormal
distribution of blood and urine measurements. Note that many of the statistics could not be calculated
for urine due to the low detection frequency.
One of the objectives of this EA was to estimate community-level exposures. While random recruitment
at the household level helps allow for such an estimation, ATSDR evaluated demographic differences
between the Westfield EA participants and all residents in the sampling frame. This was done for age,
race, and ethnicity using a two-sample test for equality of proportions. To correct for participation bias,
ATSDR also calculated geometric means adjusted to the age distribution of the sampling frame
population using 2010 Census block data.
ATSDR compared community-level statistics for PFAS in blood to national PFAS data reported by CDC in
the 2015–2016 NHANES (i.e., for the EA sample population 12 years of age and older). To control for
differences in the age distribution, the EA geometric
means were adjusted to the age distribution of the U.S.
A p-value helps determine the
population during NHANES 2015–2016. Note that
significance of the results of a statistical
NHANES 2017-2018 data were not available at the time
test, such as the difference between two
this report was originally drafted. For urine, ATSDR
means. The lower the p-value the more
compared community-level data to national-level data
likely the observed difference is not due
from the 2013–2014 NHANES compiled by Calafat et al.
chance alone. In this report, a p-value of
[2019], the only nationally representative data available
less than 0.05 (p<0.05) is described as
for PFAS in urine. ATSDR relied on two sample t-tests for
statistically significant.
these comparisons, using a p-value of less than 0.05 to
identify statistically significant differences.
ATSDR then used information gathered in the exposure questionnaire to understand and quantify how
demographic data and other exposure characteristics relate to PFAS measurements in blood. For this,
ATSDR relied on self-reported information, such as age, race/ethnicity, sex, length of residency in the
sampling frame, tap water and food consumption patterns, and work/school history. All numerical
responses were treated as continuous variables. In some cases, categorical variables were collapsed
when there were too few responses (<10) in a given category. To explore sex-specific associations (e.g.,
women having biological children [yes/no], having breastfed children [yes/no], duration of
breastfeeding), ATSDR also evaluated multivariate models for males and females only.
ATSDR did not conduct detailed statistical analyses for urine data because of low frequencies of
detection. ATSDR analyzed a subset of urine samples and found that, for all PFAS, the frequency of
detection was < 60%. The protocol specified that all urine samples would be analyzed if the geometric
mean calculated for any site exceeded the 95th percentile from NHANES. The protocol specified that
geometric means would be calculated if >=60% of samples had detections, and the rest of the samples
would be analyzed if the calculated geometric mean exceeded the NHANES 95th percentile. Since no
PFAS were detected in 60% or more of the analyzed samples, no geometric means were calculated for
any PFAS in urine and ATSDR did not analyze the remainder of the urine samples. ATSDR did calculate
the 95th percentile concentration for PFBA, the only PFAS detected in urine samples.
For tap water data, ATSDR compared PFAS levels measured with and without filtration to EPA’s health
advisory value (70 ppt for PFOA and PFOS combined) and the MassDEP public drinking water standard
for PFAS (20 ppt for the sum of PFHpA, PFHxS, PFNA, PFOA, PFOS, and PFDA) [MassDEP 2020]. For dust,
ATSDR calculated summary statistics and compared results to those in selected peer-reviewed

12

literature. ATSDR also evaluated correlations between PFAS levels measured in household dust and
blood collected from participants residing in homes where dust samples were collected.
To account for the one-stage cluster design, ATSDR conducted all statistical analyses in SAS (release 9.4,
SAS Institute, Cary, NC) using complex survey procedures (e.g., SURVEYMEANS, SURVEYREG). To do this,
ATSDR assigned household IDs to all participants and calculated summary statistics while accounting for
clustering at the household level. For blood results across all EA participants, intra-cluster correlation
coefficients ranged from 0.19 to 0.48, suggesting weak to moderate correlation of PFAS blood levels
within a household. Appendix B provides more information on clustering, as well as further details on
the statistical methods used for this EA and how results from this EA compared to the assumptions used
to estimate the target sample size of 395 participants.

Results
This section summarizes EA findings. It first profiles the Westfield EA participants and compares their
demographics to those of the entire sampling frame population, then reviews the blood, urine, tap
water, and household dust measurements that ATSDR collected. Those reviews use exposure history
questionnaire data to provide further context on the measurements. (The next section, “Discussion,”
further evaluates the observed trends using insights from the broader scientific literature on PFAS
drinking water exposures.)
Most analyses in this section reflect the entire Westfield EA participant population, but some pertain to
subsets of that population. This is because separate exposure history questionnaires were administered
to adults and children and because some questions on the adult questionnaire only applied to females.

Profile of Westfield EA Participants
EA participants responded to exposure history questions and reported information on many
characteristics, such as their age, sex, race/ethnicity, residential and occupational history, and drinking
water consumption. Table 3 summarizes this information.

13

Table 3. Characteristics of Westfield EA participants
Characteristics
Adults and children combined
Age (years)
<18
18 to 50
50+
Sex
Male
Female
Race and ethnicity†
White, non-Hispanic
Non-white or Hispanic
Adults only
Years lived at current address
<10
10 to <20
20 to <30
30+
Current primary drinking water source
Public water system
Bottled water
Average tap water consumption while living at current home (8ounce cups per day)
0
>0 to <2
2 to <4
4 to <6
6 to <8
8+
Current use of treatment or filtration device
One or more filter/treatment device(s)
None
Occupational exposures to PFAS in the past 20 years
One or more occupational exposure(s)
None

Count of EA
Participants (n)*

Percent of EA
Participants
(%)**

(mean = 49.7)
49
147
262

11
32
57

213
245

47
54

405
44

90
10

(mean = 19.9)
112
132
70
96

27
32
17
23

302
103

74
25

(mean = 5.3)
58
23
81
86
57
105

14
5.6
20
21
14
26

236
172

58
42

30
348

7.9
92

* The sums of participants for different fields in this table do not always add up to expected values, because not
every participant answered corresponding questions during the questionnaire.
** The sums of percentages for different fields in this table do not always add up to 100%, because not every
participant answered corresponding questions during the questionnaire and because of rounding.
†
ATSDR collapsed categories for race and ethnicity for all analyses because of the few responses across
categories.

14

The average age of EA participants was 49.7 years, and 90.2% of the participants identified themselves
as White non-Hispanic. Of EA participants, 53.5% identified as female, 46.5% identified as male, and
89.3% were adults, aged 18 years or older. The age cutoff is important because adults were
administered a different exposure history questionnaire with more detailed questions. Among the adult
participants, 72.7% reported living in their current homes for more than 10 years.
Adults were also asked about their current primary source of drinking water: 73.8% said Westfield’s
public water system, and 25.2% said bottled water. Adults reported drinking an average of 5.3 8-ounce
cups of water a day at home, and 57.8% said they currently use some type of filtering or treatment
device for their drinking water. Examples include filters on refrigerators, pitchers, and faucets; wholehouse carbon filtration systems; and reverse osmosis treatment systems. The questionnaire asked adults
for their occupational histories over the past 20 years; 7.9% reported holding one or more jobs with
potential PFAS exposures (e.g., firefighting, military, aviation).

Comparison of Westfield EA Participants’ Demographics to Sampling Frame
Demographics
This EA was designed to estimate PFAS levels in blood that were generalizable to the sampling frame as
a whole (i.e., Westfield households north of the Westfield River). The random sampling recruitment
method used for this EA helps ensure the absence of selection bias—that is, everyone in the sampling
frame had an equal chance of being chosen to participate. However, ATSDR also explored the potential
for participation bias—that is, substantive differences between those who chose to participate and
those who did not.
ATSDR used 2010 Census data (Table 4) [USCB 2010] to compare the EA participants’ demographic
profile with the profile of all residents in the sampling frame. ATSDR found two significant differences:
•

Age distribution. The EA participants included a higher proportion of older adults (age 50+
years) and a lower proportion of younger adults (18–50 years) and children (<18 years) than the
sampling frame population (Table 4). Specifically, 57% of the EA participants reported being 50
or older, but 34% of the sampling frame population falls in this age range. (ATSDR chose 50
years as a cutoff for older and younger adults based on the median age of menopause in the
United States, which may affect exposure profiles.) Similarly, 11% of the EA participants
reported being under 18, but 25% of the sampling frame population falls in that age range.

•

Race/ethnicity. Among the race/ethnicity characteristics, only the percent of residents who
identify as Hispanic or Latino showed a significant difference between the EA participants and
the sampling frame population (Table 4). Specifically, the EA population had statistically fewer
Hispanic or Latino participants (3.9%) than the sampling frame population (7.0%). For this
comparison, combined race and ethnicity were not available at the block level from the Census.
Therefore, only ethnicity and the race categories of White and More than one race were
compared because of the small number of respondents in other categories.

The effect of age on blood levels and its implications on community statistics is further explored
throughout this report. Refer to the “Discussion” section for ATSDR’s assessment of how these
demographic differences influence data interpretations.

15

Table 4. Demographic comparison of EA participants and the sampling frame population
Demographics
Age group (years)
<18
18 to 50
50+
Race
White
Black or African American
Am. Indian and AK Native
Asian
Nat. Hawaiian/Pacific Islander
More than one race
Ethnicity
Hispanic or Latino (of any race)

Number of
Participants
(n)*

Percent of
Participants
(%)

Sampling
Frame
Distribution
(%)†

p-Value‡

49
147
262

11
32
57

25
41
34

<0.001
<0.001
<0.001

424
<10
<10
<10
<10
11

92
—
—
—
—
2.4

93
1.5
0.1
0.8
0.04
1.8

0.55
—
—
—
—
0.42

18

3.9

7.0

0.002

* Counts may not sum to total because participants may have refused to answer questions. Counts are not shown
for categories with fewer than 10 participants.
†
Sampling frame data are based on the 2010 U.S. Census. Demographic characteristics of the sampling frame
may have changed between 2010 and 2019, the time of this EA.
‡
Two-sample test for equality of proportions with continuity correction comparing EA and 2010 Census data. A pvalue of less than 0.05 indicates a statistically significant difference between EA participants and all residents in
the sampling frame.

PFAS in Blood
This section summarizes PFAS levels that ATSDR measured from
the 459 blood samples provided by eligible participants. Results
are summarized in tables and ‘box and whisker’ plots (see text
box).

Unadjusted Community Statistics for PFAS in Blood
ATSDR first calculated the mean levels of PFAS without
accounting for the possible effect of age. Table 5 summarizes
results for the seven PFAS measured in Westfield EA
participants’ blood for all ages. Five of the seven PFAS—PFHxS,
PFOS, PFOA, PFNA, and PFDA—were detected in more than
85% of the blood samples. ATSDR’s statistical analyses
throughout this section focus on these five chemicals, and
Figure 2 shows the distributions of the individual
measurements on a log10 scale. The log10 scale allows for more
easily visualizing the wide range of serum concentrations as it
uses equal spacing for each factor of 10 increase. The PFAS
found at highest levels were PFOS (geometric mean = 5.87
micrograms per liter (µg/L)), PFHxS (4.67 µg/L), and PFOA (1.91
µg/L).

How to read a box and whisker plot:
A box and whisker plot illustrates a
summary of the data using different
statistical measures. See the image
below for how to interpret the
figures throughout this report.
O ~ - Outlier
~

Maximum observation
- below upper fence (1.5* 1QR
above 75"' percentile)
/ 75"' Percentile
.,, Mean
- Median
- 25"' Percentile

♦

''
'

1 Interquartile

: range (IQR}

''
'

♦

<- Minimum observation

16

Two PFAS—PFUnA and MeFOSAA—were detected in fewer than 60% of the samples. These low
frequencies of detection are consistent with NHANES data. Detailed statistics are not included for these
chemicals, and concentration percentiles (25th, 50th, 75th, 90th, 95th) are shown only for measurements
above the LOD.
The precision of geometric mean estimates for this EA ranged from approximately 8% to 14%,
depending on the PFAS (Appendix B, Table B2). These values are all below the desired precision of 15%
used to determine the target sample size for this EA, indicating that the collected data met the precision
target specified in the EA protocol.

Table 5. Community statistics for PFAS in blood in micrograms per liter
PFAS

FOD
(%)

PFHxS
PFOS
PFOA
PFNA
PFDA
PFUnA
MeFOSAA

99.6
NA*
NA*
98.0
85.8
58.2
54.5

Max

Geometric
Mean

95% CI for
Geometric
Mean

48.5
35.0
15.9
4.7
2.0
0.7
1.6

4.67
5.87
1.91
0.430
0.152
NA‡
NA‡

4.13–5.28
5.40–6.38
1.79–2.04
0.403–0.459
0.143–0.161
NA‡
NA‡

Percentiles
th

25th

50
(Median)

75th

90th

95th

2.20
3.60
1.27
0.246
NA†
NA†
NA†

4.53
5.88
1.86
0.389
0.105
NA†
NA†

9.76
9.90
2.72
0.610
0.179
0.159
0.143

19.4
15.5
4.01
0.823
0.269
0.265
0.363

24.9
18.6
4.88
1.08
0.347
0.348
0.556

FOD = frequency of detection, CI = confidence interval, NA = not applicable
* PFOA and PFOS are calculated sums of branched and linear subsets and are not measured directly. Linear PFOA
was detected in 99.8% of samples with a geometric mean of 1.82 micrograms per liter (µg/L); branched PFOA
was detected in 0.7% of samples. Linear PFOS was detected in 99.8% of samples with a geometric mean of 3.93
µg/L; branched PFOS was also detected in 99.8% of samples, but with a geometric mean of 1.86 µg/L.
†
Percentile is below the LOD.
‡
Per the EA protocol, geometric means were not calculated for PFAS detected in less than 60% of samples.

Figure 2. Distribution of PFAS blood levels (log scale)
::::;C)

2,

0

10

C:
0

0

◊

~
c

0

Q)
(.)

C:

0

0
(.)
"O
0

§
0

~

0

ci5
0.1

PFHxS

PFOS

PFOA

PFNA

PFDA

See 'How lo read a box and whisker plot' earlier in the PFAS in Blood section.
A log10 scale is used to allow easier visualization of/he wide range of measured blood levels, as it
uses equal spacing for each factor of 10 increase.

17

Community Statistics for PFAS in Blood Age-Adjusted to the Sampling Frame
Since the demographic profile comparison reported above showed that EA participants were
significantly older than the sampling frame as a whole, ATSDR also calculated geometric means that
were age-adjusted to the sampling frame population based on 2010 Census data for comparison.5 Ageadjusted geometric means correct for the participation bias discussed earlier and may be more
generalizable to the sampling frame community. Table 6 shows that in general, age-adjusted blood PFAS
geometric means are lower than unadjusted values. The greatest difference is observed for PFHxS and
PFOS, where age-adjusted geometric means are 20% and 18% lower than unadjusted values,
respectively. The lower values for age-adjusted geometric means reported here are consistent with
older adults having higher blood PFAS levels than younger adults. The effect of age and the implications
of these age-adjusted statistics are further discussed throughout this report.

Table 6. Geometric means for PFAS in blood in micrograms per liter, unadjusted and ageadjusted to the sampling frame
Unadjusted

Age-Adjusted to Sampling Frame

PFAS

Geometric
Mean

95% CI for Geometric
Mean

Geometric
Mean

95% CI for
Geometric Mean

PFHxS
PFOS
PFOA
PFNA
PFDA
PFUnA
MeFOSAA

4.67
5.87
1.91
0.430
0.152
NA*
NA*

4.13–5.28
5.40–6.38
1.79–2.04
0.403–0.459
0.143–0.161
NA*
NA*

3.84
4.84
1.77
0.411
0.145
NA*
NA*

3.45–4.28
4.48–5.23
1.66–1.88
0.380–0.445
0.136–0.154
NA*
NA*

CI = confidence interval
* Per the EA protocol, ATSDR did not calculate geometric means for PFAS detected in less than 60% of samples.

Comparison of EA Participants’ PFAS Blood Levels to the National Population
This section compares PFAS levels among Westfield EA participants to levels found in the U.S. general
population. To explore effects related to differences in the age distribution of EA participants vs. the
NHANES populations, ATSDR compares both unadjusted geometric means of all EA participants and
geometric means adjusted to the age distribution of the U.S. population in NHANES 2015–2016.
Table 7 shows the unadjusted comparison for the entire pool of EA participants to data from NHANES,
which are the geometric means for the 2015–2016 survey [CDC 2019]. For PFHxS, PFOS, and PFOA,
unadjusted geometric mean blood levels among Westfield EA participants were statistically (p<0.05)
higher than the national geometric mean. For PFNA, the unadjusted blood levels among Westfield EA
participants were statistically lower than the national geometric mean; for PFDA, no significant
difference was observed between Westfield EA participants and the general U.S. population.
Of the PFAS analyzed in blood, PFHxS levels had the largest elevations when compared to national
levels. The unadjusted geometric mean blood PFHxS level among Westfield EA participants was 4.0

5

One participant did not report their age and was therefore excluded from this analysis.

18

times higher than the national level. Blood PFHxS levels were above the national geometric mean for
92% of the Westfield EA participants and above the NHANES 95th percentile for 46% (Table 7). The
unadjusted geometric mean blood PFOS and PFOA levels among Westfield EA participants was 1.2 times
higher than the national level. Blood PFOS levels were above the national geometric mean for 61% of
the EA participants and above the NHANES 95th percentile for 5%. Blood PFOA levels were above the
national geometric mean for 67% of Westfield EA participants and above the NHANES 95th percentile for
9%.
On average, total PFOS measurements were composed of 67% linear PFOS (n-PFOS) and 32% branched
PFOS (Sm-PFOS). The proportion of n-PFOS found in EA participants’ blood is lower than that found in
standard PFOS products (76%–79%) [Kärrman et al. 2007] but comparable to levels found in the blood of
the general U.S. population [CDC 2019]. Measurements of total PFOA were composed of 95% linear
PFOA (n-PFOA) and 5% branched PFOA (Sb-PFOA), which is also comparable to the proportions found in
the U.S. population [CDC 2019]. All remaining statistical analyses in this report focus on total PFOA and
PFOS rather than treating the linear and branched isomers separately.
For this EA, ATSDR also calculated geometric means age-adjusted to the NHANES population. Because
the 2015–2016 NHANES survey does not report data for individuals over 12 years of age, these
geometric mean calculations are based on 436 EA participants. Table 7 and Figure 3 show that blood
PFAS geometric means adjusted to the NHANES population profile are lower than unadjusted values.
The adjusted geometric mean blood PFHxS levels among Westfield EA participants was 3.4 times the
national level. The age-adjusted geometric mean blood PFOS and PFOA levels among Westfield EA
participants was 1.1 times higher than the national levels. Even when controlling for the age-distribution
in the population, EA participants had statistically higher blood levels of PFHxS, PFOS, and PFOA than the
U.S. population.

19

Table 7. Comparison of PFAS blood geometric means (GMs) and 95th percentiles in Westfield,
Massachusetts, with the U.S. population (NHANES 2015–2016) in micrograms per liter
Westfield
Percent of
Percent of
Westfield GM
GM (CI)†:
Westfield
NHANES
Westfield
Westfield
NHANES GM
PFAS
(CI)†:
Age-Adjusted Results over
95th
95th
Results over
(CI)*
Unadjusted
to NHANES NHANES GM Percentile* Percentile NHANES 95th
2015-2016
(%)
Percentile (%)
4.67
4.02
1.18
PFHxS
(4.13–5.28)
(3.58–4.52)
91.7
4.90
24.9
46.0
(1.08–1.30)
p<0.001
p<0.001
5.87
5.29
4.72
PFOS
(5.40–6.38)
(4.89–5.73)
61.2
18.3
18.6
5.45
(4.40–5.07)
p<0.001
p=0.028
1.91
1.77
1.56
PFOA
(1.79–2.04)
(1.66–1.89)
66.9
4.17
4.88
9.37
(1.47–1.66)
p<0.001
p=0.005
0.430
0.418
0.577
PFNA
(0.403–0.459) (0.390–0.447)
36.3
1.90
1.08
1.31
(0.535–0.623)
p<0.001
p<0.001
0.152
0.148
0.154
PFDA
(0.143–0.161) (0.139–0.158)
51.9
0.700
0.347
0.44
(0.140–0.169)
p=0.777
p=0.501
PFUnA
NA‡
NA‡
NA‡
NA
0.400
0.348
2.61
‡
‡
‡
MeFOSAA
NA
NA
NA
NA
0.600
0.556
4.14
CI = 95% confidence interval, NA = not applicable
* Source: CDC 2019
†
P-values represent a t-test comparison between Westfield GM and NHANES GM.
‡
Per the protocol, geometric means were not calculated for PFAS detected in less than 60% of samples.

Figure 3. EA PFAS blood levels compared to national averages
P FHxS
PFOS

*

(/)

<{
u_
CL

PFOA
PFNA
PFDA

0

4

2

6

Geometric Mean (ug/L)
■

■

Westfield (Unadjusted)
NHANES

■

Westfield (NHANES Age-Adjusted)

Error bars represent 95% confidence inteNals. Note that overlapping confidence inteNals do not mean
that differences are not statistically significant.
•statistically Significant Difference from NHANES (p<0.05)

20

Correlations Among PFAS in Blood
ATSDR also evaluated correlations between PFAS in blood (log10). This analysis determined whether any
PFAS tended to have similar patterns in the blood of Westfield EA participants. ATSDR used Pearson
correlation coefficients (r) for this analysis. An r of 0 means two data sets are uncorrelated, and an r of 1
means two data sets are correlated (i.e., they rise and fall in proportional amounts). Table 8 shows the
Pearson correlation coefficients for the five frequently detected PFAS.
PFHxS, PFOS, and PFOA blood levels showed the strongest correlations (Table 8). All pairings of these
chemicals had Pearson correlation coefficients close to 1 (r = 0.83–0.87). On the other hand, PFNA and
PFDA had weaker correlations with each other and the other PFAS (r = 0.24–0.57).

---

Table 8. Pearson correlation coefficients between PFAS in blood (log10)*
PFHxS
PFOS
PFOA
PFNA
PFDA

PFHxS
1.00
0.87
0.83
0.39
0.24

PFOS
0.87
1.00
0.83
0.53
0.33

PFOA
0.83
0.83
1.00
0.54
0.33

PFNA
0.39
0.53
0.54
1.00
0.57

PFDA
0.24
0.33
0.33
0.57
1.00

* p<0.001 for all correlations.

PFAS Blood Levels by Demographics and Other Exposure Characteristics
This section examines how the demographic and exposure history information collected during the
questionnaire relates to blood PFAS levels. Since different questionnaires were administered to adult
and child participants, responses were analyzed separately. Additionally, some questions were
applicable only to female adult participants and are therefore also presented separately. Appendix C
(Tables C1 and C2) presents a complete summary of all adult and child questionnaire responses.
ATSDR used univariate and multivariate models to
evaluate the relationships between questionnaire data
and blood PFAS levels. This section summarizes
relationships that were found to be statistically significant.
For this EA, the following demographic and exposure
characteristics were found to be associated with at least
one PFAS:
• age,
•

sex,

•

race/ethnicity,

•

tap water consumption,

•

length of residence in the sampling frame,

•

blood donation,

•

use of stain-resistant products, and

•

breastfeeding (adult females and children only).

ATSDR created mathematical models
to identify demographic and lifestyle
characteristics associated with PFAS
blood levels.
Univariate models evaluated the
effects of one variable, or exposure
characteristic, at a time while
multivariable models evaluated the
joint effect of multiple characteristics
on blood PFAS levels at the same time.
Multivariable regression models
describe the average increases in PFAS
blood levels for each unit increase in
the exposure characteristics.

21

Table 9 summarizes the demographic and exposure characteristics that were statistically significant in
each multivariate model.

Table 9. Summary of statistically significant variables (p<0.05) in multivariate regression
models
Parameter

PFHxS
All
Adult Adult
Adult Female Male
✓
✓
✓
NA
NA
✓
✓
NA
NA

PFOS
All
Adult Adult
Adult Female Male
✓
✓
✓
NA
NA
✓
✓
NA
NA

Age (continuous)
Sex (categorical)
Age × sex (continuous)*
Years in sampling frame in the
past 20 years [Residency
✓
—
✓
—
—
duration] (continuous)
Blood donation frequency
—
✓
✓
✓
✓
(categorical)
Stain-resistant product use
—
—
—
✓
✓
(categorical)
Tap water consumption at
home in cups per day
✓
—
✓
—
—
(continuous)
Breastfeeding (categorical)
NA
✓
NA
NA
✓
Age × breastfeeding
NA
✓
NA
NA
✓
(continuous)
✓ = statistically significant, ‘—’ = not statistically significant, NA = not applicable

PFOA
All
Adult Adult
Adult Female Male
✓
✓
✓
NA
NA
✓
✓
NA
NA

—

—

—

—

—

—

—

—

—

—

—

—

—

✓

—

✓

NA

NA

✓

NA

NA

NA

✓

NA

*This variable is an interaction term, which means the effect of one variable on serum PFAS levels depends on the
value of another.

The following subsections briefly summarize results for these
topics. All other results are presented in Appendix C, as described
below.
•

•

Tables C1 and C2 present response frequencies for all
questions included in the adult and child questionnaire,
respectively. These tables also present geometric means
and 95% confidence intervals around geometric means
stratified by the response options (e.g., statistics are
presented separately for males and females) for PFHxS,
PFOS, PFOA, PFNA, and PFDA. While blood levels of PFNA
and PFDA were not found to be statistically higher than
the national geometric means, both PFAS were detected
at a high enough frequency to present meaningful results.
Summary statistics are therefore provided in Appendix C
for completeness, but not discussed below.

Variability describes the spread or
dispersion of data values. If the
values are similar to each other there
is little variability, if the values are
spread out there is more variability.
Multivariable regression can help us
understand how much of the
variability in PFAS blood levels can be
explained by the combination of
factors in the model such as age, sex,
and length of residency among
others. If the model does not explain
a large portion of the variability, that
means there are other unknown
factors influencing PFAS levels in
blood.

Tables C3 and C4 present univariate modeling results for
all questions in the adult and child questionnaire for the

22

same five PFAS, as data allow. Data are presented only when a category had at least 10
responses. Some categories were collapsed to meet this threshold.
•

Tables C5–C13 present multivariate modeling results for PFHxS, PFOS, and PFOA. Multivariate
models, including the goodness-of-fit measure, R-squared or R2, are presented separately for all
adults, male adults only, and female adults only. The closer the R2 value is to 1, the more the
variables in the model explain the variability in blood PFAS levels. Across all models, R2 values
ranged from 0.07 to 0.34. ATSDR modeled males and female adults separately to explore sexspecific differences including the potential effect of childbirth and breastfeeding on female
blood PFAS levels. The variables considered in male-only and female-only models were limited
to those that were significant in final all-adult models. ATSDR did not develop multivariate
models for children because of the small sample size for this population (n=49).

•

Figures C1–C39 present boxplots for unadjusted blood levels by each demographic and exposure
characteristic included in the statistical analyses.
Goodness of Fit Measure

R-squared or R2 is a statistical measure used
to evaluate how well a mathematical model
explains the measured data by looking at the
differences between the observed PFAS
concentrations and values predicted by the
model.
• An R2 of 1 means the model completely
predicts the observed PFAS
concentrations, so that there are no
differences between the model and the
PFAS concentrations and 100% of the
PFAS concentrations are explained by the
model.
• An R2 of less than 1 means that there are
measurements scattered higher and/or
lower than the model predictions and
there are differences between the two.

Blood PFAS Levels and Age
Because many studies have found that older people
have higher blood PFAS levels, ATSDR investigated
how Westfield EA participants’ ages related to their
blood levels. As Figure 4 illustrates, the blood levels
for PFHxS, PFOS, and PFOA increased with age in
adults, but trends were inconsistent in children.
For adults, ATSDR’s univariate analysis showed that
blood PFHxS, PFOS, and PFOA were higher in older
individuals than in younger individuals, and this
finding was statistically significant. As Figure 4
shows, PFHxS had the strongest age dependence.
The univariate analysis indicates that on average,
blood PFHxS levels in Westfield EA participants
increased 2.6% for every year of participant age in
adults. This suggests a 30% increase in blood PFHxS
levels for every 10 years of participant age in adults.
The calculated increases for PFOS (1.9% per year of
participant age) and PFOA (1.2% per year of
participant age) were lower.

ATSDR’s multivariate analysis provided further perspective on this trend, showing that the age
dependence was generally stronger for women than men among adults. For example, the all-adult
model (Appendix C, Table C5) suggests a 3.5% increase in blood PFHxS levels in adult females for every
year of participant age and a 1.1% increase in blood PFHxS levels in adult males for every year of
participant age when controlling for other characteristics; this finding was statistically significant. Similar
results were observed in the stratified male-only and female-only models. Age remained a significant
predictor of blood levels for all three PFAS in all multivariate models.
As Figure 4 shows, blood PFHxS and PFOA levels were higher in younger children for participants under
18. In univariate analyses, this trend was statistically significant only for PFOA, for which every one-year
increase in age under 18 was associated with a 2.6% decrease in blood PFOA levels. Age was not

23

statistically associated with PFHxS or blood PFOS levels in children. Note that multivariate models were
not explored for children because of the relatively small sample size.

Figure 4. PFAS blood levels in adults and children (log scale)
PFHxS*

PFOA*· t

PFOS *

100

::i'
0)

0

2c

:8

O 0

ell

10

0

0

0

~

c
(.)

C
0
(.)
"O
0
0

ffi

1

0

0

0
0

0

0

0

0

0.1

0

0

0

20

0 0

0

40

60

0

80

0

20

40

60

80

0

20

40

60

80

Age
- - - Adult - - - Child
A log10 scale is used to allow easier visualization of the wide range of measured blood levels, as it uses equal spacing for each
factor of 10 increase.
*Statistically Significant Trend (p<0.05) in Adults
tStatistically Significant Trend (p<0.05) in Children

Blood PFAS Levels by Sex
ATSDR investigated how blood PFAS levels vary between males and females because previous research
has shown that, all other factors considered equal, adult males tend to have higher blood PFAS levels
than adult females. ATSDR’s univariate and multivariate analyses both showed that PFAS levels were
higher in adult males than in adult females for PFHxS, PFOS, and PFOA. Modeled blood levels in adult
males were 19% higher for PFHxS, 43% higher for PFOS, and 18% higher for PFOA in univariate models
(Figure 5).
The all-adult multivariate models showed that the difference between males and females was larger in
younger people. For example, 30-year-old males had higher modeled blood PFHxS, PFOS, and PFOA
levels than 30-year-old females by 107%, 109%, and 57%, respectively. For 50-year-old males, this
difference was reduced to 31% for PFHxS, 58% for PFOS, and 19% for PFOA compared to 50-year-old
females. Blood levels of these three PFAS were not statistically associated with sex in children.

24

Figure 5. PFAS blood levels in adults by sex (log scale)
100

:::.

*

*

0)

0

2,
C
0

i
0

*

10

~
....

c

Q)
(.)

C
0
(.)

"O
0
0

iii

0.1
PFHxS

PFOS
Sex

■

Female
N=221

PFOA
■

Male
N=1 88

See 'How to read a box and whisk er plot' earlier in the PFAS in Blood section.
A log10 scale is used to allow easier visualiz ation of the wide range of measured blood levels, as it
uses equal spacing for each factor of 10 increase.
•statistically Significant Difference (p<0.05)

Blood PFAS Levels by Race/Ethnicity
The exposure history questionnaire asked participants to provide information about their race and
ethnicity. Because there were not enough participants in different race and ethnicity categories to
support robust statistical analyses, ATSDR focused on differences between Westfield EA participants
who self-identified as White, non-Hispanic and those who identified as Non-white, or Hispanic.
Figure 6 shows that on average, when compared to those who identified as White, non-Hispanic, blood
levels in Non-white, or Hispanic participants were 34.8% lower for PFHxS, 20.3% lower for PFOS, and
23.2% lower for PFOA in univariate models. Race and ethnicity did not remain as significant predictors of
these PFAS in multivariate analyses. This may result from age being correlated with race and ethnicity in
the U.S. population (White, non-Hispanic populations tend to be older than Non-white, or Hispanic
populations). Also, in the wider U.S. population, levels of PFAS in Hispanics tended to be lower than in
other race and ethnicity groups.

25

Figure 6. PFAS blood levels in adults by race and ethnicity (log scale)
100

§

*
0

*

10

~
....

c

Q)

u

§

1

(.)
"O

0
0

iD

0.1

PFHxS
■

PFOS

PFOA

R ace and Ethnicity
White, Non-Hispanic ■ Non-White or Hispanic
N=369

N=32

See 'How lo read a box and whisker plot' earlier in the PFAS in Blood section.
A log10 scale is used to allow easier visualization of the wide range of measured blood levels, as it
uses equal spacing for each factor of 10 increase.
•statistically Significant Difference (p <0.05)

Blood PFAS Levels and Tap Water Consumption
ATSDR investigated several questions from the adult and child questionnaires to characterize
relationships between blood PFAS levels and consumption of PFAS-contaminated drinking water. These
questions are about the drinking water source, amount of tap water consumed at home or school, and
residential history. In some cases, data trends may have been affected by subtleties in the wording of
exposure history questions, as described below.
For adults, ATSDR first considered participants’ primary drinking water source. Adult participants were
asked, “What is your current main source of drinking water in your home?” Nearly all of the responses
were tap water (74%) or bottled water (25%). There were no statistically significant differences in blood
levels between these two groups in univariate or multivariate analyses. The lack of significant
differences was an unexpected result, which ATSDR believes may be a result of how the question was
worded—particularly the word “current.” ATSDR also asked participants about any changes to their
drinking water habits in the past year; 18% reported switching from public water to bottled water in the
past year. However, since drinking water exposure in Westfield was mitigated in 2016, changes in
drinking water behavior within the past year would not affect drinking water exposure. It is possible that
participants who reported currently drinking bottled water or switching water sources in the past year
drank tap water during the period of contamination, but the extent to which that occurred is not known.
Due to these considerations, ATSDR’s data analysis did not rely on answers to these questions when
interpreting associations between PFAS levels and exposure characteristics.
ATSDR also considered participants’ self-reported tap water consumption rates. Adult participants were
asked, “During the time you lived in a home served by the water source identified above [i.e., for the
question quoted in the previous paragraph], on average how many 8-oz cups of water or beverages
prepared with tap water did you drink while at home per day?” ATSDR’s univariate analysis did not
reveal a significant linear relationship between blood PFAS levels and the amount of tap water

26

consumed.6 However, significant relationships were
observed in the multivariate analysis, which controlled for
other potential confounders. For every additional cup of
tap water an adult reported drinking at home per day,
PFHxS levels increased by 2.8% and PFOA levels increased
by 1.6% in all-adult models; both of these increases were
statistically significant. In male-only models, these
associations were significant and larger (4.4%). They were
not significant in female-only models, suggesting that the
relationship was primarily observed in male participants.
Associations between tap water consumption and blood
PFOS were not statistically significant.
For adults, ATSDR also considered the length of residency.
The exposure history questionnaire asked adults where
they had lived for the past 20 years. ATSDR calculated the
total amount of time participants reported living in
Westfield over this period. These responses can serve as a
proxy for potential exposure to PFAS-contaminated
drinking water in the community. That is, the longer the
residence within the sampling frame, the greater the
likelihood of past PFAS exposure from the Westfield
drinking water supply. Any resident reporting prior
residences in Westfield was assumed to fall within the
sampling frame.

What are confounders?
Confounding is a distortion in the
estimated relationship between a
potential predictor and measure of
exposure due to the presence of a
third variable—called a confounder.
In order for confounding to occur,
that third variable must be associated
with both the predictor (or
independent variable) and the
measure of exposure (or dependent
variable). For example, age can act as
a confounder on the estimated
strength of association between
length of residence in the sampling
frame and blood PFAS levels.
By adjusting for these types of
confounding variables in multivariate
statistical models, ATSDR can
calculate less biased estimates of the
relationships between dependent
and independent variables of
interest.

Figure 7 shows the relationship between reported residence duration in Westfield for the past 20 years
and blood PFAS levels. A consistent relationship was observed for PFHxS, PFOS, and PFOA: blood levels
increased with the number of years participants lived in the sampling frame, and this effect was most
pronounced for PFHxS. The multivariate analysis showed that only PFHxS had a statistically significant
relationship with residency duration: for every additional year that an adult participant lived in
Westfield, blood PFHxS increased by 3.6%. In male-only models, this association was significant and
larger (4.6%). The association was not significant in female-only models, suggesting that the relationship
was primarily observed in male participants. For PFOA and PFOS, the multivariate analysis did not show
statistically significant relationships with residency duration.
ATSDR also considered relationships between blood PFAS levels and current use of drinking water
filtering devices and water treatment devices but found no significant associations. While one would
expect properly maintained filtering and treatment devices to decrease PFAS drinking water exposures,
the questionnaire did not ask participants when they installed these devices. If they were installed after
PFAS mitigation was complete in January 2016, no significant relationships would be expected.

6

Initially, a significant linear relationship was observed between the reported amount of water consumed per day
among all adult participants (ranging from 0 to 64 cups) and blood levels of multiple PFAS. However, two
participants reported drinking water rates that were determined to be outliers (i.e., 48 and 64 cups per day). With
these outliers excluded, there were no significant associations between tap water consumption at home and blood
levels of any PFAS.

27

Finally, an exposure history question pertained to whether adult participants drank tap water while at
work. However, because identifying whether a participant’s place of employment was in the sampling
frame was difficult, ATSDR did not evaluate the data for drinking water consumption patterns at work.
PFHxS, PFOS, and PFOA were detected in Westfield’s drinking water sources (PFHxS at 170 ppt, PFOS at
160 ppt, and PFOA at 43 ppt). Therefore, one explanation for the high correlation among these
compounds in the blood is that the Westfield EA participants had a common exposure profile for PFHxS,
PFOS, and PFOA, such as drinking water. However, the correlations alone cannot be used to identify the
underlying source or combination of sources that contributed most to exposure.

Figure 7. PFAS blood levels in adults by length of residence in sampling frame (log scale)
PFHxS*

PFOS*

PFOA*

100
0

---

0)

2c
0

0

0

10

0

~ o0
0

oo

0

c

0

~

8

8°
0

0

o

0

Ori
0
0

00

o<#

<9
0

0

"O
0
0

8

Q>O

000

e a""
o,p

o

g

0

8O
0

0
0

O 0

o
o
O
00 0

o

0

O 0

oo

(;>

oFile Typeapplication/pdfFile TitleHampden County, MA PFAS Exposure Assessment ReportSubjectPFAS, Exposure Assessment, Per- and Polyfluoroalkyl Substance, ReportAuthorAgency for Toxic Substances and Disease RegistryFile Modified2021-11-17File Created2021-11-08