Environmental Research 128 (2014) 78–83

Contents lists available at ScienceDirect

Environmental Research
journal homepage: www.elsevier.com/locate/envres

Incidence of type II diabetes in a cohort with substantial exposure

to perfluorooctanoic acid
Conny Karnes n, Andrea Winquist, Kyle Steenland
Department of Environmental Health, Rollins School of Public Health, Emory University, 1518 Clifton Road, Atlanta, GA 30322, United States

art ic l e i nf o a b s t r a c t

Article history: Background: Research suggests an increased type II diabetes mortality risk among workers occupation-
Received 20 May 2013 ally exposed to PFOA. However, a cross-sectional study of highly exposed Mid-Ohio Valley community
Received in revised form residents did not demonstrate an association between PFOA and type II diabetes.
8 October 2013
Objectives: We examined the relationship between exposure to PFOA over time and incidence of
Accepted 11 November 2013
Available online 2 December 2013
type II diabetes in a cohort of community residents and workers exposed to high levels of PFOA via
contaminated drinking water.
Keywords: Methods: Community residents and workers were interviewed in 2008–2011 to obtain medical history
C8 and other demographic information. Cumulative serum PFOA exposure estimates were calculated based
C8 Science Panel
on residence and occupation locations, and a history of plant emissions. We estimated the risk of
Diabetes
developing type II diabetes using Cox proportional hazard models, controlling for demographic
Perfluorooctanoic acid
characteristics and family history.
Results: Out of 32,254 survey respondents, there were 4434 cases of self-reported type II diabetes, of
which 4129 were validated through medical record review. In analyses based on validated type II
diabetes, there was no trend of increased risk with increased cumulative PFOA serum levels (HRs
compared to lowest exposure decile: 0.91 (95% CI: 0.76–1.08), 1.18 (95% CI: 0.99–1.40), 0.96 (95% CI: 0.81–
1.15), 1.04 (95% CI: 0.87–1.24), 1.11 (95% CI: 0.93–1.32), 1.06 (95% CI: 0.89–1.26), 1.00 (95% CI: 0.85–1.19),
1.03 (95% CI: 0.86–1.23), 1.01 (95% CI: 0.84–1.20)). There was no association between fasting glucose level
and cumulative serum levels of PFOA, after excluding diabetics.
Conclusions: We do not find an association between PFOA exposure and incidence of type II diabetes.
& 2013 Elsevier Inc. All rights reserved.

1. Background Perfluorooctanoic acid (PFOA) is a synthetic chemical that has

Diabetes mellitus is a group of metabolic diseases characterized
by elevated blood glucose levels. Type 1 diabetes is an inflamma-
tory autoimmune disease of the pancreas in which the body does
not produce enough insulin, and accounts for 5% of diagnosed
diabetes cases. In type 2 diabetes, either not enough insulin is
produced or the body is unable to use the insulin produced
effectively. Risk factors for type 2 diabetes include obesity, lack
of physical activity, racial/ethnic background, and genetic predis-
position. Recent evidence has demonstrated that developmental
exposure to environmental chemicals is associated with over-
weight and obesity in adulthood (Newbold, 2010), and there has
been increasing interest in examining the role of such ‘non-
traditional’ risk factors in the development of diabetes and obesity
(Taylor et al., 2013).
n
Corresponding author. Fax: þ1 404 727 8744.
E-mail address: ckarnes@emory.edu (C. Karnes).
been widely used in manufacturing applications for over 60 years.
PFOA is resistant to degradation and persists indefinitely in the
environment. PFOA is not metabolized, and has an estimated
human half-life of 2.3–3.4 years (Bartell et al., 2010; Brede et al.,
2010; Olsen et al., 2007). PFOA exposure is widespread in the US
and other industrialized countries (Lindstrom et al., 2011). Low
levels of PFOA were detected in the serum of 99.9% of the
participants of the 2007–2008 National Health and Nutrition
Examination Survey (NHANES), with a geometric mean of 4.1 ng/
mL (Kato et al., 2011). Workers with occupational exposure to
PFOA at manufacturing facilities have much higher serum PFOA
levels than the general population (Steenland and Woskie, 2012).
PFOA is water-soluble and migrates readily from soil to ground-
water. Human exposure can occur via contaminated drinking
water. Other sources of exposure include food, food packaging,
treated fabrics, house dust, and air (Lau et al., 2007). Exposure to
even low levels of PFOA in drinking water has been shown to
increase serum PFOA levels by about 100 times the drinking water
concentration (Post et al., 2012). Higher levels of serum PFOA have
been found in populations where drinking water supplies were

0013-9351/$ - see front matter & 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.envres.2013.11.003
 C. Karnes et al. / Environmental Research 128 (2014) 78–83
highly contaminated from emissions from a nearby chemical plant
that used PFOA in the manufacture of Teflon (Steenland et al.,
2009a).
In animal studies, PFOA has been associated with several adverse
health outcomes including tumors and liver, immune, developmental,
metabolic, and neurobehavioral toxicity (Post et al., 2012). The USEPA
has classified PFOA as ‘likely to be carcinogenic in humans’ and recent
evidence links PFOA exposure to kidney and testicular cancer (Barry
et al., in press). Human studies conducted in various populations
(general public, communities with contaminated drinking water, and
workers) have demonstrated additional health effects of PFOA, includ-
ing elevated cholesterol (Nelson et al., 2010; Steenland et al., 2009b),
uric acid (Steenland et al., 2010), and liver enzymes (Lin et al., 2010);
and thyroid disease (Melzer et al., 2010). Due to its ubiquity,
persistence, toxicity, and associated health effects, the manufacture
and use of PFOA is presently being phased out by eight major
manufacturers in a voluntary stewardship agreement with the USEPA.
High serum PFOA concentrations have been documented
among workers at a DuPont chemical plant in the mid-Ohio valley
where PFOA was used to produce Teflon and other products
(Steenland and Woskie, 2012). Studies of workers at this plant
indicate increased risk of type II diabetes mortality among those
exposed to PFOA. One study examined a cohort of 6027 employees
who ever worked at the plant in the years 1948–2002 (Leonard
et al., 2008). These workers showed a twofold increase in diabetes
mortality in comparison to other non-exposed regional DuPont
workers, based on 22 diabetes deaths (SMR: 2.0, 1.2–3.0). A follow-
up study through 2008 of the cohort found the same twofold
increase in diabetes mortality when compared to other regional
workers (SMR: 1.9,1.4–2.6, 38 diabetes deaths, Steenland and
Woskie, 2012). However, these authors found no exposure-
response trend by quartile of estimated cumulative PFOA exposure
(SMRs: 1.8, 1.5, 2.3, 1.9 by increasing exposure quartile).
Similarly, a mortality analysis of workers at a 3M Company
facility using PFOA in Minnesota also suggested an elevated
relative risk of diabetes death among employees. A cohort of
3993 workers were classified by job into categories of never,
probably, and definitely exposed (Lundin et al., 2009). The ‘prob-
ably exposed’ workers demonstrated a twofold excess of diabetes
mortality when compared to the general Minnesota population
(SMR 2.0, 1.2–3.2, 18 diabetes deaths); however, those classified as
‘definitely exposed’ did not show a similar increase in mortality
risk. In comparison to ‘never exposed’ workers, ‘probably exposed’
and ‘definitely exposed’ workers had an increased relative risk of
diabetes death of 3.7 (1.4–10.1, 18 deaths versus 5 deaths).
Despite these findings of increased mortality risk in occupational
settings, there is little evidence of a relationship between PFOA and
diabetes among non-workers in the general population. Community
residents in the mid-Ohio valley near the DuPont plant have docu-
mented high PFOA serum concentrations due to drinking water
contaminated with PFOA (Steenland et al., 2009a,b). MacNeil et al.
(2009) conducted a cross-sectional study of serum PFOA concentra-
tions and type II diabetes in this population, which largely overlaps the
population studied here. The MacNeil et al. study was based on data
collected in 2005–2006 on 69,030 subjects in a survey (the C8 Health
Project), which included a comprehensive medical history and mea-
sured serum PFOA and blood glucose levels (Frisbee et al., 2009). The
median level of serum PFOA in this population in 2005–2006 was
28 ng/mL, compared with 4 ng/mL in the general U.S. population.
Prevalence of type II diabetes in this cohort was 8%, similar to that of
the populations of Ohio and West Virginia. In a cross-sectional
regression analysis, there was no positive trend for increased risk of
diabetes with increasing decile of measured PFOA serum concentra-
tion. In a secondary analysis, there was no association between
measured PFOA concentrations and blood glucose levels among
non-diabetics.
79
Among the general U.S. population, analyses of subjects from
NHANES did not find a relationship between PFOA levels and type
II diabetes (Melzer et al., 2010). However, prior literature has
suggested that PFOA may be associated with risk factors related to
type II diabetes, including glucose homeostasis and metabolic
disorder (Lin et al., 2008). It has also been hypothesized that
perfluouroalkyl acids may act as endocrine disrupters (White et al.,
2011), which in turn may affect body weight (Newbold et al.,
2008). While initial studies of subjects from NHANES found no
association between PFOA and body weight (Nelson et al., 2010),
early life exposure to PFOA has been associated with body weight
in adulthood in several studies. One toxicological study found that
low-dose in utero exposure to PFOA was associated with higher
body weight and increased insulin and leptin levels in female
offspring mice, while age-matched mice exposed to PFOA only
during adulthood did not have an increase in body weight (Hines
et al., 2009). In a prospective study of pregnant women,
Halldorsson et al. (2012) found a similar effect, with a positive
association between in utero exposure to PFOA and overweight at
20 years in female offspring.
The current study is based on further interviews in 2008–2011
of adult members of the original C8 Health Project, as well as
workers at the DuPont plant who did not participate in the C8
Health Project. The study features a large cohort with a wide range
of PFOA exposure levels, and uses newly developed estimates of
serum levels over time. Cases of type II diabetes were validated by
medical record review. The current longitudinal analysis examines
the association between estimated serum PFOA levels over time
and the incidence of type II diabetes.
2. Methods
2.1. Study participants
As a result of a class-action lawsuit in the mid-Ohio valley, which was settled in
2004, a survey of 69,030 mid-Ohio valley residents was conducted (the C8 Health
Project), in which demographic data and medical history was collected, and serum
PFOA concentration was measured (Frisbee et al., 2009). Subjects for the present
analyses were recruited from the adult population participating in the C8 Health
Project; approximately two-thirds of this population consented to further follow-
up studies conducted by the C8 Science Panel. The C8 Science Panel is a three-
person panel of epidemiologists created under the legal settlement of the class-
action lawsuit to determine if there was a probable link between PFOA and any
human disease (see www.c8sciencepanel.org). Fig. 1 illustrates how the cohort for
the present analysis was constructed. Details regarding cohort recruitment, study
design, and methods have been previously described and are briefly summarized
here (Winquist et al., 2013). From the original C8 Health Project participants, there
were 40,145 community members who were at least 20 years old and who
consented to contact by the C8 Science Panel for further studies. Additional study
subjects were also recruited from a cohort of 6026 DuPont employees who worked
at the Washington Works plant between the years 1948 and 2002 and who had
been included in a previous retrospective mortality study (Leonard et al., 2008).
People in these two groups were asked to complete surveys during August 2008–
May 2011. Community residents and DuPont employees who responded to the
2008–2011 surveys were combined to create a total population of 32,254 partici-
pants, of whom 12% had worked at the DuPont plant. We excluded people born
before 1920 because of concerns about the accuracy of case ascertainment in this
group. We also excluded people who self-reported type I diabetes and people who
self-reported type II diabetes but for whom we were unable to confirm the
diagnosis through medical record review as described below. After these exclusions
and exclusion of people missing data on covariates or serum estimates, 30,454
subjects remained in the final analysis.
2.2. Serum PFOA concentration estimates
Annual retrospective PFOA serum concentrations were estimated for each
participant beginning in 1951, or participant’s birth year, through 2011. Estimates
were constructed through a multi-stage modeling procedure and were based on
historic emission rates of PFOA from the DuPont facility and geologic and
meteorological data as well as estimated background PFOA levels as present in
the general US population. Exposure estimates also took into account individual
80 C. Karnes et al. / Environmental Research 128 (2014) 78–83
Fig. 1. Participant recruitment.
participants’ residential histories, drinking water sources and consumption rates,
demographic information, and self-reported body weight. Details of the estimation
procedure have been previously described (Shin et al., 2011a,b). For participants
who reported working at the DuPont plant (12% of the cohort), estimates also
incorporated assessments of occupational exposure (Woskie et al., 2012). The
estimated PFOA serum concentrations, excluding estimates for workers, were
evaluated for accuracy by comparison with measured serum PFOA concentrations
obtained in 2005–2006 during the original C8 Health Project. The Spearman’s rank
correlation coefficient for predicted values versus observed serum concentration in
2005–2006, incorporating community and occupational estimates, was 0.71, based
on 30,303 participants (Winquist et al., 2013).
2.3. Outcome
The 2008–2011 surveys asked about demographic characteristics and medical
history. Participants were asked if they had ever been told by a doctor or other
health professional that they had diabetes, and the age at which the disease was
diagnosed. Participants who reported diabetes were asked whether it was type I or
type II diabetes. People reporting a diagnosis of diabetes were asked for consent for
review of their medical records for confirmation of the diagnosis. Additional
information about diabetes diagnoses and medical record validation was obtained
from the original C8 Health Project.
Cases of ‘validated type II diabetes’ in our analysis include cases of diabetes of
any type that were self-reported on the 2008–2011 surveys for which a specific
diagnosis of type II diabetes was confirmed by medical record review. If no medical
record could be obtained or the medical record was insufficient to determine the
presence or absence of diabetes, medical record validation from the C8 Health
Project was also used to confirm cases of type II diabetes self-reported in 2008–
2011. This led to the inclusion of a number of cases (n¼ 512) that were self-reported
in 2008–2011 and would not have otherwise been included.
Analyses were restricted to validated cases. We excluded from the analysis any
people with confirmed cases of type I diabetes and people for whom no specific
type of diabetes was self-reported nor confirmed through medical record review.
People with self-reported Type II diabetes that was not confirmed via medical
record review were also excluded from the analysis. We also excluded people who
self-reported diabetes during the C8 Health Project but did not self-report diabetes
in 2008–2011.
2.4. Statistical analysis
Proportional hazard regression models were used to examine the relationship
between validated incident type II diabetes and time-varying estimated serum
PFOA concentration. Time at risk in these retrospective models began at age 20 and
ended at time of diabetes diagnosis, death, or last survey response date. Models
included demographic or socioeconomic predictors and medical history variables
that were suspected a priori to be related to diabetes. Models were stratified by
birth year, and controlled for gender, race (white versus non-white), years of
schooling, time-varying smoking and alcohol use, BMI, and use of cholesterol- and
blood-pressure-lowering medications. Survival analyses were performed using
PROC PHREG in SAS 9.3 employing the start-stop method with age as the time
variable and the EXACT method for handling ties. All confidence intervals
presented are at the 95% level.
Prospective models considered person-time since the C8 Health Project in
2005–2006; these models were run to assess the relationship between estimated
PFOA concentrations at the time of the C8 Health project and any subsequent
incidences of validated type II diabetes. The prospective analyses excluded
participants with a diagnosis of diabetes before the time of the C8 Health project.
Various serum exposure metrics were considered. To examine the effects of
lifetime exposure, we generated cumulative exposure estimates as the sum of all
yearly serum concentration estimates for each participant from birth through a
given year. Analyses used deciles of cumulative serum levels as our primary
exposure metric. Deciles were calculated using the distribution of PFOA cumulative
serum estimates of validated type II diabetes cases at the time of their diabetes
diagnosis. Analyses using deciles of cumulative exposure enabled us to assess the
shape of the dose-response curve without making prior assumptions about the
parametric form. We considered two tests for trend using both the untransformed
 C. Karnes et al. / Environmental Research 128 (2014) 78–83 81

and natural log-transformed PFOA concentration estimates, and we assessed model Table 1
fit using Akaike’s information criterion (AIC). Characteristics of participants in overall cohort and participants with validated type
We conducted several kinds of sensitivity analysis. We considered cumulative II diabetes.
serum levels with a lag of 10 years to explore possible delayed effects of past
exposure. We also conducted separate analyses restricted to those who worked at Characteristic Total N (%) Validated type II diabetes
the DuPont plant to determine whether there was a different effect of exposure for
workers. In another sensitivity analysis (‘qualifying time analysis’), the time of Participants (n) 32,254 4129
follow-up started when the participant was first qualified for the C8 Health project
(i.e., the year in which they had at least one year of residence in a PFOA- Sex
contaminated water district or first worked at the plant). Any person-time before Male 14,894 (46.2) 2066 (50.0)
the participant qualified for the C8 Health project was excluded from the analysis. Female 17,360 (53.8) 2063 (50.0)
In another sensitivity analysis (‘above background’ analysis), we used all person- Age at final interview
time (as in our main analysis), but estimated PFOA serum concentrations over time 20–29 2616 (8.1) 15 (0.4)
without the addition of background exposure, so that exposure began at year of 30–39 4964 (15.4) 132 (3.2)
first exposure to contaminated water or work in the DuPont plant. 40–49 6400 (19.8) 437 (10.6)
In another sensitivity analysis, we restricted the cohort to those who partici- 50–59 7338 (22.8) 1004 (24.3)
pated in the C8 Health Project (94%), and included family history of diabetes 60–69 6096 (18.9) 1352 (32.7)
(collected in the C8 Health project but not the 2008–2011 surveys) as a covariate in 70–79 3483 (10.8) 915 (22.2)
the model (results not shown). While this variable was a strong predictor of 80–89 1166 (3.6) 243 (5.9)
diabetes, it did not act as a confounder and results for exposure-response were very 90–99 179 (0.55) 29 (0.7)
similar to those presented here. 100–110 12 (0.04) 2 (0.05)
Finally, as a secondary analysis, we examined the relationship between serum
PFOA concentration estimates and fasting serum glucose which was measured BMI category
during the C8 Health Project. This analysis was restricted to participants in the C8 Underweight ( o 18.5) 499 (1.6) 23 (0.6)
Health Project (94% of the cohort). A linear regression analysis was conducted using Normal (18.5–o 25) 8665 (26.9) 438 (10.6)
estimated cumulative serum PFOA concentration at the time of a person’s Overweight (25–o 30) 11,307 (35.1) 1125 (27.3)
participation in the C8 Health Project in relation to the outcome of fasting serum Obese ( Z 30) 11,751 (36.4) 2540 (61.6)
glucose measured at that same time. This analysis controlled for gender, race, age Race/ethnicitya
group, BMI, and family history of diabetes. Subjects who reported eating in the 6 h White 31,185 (97.6) 3992 (96.7)
before the blood draw or who self-reported any type of diabetes were excluded Non-white 774 (2.4) 137 (3.3)
from the analysis.
Educationb
Less than highschool 3063 (9.5) 609 (14.8)
Highschool or GED 12,971 (40.2) 1851 (44.8)
3. Results Some college 10,522 (32.6) 1244 (30.1)
Bachelor or higher 5694 (17.7) 425 (10.3)
3.1. Descriptive statistics Smoking at 1st interview
Smoked and did not quit 6923 (22.8) 660 (16.5)
There were 5141 cases of self-reported diabetes. Of these, 344 Smoked and quit 8478 (27.9) 1536 (38.4)
Never smoked 15,030 (49.4) 1809 (45.2)
participants self-reported type I diabetes, and 4434 self-reported
type II diabetes, while 363 reported diabetes without specifying a Cholesterol-lowering medications
Yes 9909 (30.7) 2682 (65.0)
type. There were 4129 validated cases of type II diabetes con-
No 22,345 (69.3) 1447 (35.0)
firmed through medical record review as described above (86% of
those reporting type II diabetes or diabetes without type speci- Blood-pressure-lowering medications
Yes 12,325 (38.21) 3060 (74.1)
fied). Table 1 presents summary statistics for the 32,254 partici- No 19,929 (61.8) 1069 (25.9)
pants and for those with validated type II diabetes.
Family history of type II diabetesc
Yes 9701 (31.9) 2250 (56.2)
3.2. Survival analyses No 20,730 (68.1) 1755 (43.8)

Table 2 presents the results of the retrospective survival
analysis of validated type II diabetes in relation to cumulative
estimated serum PFOA concentration for exposure deciles. The
results do not show an association between estimated PFOA
exposure and type II diabetes. The hazard ratios (HRs) and 95%
CIs comparing higher deciles to the lowest decile (referent group)
were: 0.91 (95% CI: 0.76–1.08), 1.18 (95% CI: 0.99–1.40), 0.96 (95%
CI: 0.81–1.15), 1.04 (95% CI: 0.87–1.24), 1.11 (95% CI: 0.93–1.32),
1.06 (95% CI: 0.89–1.26), 1.00 (95% CI: 0.85–1.19), 1.03 (95% CI:
0.86–1.23), 1.01 (95% CI: 0.84–1.20), AIC ¼34,405.97. Trend tests
did not show an increase in hazard with increasing log cumulative
serum PFOA serum concentration (HR ¼1.00 per log unit of PFOA
exposure (μg/mL), p¼ 0.84, AIC ¼ 34,405.75) or cumulative serum
PFOA serum concentration (HR¼1.00 per unit of PFOA exposure
(μg/mL), p¼ 0.60, AIC ¼34,405.50).
Table 3 presents the results of the prospective analysis of
confirmed new cases occurring after the C8 Health Project in
2005–2006 for the overall cohort. There was no trend of increased
risk of type II diabetes with increased cumulative PFOA exposure.
Analyses using exposures estimates lagged by 10 years and
sensitivity analyses beginning at year of first qualifying for the C8
Health Project or without background exposure also failed to find an
association between PFOA exposure and diabetes incidence (results
a
295 missing race/ethnicity information.
b
4 missing education information.
c
Available for participants of the C8 Health Project survey (n¼ 30,431).
not shown). Models using self-reported cases found similar results
(results not shown). Analyses restricted to workers only had some
elevated HRs in upper quintiles (HRs and 95% CIs relative to lowest
quintile of PFOA exposure: 1.14 (95% CI: 0.53–2.47), 0.864 (0.41–1.83),
1.19 (0.56–2.54), 1.30 (0.61–2.80), AIC¼2940.42, 423 cases)), but
neither trend tests for log cumulative serum level or untransformed
cumulative serum level were close to significance (p¼ 0.46, AIC¼
2941.50 and p¼0.96, AIC¼ 2942.04).
The median fasting serum glucose level for the cohort, excluding
diabetics, was 94.0 ng/mL (mean 95.5, standard deviation 14.1, range
39–424 ng/mL). The results of a linear regression with the outcome of
fasting glucose in relation to deciles of cumulative estimated serum
PFOA concentration are presented in Table 4. There was no positive
association between fasting serum glucose and estimated cumulative
PFOA exposure (parameter estimates compared to the lowest decile as
the referent group: 0.0013 (p¼0.80),  0.0022 (p¼0.67),  0.0014
(p¼0.79),  0.0033 (p¼ 0.53), 0.0033 (p¼0.53),  0.0046 (p¼0.39),
 0.0054 (p¼0.30),  0.0086 (p¼ 0.10),  0.0053 (p¼0.31), R2 ¼0.12).
82 C. Karnes et al. / Environmental Research 128 (2014) 78–83

Table 2 an association between PFOA and type II diabetes. In addition,

Results of retrospective survival analysisa of type II diabetes in relation to there was no clear trend observed between estimated PFOA
cumulative serum PFOA concentrations with no lag (N ¼ 30,424; 3756 cases of
exposure and fasting serum glucose levels.
type II diabetes).
The current findings are in contrast to other studies that have
Metric HR (95% CI) p-Value AIC shown increased mortality risk for type II diabetes among workers
exposed to PFOA (Leonard et al., 2008; Lundin et al., 2009).
Decile 1b ( o 0.105) 1.00 34,405.97 However, it is important to note that mortality is not the best
Decile 2 (0.105–0.146) 0.91 (0.76, 1.08) 0.29
Decile 3 (0.147–0.177) 1.18 (0.99, 1.40) 0.07
way to studies diabetes, which is typically a non-fatal disease.
Decile 4 (0.178–0.242) 0.96 (0.81, 1.15) 0.66 Mortality data depends on the effectiveness of treatment and
Decile 5 (0.243–0.363) 1.04 (0.87, 1.24) 0.69 survival patterns, and may not be an accurate reflection of the
Decile 6 (0.364–0.619) 1.11 (0.93, 1.32) 0.26 underlying incidence of disease. Further, type II diabetes is
Decile 7 (0.620–1.406) 1.06 (0.89, 1.26) 0.52
typically underreported as a cause of death, and death certificates
Decile 8 (1.407–3.79) 1.00 (0.85, 1.19) 0.96
Decile 9 (3.80–7.73) 1.03 (0.86, 1.23) 0.75 underestimate the prevalence of diabetes among the deceased
Decile 10 (4 7.73) 1.01 (0.84, 1.20) 0.95 (Will et al., 2001, McEwen et al., 2006). It is estimated that
diabetes incidence rates are 30–40 times higher than mortality
Continuous (linear) 1.00 (0.99, 1.00) 0.60 34,405.50
Continuous (log-linear) 1.00 (0.98, 1.02) 0.84 34,405.75 rates (Geiss et al., 2006, Gregg et al., 2012). In addition, while
earlier studies demonstrated evidence of increased mortality risks
a
All models are stratified by birth year and are adjusted for sex, non-white among workers, such findings were based on analyses of exposed
race, education, current and former smoking and alcohol consumption (time versus non-exposed groups (Leonard et al., 2008, Lundin et al.,
varying), BMI, and use of cholesterol- and blood-pressure-lowering medications.
2009). The most recent mortality study of exposed workers used
Results are for exposure estimates with addition of background exposure, lag 0,
starting in 1952 or at qualifying year or at age 20 years (whichever is latest), estimates of cumulative exposure to PFOA generated for each
excluding people born before 1920. participant based on occupational history and found no trend of
b
Deciles (mg/mL  year) were defined by the estimated cumulative PFOA serum increased risk of diabetes death with increasing exposure
concentration among cases of type II diabetes at the time of diagnosis.
(Steenland and Woskie, 2012).
This study is in agreement with other previous research that
Table 3 failed to find an association between PFOA and type II diabetes in
Results of prospective survival analysis of type II diabetes in relation to cumulative cross-sectional analyses of this same mid-Ohio valley population
serum PFOA concentrations with no lag (N ¼27,921; 814 cases of type II diabetes). (McNeil et al., 2009) or in the general US population (MacNeil
et al., 2009).
Metric HR (95% CI) p-Value AIC
This study has several important strengths. While most prior
Decile 1 ( o0.172) 1.00 6818.80 studies of PFOA health effects are based on small samples with
Decile 2 (0.173–0.215) 0.96 (0.70, 1.32) 0.80 very little contrast in exposure, the current study examines a large
Decile 3 (0.216–0.286) 0.79 (0.57, 1.08) 0.13 cohort with a wide range of PFOA exposure levels. In addition, we
Decile 4 (0.287–0.406) 1.02 (0.74, 1.41) 0.90
use estimates of serum PFOA concentrations through participants’
Decile 5 (0.407–0.608) 1.16 (0.84, 1.59) 0.36
Decile 6 (0.609–1.15) 0.92 (0.67, 1.27) 0.63 lifetimes generated from a sophisticated fate-transport model that
Decile 7 (1.16–2.79) 0.95 (0.69, 1.31) 0.76 was validated against observed PFOA serum levels in 2005–2006.
Decile 8 (2.80–5.65) 1.10 (0.80, 1.52) 0.57 These data enable us to perform a longitudinal analysis, which
Decile 9 (5.66–9.61) 1.18 (0.86, 1.62) 0.32
allows for stronger causal inference than is possible with cross-
Decile 10 (4 9.61) 1.00 (0.72, 1.39) 0.99
sectional analyses. Our data also include several thousand cases of
Continuous (linear) 1.00 (1.00, 1.01) 0.31 6811.65 diabetes that have been validated through medical record review.
Continuous (log-linear) 1.03 (0.98, 1.08) 0.18 6810.73 This study also has several limitations. One limitation of this
study is the use of a survivor cohort. Community members must
have been alive in 2005–2006 in order to participate in the
Table 4 original C8 Health Project study. This could potentially bias our
Results of linear regression analysis of fasting glucose in relation to cumulative results towards the null if diabetes cases were less likely to have
estimated serum PFOA concentration among non-diabetics at the time of the C8 survived, and if these cases had a different exposure profile than
Health project (N ¼ 11,309).nn cases which were included. However, because diabetes is typically
Decilea Estimate p-Value
not a fatal disease this is not likely.
Another potential source of bias may arise from inaccuracies in
Decile 1 ( o155.4) 1.00 the estimated exposures to PFOA. Annual exposure levels were
Decile 2 (155.4–190.4) 0.0013 0.80 estimated for each participant based on self-reported residential
Decile 3 (190.5–238.5)  0.0022 0.67
history and drinking water intake, as well as DuPont emission
Decile 4 (238.6–310.3)  0.0014 0.79
Decile 5 (310.4–480.7)  0.0033 0.53 patterns and other environmental factors. While there is potential
Decile 6 (480.8–797.8)  0.0033 0.53 for misclassification in the estimates, we expect the estimates to
Decile 7 (797.9–1544.7)  0.0046 0.39 be reliable in the relative ranking of participants’ exposures.
Decile 8 (1544.8–4037.8)  0.0054 0.30
Another limitation of this study is that it does not examine the
Decile 9 (4037.9–8499.3)  0.0086 0.10
Decile 10 (8499.3)  0.0053 0.31
potential effects of early life exposure to PFOA on diabetes
observed in adulthood. PFOA has been detected in cord blood
nn
Deciles of estimated cumulative PFOA serum concentration (ng/mL). (Apelberg et al., 2007) and breastmilk (Von Ehrenstein et al.,
a
R2 ¼ 0.12. 2009). Further research is needed to understand the possible
effects of early life exposure to PFOA.
4. Discussion To summarize, while research suggests an increased mortality
risk for type II diabetes among workers exposed to PFOA, a
We did not find evidence of an association between estimated previous cross-sectional analysis failed to find an association
cumulative PFOA exposure and incidence of type II diabetes. between PFOA and diabetes. The current study examined a large
Further examination of the worker cohort also does not suggest cohort with validated cases of type II diabetes and estimates of
 C. Karnes et al. / Environmental Research 128 (2014) 78–83
PFOA exposure over time. We found no trend of increased risk of
type II diabetes with increasing exposure to PFOA.
Acknowledgments
This work was supported by the C8 Class Action Settlement
Agreement (Circuit Court of Wood County, West Virginia) between
DuPont and Plaintiffs, which resulted from releases of the chemi-
cal perfluorooctanoic acid (PFOA, or C8) into drinking water. Funds
are administered by an agency that reports to the court. Our work
and conclusions are independent of either party to the lawsuit.
References
Apelberg, B.J., Witter, F.R., Herbstman, J.B., Calafat, A.M., Halden, R.U., Needham, L.L.,
Godman, L.R., 2007. Cord serum concentrations of perfluorooctane sulfonate
(PFOS) and perfluorooctanoate (PFOA) in relation tho weight and size at birth.
Environ. Health Perspect. 115, 1670–1675.
Barry, V., Winquist, A., Steenland, K., Perfluorooctanoic acid (PFOA) exposures and
incident cancer among adults living near a chemical plant. Environ. Health
Perspect., http://dx.doi.org/10.1289/ehp.1306615, in press.
Bartell, S.M., Calafat, A.M., Lyu, C., Kato, K., Ryan, P.B., Steenland, K., 2010. Rate of
decline in serum PFOA concentrations after granular activated carbon filtration
at two public water systems in Ohio and West Virginia. Environ. Health
Perspect. 118, 222–228.
Brede, E., Wilhelm, M., Goen, T., Muller, J., Rauchfuss, K., et al., 2010. Two-year
follow-up biomonitoring pilot study of residents' and controls' PFC plasma
levels after PFOA reduction in public water system in Arnsberg, Germany. Int. J.
Hyg. Environ. Health 21, 217–223.
Frisbee, S.J., Brooks Jr., A.P., Maher, A., Flensborg, P., Arnold, S., Fletcher, T., et al.,
2009. The C8 Health Project: design methods, and participants. Environ. Health
Perspect. 117, 1873–1882.
Geiss, L.S., Pan, L., Cadwell, B., Gregg, E., Benjamin, S.M., Engelgau, M.M., 2006.
Changes in incidence of diabetes in U.S. adults, 1997–2003. Am. J. Prev. Med. 30,
371–377.
Gregg, E.W., Cheng, Y.J., Saydah, S., Cowie, C., Garfield, S., Geiss, L., et al., 2012.
Trends in death rates among U.S. adults with and without diabetes between
1997 and 2006: findings from the National Health Interview Survey. Diabetes
Care 35, 1252–1257.
Halldorsson, T.I., Rytter, D., Haug, L.S., Bech, B.H., Danielsen, I., Becher, G., et al.,
2012. Prenatal exposure to perfluorooctanoate and risk of overweight at 20
years of age: a prospective cohort study. Environ. Health Perspect. 129,
668–673.
Hines, E.P., White, S.S., Stanko, J.P., Gibbs-Flournoy, E.A., Lau, C., Fenton, S.E., 2009.
Phenotypic dichotomy following development exposure to perfluorooctanoic
acid (PFOA) in female CD-1 mice: low doses induce elevated serum leptin and
insulin, and overweight in mid-life. Mol. Cell. Endocrinol. 304, 97–105.
Kato, K., Lee-Yang, W., Jia, L.T., Kuklenyik, Z., Calafat, A.M., 2011. Trends in exposure
to polyfluoroalkylchemics in the U.S. population: 1999–2008. Environ. Sci.
Technol. 45, 8037–8045.
Lau, C., Anitole, K., Hodes, C., Lai, D., Pfahles-Hutchens, Seed, J., 2007. Perfluoroalkyl
acids: a review of monitoring and toxicological findings. Toxicol. Sci. 99,
366–394.
Leonard, R.C., Kreckmann, K.H., Sakr, C.J., Symons, J.M., 2008. Retrospective cohort
mortality study of workers in a polymer production plant including a reference
population of regional workers. Ann. Epidemiol. 18, 15–22.
Lin, C.Y., Chen, P.C., Lin, Y.C., Lin, L.Y., 2008. Association among serum perfluoroalkyl
chemicals, glucose homeostasis and metabolic syndrome in adolescents and
adults. Diabetes Care 32, 702–707.
Lin, C.Y., Lin, L.Y., Chaing, C.K., Wang, W.J., Su, Y.N., Hung, K.Y., et al., 2010.
Investigation of the associations between low-dose serum perfluorinated
chemicals and liver enzymes in US adults. Am. J. Gastroenterol. 105, 1354–1363.
83
Lindstrom, A.B., Strynar, M.J., Libelo, E.L., 2011. Polyfluorinated compounds: past,
present, and future. Environ. Sci. Technol. 45, 7954–7961.
Lundin, J., Alexander, B., Olsen, G., Church, T., 2009. Ammonium perfluorooctanoate
production and occupational mortality. Epidemiology 20, 921–928.
MacNeil, J., Steenland, N.K., Shankar, A., Ducatman, A., 2009. A cross-sectional
analysis of type II diabetes in a community with exposure to perfluorooctanoic
acid (PFOA). Environ. Res. 109, 997–1003.
Melzer, D., Rice, N., Depledge, M.H., Henley, W.E., Galloway, T.S., 2010. Association
between serum perfluorooctanoic acid (PFOA) and thyroid disease in the US
National Health and Nutrition Examination Survey. Environ. Health Perspect.
118, 686–692.
McEwen, L.N., Kim, C., Haan, M., Ghosh, D., Lantz, P.M., Mangione, C.M., et al., 2006.
Diabetes reporting as a cause of death—results from the Translating Research
into Action for Diabetes (TRIAD) study. Diabetes Care 29, 247–253.
Nelson, J.W., Hatch, E.E., Webster, T.F., 2010. Exposure to polyfluoroalkyl chemicals
and cholesterol, body weight, and insulin resistance in the general U.S.
population. Environ. Health Perspect. 188, 197–203.
Newbold, R.R., 2010. Impact of environmental endocrine disrupting chemicals on
the development of obesity. Hormones 9, 206–217.
Newbold, R.R., Padilla-Banks, E., Jefferson, W.N., Heindel, J.J., 2008. Effects of
endocrine disruptors on obesity. Int. J. Androl. 31, 201–208.
Olsen, G.W., Burris, J.M., Ehresman, D.J., Froehlich, J.W., Seacat, A.M., Butenhoff, J.L.,
et al., 2007. Half-life of serum elimination of perfluorooctanesulfonate,per-
fluorohexanesulfonate, and perfluorooctanoate in retired fluorochemical pro-
duction workers. Environ. Health Perspect. 115, 1298–1305.
Post, G.B., Cohn, P.D., Cooper, K.R., 2012. Perfluorooctanoic acid (PFOA), an emerging
drinking water contaminant: a critical review of recent literature. Environ. Res.
116, 93–117.
Shin, H.M., Vieira, V.M., Ryan, P.B., Detwiler, R., Sanders, B., Steenland, K., Bartell, S.
M., 2011a. Environmental fate and transport modeling for perfluorooctanoic
acid emitted from the Washington Works facility in West Virginia. Environ. Sci.
Technol. 45, 1435–1442.
Shin, H.M., Vieira, V.M., Ryan, P.B., Steenland, K., Bartell, S.M., 2011b. Retrospective
exposure estimation and predicted versus observed serum perfluorooctanoic
acid concentrations for participants in the C8 health project. Environ. Health
Perspect. 119, 1760–1765.
Steenland, K., Jin, C., MacNeil, J., Lally, C., Ducatman, A., Viera, V., et al., 2009a.
Predictors of PFOA levels in a community surrounding a chemical plant.
Environ. Health Perspect. 117, 1083–1088.
Steenland, K., Tinker, S., Frisbee, S., Ducatman, A., Vaccarino, V., 2009b. Association
of perfluorooctanoic acid and perfluorooctane sulfonate with serum lipids
among adults with elevated community exposure to PFOA. Am. J. Epidemiol.
(170), 1268–1278.
Steenland, K., Tinker, S., Shankar, A., Ducatman, A., 2010. Association of perfluor-
ooctanoic acid (PFOA) and perfluorooctanesulfonate (PFOS) with uric acid
among adults with elevated community exposure to PFOA. Environ. Health
Perspect. 118, 229–233.
Steenland, K., Woskie, S., 2012. Cohort mortality study of workers exposed to
perfluorooctanoic acid. Am. J. Epidemiol. 176, 909–917.
Taylor, K.W., Novak, R.F., Anderson, H.A., Birnbaum, L.S., Blystone, C., et al., 2013.
Evaluation of the association between persistent organic pollutants (POPs) and
diabetes in epidemiological studies: a national toxicology program workshop
review. Environ. Health Perspect. 121, 774–783.
Von Ehrenstein, O.S., Fenton, S.E., Kato, K., Kuklenyik, Z., Calafat, A.M., Hines, E.P.,
2009. Polyfluoroalkyl chemicals in the serum and milk of breastfeeding
women. Reprod. Toxicol. 27, 239–245.
White, S.S., Fenton, S.E., Hines, E.P., 2011. Endocrine disrupting properties of
perfluorooctanoic acid. J. Steriod Biochem. Mol. Biol. 127, 16–26.
Will, J.C., Vinicor, F., Stevenson, J., 2001. Recording of diabetes on death certificates:
has it improved? J. Clin. Epidemiol. 54, 239–244.
Winquist, A., Lally, C., Shin, H.M., Steenland, K., 2013. A study of the health effects of
PFOA among a cohort of highly exposed community residents and workers:
design, methods, and participants. Environ. Health Perspect. 121, 893–899.
Woskie, S.R., Gore, R., Steenland, K., 2012. Retrospective exposure assessment of
perfluorooctanoic acid serum concentrations at a fluoropolymer manufacturing
plant. Ann. Occup. Hyg. 56, 1025–1037.