ARTICLE pubs.acs.org/IECR
Determinants of Plasma PFOA and PFOS Levels Among 652 Danish Men Kirsten T. Eriksen,*,† Mette Sørensen,† Joseph K. McLaughlin,‡ Anne Tjønneland,† Kim Overvad,§ and Ole Raaschou-Nielsen† †
Institute of Cancer Epidemiology, Danish Cancer Society, Copenhagen, Denmark International Epidemiology Institute, Rockville, Maryland, United States § Department of Epidemiology, School of Public Health, Aarhus University, Denmark ‡
ABSTRACT: Perfluorooctanoate (PFOA) and perfluorooctane sulfonate (PFOS) are used in a variety of industrial and consumer products and have been detected worldwide in human blood. The sources for human exposure are not well described, but dietary intake is suggested as an important source. In this study of 652 Danish men from the Diet, Cancer and Health cohort, we examined intake of 10 major dietary groups, tap water drinks, alcohol consumption, cooking method, geographical area, age, smoking status, and BMI as potential determinants of PFOA and PFOS plasma levels. Living in the Aarhus area was associated with higher PFOA and PFOS plasma levels compared with living in the Copenhagen area, and never smokers had higher levels than current smokers. Frying as compared with other cooking methods was a determinant of PFOA and PFOS levels. BMI and alcohol consumption were inversely associated with both compounds. Among the dietary groups, only intake of eggs was significantly positively associated with PFOS plasma levels. In future studies, PFOA and PFOS levels in air, dust and water samples should be measured to elucidate further the sources of exposure; exposure through diet needs to be studied in greater detail. Our finding of a higher body burden of PFOA and PFOS among never smokers also warrants further evaluation.
’ INTRODUCTION Perfluorooctanoate (PFOA) and perfluorooctane sulfonate (PFOS) are synthetically produced perfluorinated chemicals (PFCs) known for their antiwetting and surfactant properties. They are used in a variety of industrial and consumer products, such as nonstick cookware, waterproof, breathable textiles, fire fighting foams, and protective coatings for paper, food packing materials, and carpets 1-3. They have no known natural sources 4. PFOA and PFOS are resistant to metabolic and environmental degradation, are bioaccumulative, and have been found in wildlife 5 and in human blood and tissue samples from occupationally exposed workers and in the general population worldwide 6-8. Generally, people in the Western world have higher levels than people in developing countries 8. The entry routes of perfluorinated chemicals are by inhalation, oral intake, and dermal absorption. Elimination half-lives in humans have been estimated at 4 years for PFOA and 5 years for PFOS 9. PFOA and PFOS are suspected of hormonal disruption and carcinogenic activity 10,11. Animal studies have shown that PFOA and PFOS exposures at high levels are associated with delayed physical development, endocrine disruption, neonatal mortality, and liver, pancreatic, and testicular tumors in animals 12-15. In a recent epidemiological investigation, we observed no associations between PFOA and PFOS at plasma levels seen in the general population and risk of prostate, bladder, pancreatic, or liver cancers 16. Other human studies have reported associations of PFOA or PFOS with reduced fecundity, low semen quality, and reduced birth weight 17-19. Although these chemicals are ubiquitous in the environment 15, and the potential health effects are of public concern, the major sources and pathways of exposure for the general population still need to be clarified 20. r 2010 American Chemical Society
A review of studies on PFC exposure indicates that most of human PFOA and PFOS exposure may be attributed to the oral route, mainly diet 21. These compounds have been detected in foods such as red meat, fish, eggs, and potatoes 21-28. Popcorn and fast food may be indirect dietary sources due to contamination from leaching of paper coatings 25. Foods prepared on polytetrafluoroethylene (PTFE)-coated cookware are also potential exposure sources, though it is still uncertain whether transport of PFCs during cooking is important 29. Inhalation may also be an important exposure route, as considerable amounts of PFOA and PFOS have been found in indoor dust samples 30. Further, tap water and drinks made from tap water are also suggested as possible sources, particularly for people living close to contaminated areas 4. The aim of the present study was to identify dietary and other determinants of PFOA and PFOS plasma levels among 652 Danish men from the prospective Diet, Cancer, and Health (DCH) cohort study 31 to further refine identification of PFC exposure sources in humans.
’ MATERIALS AND METHODS Study Population. Between December 1993 and May 1997, 57 053 individuals (27 178 men and 29 875 women) 50-65 years of age, born in Denmark and with no previous cancer Special Issue: Perfluoroalkyl Acid Received: February 25, 2010 Accepted: September 24, 2010 Revised: August 19, 2010 Published: October 12, 2010 8137
dx.doi.org/10.1021/es100626h | Environ. Sci. Technol. 2011, 45, 8137–8143
Environmental Science & Technology
ARTICLE
Table 1. Baseline Characteristics and PFOA and PFOS Plasma Levels of the Study Population (N = 652) at Enrolment characteristics
distribution
geographical area, N (%) Copenhagen
177 (27)
suburban Copenhagen
282 (43)
Aarhus area
193 (30)
age (years), median (5th-95th percentile)
55 (50-64)
smoking status, N (%) current
255 (39)
former never
230 (36) 163 (25)
BMI (kg/m2), median (5th-95th percentile)
26.4 (22.0-33.2)
alcohol intake (g/day), median (5th-95th percentile)a
19.6 (1.8-80.6)
selected dietary groups, median (5th-95th percentile) fish (g/day)b
42.8 (13.0-103.3)
red meat (g/day)c
134.9 (65.5-249.5)
poultry (g/day)d
19.6 (4.6-64.3)
all fruits incl. juices (g/day) all vegetables incl. juices (g/day)
144.0 (25.0-482.1) 148.9 (47.7-359.7)
cereals (g/day)e
212.3 (101.6-354.0)
dairy products (g/day)f
293.6 (52.1-1068.0)
eggs (g/day)
23.6 (7.4-67.7)
potatoes (g/day)
141.5 (52.3-325.3)
snacks (g/day)g
6.5 (0-28.7)
tap water (L/day)h
1.5 (0.5-2.8)
PFOA plasma level (ng/mL), median (5th-95th percentile) PFOS plasma level (ng/mL), median (5th-95th percentile)
6.8 (3.1-13.3) 34.9 (16.4-62.1)
Excluding nondrinkers in the calculation (N = 12). b Sum of fresh and processed fish. c Beef, veal, pork and lamb (including processed meat and offal). Chicken and turkey. e Sum of whole-grain and refined-grain cereals. f Milk, cheese, cream, yogurt, ice cream, and other cultured milk products. g Peanuts, bacon rinds, French fries and potato chips. h Sum of intake of tap water, coffee, tea and fruit syrup diluted with tap water. a
d
diagnosis, were enrolled in the prospective DCH cohort 31. Cohort members were from 16 municipalities in the Copenhagen area and three municipalities in the Aarhus area. Blood samples were drawn from each participant at enrolment and stored at -150 °C. Relevant Danish ethical committees approved the study, and written informed consent was obtained from all participants. In a previous study 16, PFOA and PFOS plasma levels were measured for 1240 individuals from the DCH cohort diagnosed with either prostate, bladder, pancreatic, or liver cancer and for a randomly selected comparison group of 772 individuals in the same cohort. The present analysis was based on the men included in the comparison group of the described study; we only included men since they made up almost 90% of the subcohort group and since a gender difference in PFC metabolism is proposed. Among the 680 men we excluded 28 who were also assigned as a case (to minimize the possibility that a subclinical cancer could affect the PFOA and PFOS plasma levels), resulting in a study group of 652 men. Measurement of PFOA and PFOS Plasma Levels. Concentrations of PFOS and PFOA in plasma were measured using high performance liquid chromatography/tandem mass spectrometry at the 3M Toxicology Laboratory, using the methods described by Ehresman et al. 32. The instrument used for analysis was API 4000 mass spectrometer (Applied Biosystems/MDS-Sciex Instrument Corporation, Forest City, CA). The type of HPLC column used was Mac-Mod Analytical (Mac-Mod, Chads Ford, PA) ACEs C-18, 5 μm, 100 2.1 mm i.d. Stable labeled analogs of
PFOS (18O2 PFOS) and PFOA (13C2 PFOA) were used in all extracted procedures. For all samples, solid phase extraction was performed using Waters (Milford, MA) Oasiss hydrophiliclipophilic balance 3.0 mL cartridges. The plasma concentrations were evaluated using standard curves based on extracted spiked calf serum. The use of calf serum for the standard curve matrix has been shown to be appropriate for quantification of plasma samples. The lower limit of quantification (L) for the method was 1.0 ng/mL for both PFOS and PFOA. In our analysis, PFOS was measured above L in all samples and two PFOA values were below the L and√were assigned the value 0.71 ng/mL, according to the formula L/ 2 for replacement of nondetectable values 33. As a quality control measure, 50 blind samples were selected for repeated analysis of PFOA and PFOS concentrations to ensure the accuracy and reliability of the data. The 3M Toxicology Laboratory was blinded to this series of parallel measurements. Mean coefficients of variation were low (5.9% for PFOA and 1.8% for PFOS). Diet and Lifestyle Questionnaires. Information on lifestyle factors, including dietary intake, alcohol consumption, and smoking habits, as well as sociodemographic characteristics, was obtained at enrolment into the DCH cohort using detailed self-administered, interviewer-checked, questionnaires. For each participant, mean daily intakes of specific food items and nutrients were calculated using the software program Food Calc Version 1.3 34. Validation of the food frequency questionnaire has been described previously 35-37. For this investigation, 17 questionnaire or questionnaire-derived items were determined as potential determinants of PFOA and 8138
dx.doi.org/10.1021/es100626h |Environ. Sci. Technol. 2011, 45, 8137–8143
Environmental Science & Technology
Figure 1. Correlation between PFOA and PFOS plasma levels (ng/mL) of the study population at enrolment.
PFOS plasma levels. Dietary variables were daily intake of red meat (beef, veal, pork and lamb (including offal and processed meat)); poultry (chicken and turkey); fish (total of fresh and processed fish); eggs; dairy products (milk, cheese, cream, yogurt, ice cream, and other cultured milk products); fruits (including fruit juices); vegetables (including vegetable juices); cereals (total of wholegrain and refined cereals); potatoes; snacks (peanuts, bacon rinds, French fries, and potato chips), tap water drinks (total of intake of tap water, coffee, tea, and fruit syrup diluted with tap water); and most frequent cooking method for meat and fish (frying versus nonfrying (boiling, oven use and microwave use)). Further, we examined geographical area (Aarhus area, suburban Copenhagen, Copenhagen), age, smoking status (current, former, never), BMI, and alcohol intake. Statistical Analyses. Potential determinants of the plasma PFOA and PFOS levels were analyzed by generalized linear models. The analyses were conducted using the PROC GLM procedure of SAS 9.1 (SAS Institute Inc., Cary, NC). Distributions of PFOA and PFOS plasma levels were only slightly skewed, and comparable results were observed for untransformed and log-transformed PFOA and PFOS levels. Regression estimates for the untransformed data are presented in this paper. Residual plots were randomly distributed. The following approaches were used: (i) univariate regression analyses, in which each potential determinant was included individually in the model; and (ii) multivariate regression analyses, in which all potential determinants were included and thereby mutually adjusted. The multivariate regression analyses were based on 646 subjects due to missing data from six subjects for two variables.
’ RESULTS PFOA and PFOS plasma levels, age, baseline characteristics and daily intake of major dietary groups are presented in Table 1. Concentrations of PFOA and PFOS were strongly correlated (Spearman’s correlation coefficient, rs = 0.71 and P < .0001) (Figure 1). Similar determinants were found for PFOA and PFOS plasma levels and multivariate regression analysis results are shown in Table 2. Similar results were found with univariate regression analyses (data not shown). Excluding outliers in the models did not change results (data not shown). Both the crude and adjusted analyses showed that living in the Aarhus area was associated with
ARTICLE
higher plasma levels of both PFOA and PFOS compared with living in Copenhagen. People living in suburban Copenhagen had intermediate concentrations. We also found that never smokers had significantly higher levels than current smokers, whereas former smokers had intermediate concentrations. BMI and alcohol intake were inversely associated with both compounds. Among the major dietary groups, intake of eggs was positively associated with PFOS plasma levels, and intake of dairy products and cereals was slightly inversely associated with PFOS levels (Table 2); neither was significantly associated with PFOA levels (Table 2). Positive associations between intake of potatoes and plasma levels of both PFCs were observed, though results were not significant. No significant associations were found for the other dietary groups in our study. Frying was associated with higher levels of both PFCs. The models including all potential determinants accounted for 14% and 24% of the variation in PFOA and PFOS plasma level, respectively (R2 values).
’ DISCUSSION Our results showed that geographical area and smoking status were determinants of PFOA and PFOS plasma levels. BMI and alcohol intake were inversely associated with both compounds. Intake of eggs was positively associated with PFOS but not PFOA plasma level, and a slight inverse association with PFOS level was found for intake of dairy products and cereals. Study participants reporting frying as the most frequent cooking method had higher plasma levels of both PFCs. Geographic differences in PFOA and PFOS plasma levels have been reported in nonoccupationally exposed persons worldwide 8,38 most likely reflecting variation in lifestyle, the occurrence of dietary sources with varying PFC levels, and differences in local environmental contamination. In our study, people living in the Aarhus area of Denmark had higher levels of plasma PFOA and PFOS than people living in Copenhagen, whereas those in suburban Copenhagen had intermediate levels. A recent study found widespread PFC contamination, mainly PFOA and PFOS, of the Danish environment levels similar to other countries with no PFC manufacturing 39. Since perfluorinated chemicals are not manufactured in Denmark and since all plasma samples in our study were analyzed at the same time in the same laboratory 16, our results suggests that there is a regional variation in exposure to these chemicals even in a small and relatively homogeneous country as Denmark. The regional variation of PFOA and PFOS plasma levels could be due to differences in local contamination of air and drinking water but we were not able to explore this further in the present study. The regional variation could also be due to underlying factors such as variation in lifestyle associated with exposure. Adjusting for socioeconomic status, measured as years of school attendance (less than 7 years, 8-10 years or more than 10 years), did not influence the findings (data not shown). Active smokers had lower plasma levels of PFOA and PFOS than never smokers. Similarly, a previous Danish cohort study of 1076 pregnant women reported daily maternal smoking during pregnancy to be associated with lower maternal PFOA and PFOS plasma levels compared with never smoking 25. These results could reflect differences between smokers and nonsmokers in lifestyle patterns associated with sources of PFOA and PFOS exposure or an enhanced elimination rate of these compounds in smokers, though to our knowledge this has not been examined. 8139
dx.doi.org/10.1021/es100626h |Environ. Sci. Technol. 2011, 45, 8137–8143
Environmental Science & Technology
ARTICLE
Table 2. Associations Between PFOA and PFOS Plasma Levels (ng/mL) and Study Variablesa PFOA study variables
N
b
difference (ng/mL)
PFOS
95% CI
p value
difference (ng/mL)
95% CI
p value
geographical area Copenhagen
177
ref
suburban Copenhagen
282
1.60
(1.02;2.19)
