Dietary Intake Is Associated with Phthalate Body Burden in a Nationally Representative Sample

Background Phthalates are compounds that are used in a wide range of consumer products. However, the contribution of dietary intake to phthalate exposure has not been well defined. Objective The objective of this study was to assess the contribution of different food types to phthalate exposure. Phthalates are chemicals of concern because of the high levels measured in people and the environment, as well as the demonstrated toxicity in animal studies and limited epidemiological studies. Previous research, although limited, has suggested that phthalates contaminate food in various countries. Methods We conducted an exploratory analysis of data collected as part of the 2003–2004 National Health and Nutrition Examination Survey (NHANES). Associations between dietary intake (assessed by a 24-hr dietary recall) for a range of food types (meat, poultry, fish, fruit, vegetable, and dairy) and phthalate metabolites measured in urine were analyzed using multiple linear regression modeling. Results We found that metabolites of di-(2-ethylhexyl) phthalate (DEHP) and high-molecular-weight phthalate metabolites were associated with the consumption of poultry. Monoethyl phthalate, the metabolite of diethyl phthalate (DEP), was associated with vegetable consumption, specifically tomato and potato consumption. Discussion These results, combined with results from previous studies, suggest that diet is an important route of intake for phthalates. Further research is needed to determine the sources of food contamination with these toxic chemicals and to describe the levels of contamination of U.S. food in a large, representative U.S. sample.


Research
Phthalates are a family of chemicals that have been approved for use in a wide range of con sumer products, including building materials, personal cosmetics, pharmaceuticals, nutritional supplements, solvents, adhesives, paints, lac quers, insecticides, cleaning materials, children's toys, and food packaging (Schettler 2006). The annual global use of phthalates is estimated to exceed 3 million metric tons per year (Schettler 2006). Phthalates have been used for > 50 years to increase the plasticity of rigid materials such as polyvinyl chloride (PVC).
Toxicological studies of phthalates and related compounds have established potential health effects of exposure, particularly endo crine disruption. In animal studies, researchers have found an association between phthalate exposure and the induction of testicular toxicity in rats and guinea pigs as well as male repro ductive tract lesions such as seminiferous epi thelial degeneration (Barlow et al. 2004;Gray et al. 1982). Phthalate exposure also has effects on the female reproductive system, includ ing decreased estradiol production, prolonged estrous cycles, and no ovulation in adult rats (Davis et al. 1994;LovekampSwan and Davis 2003). In a threegeneration phthalate exposure study, Pereira et al. (2007) found significant increases in liver enzymes [alanine aminotrans ferase (ALT) and aspartate aminotransferase (AST)] over the three generations and decreases in serum cholesterol, liver glutathione, and glutathione reductase, especially in the F2 generation. Effects on sexual health and devel opment have been observed in recent human studies. Phthalate exposure was associated with decreased anogenital distance in male infants, which was found to be significantly correlated with penile volume and cryptorchidism (Swan et al. 2005). A study using the National Health and Nutrition Examination Survey (NHANES) data from 1999-2002 found an association between urinary phthalate metabolites and increased waist circumference and insulin resistance (Stahlhut et al. 2007). Additionally, increased phthalate body burden, particularly of one metabolite, mono2ethylhexyl phtha late (MEHP), was associated with decreases in testosterone and estradiol in adult men (Meeker et al. 2009). Levels of phthalates in house dust have been associated with rhinitis, eczema, and asthma in humans (Bornehag et al. 2004).
Typically, phthalate body burden is esti mated through measurement of phthalate metabolites in the urine. The halflife of the parent phthalates is very short, and high lev els of contamination in laboratories present technical difficulties when analyzing for phtha lates (Lorber et al. 2010). Silva et al. (2004), assessed body burdens of phthalate metabo lites for a portion of the U.S. population in NHANES. In this study, they found that chil dren had significantly higher levels of certain phthalate metabolites, monobutyl phthalate, monobenzyl phthalate (MBzP), and MEHP than did adults; however, they had significantly lower levels of monoethyl phthalate (MEP), likely because of the use of MEP in consumer products such as cosmetics, which are used more by adults than by children. Another study compared the levels of phthalate metab olites in infants and their mothers and found that, on average, creatinineadjusted levels of phthalate metabolites were higher in infants than in their mothers (Sathyanarayana et al. 2008). Creatinine adjustment has been used to normalize excretion of phthalate metabo lites and many other contaminants (Jackson 1966). Given a daily rate of elimination of creatinine, one can estimate a full day's elimi nation of the phthalate metabolite. With other considerations, this daily elimination can be used to estimate the intake that may have been associated with this measurement. A study of a Danish-Finnish cohort of mothers found phthalate metabolites in all samples of breast milk that were tested (Main et al. 2006).
Phthalates have been used widely as plas ticizers in food packaging, and as such, diet is considered a major route of exposure (Schettler 2006). Research on the levels of phthalates in food is limited because the accurate determi nation of phthalate levels by a laboratory is severely complicated by the ubiquitous nature of phthalate contamination (Fromme et al. 2007). Phthalates have been measured previ ously in NHANES, beginning in NHANES III, where levels of phthalate metabolites were measured in samples collected from 289 adults in 1988289 adults in -1994289 adults in (Blount et al. 2000, to more recent studies that have greatly expanded the sample population [Centers for Disease Control andPrevention (CDC) 2005, 2009;Silva et al. 2004].
Although research suggests that dietary intake is a significant route of exposure for phthalates, because of elevated levels detected in food samples (Fromme et al. 2007;Tsumura et al. 2001Tsumura et al. , 2003, few research studies have been conducted to quantify how dietary intake correlates with body burden of these toxic chemicals.
The primary aim of this study was to ana lyze the relationship between meat, dairy, fish, fruit, and vegetable consumption and total phthalate body burden in the United States using data from NHANES 2003(CDC 2010, an ongoing health and nutri tion survey of the U.S. population. The sec ondary aim of this study was to analyze the relationship between meat, dairy, fish, fruit and vegetable consumption and body burden of individual phthalate metabolites, as well as the sums of metabolites of high and low molecularweight phthalates.

Materials and Methods
Data set. The NHANES is an ongoing survey of the civilian, noninstitutionalized U.S. pop ulation designed to monitor the health status of adults and children [Centers for Disease Control and Prevention (CDC) 2010]. The survey combines questionnaire data with physical examinations as well as biomoni toring of chemical contaminants in blood and urine. NHANES is administered by the National Center for Health Statistics, and the biomonitoring is performed by the National Center for Environmental Health, both divi sions of the CDC.
NHANES typically samples around 5,000 individuals each year using a complex, multistage probability sampling design with oversampling of populations that might be at greater health risk. Although data are collected continuously, data sets are reported in 2year cycles to ensure that the sample size is suf ficiently large. In the NHANES 2003-2004 survey, phthalate metabolites were measured in urine from 2,697 individuals ≥ 6 years old. Data for both measurements of phthalate metabolites and dietary intake were available for 2,374 individuals. There was no significant difference in age or sex between the individu als who had dietary intake data and those who did not. The proportion of white individuals was higher in the group that had dietary data than in the group that did not.
Dietary intake. Dietary intake was assessed in NHANES using two 24hr dietary recall facetoface interviews, a 24hr dietary recall conducted by telephone 3-10 days after the interview, and a food frequency questionnaire. Instructions for the interviews were provided orally during the first interview session in English, Spanish, or both languages. During both reporting sessions, information related to types of foods and beverages consumed, por tions, and meal times was collected. The 24hr dietary recall questionnaire gathered informa tion about food type and portions consumed, and the food frequency questionnaire asked about frequency of food consumption but not about portion size. The food frequency ques tionnaire was distributed to enrolled partici pants who completed at least one 24hr dietary recall interview. Because of the short halflife of phthalates in the human body, which are estimated to be on the order of hours, dietary intake was estimated using only the first 24hr dietary recall, which assessed dietary intake for the day before urine was collected for phthalate measurements.
The individual foods measured in the dietary intake interview were grouped based on the MyPyramid equivalents [U.S. Department of AgricultureAgricultural Research Service (USDAARS) 2010]. The groups of interest for this analysis were milk, meat and beans, vegeta bles, and fruit. Subgroups of interest were total dairy foods (milk, yogurt, and cheese); meat (beef, pork, veal, lamb, and game); frankfurt ers, sausage, and luncheon meats (made from meat or poultry); poultry (chicken, turkey, and other); fish and shellfish high in n3 fatty acids; fish and shellfish low in n3 fatty acids; eggs; vegetables; legumes; and fruit.
Chemical analyses. Chemical analyses of phthalates were conducted as reported previ ously (Kato et al. 2005;Sjodin et al. 2004). Phthalate metabolites are reported individu ally in the NHANES data set. Any phthalate metabolites that were detected in fewer than 60% of the samples were excluded from future analyses, because of the inability to calculate geometric means for these congeners, as has been reported previously in NHANES analyses (Sjodin et al. 2008). Total phthalate metabo lite levels were calculated for each individual by summing the levels of the analytes that met this condition. Levels that were below the limit of detection were estimated by dividing the limit of detection for that specific analyte by the square root of two. Additionally, high and lowmolecularweight phthalate metabolite totals were calculated, because low and high molecularweight phthalates are used in dif ferent classes of commercial products and thus may represent different routes of exposure. Lowmolecularweight phthalates are defined as those with molecular weight < 250 g/mol (Wolff et al. 2008). Additionally, associations between the sum of the four di(2ethylhexyl) phthalate (DEHP) metabolites [mono2ethyl 5carboxypentyl phthalate (MECPP); mono (2ethyl5hydroxyhexyl) (MEHHP); MEHP; and mono(2ethyl5oxohexyl) phthalate (MEOHP)] and dietary intake were also cal culated. Phthalate metabolites were analyzed on a creatinineadjusted basis.
Demographic variables. The demographic variables we analyzed included age, sex, ethnic ity, and body mass index (BMI; the weight of an individual in kilograms divided by the square of their height in meters). A BMI of < 18.5 is considered underweight, 18.5-25 is considered normal weight, 25-30 is over weight, and > 30 is obese. In the multiple regression analyses used in this study, we treated BMI as a continuous variable.
Data analysis. Food consumption, using the MyPyramid equivalents (USDAARS 2010), was correlated with total phthalate metabolite concentrations, total low and high molecularweight phthalate metabolite con centrations, and levels of individual phthalate metabolites. Three stages of data analyses were conducted. First, preliminary (or univariate) analysis was conducted to find mean, median, and ranges of the sample's demographics, dietary characteristics, and body burden of phthalate metabolites based on sample weights. Means, geometric means, and medians of indi vidual and total phthalate metabolites were also calculated based on weighted data. Phthalate metabolite levels were found to be highly skewed, thus logtransformed values of these variables were used in all subsequent bivariate and multivariable analyses.
Second, we assessed relationships between dietary intake of individual food types and measured levels of individual phthalate metabolites with bivariate analyses, includ ing correlations and bivariate linear regres sion models. The dependent variable in these analyses was a measured chemical level, that is, individual phthalate metabolites, low or highmolecularweight metabolites, DEHP metabolites, or total metabolites. The total servings of one MyPyramid food subgroup (USDAARS 2010) were included in each model. All regression models were built using weighted data. Finally, for each association that was statistically significant in the bivariate analyses (p < 0.05), we constructed a multiple linear regression model, including that food type and the covariates age, BMI, sex, and ethnicity. Age and BMI were modeled as con tinuous variables, and sex and ethnicity were modeled categorically (Table 1).
We performed the data analysis with the survey procedures using SAS software (version 9.2; SAS Institute, Cary, NC), with the appro priate sample characteristics (strata, clusters, and weights), as described in the NHANES documentation (CDC 2010).

Results
Univariate analysis. Most individuals (52.4%) in the NHANES sample with food and phthalate data were female. The demographic distribution of the NHANES participants is dis played in Table 1. The age of NHANES partic ipants ranged from 6 to 85 years old (weighted mean = 39.0 years). The BMI ranged from 13.0 to 58.6 (weighted mean = 26.8 kg/m 2 ). The percentage of individuals who were clas sified as underweight, normal weight, over weight, and obese is shown in Table 2. Table 3 presents the dietary intake of the food groups we analyzed. There was a wide range of meat consumption, from 0 ounce equivalents to 25.5 ounce equivalents con sumed. The consumption of fish and poultry was also quite variable, with ranges of 0-14.3 and 0-17.7 ounce equivalents consumed, respectively. The median cups of fruit con sumed was 0.6, whereas the median consump tion of vegetables (excluding legumes) was 1.1.
Adjusting for creatinine concentration did not affect which phthalate metabolites were detected at the highest levels. MEP still had the highest geometric mean concentration at 184.3 µg/g creatinine (range, 1.9-29,986). The lowmolecularweight phthalates had a geo metric mean creatinineadjusted concentration of 190.5 µg/g creatinine (range, 2.1-29,987), whereas the highmolecularweight phthalates had a geometric mean creatinine adjusted concentration of 129.7 µg/g creatinine (range, 0.7-4,901). Total phthalates had a geo metric mean creatinineadjusted concentration of 413.3 µg/g creatinine (range, 2.8-30,119).
Bivariate analysis. Scatterplots with LOESS curves did not show outliers or curvi linear relationships. Bivariate linear regres sion models estimated the relationship between consumption of single food groups and phthalate levels. Food types that had a statistically significant (p < 0.05) association were included in multiple regression analyses that adjusted for demographic covariates. The statistically significant results of the bivari ate and multiple regression analyses can be seen in Table 6. Nonsignificant results of bivariate and multiple regression analyses can be seen in Supplemental Material, Table 1 (doi:10.1289/ehp.0901233). The regression coefficients shown describe the change in log transformed creatinineadjusted phthalate metabolite level associated with consumption of one additional serving of food.
In bivariate analyses, total dairy consump tion was significantly associated with MCPP. Poultry consumption was a significant predic tor of levels of all of the individual DEHP metabolites, MEHP, MEHHP, MEOHP, MECCP, total DEHP metabolites, as well as highmolecularweight phthalate metabolites and total phthalate metabolites. Egg consump tion was also significantly associated with uri nary MEHP levels. Highmolecularweight phthalates, particularly DEHP, are lipid   soluble (Dickerson 1997). Despite the high fat content in meat, consumption was associated with MEP levels, the metabolite of DEP, as well as with lowmolecularweight phthalate metabolites, likely because of the association with MEP. Vegetable consumption was also significantly associated with MEP levels, which was one of the strongest effects measured.
To further examine the relationship between vege table intake and MEP levels, we calculated the association between tomato and potato consumption and MEP levels. Consumption of both of these vegetables was associated with MEP levels. Additionally, tomato consump tion was associated with lowmolecularweight phthalate metabolites, likely because of the strong association between tomato consump tion and MEP levels. Fruit consumption was associated with levels of MnMP, a metabolite of dimethyl phthalate, a lowmolecularweight phthalate. Additionally, fruit consumption was found to be inversely associated with high molecularweight phthalate metabolites and DEHP metabolites. Fish consumption was associated with levels of MiBP, a metabolite of diisobutyl phthalate, a highmolecularweight phthalate. Levels of MnBP were not found to be significantly associated with any of the dietary variables.

Multivariable analysis.
To adjust for pos sible covariates that may have confounded the effects observed in the bivariate regres sion analyses, we also performed multiple regression analyses. In these analyses, other variables included in the model were age, eth nicity, sex, and BMI. Phthalate levels were previously shown to be associated with BMI (Hatch et al. 2008). The regression coefficient for the association between the logadjusted creatinineadjusted DEHP metabolites and poultry consumption in the multivariable model was 0.055, suggesting that an increase of one ounce of poultry per day is associated with an increase in DEHP metabolite levels of approximately 5.7%.

Discussion
This analysis demonstrates that there is an association between the reported consumption of different food types and the levels of phtha late metabolites in urine of study participants. Previous research has established that many commonly consumed food types are con taminated with persistent organic pollutants, although the literature is sparse for phthalate levels in food, because of the difficulties of analyzing parent phthalate compounds.
A study of homogenized duplicated diet samples collected over the course of 7 days from 50 total study subjects (27 women, 23 men) age 14-60 years old from Germany found that dietary intake of DEHP was sig nificantly correlated with levels of DEHP metabolites measured in urine (Fromme et al. 2007). The median daily intake of DEHP was estimated to be 2.4 µg/kg body weight (95th percentile estimate: 4.0 µg/kg body weight), whereas the median daily intake of di-n-butyl phthalate (DnBP) was estimated to be 0.3 µg/kg body weight (95th percen tile estimate: 1.4 µg/kg body weight). The median daily intake of diisobutyl phthalate (DiBP) was estimated to be 0.6 µg/kg body weight (95th percentile estimate, 2.1µg/kg body weight). However, although DEHP intake was significantly correlated with lev els of DEHP metabolites measured in study participants, similar correlations were not observed for DnBP and DiBP metabolites. In the current study, creatinineadjusted levels of MEHP (a metabolite of DEHP) were significantly associated with poultry consumption; total meat, poultry, and fish consumption; and discretionary solid fat con sumption in multivariate regression models. Creatinineadjusted levels of another DEHP metabolite, MECPP, were also significantly associated with discretionary solid fat con sumption in multivariate regression models. Creatinineadjusted levels of MEHHP, also a DEHP metabolite, were significantly asso ciated with poultry and discretionary solid fat consumption. Similar results were also observed for creatinineadjusted MEOHP, a DEHP metabolite as well. The results from the Fromme et al. study (2007), as well as this study, point to diet as a significant route of exposure to DEHP, mostly likely from meat consumption or consumption of solid fats.
A research group from Japan examined the levels of phthalate esters in duplicate diet sam ples on two separate occasions (Tsumura et al.  2001,2003). In the first study conducted in 1999, 1week duplicate diet samples were col lected for 63 patients in three hospitals from three geographically distinct prefectures in Japan (Tsumura et al. 2001). DEHP was also detected at the highest levels in these samples, with concentrations ranging from 10-4,400 ng/g wet weight (ww). There was a signifi cant difference between mean levels of DEHP detected in the three hospitals, with two of the hospitals having considerably higher lev els (384 ng/g and 478 ng/g) compared with 46 ng/g. The authors hypothesized that the elevated levels in two of the hospitals were due to the use of disposable PVC gloves, which used DEHP in manufacturing, during food preparation. In a followup study, Tsumura et al. (2003) examined levels of phthalates in duplicate diet samples that were collected in 2001after the use of DEHP in PVC gloves was banned. Once again, the researchers col lected weeklong duplicate diet samples from 63 patients at the same 3 hospitals. DEHP levels in these samples were detected at signifi cantly lower levels than in the previous study, with a range of 6-675 ng/g ww compared with 10-4,400 ng/g ww in the previous study. The lower levels of phthalates detected in this study support the hypothesis of Tsumura et al. (2003) that the PVC gloves used in food prepara tion were a substantial source of phtha lates in the total diet samples. The estimated daily average intake of DEHP was 160 µg/day. Nondetected levels in this study were esti mated using half of the detection limit, which ranged from 0.1-15.6 ng/g ww, which could overestimate the daily phthalate intake because of the high detection limit. Tsumura et al. (2003) also measured levels of other phtha lates whose metabolites were measured in the present study: dinbutyl phthalate (DBP), benzylbutyl phthalate (BBP), dicyclohexyl phthalate (DCHP), and diisononyl phthalate (DINP). DBP was detected above the limit of quantification (LOQ) in 12 of 63 samples tested, with detected values ranging from 4.5-27.1 ng/g ww. BBP was detected above the LOQ in 29 of 63 samples, with detected values ranging from 0.9 to 27.1 ng/g ww. DCHP was not quantified in any samples. DINP was detected above the LOQ in 2 of 63 samples; one sample was 10 ng/g ww, the other 24 ng/g ww. The results of both stud ies suggested that diet is likely a significant route of exposure to phthalates (Tsumura et al. 2001(Tsumura et al. , 2003. A study conducted in Denmark examined the levels of DEHP, DBP, and BBP in total diet samples, infant formula, and baby food (Petersen and Breindahl 2000). DEHP was detected in 11 of 29 total diet samples above the LOQ, with an estimated average concen tration in all samples ranging from 110 to 180 ng/g ww. DBP was detected above the LOQ in 6 of 29 total diet samples, with an estimated average concentration in all sam ples ranging from 90-190 ng/g ww. BBP was detected above the LOQ in 8 of 29 diet sam ples, with an estimated average concentration in all samples ranging from 17-19 ng/g ww. Plasticizers were detected in 6 of 11 baby food samples tested as well as 8 of 11 infant formula samples. DEHP was measured at the highest levels in both individual baby food and infant formula samples, 0.63 µg/g ww and 0.06 µg/g ww, respectively. Overall, DEHP was estimated to be consumed at the highest levels through diet, with an estimated intake of 0.30 mg/day, whereas the estimated daily consumption of DBP was 0.29 mg/day and 0.03 mg/day for BBP. Wormuth et al. (2006) explored routes of exposure of the European populations to phthalates via modeling. Potential routes con sidered in the study included consumption of contaminated food, ingestion of contam inated soil and dust, ingestion of personal care products containing phthalates, dermal exposure to personal care products, and inha lational exposure. The authors estimated that infants and small children were exposed at the highest levels to all of the eight differ ent phthalates studied when normalizing by body weight. The study found that dietary intake was the primary route of exposure to DEHP, DiBP, and DnBP. The main route of exposure to DEP was estimated to be dermal exposure to personal care products. This study emphasized the importance of dietary intake in total phthalate exposure as well as the wide range of routes that should be considered when assessing phthalate exposure.
The findings from our study suggest that there is an association between dietary consumption of certain food types and lev els of chemical contaminants measured in the NHANES study population. Poultry con sumption was significantly associated with cre atinineadjusted DEHP metabolites MEHP, MEHHP, MEOHP, and MECPP as well as highmolecularweight phthalate metabolites. Additionally, the finding that egg consump tion is significantly associated with levels of MEHP suggests that chickens themselves may be contaminated with phthalates and that food is not being contaminated just through packaging and processing. Fruit and vegetable consumption was associated with metabo lites of the lowmolecularweight phthalates, such as DEP and DMP. Fruit consumption was inversely associated with metabolites of highmolecularweight phthalates and DEHP metabolites. DEP and DMP are much more water soluble than the higher weight phtha lates, which could explain their presence in fruit and vegetables (Stales et al. 1997). Meat intake was associated with the metabolite of DEP, but not with DEHP metabolites as poultry was. This finding suggests a potentially different source of contamination in poultry compared with meat. Routes of phthalate contamination for food could vary between food types, and could be from the food itself, during processing and packaging, or during storage. Because of this current uncertainty in sources of phthalate contamination in food, future studies of chemical contamination of U.S. food should involve experts from the food industry who could assist in determining at what step in the production process contami nation is occurring. Studying human exposure to phthalates is complicated by the difficulty in analyzing the parent compounds because of widespread environmental contamination, including contamination in laboratories, which makes determining how much of each phtha late is present in various products challenging. Additionally, various countries have differ ing regulations about how much phthalate is allowed in a product, which can complicate exposure estimation. We recently described a mixture of chemicals, many with unknown or incompletely described human toxicity, found in U.S. food (Schecter et al. 2009(Schecter et al. , 2010. The potential exists for additive, synergistic, or multiplicative toxic effects for chemicals found in food or food packaging, particularly for endocrinedisrupting compounds.
Although dietary intake was significantly associated with levels of phthalate metabolites measured in urine, dietary intake, as meas ured in this study, cannot solely explain all of phthalate body burden. The potential routes of phthalate exposure are diverse, because of the wide range of commercial uses of phtha lates. The main strengths of this study are the large sample sizes measured by NHANES, as well as the comprehensive data collected about each study participant that included both sur vey questions and chemical analyses of phtha late levels. The main weakness of the current approach is the crosssectional design of the NHANES study, which makes determining causation impossible. Dietary intake in this study was estimated by a single 24hr dietary recall that estimated the food consumed the day before urine was collected, and chemi cal levels were measured on only one occa sion. A followup study could bene fit from multiple phthalate metabolite meas urements to provide an average level. The possibility exists that these data may not be truly repre sentative of either dietary intake or chemical levels. However, the results from this study, combined with results from previous studies, suggest that the food supply is contaminated with phthalates, among other chemicals. A large study, possibly government sponsored, of chemical contamination of food is likely necessary to accurately assess the amount of and sources of contamination to protect the health of the public.