Body Burdens of Brominated Flame Retardants and Other Persistent Organohalogenated Compounds and their Descriptors in U.S. Girls

Abstract

Background

Levels of brominated flame retardants are increasing in U.S. populations, yet little data are available on body burdens of these and other persistent hormonally-active agents (HAAs) in school-aged children. Exposures to such chemicals may affect a number of health outcomes related to development and reproductive function.

Objective

Determine the distribution of biomarkers of polybrominated diphenyl ethers (PBDEs), polychlorinated biphenyls (PCBs), and organo-chlorinated pesticides (OCPs), such as DDT/DDE, in children, and their variation by key descriptor variables.

Methods

Ethnically diverse cohorts of girls 6-to-8 years old at baseline are being followed for growth and pubertal development in a multi-site, longitudinal study. Nearly 600 serum samples from the California and Ohio sites were analyzed for lipids, 36 PCB congeners, 11 PBDE congeners, and 9 OCPs. The biomarker distributions were examined and geometric means compared for selected analytes across categories of age, race, site, body mass index (BMI), parental education, maternal age at delivery, and breastfeeding in adjusted models.

Results

Six PBDE congeners were detected among greater than 70% of samples, with BDE-47 having the highest concentration (median 42.2, range 4.9–855 ng/g lipid). Girls in California had adjusted geometric mean (GM) PBDE levels significantly higher than girls in Ohio. Furthermore, Blacks had significantly higher adjusted GMs of all six PBDE congeners than Whites, and Hispanics had intermediate values. GMs tended to be lower among more obese girls, while other variables were not strongly associated. In contrast, GMs of the six PCB congeners most frequently detected were significantly lower among Blacks and Hispanics than Whites. PCBs and the three pesticides most frequently detected were also consistently lower among girls with high BMI, who were not breastfed, whose mothers were younger, or whose caregivers (usually parents) were less educated. Girls in California had higher GMs than in Ohio for the pesticides and most PCB congeners, but the opposite for CB-99 and −118.

Conclusions

Several of these potential HAAs were detected in nearly all of these young girls, some at relatively high levels, with variation by geographic location and other demographic factors that may reflect exposure pathways. The higher PBDE levels in California likely reflect differences in fire regulation and safety codes, with potential policy implications.

During the last few decades, concern has grown about the health effects of exposure to environmental contaminants that have hormonally active properties. While some of the chemicals of original concern were banned in many Westernized countries (such as DDT/DDE and PCBs), they are persistent in the environment. More recently, levels of a class of brominated flame retardants, the polybrominated diphenyl ethers (PBDEs), have been found to be increasing in the environment and subsequently in human biospecimens, and are present at higher concentrations in human serum and milk in the U.S. than in Europe (Hites 2004; Norstrom et al. 2002; Sjodin et al. 2004a). In contrast, levels of PCBs have been declining since the 1970s (Sjodin et al. 2004a). The PBDE studies have tended to include only adults (Bradman et al. 2007a and b; Hites 2004; Petreas et al. 2002; Sjodin 2004a; Schecter et al. 2005; Thomsen et al. 2002) and although there is much data on PCBs and DDT/DDE in adults and newborns, levels of these compounds are not routinely monitored in children under age 12y. The Centers for Disease Control and Prevention (CDC 2005) reports concentrations of these potential hormonally active agents (HAAs) in older children (12–19y) in the National Health and Nutrition Examination Survey (NHANES); more recent data on PBDEs were published separately (Sjodin et al. 2008) and in the very recent new report (CDC 2009).

Exposure to these persistent compounds in children is primarily through the diet (including breast feeding), as well as dust or inhalation for PBDEs (Lorber 2008). As the reproductive system does not fully mature until later puberty during adolescence, chemicals that affect the endocrine system might potentially alter such development. The observed trend towards pubertal development occurring at younger ages (Hermann-Giddens et al. 1997; Euling et al, 2008) has raised the possibility that exogenous exposures may be contributing factors. Due to these concerns and the association of early menarche with increased risk of breast cancer, a study of determinants of puberty in girls is being conducted within the NIEHS/NCI-sponsored Breast Cancer and the Environment Research Centers (BCERC), which include transdisciplinary research collaborations integrated across biologic, epidemiologic, and community outreach projects. Of primary interest are exposures to potential HAAs in young girls before breast development (Hiatt et al. 2009). Working with community representatives, laboratory staff at the National Center for Environmental Health of the Centers for Disease Control and Prevention (CDC), and the relevant literature, we identified numerous compounds for evaluation in a small pilot sample of girls who provided urine and serum specimens (Pinney et al. 2008, Windham et al. 2008, Wolff et al. 2007a). Factors considered in determining which chemical families were subsequently measured in the entire cohort included high detection frequency and adequate variability. This current analysis focuses on persistent organo-halogenated compounds measured in serum, which was collected at two sites. To provide data lacking on children of this age (6–9y), our goal was to characterize the distribution of these compounds and to compare them by geographic and other common demographic variables.

Materials and Methods

Study Population

In the BCERC epidemiologic studies, girls ages 6–8 years at enrollment are being followed to measure onset and progression of pubertal maturation, including breast and pubic hair development. This longitudinal study began in 2005 at three sites: 1) Mount Sinai School of Medicine, which recruited girls through clinics, schools, and neighborhood centers in East Harlem in New York City; 2) Cincinnati Children’s Hospital/ University of Cincinnati, Ohio (abbreviated as OH site), which recruited through public and parochial schools in the greater Cincinnati metropolitan area and through the Breast Cancer Registry of Greater Cincinnati; and 3) the Kaiser Permanente Northern California group, which recruited members of the Kaiser Foundation Health Plan born and still residing in the San Francisco Bay area (abbreviated as the CA site). Eligibility criteria for the study included age 6–8 years, female sex, and no underlying endocrine-associated medical conditions. All sites obtained informed consent from parent or guardian and child assent, administered by each institution’s IRB. Study designs and methods were standardized for most but not all components, as each Center retained some unique scientific aims or modified procedures for the unique circumstances of its population.

Sources of Data

At the time of the baseline and subsequent annual visits, girls had a physical exam and primary care givers (usually a parent) completed detailed questionnaires on demographic and a wide variety of other factors potentially related to puberty. Girl’s race and Hispanic ethnicity, highest education of a primary provider (usually a parent), household income, history of breast-feeding and maternal age at delivery were obtained from baseline questionnaires. For the purposes of this report, girl’s race/ethnicity is classified into mutually exclusive categories in the following priority order: Black (regardless of ethnicity), Hispanic (including any race other than Black), non-Hispanic Asian or Pacific Islander, non-Hispanic White, or Other (the last three are summarized without the term “non-Hispanic” in the text). Height and weight were measured during the physical exam using calibrated scales and stadiometers by research staff that had been trained and certified uniformly across all three sites. Body mass index (BMI) was calculated as weight/height-squared (kg/m²) and then classified as <85th national percentile or ≥ 85th percentile, indicating risk for obesity (Himes and Dietz 1994), using age and sex-specific CDC growth charts (CDC 2000). Age and BMI were calculated for the time of the serum draw (in months and years).

Blood samples were collected only at the CA and OH sites using materials and procedures provided by the CDC, after which they were processed to serum and stored frozen at −80°C until shipment. Samples were collected over the course of a few years as recruitment continued (in OH) or some girls only provided serum at follow-up visits (in CA). Ohio generated 270 samples and California 348, with a mean sample volume of 1.9 mls, for determining the concentrations of 11 PBDE “congeners” (including 2,2’,4,4’,5,5’-hexabromobiphenyl), 35 PCB congeners, and nine organochlorine pesticides (o,p’ and p,p’-DDT, the metabolite p,p’-DDE, hexachlorobenzene (HCB), beta and gamma-hexachlorocyclohexane (HCCH), oxychlordane, trans-nonachlor, and mirex). The highly sensitive laboratory techniques used by the CDC for measuring the selected exposure biomarkers have been published (Sjodin et al. 2004b). Briefly, the methodology includes automatic fortification of the samples with internal standards as well as formic acid and water for denaturation and dilution of the samples. The samples were extracted by solid phase extraction; removal of co-extracted lipids was performed on a silica:silica/sulfuric acid column. Final quantitative determination of the target analytes was performed by gas chromatography isotope dilution high resolution mass spectrometry (GC-IDHRMS). Blank samples (n=3) were included in every set of 24 unknowns and 3 quality control samples. Reported concentrations were blank-corrected. The limit of detection was determined as the highest of (i) the instrumental detection limit or (ii) 3xSD of the blank samples. Hence, the limit of detection is directly proportional to the sample volume and the lipid concentration of the serum and was determined individually for each serum sample and analyte. The available amount of sample ranged from 0.2g to 2g. Detection limits were typically in the low pg/g serum range or sub to low ng/g lipid. As an additional quality control measure, a pool of serum was purchased (from the Interstate Blood Bank in Ohio, assembled from females of mixed race in their 20’s) and samples interspersed with approximately every 10 valid girls’ samples at each site. Concentrations in this report are lipid-adjusted (ng/g lipid weight), with lipids determined using commercially available kits (Roche Diagnostic Corp) for measurement of total triglycerides and total cholesterol.

Data Analysis

For each analyte, up to 618 girls provided samples, but seven were missing lipid values, three were not reportable by the lab and nine did not have questionnaire data, so they were excluded from further analyses. The distribution of each analyte (mean, median, standard deviation, range, etc) was examined and is reported for those in which more than 60% of samples were above the detection limits. In addition, geometric means (GMs) were calculated because of the non-normal distribution. Following prior practice (Wolff et al. 2005), for measurements below the limit of detection (LOD), we substituted the value LOD/√2 for calculation of descriptive statistics. We summed the individual concentrations for the six most frequently detected PCB and PBDE congeners, separately (CBs 99, 118, 138/158 (co-eluted), 153, 170, and180, labeled PCBsum; and BDEs 28, 47, 99, 100, 153, 154, labeled PBDEsum) and examined these variables similarly. We calculated a correlation matrix and Pearson correlation coefficients for the 17 compounds (including sums). Coefficients of variation (CVs) were calculated in the 52 quality control samples as SD/mean × 100.

Analysis by co-factors (categories of site, child’s race/ethnicity, age and BMI, caregiver education and income, maternal age at delivery and menarche, and history of breast feeding) was conducted comparing primarily GM levels by t-tests and ANOVA. Twenty-five girls were missing data on one or more of these covariates. We then performed multivariate analyses using the General Linear Model (GLM) procedure that accommodates unbalanced designs including those factors significantly associated with body burden levels, except income because of its correlation with education. Because breast feeding may represent a direct source of exposure, we present results for most variables without breast feeding in the model, adding it to report results only for this variable. Generally, inclusion of breast feeding did not affect the other associations. The model yields adjusted geometric means and their differences from a selected reference category for each variable.

Results

Age at blood draw ranged from 6–10 years (the only 10-year old had recently turned ten), with the majority aged 7–8 years old (Table 1). The sample was racially diverse with 52% white overall, and of fairly high socioeconomic status (SES) based on education level of the care-giver. About one-third of the girls had a BMI indicating risk for obesity and the majority had been breastfed (Table 1). There were some differences between the two sites with girls from CA being older at blood draw, having older mothers, more likely to be Asian or Hispanic, and breast-fed, than in the OH sample, but with a slightly lower level of care-giver education. The proportion of girls at risk for obesity was very similar in the two sites.

Table 1.

Characteristics of Girls in the BCERC Study of Serum Biomarkers, by Site, 2005–07

Characteristic		California (348)		Ohio (270)		Total
		N	%	N	%	%
Age
6 – 6.9 yrs		50	14.6	62	24.2	18.7
7 – 7.9 yrs		151	44.0	111	43.4	43.7
8 – 8.9 yrs		94	27.4	81	31.6	29.2
≥9 yrs		48	14.0	2	0.8	8.4
Race
White		144	42.0	168	65.6	52.1
Black		75	21.9	75	29.3	25.0
Hispanic (any race)		80	23.3	9	3.5	14.9
Asian		37	10.8	4	1.5	6.8
Other		7	2.0	0		1.2
Body Mass Index
<85th percentile		246	71.7	186	72.7	72.1
≥85th percentile		97	28.3	70	27.3	27.9
Breastfed
Yes		321	93.6	167	67.9	82.9
No		22	6.4	79	32.1	17.1
Education (Primary Provider)
≤ High School		61	18.0	23	9.4	14.4
Some College		108	31.9	75	30.6	31.3
≥ College Degree		170	50.1	147	60.0	54.3
Maternal Age at Delivery
<25 yrs		38	11.1	46	18.9	14.3
25–29 yrs		60	17.5	61	25.1	20.7
30–34 yrs		108	31.5	80	32.9	32.1
≥35 yrs		137	39.9	56	23.1	32.9
Maternal Age at Menarche
<12 yrs		79	24.6	49	21.1	23.1
12–13.9 yrs		164	51.1	136	58.6	54.3
≥14 yrs		78	24.3	47	20.3	22.6

Seven of 11 PBDE congeners were detected among more than 60% of samples, five of these in more than 80%. BDE-47 had the highest concentration (GM 42.8 ng/g lipid), followed by BDE-153, then BDE-99 and BDE-100 (Table 2). Ten of 35 PCB congeners and three pesticides were also detected in more than 60% of the girls (Table 2). Among the pesticides, p,p’-DDE was measured in nearly all samples and at a factor of 10–50 times higher (GM 107.3 ng/g lipid) than the other pesticides. Ten PCB congeners were detected in more than 60% of samples with five detected in nearly all girls. PCB congeners 153 and 138/158 were measured at the highest concentrations (GM 8.1 and 6.1 ng/g lipid, respectively). Fifty-two similarly-labeled QC samples (non-fortified human serum) were included intermixed with the girls’ samples. After excluding one QC outlier result, the coefficients of variation (CVs) for the blind QCs for those analytes detected in more than 75% of samples ranged from 4% (p,p’-DDE) to 36% (BDE-99). When the study samples from the same batch as the outlier blind QC sample were re-analyzed, the correlation between the original and repeated measurements ranged from 0.68 (CB-74) to 0.9996 (BDE-99), with an average of 0.95, so the original values were considered accurate for use.

Table 2.

Distribution of Polybrominated Diphenyl Ethers (PBDEs), Polychlorinated Biphenyls (PCBs) and Organochlorine Pesticides, by lipid-weight values (ng/g), in Serum of Young Girls (6–9y), 2005–07

Com- pound	LOD Rangea	% > LOD	Minb	95th Percentile	Max	Mean (SD)	Geometric Mean (GSD)	Median	NHANES Medianc
PBDE Congener
28	0.3–15.2	81.3	0.5	6.1	33.2	2.2 (2.4)	1.5 (2.3)	1.6	1.2
47	0.3–40.8	99.2	4.9	170.0	855.0	60.4 (67.1)	42.8 (2.2)	42.2	27.2
85	0.3–15.2	66.6	0.4	3.5	16.5	1.2 (1.4)	0.9 (2.1)	0.8	<LOD
99	0.3–22.7	98.7	0.8	35.3	154.0	13.4 (14.9)	9.6 (2.2)	9.0	5.7
100	0.3–15.2	99.3	0.8	44.0	382.0	15.0 (23.1)	9.8 (2.3)	9.6	4.9
153	0.3–15.2	98.8	1.3	54.9	220.0	19.2 (20.3)	13.7 (2.3)	13.6	7.5
154	0.3–15.2	73.1	0.4	4.4	31.8	1.5 (2.0)	1.0 (2.2)	1.0	<LOD
Sumd	N/A	N/A	31.4	300.6	1398.6	112.8 (117.9)	83.6 (2.1)	79.3	NA
OC Compound
p,p-DDE	0.3–208.0	99.2	11.0	583	8010.0	191.1 (429.3)	107.3 (2.6)	101.5	93.6
T-nonacl	0.3–75.8	69.4	1.1	29.3	66.1	9.0 (9.3)	5.9 (2.5)	6.0	<LOD
HCB	0.3–88.3	95.8	2.6	19.0	62.4	9.8 (5.7)	8.7 (1.6)	8.7	13.4
PCB Congener
74	0.8–39.5	69.9	1.1	10.7	36.5	3.8 (3.8)	2.7 (2.2)	2.6	2.2
99	0.3–15.1	95.5	0.5	7.6	57.4	3.0 (3.5)	2.2 (2.1)	2.1	2.3
105	0.3–15.1	69.3	0.5	2.8	69.3	1.3 (3.5)	0.9 (2.0)	0.8	0.7
118	0.3–17.5	97.3	0.6	11.8	205.0	5.1 (11.0)	3.5 (2.2)	3.4	2.8
138/158	0.3–15.1	96.8	0.9	29.1	155.0	10.0 (13.1)	6.1 (2.7)	5.6	4.6
153	0.3–15.1	98.7	1.0	44.8	127.0	14.3 (18.4)	8.1 (2.9)	7.4	5.4
156	0.3–15.1	61.0	0.4	8.5	60.5	2.5 (4.0)	1.3 (3.0)	1.1	0.5
170	0.3–15.1	76.5	0.4	13.7	47.4	3.8 (5.3)	1.9 (3.3)	1.7	1.1
180	0.3–15.1	94.3	0.5	34.9	133.0	9.3 (13.4)	4.4 (3.5)	3.8	3.0
187	0.3–15.1	64.9	0.4	8.2	36.4	2.4 (3.3)	1.3 (3.0)	1.2	1.0
Sume	N/A	N/A	15.7	164.7	708.9	55.3 (67.7)	34.9 (2.5)	32.0	NA

^aLOD (Limit of Detection) values tended to be at bottom of these ranges, but were dependent on sample volume and blanks, and calculated for each participant.

^bThe minimum represents the lowest value measured that was >LOD and thus overlaps with the LOD of some samples. The true minimums are <LOD for compounds detected in <100% of samples.

^cMedian values from NHANES for children 12–19y in 2003–2004 (CDC 2009).

^dPBDE sum includes the six congeners detected in >70%, e.g. all listed except BDE-85. Other congeners measured (and the detection frequency in %) were: BDE-17 (11.4%), −66 (22.8), −183 (27.0), bb-153 (42.2).

^ePCB sum includes CBs – 99, 100, 138/158, 153, 170 and 180. Other congeners measured (and the detection frequency in %) were: CB-28 (24.3%), −87 (20.5), −101 (15.9), −110 (13.9), −146 (56.6), −157 (31.5), −167 (25.5), −172 (18.3), −177 (25.4), −178 (29.3), −183 (39.6), −194 (47.6), −195 (16.8), −196_203 (53.6), −199 (47.6), and −206 (29.8). In addition, CB-44, 49, 52, 66, 128, 149, 151, 189, and 209 were detected in <10% of samples (N=596−599).

The primary PBDE congener concentrations were significantly correlated with each other (Pearson r’s > 0.6), with BDE-153 slightly less correlated than others but still significant. The primary PCB congeners were also significantly correlated with each other, with CB-153,-170,-180 and −138/158 particularly highly correlated (r>0.9). The pesticides were moderately correlated with each other (r’s = 0.2–0.5, p<0.001). Across groups, the PCB sum was most strongly correlated with HCB and t-nonachlor levels, and somewhat less so with p,p’-DDE, with CB-118 the least correlated of the PCB congeners with pesticides. PBDE levels were not at all correlated with the PCB sum or individual congeners, except for BDE-153, which was the only PBDE congener significantly correlated with the pesticides.

The crude GM of the PBDE sum and all congeners except BDE-85 were significantly lower in girls from the OH than the CA site. Blacks tended to have the highest PBDE GM concentrations, followed closely by Hispanics, compared to Whites with the lowest. The PBDEsum was also lower in girls who were older, of high BMI, and with higher educated care-givers. Levels did not vary consistently by mother’s age at delivery and whether the girl was breast-fed or by duration. In adjusted models including age, BMI, race, site, maternal age and care-giver education, most crude associations remained significant (table 3). CA girls had adjusted GMs that were significantly on the order of 30% higher than OH girls for all six frequently detected PBDE congeners. Blacks had significantly higher GMs than Whites for most congeners whereas adjusted levels in Hispanics were similar to Whites. There was a tendency to lower mean levels with high BMI or higher education, but not consistently statistically significant. Age of child or mother and breast feeding were generally not associated after adjustment.

Table 3.

Adjusted^a Geometric Means (ng/g lipid) of Polybrominated Diphenyl Ethers (PBDEs) in Serum of Young Girls (6–9y), 2005–07

		PBDE Congener
		−28	−47	−99	−100	−153	−154	Sum
Study Site
California		1.9*	47.7*	11.1*	10.4*	13.5*	1.1*	89.8*
Ohio		1.2	35.0	8.2	7.4	9.9	0.8	65.9
Child’s Race/Ethnicity
Asian		1.4	36.2	8.0	7.3	8.7	0.8	65.8
Black		1.8	48.3*	12.3*	11.5*	16.3*	1.3*	95.5*
Hispanic		1.5	40.0	9.0	8.6	11.6	0.9	75.9
White		1.5	39.8	9.3	8.4	10.8	0.9	73.4
BMI percentile of Child
<85th percentile		1.6	44.2	10.0	9.9	15.5	1.0	87.1
≥85th percentile		1.5	37.8*	9.1	7.8*	8.6*	0.9*	67.9*
Age of Child
≤6.9 yrs		1.6	41.8	9.8	9.6	12.2	1.0	80.8
7.0–7.9 yrs		1.4	42.6	9.9	9.6	12.4	1.0	81.0
8.0–8.9 yrs		1.6	42.6	9.7	9.0	11.5	1.0	78.5
≥9.0 yrs		1.6	36.7	8.7	7.3*	10.3	0.8†	68.1
Education of Primary Provider
≤High School		1.5	41.9	10.2	9.3	11.4	1.0	78.6
Some College		1.7*	43.5†	9.8	9.1	12.1	1.0	81.2†
≥College grad		1.4	37.4	8.6	8.1	11.2	0.9	71.3
Maternal Age at Delivery
<25 yrs		1.7	46.7	10.8	10.0	11.8	1.1	86.1
25–29 yrs		1.6	42.1	9.8	9.1	12.2	1.0	79.6
30–34 yrs		1.4	34.7*	8.1†	7.7†	11.8	0.9	68.0†
≥35 yrs		1.5	40.9	9.5	8.5	10.5	0.9	75.1

^aAdjusted for variables listed, for each compound separately.

^*Significant difference in means (p<0.05) within each variable, with reference group being White for race/ethnicity, 7.0–7.9 yrs for age of child, ≥college grad for education of primary care-giver, and 25–29 yrs for maternal age at delivery.

^†0.05 ≤p < 0.10.

For the PCBsum, the crude GM was significantly lower in girls from OH, but this varied by congener; the lower chlorinated congeners (CB-74, 99, 105, 118) showed higher GMs in OH and comprised a larger proportion of the PCB sum than in CA. This may be reflective of race/ethnicity differences as Asians tended to have the highest GMs of the more highly chlorinated congeners (CB-170, 180, and 187). Crude levels were higher in girls with older mothers or who were breastfed, and by longer duration of breast-feeding. After adjustment (table 4), GMs of several congeners and the sum were still higher in CA girls, but CB-99 and CB-118 were significantly lower in CA. Black and Hispanic girls had lower adjusted GMs for all six frequently-detected congeners, whereas Asian girls had similar GMs, as Whites. History of breastfeeding was consistently associated with higher GM PCB levels, approximately doubled for several congeners (CB-153, 170, 180), as was older maternal age, adjusting for each other. The association of higher PCB levels with higher education appeared to be independent of race. Adjusted GMs were lower among heavy girls, and somewhat so among older girls, although the later was not consistently statistically significant (table 4).

Table 4.

Adjusted^a Geometric Means (ng/g lipid) of Polychlorinated Biphenyls (PCBs) and Organochlorine Compounds in Serum of Young Girls (6–9y), 2005–07

		PCB Congener							OC Compound
		−99	−118	−138/158	−153	−170	−180	Sum	HCB	T-Nonachlor	p,p-DDE
Study Site
California		1.6*	2.3*	4.4*	5.9*	1.5*	3.5*	20.5*	8.0*	4.8*	149.1*
Ohio		1.9	3.1	3.8	4.6	0.9	2.0	16.9	6.9	3.6	50.4
Child’s Race/Ethnicity
Asian		2.1	3.0	4.9	6.7	1.4	3.2	22.9	7.5	4.4	102.3*
Black		1.6†	2.4*	3.6*	4.2*	1.0*	2.2*	15.8*	6.6*	3.4*	69.1
Hispanic		1.5†	2.4*	3.6*	4.4*	0.9*	2.1*	15.5*	7.8	4.3	110.7*
White		1.8	3.0	4.5	6.0	1.4	3.2	21.2	7.8	4.7	72.1
BMI percentile of Child
<85th percentile		2.2	3.5	5.6	7.3	1.6	3.7	25.3	9.1	5.2	107.8
≥85th percentile		1.3*	2.0*	3.0*	3.8*	0.9*	1.8*	13.6*	6.0*	3.3*	69.7*
Age of Child
≤6.9 yrs		2.0	3.2	4.5	5.8	1.3	2.8	20.6	8.4*	4.7	98.9
7.0–7.9 yrs		1.9	3.0	4.3	5.4	1.2	2.6	19.3	7.3	4.5	89.9
8.0–8.9 yrs		1.8	2.8	4.3	5.4	1.2	2.6	19.2	7.1	4.1	89.7
≥9.0 yrs		1.4*	1.9*	3.4†	4.4	1.0	2.4	15.6†	7.0	3.4*	70.7*
Education of Primary Caregiver
≤High School		1.4*	2.2*	2.8*	3.4*	0.8*	1.7*	12.8*	6.9*	3.0*	78.8*
Some College		1.8†	2.8†	4.3*	5.4*	1.1*	2.5*	19.0*	7.4†	4.2*	83.5*
≥College grad		2.1	3.2	5.7	7.7	1.8	4.1	26.3	8.0	5.8	98.9
Maternal Age at Delivery
<25 yrs		1.4*	2.3	2.8*	3.5†	0.8	1.6	13.3†	6.9	3.5	62.9*
25–29 yrs		1.7	2.6	3.7	4.5	0.9	2.0	16.0	7.3	3.7	80.7
30–34 yrs		1.8	2.8	4.8*	6.1*	1.4*	3.2*	21.3*	7.8	4.3	97.7*
≥35 yrs		2.0†	3.0*	5.8*	7.8*	1.8*	4.3*	26.3*	7.7	5.3*	113.8*

^aAdjusted for variables listed, for each compound separately.

^*Significant difference in means (p<0.05) within each variable, with reference group being White for race/ethnicity, 7.0–7.9 yrs for age of child, ≥college grad for education of primary care-giver, and 25–29 yrs for maternal age at delivery.

^†0.05 ≤p < 0.10.

Crudely, the pesticides had similar descriptor patterns as the PCBs except by race. With adjustment, girls in CA again had higher levels, particularly of p,p’-DDE. The adjusted GM of p,p’-DDE was higher in Asians and Hispanics, compared to Whites, but not of HCB and t-nonachlor, while Blacks had the lowest GMs of all three compounds (table 4). GM levels of the pesticides decreased with child’s age but increased with maternal age and somewhat so with care-giver education (attenuated by adjustment for breast-feeding). Girls who were breast-fed had significantly higher GM levels of all pesticides, up to 50% higher for p,p’-DDE.

Discussion

Our results show that many of these potential hormonally-active agents, most of which have been little studied previously in children of this age, were detectable in a large proportion of young girls, some at relatively high levels. Concentrations of many, most consistently the PBDEs, varied by geographic location with higher levels found in the California site. There were variations by race, with Blacks having the highest mean PBDE levels, but about 50% lower mean PCB levels and slightly lower pesticide levels, than Whites, even after adjustment for other factors. Hispanics had significantly lower levels of PCBs as well, but higher levels of p,p’-DDE. Asians had similar levels to Whites of all compounds except p,p’-DDE, for which they had the highest GM. As compared to the other compounds, patterns for PBDE levels by the other descriptors differed, including for BMI, SES (as measured by care-giver education), maternal age at delivery and history of being breast-fed. The girls who provided serum specimens were similar demographically to those who did not, so our results for these characteristics should be generalizable to the entire sample.

This is the first study to examine individual PBDE levels in elementary school-aged children in the U.S., reporting mean and median values generally 2–10 times greater than earlier published data (Hites 2004; Petreas et al. 2002; Thomsen et al. 2002). Sjodin et al (2004a) reported increasing PBDE concentrations in pooled U.S. serum samples from 1985 to 2002 and Hites (2004) estimated an exponential increase by a factor of 100 in the last 30 years, or a doubling every 4–6 years. Furthermore, studies from the U.S. generally report higher levels than those from Japan and Europe, particularly for BDE-47. Our GM for BDE-47 (42.8 ng/g lipid) is 50 percent higher than that of recent (2003–2004) NHANES data for 12–19 year olds, with differences of a similar magnitude for BDEs- 99, 100, and 153 (table 2 and CDC 2009). The lack of detection of some congeners in a large proportion of girls should not be interpreted to suggest absence of the compound, but is dependent on assay sensitivity and sample volume available.

The higher PBDE levels in CA than OH suggest greater exposure, as the difference was not explained by other co-variates examined. A recent study of dust samples taken from homes in California and Massachusetts found PBDE levels were 3–9 times higher in CA (Zota et al. 2007). Unlike other states, California has a history of stringent regulations (CA 2000) regarding flammability of upholstered furniture that can only be met by the addition of flame retardants, such as PBDEs. The primary commercial BDE mixture used to meet the polyurethane standard has been pentaBDE, which contains mainly BDE-47 and BDE-99, but also BDE-100, BDE-153, and BDE-154; the congeners we detected most frequently.

House dust and diet are considered the primary sources of PBDE exposure and young children are more likely to have higher dust exposure due to mouthing and playing activities, as well as to be exposed via breast milk (Lorber 2008). Our generally higher levels may therefore also indicate young children at greater risk of exposure, as with some other environmental chemicals and supported by exposure pathway modeling (Lorber 2008). Confirming this likelihood, the youngest age group in the NHANES data (12–19y) generally had higher mean PBDE levels than those of older age groups (Sjodin et al. 2008) and in pooled serum samples, recent data from Australia showed children 7–12y had higher PBDE levels than adults over 30y, but lower than children 2–6y (Toms et al. 2009). An increase in levels from newborn to early childhood (up to age 3 or 4y) has been attributed to breast-feeding (Carrizo and Grimalt, 2007; Toms et al. 2009). However, we did not find significant differences by whether or duration girls were breastfed, nor by mother’s age at delivery or child’s age within our relatively narrow range. Thus by age 6 and older, differences in PBDE body burdens may be most reflective of post-natal environmental exposures that are continuing, including from dust and diet.

By race, we found significantly higher PBDE levels in Black than in White girls, with intermediate levels in Hispanics before adjustment, suggesting possible differences in exposure. There are few comparison data available; NHANES data are for all ages combined and are not adjusted, but do indicate that Mexican and African Americans had higher GMs than non-Hispanic Whites for BDE-47 and BDE-99, and the 95^th percentile levels were particularly high for African Americans (Sjodin et al. 2008). A study of cord blood levels in Baltimore found Asians had lower mean BDE-47 and BDE-153 levels than non-Hispanic Whites, and African Americans initially had higher PBDE levels, but not after adjustment (Herbstman et al. 2007). The reasons for potentially higher levels in Blacks should be explored further, perhaps including differences in metabolism and environmental conditions, such as condition of housing and furniture, or take-home exposures from parental occupations.

Examining the organo-chlorine compounds, the median we report for p,p’-DDE (101.5 ng/g lipid) is similar to that reported in NHANES for 12–19 year olds during 1999–2002 and 2003–04 and that for hexachlorobenzene is slightly less (CDC 2009). For the other pesticides we examined, the majority (75–90%) of 12–19 year olds in NHANES data had levels below the detection limit, so GMs were not calculated. We thus had higher detection rates for trans-nonachlor (69%), B-hexachlorocyclohexane (36%), and p,p’-DDT (30%). The mean and maximum p,p’-DDE levels we report are higher than recently reported in pregnant women (88 and 622 ng/g respectively) in Sweden (Glynn et al. 2007), but as expected are much lower than in earlier cohorts of pregnant women (Hertz-Picciotto et al. 2005) and do not represent a comparable population.

For PCBs, our median levels tended to be similar, but slightly higher than, those in the recent NHANES data for 12–19 year olds (table 2, CDC 2009). Comparing studies of neurodevelopment in children, Longnecker et al. (2003) reported much higher median CB-153 levels in maternal serum ranging from about 30–140 ng/g in the U.S. and up to 450 ng/g in the Faroe Islands. Most of these samples are from much earlier cohorts or have established exposure sources.

GM levels were higher in CA than OH for the three pesticides most frequently detected and four of the six frequently detected PCB congeners, similar to PBDEs. Consistent with these results, in the Nurses Health Study, Laden et al. (1999) reported that mean DDE levels were higher in participants from the West than the Midwest. Our results were not explained by race or SES, but race categories do not describe immigrant status. Pesticide exposure patterns differ in countries outside the U.S. (Bradman et al. 2007b, Smith 1999), hence a larger influx of immigrants (e.g. the parents) to CA from developing countries in Latin American and Southeast Asia, where some of these pesticides are still used, may explain higher levels of persistent pesticides in their offspring, via placental transport or breast-feeding.

In NHANES data combining all ages (CDC 2009), levels of p,p’-DDE were higher among Mexican Americans than Blacks or Whites, consistent with our findings for Hispanic girls. As noted above, the higher DDE levels we find in Asians and Hispanics may reflect a higher proportion of immigrants. Also similar to our results, some of the PCB congener levels in NHANES were lower among Mexican Americans. In contrast to the PBDEs, we found lower PCB levels in Blacks, as well as lower pesticide levels, which were not evident in NHANES data. Other studies have suggested lower PCB levels in Black than in White gravidas, but results were inconsistent by adjustment and these are not comparable populations (Herbstman et al. 2007, Wolff et al. 2007b). Nor are the NHANES data, which are for all ages (>12) and genders combined, and Asians are not separated, so our data thus add important detail to the literature.

Our finding of an inverse relationship between BMI and several pesticides or PCBs is difficult to interpret as either a contributor to or consequence of chemical levels. We examined BMI because of changing growth and fat stores in children of this age (Wolff and Landrigan, 2002). Exposure to some of the compounds occurs through breast-feeding, which may affect later BMI. We observed a slight trend towards lower BMI among breast-fed girls, which was, however, not statistically significant. Furthermore, in models that control for breast-feeding, the inverse association of the chemicals with BMI persisted. To explore this further, we examined the relationship between height and the pesticides or the PCBsum (adjusting for the other co-variates) and also found an inverse relationship. Similarly, a study from Germany found that girls with higher DDE levels during childhood had reduced growth at age 10 and shorter height at age 8, but reported no associations with PCBs (Karmaus et al. 2001, 2002). Levels of these persistent chemicals in girls primarily represent in utero and lactational exposures from the mother, as well as possibly later childhood exposure from diet and other sources. Studies with data on maternal or early childhood PCB levels show inconsistent results with later childhood anthropometric measures (Blank et al. 2000; Hertz-Picciotto et al. 2005, Gladen et al. 2000; Guo et al. 1995), likely further complicated by pubertal status at the time of anthropometric measurement.

In adults, a complex pharmacokinetic model has been proposed to explain interaction between BMI, time since exposure and organochlorine concentrations in adipose tissue (Wolff and Anderson 1999), which may also explain some of the findings in girls. DDE levels were inversely related to BMI in nulliparous Chinese women (Perry MJ et al. 2005), but varied in other studies. PCB levels were also lower in obese women in at least one study (Herbstman et al. 2007), but not in another (Wolff et al. 2007b). These observations seem less likely to support an effect on growth in children and perhaps suggest a “dilution” effect, which we can examine further when the longitudinal data collection is complete. Unlike typical findings in adults and with maternal age in our study, we found that some of the PCBs and pesticide levels were lower in older children, likely reflecting tapering effects of exposure occurring trans-placentally or through breast-feeding.

In summary, our study provides data that was previously lacking on body burdens of HAAs in young children from a large, racially diverse population using the most sensitive assays now available. Similar to some studies (Bradman et al. 2007a, Herbstman et al. 2007), but not all, we did not see strong correlations between PBDEs and PCBs, nor were their descriptors consistent. The variation we found by race and BMI, which are both relevant to puberty, as well as by site, warrant further investigation, both with respect to puberty and to identify factors contributing to exposure.