Polybrominated Diphenyl Ether (PBDE) Flame Retardants and Thyroid Hormone during Pregnancy

1Polybrominated diphenyl ethers (PBDEs) are synthetic chemicals used as flame retardants in a variety of consumer products such as electronics, furniture, textiles, and construction materials. The chemical structure and properties of PBDEs are similar to those of polychlorinated biphenyls (PCBs), which were banned in the United States in 1978. Theoretically, a total of 209 PBDE congeners may be produced depending on the number and position of bromine atoms on the diphenyl ether structure, although only congeners with more than four bromines are used commercially. PBDEs are lipophilic, bioaccumulate in wildlife and humans, and biomagnify up the food chain. Congeners with lower bromine contents are particularly persistent, with estimated half-lives ranging between 2 and 12 years in humans. PBDEs are global contaminants that have been detected in human adipose tissue, serum, and/or breast milk samples collected in Asia, Europe, North America, Oceania, and the Arctic. The concentration of these chemicals in human serum and breast milk has exponentially increased in the last three decades.

2Prenatal exposure to PBDEs has been reported to alter spontaneous motor behavior, memory, and learning in rats and mice. In humans, higher maternal serum PBDE levels were found to be related with lower scores on measures of intelligence and attention and increased time to pregnancy. Maternal thyroid hormones (THs) play an essential role in fetal brain development and modulate menstrual cycle characteristics. It has thus been suggested that PBDEs may affect neurodevelopment and fertility by disrupting THs.

3Similarly to studies on PCBs, most studies conducted in nonpregnant rats and mice report that exposure to PBDEs lowers free and/or total thyroxine (T4) in a dose-dependent fashion and thus has a hypothyroxinemic effect; PBDE exposure generally did not affect thyroid-stimulating hormone (TSH) levels [see Supplemental Material, Table 1a (doi:10.1289/ehp.0901905)]. Only two experimental studies have been conducted in pregnant animals. exposed rat dams daily to the PBDE commercial mixture DE-71 at doses of 0, 1, 10, and 30 mg/kg from gestation day (GD) 6 to postnatal day (PND) 21 and found a 48% and 44% decrease in total T4 in the high-exposure group relative to controls at GD20 and PND22, respectively (free T4 not measured)., however, did not find daily exposure to PBDE congener BDE-99 or the commercial mixture Bromkal 70-5DE beginning on GD4 to affect total or free T4 levels in mice at GD17.

4In contrast to animal experimental studies, human epidemiologic studies conducted in nonpregnant adults generally reported lower TSH and higher free and total T4 in relation with higher PBDE concentrations in serum and house dust, suggesting that these chemicals may exert a hyperthyroidic effect [see Supplemental Material, Table 1b (doi:10.1289/ehp.1001905)]., for instance, found positive associations between the sum of eight PBDE congeners and both free and total T4, and an inverse association with TSH in 308 male Great Lakes fish consumers. Participants in the fourth quartile of exposure to BDEs 99 and 153, but not BDEs 47 and 100, had modestly increased free and total T4 levels relative to those in the first quartile of exposure. In addition, odds of hyperthyroidism were 5.7 times higher [95% confidence interval (CI), 0.9–36.4] in men with total PBDEs (∑PBDEs) above versus below the 90th percentile (0.78 ng/g serum). In the largest study to date (n = 623), found a positive relation between plasma BDE-153, but not BDE-47, and total triiodothyronine (T3) in Inuits; they also reported nonstatistically significant inverse relationships between BDEs 47 and 153 and TSH. Contrary to the above studies, found higher serum TSH in 23 highly exposed (median ∑PBDEs, 382 ng/g lipids) Chinese electronic-waste workers relative to 26 controls who also had elevated serum ∑PBDE concentrations (median, 158 ng/g lipids).

5Only one small study (n = 9) examined associations between serum PBDE and TH concentrations in women during pregnancy; no associations were found between ∑PBDEs and free or total T4 in serum samples collected shortly before delivery. measured PBDEs and TH in cord serum and reported that higher concentrations of BDE-100, but not BDEs 47 and 153, were associated with lower (< 20th percentile) total T4 [odds ratio (OR) = 2.1; 95% CI, 1.1–4.2] among women who had spontaneous unassisted vaginal deliveries only (n = 92). BDEs 100 and 47, but not BDE-153, were associated with a reduced likelihood of high (> 80th percentile) TSH levels (OR = 0.4; 95% CI, 0.2–0.8 for both chemicals) among the same women.

6We have previously reported a positive association between PCBs grouped according to their potential to induce uridine diphosphate glucuronosyltransferase (UDP-GT) in rodents and neonatal TSH and between total PCBs and free T4 in pregnant women. The purpose of the present investigation was to determine the relation between serum PBDE concentrations and thyroid function in the same population of pregnant low-income Latina women living in California.

7Data from the Center for the Health Assessment of Mothers and Children of Salinas (CHAMACOS), a birth cohort study of health and environmental exposures, were used for this study. Pregnant women who sought prenatal care at one of six participating health clinics between October 1999 and October 2000 and were < 20 weeks gestation, ≥ 18 years of age, eligible for state-sponsored health care (Medi-Cal), and intended to deliver at Natividad Medical Center (Monterey County, CA, USA) were enrolled in CHAMACOS (n = 601). Women were excluded from the present analyses if they did not participate through delivery or did not give birth to a live child (n = 64), bore twins (n = 5), refused to give a blood sample or gave a sample of insufficient volume for PBDE (n = 168) or TH (n = 69) analyses, or took medication that could affect TH levels (n = 1). In addition, data were not reported for 24 serum samples that failed to meet quality assurance standards for PBDE measurement. A total of 270 women were thus included in this analysis. Study participants provided written informed consent, and all research activities were approved by the University of California–Berkeley Committee for the Protection of Human Subjects.

8Participants were interviewed at enrollment (mean ± SD, 14.0 ± 5.0 weeks gestation) and at the end of the second trimester of pregnancy (26.5 ± 2.4 weeks gestation) using structured questionnaires. Information about demographics, country of birth, time lived in the United States, parity, and health-related behaviors was collected. Medical records were abstracted by a registered nurse to obtain data on thyroid-related diseases, medication use, and general health status.

9Blood samples were collected around the time of the second interview (mean ± SD, 27.3 ± 3.1 weeks’ gestation) and were immediately processed and stored at −80°C. PBDEs were measured in serum at the Centers for Disease Control and Prevention (Atlanta, GA, USA) using gas chromatography/isotope-dilution high-resolution mass spectrometry (GC-IDHRMS). PBDE concentrations are expressed on a blood lipid basis. Total lipids were determined based on the measurement of triglyceride and total cholesterol in serum using standard enzymatic methods (Roche Chemicals, Indianapolis, IN, USA). Limits of detection (LODs) ranged between 0.2 and 1.6 ng/g lipids except for BDE-47 (range, 0.8–5.6 ng/g lipids). Quality control samples were included in each run. THs were measured in serum by Quest Diagnostics’ Nichols Institute (San Juan Capistrano, CA, USA). Free T4 was measured by direct equilibrium dialysis followed by radioimmunoassay. Although results obtained by commonly used immunoassays may be affected by the blood concentration of T4-bound proteins, which increases during pregnancy, equilibrium dialysis physically separates the free from the bound hormone before measuring it with a highly sensitive immunoassay. This method yields accurate measurements in samples with normal and elevated T4-bound protein concentrations. Total T4 was measured by solid-phase immunochemiluminometric assay, and TSH by ultrasensitive third-generation immunochemiluminometric assay. LODs were 0.1 ng/dL, 0.1 μg/dL, and 0.01 mIU/L for free T4, total T4, and TSH, respectively.

10To control for other environmental exposures that may affect TH, we measured PCBs and organochlorine pesticides [including hexachlorobenzene (HCB), p,p′-dichlorodiphenyl trichloroethane (DDT), o,p′-DDT, p,p′-dichlorodiphenyl dichloroethylene (DDE), γ-hexachlorocyclohexane, dieldrin, mirex, and trans-nonachlor] in serum samples using GC-IDHRMS, and lead in a subset of maternal (n = 70) and umbilical cord (n = 161) blood samples using graphite furnace atomic absorption spectrophotometry. PCBs and organochlorine pesticides were expressed on a serum lipid basis.

11We log10-transformed the serum concentrations of PBDE congeners to reduce the effect of outliers. We used Pearson’s correlations to evaluate the interrelationship of PBDE congeners and analysis of variance (ANOVA) to examine associations between demographic characteristics and PBDE serum concentrations. We used multiple linear regression models to investigate the relationship between maternal PBDE and TH serum concentrations. Free and total T4 were normally distributed, whereas TSH was right-skewed and was thus log10-transformed to approximate a normal distribution. We ran separate models for each congener with a detection frequency > 50% (BDEs 28, 47, 99, 100, and 153) and for their sum. We ran models expressing exposure continuously and also categorically for each quartile of PBDE to investigate the possibility of a threshold for effect or other nonmonotonic exposure–response relationship. In addition, we fit generalized additive models with 3-degrees-of-freedom cubic splines to evaluate the shape of exposure–response curves and to test for digression from linearity while controlling for covariates. Using multiple logistic regression, we investigated associations between PBDE serum concentrations and maternal hyperthyroidism based on laboratory reference ranges for women in their second (TSH < 0.5 mIU/L) or third (TSH < 0.8 mIU/L) trimester of pregnancy.

12Potential confounders considered for inclusion in models (categorized as shown in Table 1 or expressed as indicated in parentheses) comprised maternal age (continuously), race/ethnicity, education, family income, country of birth, number of years spent in the United States, parity, body mass index, gestational age at the time of blood collection (in weeks, continuously), and smoking, alcohol, and drug consumption during pregnancy. We also considered environmental exposures such as blood lead, serum PCB [the sum of congeners with detection frequencies > 75% and of potential UDP-GT inducers], and organochlorine pesticide concentrations, and we examined the possibility of an interaction between PBDEs (continuously) and PCBs (continuously, and dichotomized at the 75th and 90th percentiles), which was previously reported in rats. All environmental exposures were log10-transformed. We had complete data on most covariates. Values of missing covariates were imputed at random based on observed probability distributions (< 2% missing) or on prediction models using nonmissing variables (≥ 2% missing). Covariates associated with any of the outcomes (p < 0.20) were included in all models. Final models comprised maternal age at enrollment, education, country of birth, gestational age at the time of blood collection, and family income as well as maternal HCB and total PCB serum concentrations.

13Signals below instruments’ LODs may yield better estimates of true concentrations than imputed values. When possible, we thus used values < LOD as measured by instruments. Undetected values were imputed based on a log-normal probability distribution whose parameters were estimated by maximum likelihood estimation. This procedure has been reported to perform better than simple substitution methods using LOD/2 or. Undetected TSH levels were assigned a value equal to half the LOD of 0.01 mIU/L.

14We conducted sensitivity analysis to evaluate the robustness of our results. We re-ran models excluding outliers with externally studentized residuals > 3. We also applied separate models expressing PBDEs on a total lipid basis, and on a serum basis while including triglycerides and total cholesterol as covariates in models, and by expressing ∑PBDEs on a weight and molar basis. In addition, we ran models with untransformed PBDE values and outliers excluded.

15Because findings were similar for all the models described above, we present only results from regressions with potential outliers included, log10-transformed exposures expressed on a serum lipid basis (nanograms per gram lipids), and ∑PBDEs as well as individual congeners expressed on a weight basis. Statistical significance was set at p < 0.05 for main effects and p < 0.10 for interactions based on two-sided tests. Statistical analyses were performed using STATA/IC (version 10.1; StataCorp LP, College Station, TX, USA) and R (version 2.7.1; R Foundation for Statistical Computing, Vienna, Austria).

16Participants were mostly young (mean ± SD, 25.5 ± 5.0 years), low income (62% ≤ federal poverty threshold), Latina (95%) women who had emigrated from Mexico within 10 years at the time of enrollment (75%; Table 1). Most women were multiparous (67%) and did not have a high school diploma (77%). Although a small percentage of women reported that they smoked (7%) or used illegal drugs (2%), almost a quarter (24%) consumed some alcohol during pregnancy, and most (61%) were overweight or obese before pregnancy.

17Table 2 shows the serum concentrations and detection frequencies of individual PBDE congeners and of the sum of congeners with detection frequencies > 50% (∑PBDEs). PBDE levels were lower than those reported in a nationally representative sample of nonpregnant adults. However, levels in CHAMACOS women were likely higher before pregnancy because changes in fat mass alter the serum concentration of persistent organic pollutants. As reported in previous studies, BDEs 47, 99, 100, and 153 had the highest detection frequencies (> 98% detection) and accounted for virtually all ∑PBDEs. BDE-47 contributed more than half of ∑PBDEs, followed by BDEs 99, 153, and 100. We detected BDEs 17, 66, 85, 154, and 183 in < 50% of samples and did not consider them in further analyses. PBDE congeners were moderately to strongly intercorrelated (r = 0.6–0.9, p < 0.001); BDEs 47, 99, and 100, which are the main components of penta-BDE commercial mixtures, were strongly intercorrelated [r > 0.9, p < 0.001; see Supplemental Material, Table 2 (doi:10.1289/ehp.1001905)].

18Mean (± SD) serum concentrations of free and total T4 were 0.83 ± 0.24 ng/dL and 10.7 ± 1.6 μg/dL, respectively; the geometric mean for TSH was 1.2 (geometric SD = 1.7) mIU/L. Six participants had low free T4 (< 0.5 ng/dL), 10 had low total T4 (< 8.0 μg/dL), and one had high TSH concentrations (> 4.6 and > 5.2 mIU/L in second and third trimesters, respectively) based on laboratory reference ranges; TSH was elevated in 14 women based on National Academy of Clinical Biochemistry guidelines (> 2.5 mIU/L). TSH was suppressed (< 0.01 mIU/L) in two women and was low (< 0.5 and < 0.8 mIU/L in second and third trimesters, respectively) in 35 women; four women had high free T4 (> 1.6 ng/dL), whereas none had elevated total T4 (> 17.8 and > 20.1 μg/dL in second and third trimesters, respectively). Free and total T4 decreased with age in a linear fashion (r = −0.21 and −0.25, respectively; p < 0.001); TSH was not associated with age. Other demographic characteristics were not significantly associated with TH levels (data not shown).

19Table 3 shows that none of the PBDE congeners were significantly associated with free or total T4 concentrations. Despite small coefficients of determination (R2), all PBDE congeners were significantly inversely associated with TSH. Associations ranged between a 10.9% (95% CI, −20.6 to 0.0%) and a 18.7% (95% CI, −29.2 to −4.5%) decrease in TSH for every 10-fold increase in the serum concentration of individual congeners (computed from Table 3). A 10-fold increase in ∑PBDE was associated with a 16.8% (95% CI, −27.6 to −2.3%) decrease in TSH, corresponding to a 37.7% decrement over the full range of ∑PBDEs. Furthermore, although tests for digression from linearity were not statistically significant after the exclusion of the two participants with suppressed TSH (data not shown), categorizing PBDEs into quartiles provided some evidence suggestive of nonmonotonic exposure–response relationships (Figure 1).

20Given these results, we also investigated associations between PBDE serum concentrations and maternal hyperthyroidism using laboratory reference ranges for women in their second or third trimester of pregnancy. Clinical hyperthyroidism is characterized by depressed TSH and elevated free T4. Except in one case, all women with low TSH in this population had normal free T4 levels, corresponding to the definition of subclinical hyperthyroidism. Odds of subclinical hyperthyroidism were nonsignificantly increased 1.9 times (95% CI, 0.8–4.5) for each 10-fold increase in ∑PBDEs and ranged between 1.6 (95% CI, 0.7–3.7) for BDEs 99 and 2.4 (95% CI, 0.9–6.1) for BDE-153 (Table 4). Women in the highest quartile of ∑PBDEs and BDEs 100 and 153 had significantly increased odds of subclinical hyperthyroidism relative to women in the first quartile.

21The serum concentration of other chemicals did not confound associations, and in contrast to results observed in rats, we found no effect modification by PCBs (medians = 65.4 ng/g lipids for ∑PCBs and 19.6 ng/g lipids for enzyme inducers; data not shown).

22We report significant inverse associations between TSH concentrations and serum measurements of ∑PBDEs and BDEs 28, 47, 99, 100, and 153 in pregnant women. The odds of subclinical hyperthyroidism were also elevated in relation to ∑PBDEs and BDEs 100 and 153. Associations appeared to be primarily due to a decrease in TSH in participants in the highest quartile of PBDE serum concentrations. Relationships between ∑PBDEs and individual PBDE congeners and free T4 were generally null, and associations with total T4 were mostly inverse, but none were statistically significant.

23This is the largest study to investigate associations between PBDEs and TH serum concentrations in pregnant women. Only one small study (n = 9) previously examined the question and found no association between the sum of BDEs 47, 99, 100, 153, 154, and 183 and free or total T4 but did not measure TSH. Contrary to most studies conducted in nonpregnant adults, we did not find positive trends between PBDE exposure and free T4. Although this discrepancy may be explained in part by differences in the methods used to measure free T4 by prior studies (immunoassays) relative to the present study (direct equilibrium dialysis), elevated free T4 suggests that exposure to PBDE may have a hyperthyroidic effect, which is consistent with our results of decreased TSH. For the most part, previous studies of nonpregnant adults did suggest reduced TSH serum concentrations in relation to higher PBDE exposure. Lending support to our results, a recent study found that men with ∑PBDEs > 95th percentile (193 ng/g lipids) had substantially increased odds of having detectable serum thyroglobulin antibodies (OR = 6.1; 95% CI, 1.9–19.2), which are found in 80% of Graves disease patients. Graves disease is believed to be the major cause of hyperthyroidism during pregnancy, accounting for > 85% of cases, and may play a role in subclinical hyperthyroidism.

24It is unclear whether low maternal TSH affects fetal health because in vitro studies suggest that human placental permeability to TSH is limited. Only one study has investigated the relation between maternal subclinical hyperthyroidism and adverse pregnancy outcomes in humans. The authors found no increase in low birth weight, major malformations, or fetal, neonatal, or perinatal mortality in infants of 433 women with TSH levels ≤ 2.5th percentile for gestational age and nonelevated free T4 levels (≤ 1.75 ng/dL) relative to 23,124 women with normal TSH levels. Nevertheless, subclinical hyperthyroidism may lead to clinical hyperthyroidism, and hyperthyroidism during pregnancy has been linked with increased risks of miscarriage, premature birth, and intrauterine growth retardation. No studies have investigated the latent effects on subsequent child health or development., however, reported that brain neuronal and glial cell differentiation is affected in offspring of partially thyroidectomized rats rendered moderately hyperthyroidic by daily infusion of T4, suggesting that maternal hyperthyroidism may affect fetal neurodevelopment.

25There are no data regarding associations between subclinical hyperthyroidism during pregnancy and maternal health, although clinical hyperthyroidism has been related to preeclampsia. It is also unclear whether thyroid dysfunction during pregnancy is related to pre- or postpartum TH status. In the nonpregnant state, however, depressed TSH suggests that a woman’s free T4 and/or T3 is above her own individual set point, which can be indicative of mild thyroid failure. Studies conducted in nonpregnant adults report that subclinical hyperthyroidism may be associated with all-cause mortality, cardiovascular mortality, cardiac dysfunction, reduced bone mineral density, and increased fracture risk.

26The present study has a number of strengths. We used state-of-the-art methods to measure TH, including equilibrium dialysis for free T4 and an ultrasensitive third-generation assay with low LODs to measure TSH. We also had information on a large number of potential confounders, including demographic characteristics and environmental exposures to other endocrine disruptors such as lead, PCBs, and organochlorine pesticides. In addition, these results were unchanged after the exclusion of outliers and were robust to the lipid-adjustment method and to the summation method for ∑PBDEs (weight or molar basis).

27The strong correlation among PBDE congeners, however, hampered our ability to distinguish their independent association, and the cross-sectional nature of this study limits causal inference. Reverse causation, for instance, cannot be excluded because TH regulates a number of metabolic pathways, including lipid metabolism and the activity of some cytochrome P450 enzymes, which may alter PBDE serum concentrations. In addition, the mechanism of action for reduced TSH has not been clearly established. Possibly because of their structural similarity with T4 and T3, hydroxylated PBDEs (OH-PBDEs) have been shown to bind to thyroid receptors α1 and β and may thus inhibit the release of TSH by the pituitary. Exposure of human hepatocytes to BDE-99 in vitro has also been shown to up-regulate type I deiodinase, which is involved in the deiodination of T4 to T3 and reverse-T3. Elevated T3 would result in decreased TSH levels, but we did not measure T3 in this study because of limited sample volume. In addition, other chemicals have been shown to lower TSH through binding to the retinoid X receptor or interference with neuroendocrine signaling pathways, but few studies have investigated whether PBDEs act through these mechanisms.

28It is noteworthy that studies conducted in rodents generally reported a hypothyroxinemic effect of exposure to PBDEs whereas human studies suggest a hyperthyroidic effect. Discrepancies between human and animal studies may be due to the high doses used in animal studies and physiologic differences. For instance, OH-PBDEs have been shown to competitively bind to human transthyretin (TTR), possibly resulting in increased T4 clearance. Although TTR binds 75% of the circulating T4 in rats, it only binds 10–15% in humans, and thus effects of PBDEs through this mechanism may be stronger in rats than in humans. Animal studies have also reported that the PBDE commercial mixtures DE-71, DE-79, and Bromkal 70-5DE induce UDP-GT, which catalyzes the glucuronidation of T4, the rate-limiting step in T4 elimination. It is, however, unclear whether PBDEs induce UDP-GT in humans.

29We report an inverse association between TSH and ∑PBDEs and BDEs 28, 47, 99, 100, and 153 serum concentrations in pregnant women around the 27th week of gestation. Odds of subclinical hyperthyroidism were also elevated in association with increased exposure to some of these chemicals. We observed these findings in a population with median serum PBDE concentrations within the range of a nationally representative sample. Although maternal clinical hyperthyroidism has been associated with adverse pregnancy outcomes such as preeclampsia, premature births, and low birth weight, few data are available on the direct effects of maternal subclinical hyperthyroidism on fetal and child development. In future analyses, we thus intend to examine whether subclinical hyperthyroidism and maternal exposure to PBDEs are associated with these outcomes.