Fire safety standards promulgated in California in the late 1970’s have led to the proliferation of chemical flame retardants (FRs) used in consumer products worldwide [1]. Brominated flame retardant compounds are used as additives in products at up to 20 % by weight of the product [2]. Precisely because they are used as additives, rather than covalently bound to a polymer matrix, these halogenated FRs can and do migrate from their source products into the indoor and outdoor environment [3]. In fact, the global use of one class of environmentally-persistent flame retardant chemicals over a 35-year period, polybrominated diphenyl ethers (PBDEs), has led to their near ubiquitous presence in animals, abiotic matrices and food, as well as in air, dust and surfaces of indoor environments [4–17].

PBDEs share a similar structure to polychlorinated biphenyls (PCBs), polybrominated biphenyls (PBBs), tetrabromobisphenol A (TBBPA), triclosan, and the halogenated-phenyl ring of thyroid hormone thyroxine (T₄), with the potential for 209 different congeners based on the number and position of the bromine within the aromatic rings. PBDEs were manufactured in three commercial products—PentaBDE, OctaBDE, DecaBDE—each of which is dominated by one or more congeners (BDEs 47 (2,2′4,4′-tetra-bromodiphenyl ether), 99 (2,2′4,4′5-penta-bromodiphenyl ether), 100 (2,2′,4,4′,6-penta-bromodiphenyl ether); BDE 183 (2,2′,3,4,4′,5′,6-hepta-bromodiphenly ether); BDE 209 (deca-bromodiphenyl ether), respectively). The use of PBDEs as flame retardants have been phased out or banned in some countries, and updated California fire safety standards no longer require PBDEs or other flame retardant chemicals to be used in furniture [18]. However, exposure is expected to continue for several decades because of the reservoir of these chemicals that exist in consumer products that have long durations of use (e.g. couches), and the environmental stability of PBDEs. The most recent data available from NHANES shows that a reduction in PBDEs in serum was not yet observed on a national level following the phase-out of several PBDEs in 2004 [19], although more current biomonitoring data suggest that levels are decreasing [20, 21].

We have known for over 20 years that some PBDEs are carcinogenic. A study by the National Toxicology Program reported in 1986 found that DecaBDE causes thyroid and liver tumors in rats and mice [22]. More recently, a NTP study on PBDEs found that DE71, the major commercial mixture of PentaBDE, is a two-species, two-sex carcinogen [23]. Epidemiological evidence demonstrates that PBDEs are also endocrine-disrupting compounds that interfere with thyroid hormone action [24], reproduction, and neurodevelopment [25–30]. Subtle changes in circulating thyroid hormone concentrations (e.g., free and total T₃ (triiodothyronine), T₄, TSH (thyroid stimulating hormone)) are associated with many adverse health effects. Thyroid hormones regulate the growth and differentiation of the brain and are therefore critical for neurodevelopment [31]; the developing fetus relies on maternal thyroid hormone until week 11 of gestation [32] before beginning to produce endogenous hormones, and doesn’t produce sufficient hormone for its own needs until approximately week 20 [33]. Further, because the thyroid is a modulator for many of the body’s organs, studies have linked thyroid conditions with other diseases like kidney [34], liver [35], and heart [36].

Thyroid disorders disproportionately impact women [37], and the rates of thyroid cancer are increasing globally, with women having higher rates of thyroid cancer than men, and older women having higher rates than younger women [38]. We know of only one study that evaluated PBDE-associated health effects in a representative U.S. population sample (NHANES) and those researchers found associations with diabetes and metabolic syndrome [39], but did not evaluate thyroid disease as an outcome. Further, we know of no research related to the potential sensitivity of menopausal women to thyroid disease based on PBDE concentration, despite the well-described role of estrogen on thyroid function [40] and the ability of PBDEs to exert estrogenic effects [41]. The objective of our research was to explore the relationship between PBDE exposures in women and thyroid disease using nationally-representative (U.S.) data from the NHANES.

For all PBDE congeners except for BDE 153, the odds ratios for women in the highest exposure category were higher when the analysis was restricted to postmenopausal women. For example, for BDE 47, when all women were included in the analysis, the odds of having a current thyroid problem was 1.5 compared to 2.1 for just postmenopausal women. For BDE 99, this difference was two-fold, with postmenopausal women in the highest exposure category having 3.6 times the odds of having a current thyroid problem (95 % CI: 1.1–11.8), compared to 1.5 times the odds observed for all women (95 % CI: 0.97–2.3). The analysis for a current thyroid problem post-menopause showed odds ratios that were less than unity for the 3rd quartile.

The NHANES dataset did not contain enough men with current thyroid disease for our models to converge in a men-only analysis. In an analysis with all men and women combined, we did not find significant associations between the exposure categories and current thyroid disease (Table 1). Similarly, we originally attempted to explore thyroid cancer specifically, but this NHANES cycle only had nine participants reporting thyroid cancers, precluding us from performing regression analysis.

Our study found a positive association between PBDEs in serum and thyroid disease in U.S. women, a finding that suggests that effects of PBDEs on thyroid hormones, documented extensively in toxicological and epidemiological studies, is leading to significant downstream impacts. For example, increased PBDE levels in animals, primarily in mice and rats studies, have been associated with altered T₄ and T₃ levels [46–50] suggesting thyroid dysregulation in vivo, with several proposed mechanisms of action [51–54]. Associations between thyroid hormone levels and PBDEs have also been observed for animals in the wild, including birds, fish, and polar bears [55–57]. In humans, there are many studies that describe similar findings of associations between PBDE concentrations in serum and thyroid hormone concentrations [21, 25, 26, 29, 58–61].

One possible mechanism accounting for the PBDE-induced reduction in serum T₄ is displacement of T₄ from the serum binding protein transthyretin (TTR) or thyroxine binding globulin (TBG) [54]—similar to what has been previously observed for polychlorinated biphenyls (PCBs) [62]. (Significantly, a study examining the effects of chemical mixtures observed a potentially synergistic effect on T₄ levels with co-exposures to BDE 47 and PCBs [47]). T₃ and T₄ travel in blood bound to plasma proteins; only 0.04 % of T₄ and 0.4 % of T₃ are unbound (or “free”), and consequently, available for entry and action in target tissues [32]. The tetra-brominated congeners, BDE 47 in particular, share a similar structure to T₄, with both having the diphenyl ether structure and four halogens (iodine for T₄; bromine for BDE47). A major difference is that T₄ has a hydroxyl group on one of the rings, positioned between the two halogens. However, PBDEs are hydroxylated in vivo and, when this occurs in the meta position, it yields 3-OH-BDE47, which also has the hydroxyl group positioned between two halogens [53].

Along with epidemiological and toxicological evidence showing perturbations in thyroid hormone concentrations, we are now beginning to understand plausible mechanisms by which this disruption is occurring; the similarity of structures of exogenous chemicals to endogenous hormones may be leading to competition for receptor binding sites, causing inhibition or amplification. For thyroid hormones, a study by Cao et al. [52] found that hydroxylated PBDEs bind to transthyretin (TTR) and thyroxine-binding globulin (TBG) at the same sites as the target hormone, T₄. Hamers et al. measured binding affinities for PBDEs and metabolites to TTR and reported that the hydroxylated metabolites bound 160–1600 times more avidly than the parent compounds [63]. Further, a study by Butt and Stapleton [51] found that hydroxylated PBDEs are potent inhibitors of thyroid hormone sulfation. Also relevant to our study that focused on menopausal status, hydroxylated PBDEs have been shown to compete with estrogen enzymes. A crystallographic analysis examining the binding affinity of 3-OH-BDE47 to a hormone enzyme, estrogen sulfotransferase, demonstrated that the first phenolic group positions the compound within the enzyme similarly to how 17βestradiol would situate [41]. Further, the authors demonstrated that this positioning creates additional hydrogen bonding via the hydroxyl group, also similar to 17βestradiol, while accommodating the two halogens.

Perhaps the most striking and unique finding in this study is that the odds of having a current thyroid problem associated with PBDEs are so much higher in postmenopausal women. One hypothesis is that this is related to the change in hormone concentrations in postmenopausal women and the affinity of PBDEs to binding sites for both estrogen and thyroid hormones. Menopause is initiated when the ovaries discontinue producing two hormones, estrogen and progesterone. In at least two ways, the strong binding affinity of OH-PBDEs with estrogen sulfotransferase can interfere with the change in estrogen levels that occur during menopause. First, with less estrogen in the system, there may be increased metabolism of PBDEs by this estrogen enzyme [41]. Second, the OH-PBDEs may be competing for these binding sites that are critical for the removal of circulating estrogen produced by other tissues, leading to higher than expected levels of circulating estrogen. This potential interference of the estrogen clearance pathway in the liver via estrogen sulfotransferase, in and of itself, would not explain the higher odds of thyroid problems. However, estrogen (and androgen) can increase the levels of serum thyroxine binding globulin (TBG), interfering with one of the three primary proteins responsible for thyroid hormone transport in serum. In addition to these potential impacts on thyroid hormone via estrogen-mediated pathways, PBDEs can also have a direct impact on thyroid hormones through serum binding proteins outside of the mechanism that involves estrogen; PBDEs show a strong affinity for both TBG and another serum binding protein, transthyretin, and therefore can have a more direct impact on circulating thyroid hormone levels [41, 52]. All of this suggests the disruption of thyroid signaling by PBDEs that may be enhanced by the altered estrogen levels during menopause.

Results from epidemiologic studies on the effects of PBDEs on thyroid hormones are consistent with the toxicological evidence, in that they suggest interference with thyroid hormone regulation. However, in some epidemiologic studies, estimates of PBDE exposure are positively associated with measures of thyroid function, and in others the opposite pattern is observed. For example, in a study of 297 infants, Herbstman et al. observed that cord blood serum PBDE levels were associated with reduced total and free T₄ levels [27]. Conversely, Turyk et al. (2008) conducted a study of 405 adult male sport fish consumer and observed a positive trend between free T₄ and PBDE serum levels [61]. A more recent study found a positive association between PBDEs in serum of pregnant women and serum total and free T₄ and T₃ [29].

A limitation of our study is the inability of this analysis to determine causality because the data are cross-sectional; thus, we do not have temporality captured in the data to know if the exposure preceded the outcome. Also, the standard approach to analysis of serum concentrations for lipophilic compounds is to adjust the concentrations based on lipid-content in the serum and report the results as ‘lipid-adjusted’. This approach, however, may have a critical limitation. Thyroid hormones affect lipid metabolism at several points—synthesis, mobilization and degradation [66]. If a person has thyroid disease not caused by PBDE exposure, and the disease increases lipid metabolism leading to lower lipids in the bloodstream, then adjusting PBDEs by this lower lipid concentration yields a higher lipid-adjusted PBDE concentration, creating the appearance of a positive association between thyroid disease and PBDEs. Therefore, it is possible that women who self-report current thyroid disease have elevated lipid-adjusted PBDE concentrations in serum as a result of having thyroid disease. Further, even if serum concentrations are not adjusted for lipid content, the concentration in serum is still impacted by lipid content in the bloodstream, so a simple solution cannot be using serum concentrations that are not lipid-adjusted (we performed an analysis using wet-weight concentrations and found similar results as when we used lipid-normalized concentrations; Additional file 1: Table S1). Considering these points, it is plausible then that the relationship of thyroid dysfunction to PBDEs could be a function of reverse causality, with PBDE concentrations, lipid-adjusted or not, simply an outcome of thyroid dysfunction and altered lipid metabolism.

Another potential limitation of this study is the lack of specificity in the NHANES questionnaire regarding thyroid problems. NHANES is a nationally representative health survey (U.S.) that, due to its breadth, only covers thyroid-related disease with three non-specific questions: 1) Has your doctor ever told you that you have a thyroid problem, 2) Do you still have a thyroid problem, and 3) Do you have thyroid cancer. Yet, ‘thyroid problem’ and ‘thyroid cancer’ encompass a wide range of specific illnesses (e.g., hyperthyroidism, hypothyroidism, nodules, thyroiditis, goiter; papillary thyroid cancer, follicular thyroid cancer, medullary thyroid cancer), each with their own potential etiology (e.g., iodine deficiency, Graves’ disease, Hashimoto’s disease, radiation, environmental chemicals). Despite the lack of specificity in the NHANES questions on the outcome variables, which would likely bias the results toward a null finding, we still observed strong and consistent associations between PBDEs in serum and thyroid problems. Last, the congeners available for analysis in NHANES are also from the same commercial product and are correlated in environmental and body burden samples [5, 42]. Therefore, we cannot rule out that one congener is driving the observed associations seen for several congeners.

Overall, the findings from this analysis of data reported in NHANES provide additional evidence of the impact of PBDEs on the thyroid. We also observed a stronger effect for postmenopausal women. Additional research is warranted to confirm these findings, particularly when the next cycle of NHANES data with PBDE serum concentrations is made available.

We found that women with the highest flame retardant concentrations in their blood were significantly more likely to have a thyroid problem, with stronger associations found for post-menopausal women. Specifically, women with lipid-adjusted serum concentrations of BDE 47, 99 and 100 in the highest quartile of exposure had higher odds of having a current thyroid problem compared to women with lower serum concentrations. These associations were much stronger in an analysis that was restricted to postmenopausal women.