Hormone Disruption by PBDEs in Adult Male Sport Fish Consumers

Background Persistent pollutants, such as polychlorinated biphenyls (PCBs), affect endocrine function. Human exposure to polybrominated diphenyl ethers (PBDEs), which are similar in structure to PCBs, has increased recently, but health effects have not been well studied. Objectives Our goal in this study was to determine whether PBDE body burdens are related to thyroid and steroid hormone levels, thyroid antibodies, and thyroid disease in a cohort of frequent and infrequent adult male sport fish consumers. Methods We tested serum from 405 adult males for PBDE congeners, PCB congeners, testosterone, sex-hormone–binding globulin (SHBG), SHBG-bound testosterone, thyroglobulin antibodies, and the thyroid hormones thyroxine (T4), triiodothyronine (T3), thyroid-stimulating hormone (TSH), and T4-binding globulin (TBG). We collected data on demographics, fish consumption, medical diseases, and medication use. Results The median sum of PBDEs was 38 ng/g lipid. In 308 men without thyroid disease or diabetes, PBDEs were positively related to measures of T4 and reverse T3 and inversely related to total T3 and TSH. PBDEs were positively related to the percentage of T4 bound to albumin, and inversely related to the percentage of T4 bound to TBG. Associations of BDE congeners with hormones varied. BDE-47 was positively associated with testosterone levels. Participants with PBDEs over the 95th percentile were more likely to have thyroglobulin antibodies, although high PBDE exposure was not associated with thyroid disease. PBDE effects were independent of PCB exposure and sport fish consumption. Conclusions PBDE exposure, at levels comparable with those of the general U.S. population, was associated with increased thyroglobulin antibodies and increased T4 in adult males.

Polybrominated diphenyl ethers (PBDEs) are used as flame retardants in electronic equipment, home furnishings, textiles, and construction materials. They are similar to polychlorinated biphenyls (PCBs) in structure and in their persistence and bioaccumulative properties (Birnbaum and Staskal 2004). Over the last 20 years, PBDE levels have increased in human samples, whereas PCBs have declined (Schecter et al. 2005).
Because PBDEs are similar in structure to thyroxine (T 4 ) and triiodothyronine (T 3 ) (Hamers et al. 2006), concerns have been raised regarding their effect on thyroid function, which is regulated by the hypothalamopituitary-thyroid axis and influences development and gene expression in vertebrates (Zoeller et al. 2007). Reduction of circulating thyroid hormone is compensated for by release of thyroid-releasing hormone from the hypothalamus, which in turn increases thyroidstimulating hormone (TSH) release from the pituitary, ultimately stimulating thyroid hormone production. T 4 and T 3 are transported to peripheral tissues bound to proteins, primarily T 4 -binding globulin (TBG), but also to albumin and transthyretin (TTR). TBG production is stimulated by estrogen and inhibited by testosterone. T 4 is the major hormone produced by the thyroid. Some T 3 is produced directly by the thyroid, but most is derived from peripheral deiodination of T 4 . T 3 and T 4 are primarily metabolized by deiodination to diiodothyronine and reverse T 3 (rT 3 ), with some metabolism through glucuronidation, sulfonation, and other pathways. This complex system is vulnerable to disruption by a variety of chemicals through changes in hormone production, transport, and/or metabolism (Zoeller et al. 2007).
Biologic effects of PBDEs in rodents are similar to those of PCBs, with increased risks for reproductive and endocrine disruption (Ellis-Hutchings et al. 2006;Lilienthal et al. 2006;Stoker et al. 2004;Zhou et al. 2002), and neurodevelopmental problems (Kuriyama et al. 2005). In humans, PCBs have been associated with disruption of thyroid hormone homeostasis (Langer et al. 2007;Persky et al. 2001;Turyk et al. 2007), but the effects of PBDEs on thyroid hormones have been investigated only in a few smaller studies (Bloom et al. 2008;Hagmar et al. 2001;Julander et al. 2005;Yuan et al. 2008).
In 2001, we reported that PCBs were associated with lower levels of T 4 and free T 4 index in women and T 4 and sex-hormone-binding globulin (SHBG)-bound testosterone in men from a cohort of frequent and infrequent Great Lakes fish consumers (Persky et al. 2001). In 2003, we invited participants from the original cohort to participate in a follow-up study to explore potential mechanisms by which PBDEs, PCBs, and p,p´-diphenyldichloroethene (DDE) might be affecting thyroid hormone balance. In addition to the standard hormones (free and total T 4 and T 3 , as well as TSH), we explored via additional laboratory parameters specific mechanisms of action suggested by laboratory studies, such as changes in transport by serum-binding proteins (Hallgren et al. 2001;Hamers et al. 2006) and increase in thyroglobulin antibodies (Langer et al. 2007). In this study we explored the relationship of PBDE exposure with hormone homeostasis, thyroglobulin antibodies, and thyroid disease in men. Associations of thyroid hormones with PCB congeners and DDE will be reported separately.

Materials and Methods
Study participants. We invited a cohort of 4,206 frequent and infrequent consumers of Great Lakes fish established during the early 1990s (Hanrahan et al. 1999) to participate in a follow-up study. Information on fish consumption, medical diseases, and use of prescription and over-the-counter medications and vitamin supplements, and blood and urine samples were collected from 354 men during 2003-2004. Blood was collected into red-top tubes and allowed to clot for 20-30 min. Serum for exposure analyses was transferred to hexane-rinsed glass tubes and frozen. Serum and urine samples for hormone assays were frozen in polypropylene tubes. In addition, stored serum samples collected from 51 men in 2001-2003 were analyzed for hormone and exposure levels. The study protocol was approved by institutional review boards at the University of Wisconsin, Madison, and the University of Illinois at Chicago, and all subjects gave written informed consent before participation.
Exposure analyses. Serum samples were tested for PBDEs, PCBs, and DDE by the Wisconsin State Laboratory of Hygiene as previously described (Anderson et al. 2008). Briefly, sera were extracted with hexane/ethyl ether, with cleanup and fractionation using Florisil, silica gel, and concentrated sulfuric acid. PBDEs were analyzed by gas chromatographymass spectrometry (GC-MS), and PCBs and DDE by GC. Quality control was monitored by the use of method blanks, spikes of bovine serum, duplicates of bovine serum spikes or sample duplicates, surrogate spikes, and confirmation of the analytes by second column or GC-MS, as appropriate. Mean recoveries were 76-91% for 24 tri-to decaBDE congeners, 97% for DDE, and 81-94% for di-to hexaPCB congeners.
Hormone analyses. Hormone assays were performed on serum and urine samples at Northwestern University in R.C.'s laboratory. Total T 4 (serum and urine), total T 3 , and the free unbound concentrations of these thyroid hormones were measured by radioimmunoassay (Diagnostic Products Corporation, Inc., Los Angeles, CA). Specificity was > 99%. Interassay and intraassay coefficient of variations (CVs) were, respectively, 3.0% and 3.3% for total T 3 , 4.0% and 5.0% for total T 4 , 6.9% and 4.3% for free T 4 , 28.8% and 7.9% for free T 3 , and 5.6% and 14.9% for urinary total T 4 . We measured rT 3 in a competitive radioimmunoassay with a sensitivity of 7.0 ng/dL (ALPCO Diagnostics, Windham, NH). The antiserum used was highly specific: T 3 and T 4 cross-react by < 0.1%. Interassay and intraassay CVs were 13.4% and 5.5%, respectively.
We measured TSH and TBG in the Immulite System (Diagnostic Products). The TSH assay had a sensitivity of 0.002 µIU/mL and was highly specific, with < 0.1% crossreaction with other glycoprotein hormones. The TBG assay was also highly specific, with a sensitivity of 1.1 µg/mL. Interassay CVs were 14.7% for TSH and 9.7% for TBG.
We examined the distribution of T 4 binding in plasma by radioelectrophoresis (Borst et al. 1982;Leopold et al. 1987). We separated albumin-and TBG-bound 125 I-T 4 on agarose gels after incubation of 125 I-T 4 with the serum for 2 hr at 37°C. TTR, which we did not quantify in this analysis, is clearly separated from TBG in this system. The gels were stained with bromothymol blue to identify albumin in the samples. Standards of TBG and TTR were run in parallel to determine the location relative to albumin on the gel. The areas corresponding to TBG and albumin were cut out of the gel and counted in a gamma counter, and the percentage of the total 125 I-T 4 in each fraction was determined.
We measured urine creatinine spectrophotometrically by the Jaffe reaction after ethyl ether extraction. Interassay and intraassay CVs were 9.3% and 5.6%, respectively.
We measured testosterone in serum using a coated tube assay that employs 125 I-labeled testosterone as the tracer (Diagnostic Systems Laboratories, Webster, TX). The antiserum cross-reacted < 0.9% with androstenedione and androstenediol and 5.8% with dihydrotestosterone. Interassay and intraassay CVs were 17.0% and 6.6%, respectively. We measured SHBG using a competitive radioimmunoassay with a sensitivity of 5 nmol/L (Diagnostic Systems Laboratories). The interassay and intraassay CVs were 15.7% and 6.6%, respectively. SHBG-bound testosterone was determined as described by Bonfrer et al. (1989). We equilibrated a 0.2-mL volume of serum diluted 1/8 with buffer with 3 H-estradiol overnight at 4°C. A 0.10 mL suspension of a conconavalin-A (Con-A) Sepharose conjugate was added to the serum. SHBG was allowed to bind to the Con-A during a 30-min incubation period at room temperature. Testosterone in the serum maintains its equilibrium concentration with SHBG in the presence of endogenous factors such as other androgens, estrogens, and free fatty acids (Bonfrer et al. 1989;Street et al. 1989). Separation of unbound 3 H-testosterone from that bound to the Sepharose Con-A was achieved by centrifugation at 0°C to minimize dissociation of bound estradiol. The interassay and intraassay CVs were 4.5% and 3.6%, respectively.
We explored associations of thyroglobulin antibodies and thyroid disease with ΣPBDEs greater than the 90th or 95th percentiles in the full cohort of 405 men using logistic regression models, with adjustment for age.
Participants were excluded from the hormone analyses if they reported medical conditions or medication use known to affect thyroid hormone levels (Meier and Burger 2005). Complete data for exposure and hormone measures were available for 308 men for the hormone analysis after excluding participants missing data for lipids (n = 12); using thyroid hormones or having thyroid disease (n = 21); using blood-glucose-regulating medications or having diabetes (n = 60); using other hormones (n = 11; testosterone, systemic corticosteroids, melatonin, human growth hormone); or using other medications known to affect thyroid hormones (n = 4; dilantin, tegretol, lithium, carbodopa).
Associations of hormones with ΣPBDEs and BDE-47 were modeled using linear regression, and Pearson's partial correlation coefficients for associations of hormones with exposures were estimated with the same variables used in the linear regression models. We considered age, body mass index (BMI), and serum lipids to be important covariates and included them in all multivariate models. Other potential confounding variables were added individually to these models to determine if their inclusion affected the conclusion about the significance of the PBDE/hormone association (p < 0.05 or p > 0.05). Factors that were evaluated as potential confounders included smoking, alcohol use, medication use (antilipids, beta blockers, furosamide), Great Lakes sport fish meals in the past year, sport fish meals in the past year, ΣPCBs, DDE, years consuming sport fish meals, years consuming Great Lakes sport fish meals, and HA1c level. We also considered measured levels of testosterone, SHBG, and SHBGbound testosterone as potential confounders for thyroid hormone analyses.
We examined modification of the effect of ΣPBDEs on hormones by other exposure covariates (all potential confounding variables noted above) in linear regression models that included multiplicative interaction terms for ΣPBDEs and the potential effect modifier, adjusting for age, BMI, and lipids. We did not evaluate covariates identified as effect modifiers (p < 0.05 for interaction term) as potential confounders, but we stratified models of the effects of ΣPBDEs on hormones by above and below median levels of the effect modifier.
To determine if results were affected by extreme hormone values, we estimated models after exclusion of participants with values more than three interquartile ranges above the 75th percentile or below the 25th percentile for hormone measures. Models were also reestimated using a variable for ΣPBDEs where congeners below the LOD were imputed as 0, but this did not affect our findings.
We designed this study to explore associations of PBDEs with standard thyroid hormone parameters, free and total T 4 and T 3 , as well as TSH, and also with additional laboratory parameters to test specific mechanisms of action. We explored patterns in the associations of PBDEs with thyroid hormones regarding congener-specific associations and independence of associations to examine mechanistic hypotheses.
Because an effect of PBDEs at one point in thyroid homeostasis could potentially change other related thyroid hormone parameters, we examined the independence of significant associations of thyroid hormones with ΣPBDEs regarding other measured thyroid hormones. When we identified a significant association between ΣPBDEs and a thyroid hormone, we further adjusted the linear regression model for other thyroid hormone levels individually. When the β-coefficient for ΣPBDEs changed by > 20% after adjustment for a second hormone, this suggested that the effect of ΣPBDEs on the original hormone may be related to or mediated by the second hormone.

Results
Characteristics of the cohort included in the hormone analysis are shown in Table 1. Most men drank alcohol at least once a month (78%), but few smoked cigarettes (11%), and medication use varied, with 33% using antilipidemics, 17% using beta blockers, and 3% using furosamide (data not shown). Levels of ΣPBDEs in the men were similar to those found for a large sample representative of the U.S. population of similar age and ethnicity: in this study, geometric mean = 27 ng/g lipid [95% confidence interval (CI), 24-30 ng/g lipid]; in the National Health and Nutrition Examination Survey, geometric mean = 34 ng/g lipid (95% CI, 27-43 ng/g lipid) (Anderson et al. 2008). However, levels of ΣPCBs were somewhat higher in the present study than in the National Health and Nutrition Survey (Anderson et al. 2008). Because we excluded men with thyroid disease, thyroid hormone levels were predominantly within normal ranges (Table 1).
ΣPBDEs was significantly and positively associated with several thyroid hormones, including total T 4 , free T 4 , urinary T 4 , rT 3 , and albumin-bound T 4 (only after exclusion of two extreme outliers) and was negatively associated with TSH, but only after control for sport fish consumption (Table 2). We found generally similar associations for these thyroid hormones with BDE-47, the dominant BDE congener (Table 2). Figure 1 shows dose-response models for quartiles of ΣPBDEs. We saw the strongest dose response for urinary T 4 , whereas only the highest ΣPBDE quartile was elevated for free T 4 and rT 3 . Total T 3 , which was not significantly associated in the continuous analysis (Table 2), was significantly negatively associated with ΣPBDE quartiles. On the other hand, total T 4 and TSH were not significantly associated with ΣPBDEs in the ordinal dose-response models. The effect of ΣPBDEs on T 4 binding to serum proteins was limited to the highest exposure quartile.
Urinary T 4 was the only hormone associated with all four BDE congeners (Table 3). rT 3 , total T 4 , and free T 4 were positively associated with BDE-99 and BDE-153, total T 3 was negatively associated with BDE-47 and BDE-153, and free T 3 was negatively associated with BDE-153. BDE-100 was negatively associated with TBG-bound T 4 and positively associated with albumin-bound T 4 , with similar associations for BDE-153, but only in the highest tertile (Table 3).
We found significant associations among many of the thyroid hormone measurements (data not shown). Because an effect of PBDEs at one point in thyroid homeostasis could potentially change other related thyroid hormone parameters, we examined the independence of significant associations of thyroid hormones with ΣPBDEs regarding other measured thyroid hormones (Table 4). The associations of urinary T 4 and albumin-bound T 4 with ΣPBDEs were independent of other thyroid hormones (Table 4). However, associations of ΣPBDEs with rT 3 , free T 4 , total T 4 ,

PBDEs and hormones in adult males
Environmental Health Perspectives • VOLUME 116 | NUMBER 12 | December 2008 total T 3 , and TSH were independent of many but not all other thyroid hormone levels.
The effect of PBDEs on rT 3 , free T 4 , TBGbound T 4 , and albumin-bound T 4 was significantly modified by HA1c levels, with stronger associations in persons with higher HA1c levels (data not shown). Effects of PBDEs on rT 3 and albumin-bound T 4 were stronger among infrequent sport fish consumers, whereas effects on TSH were stronger among frequent consumers (data not shown). We found no evidence of effect modification by PCBs or DDE, years of sport fish consumption, medication use, age, BMI, serum lipids, smoking, alcohol use, or steroid hormone levels.
Testosterone, SHBG, and SHBG-bound testosterone were not associated with ΣPBDEs or BDE-47 as a continuous variable (Table 2), but the ordinal variable for BDE-47 tertiles was positively associated with testosterone (Table 3).
We studied thyroid disease and thyroglobulin antibodies in the entire cohort of 405 men. Thyroglobulin antibodies were present in 7.8% of the full cohort and in 31.3% of those whose ΣPBDEs exceeded the 95th percentile [odds ratio (OR) = 6.1; Table 5]. High PBDE exposure was not significantly associated with a diagnosis of thyroid disease (Table 5).

Discussion
Exposure to PBDEs at levels comparable with those in the general U.S. population was associated with thyroid and steroid hormone levels in adult men without thyroid disease or diabetes. PBDEs were positively related to measures of T 4 (total T 4 , free T 4 , urinary T 4 ) and rT 3 , and inversely related to total T 3 and TSH. PBDEs were positively related to the percentage of T 4 bound to albumin and inversely related to the percentage of T 4 bound to TBG. Associations of BDE congeners with thyroid hormones varied. BDE-47 was positively associated with testosterone levels.
Our finding of increased thyroglobulin antibodies in 31% of participants with the highest PBDE body burdens is potentially of biologic significance because thyroglobulin antibodies are found in 80-90% of patients with chronic autoimmune thyroiditis and 50-60% of patients with Grave's disease (Marcocci and Marino 2005). The 8% prevalence of antibodies in the entire cohort is similar to rates seen in normal adult male populations (Hollowell et al. 2002). Exposure to PCBs, which are similar in structure to PBDEs, has been associated with increased antithyroperoxidase antibodies (Langer et al. 2007). The small number of cases of hypo-and hyperthyroid disease limit our ability to draw conclusions on effects of PBDEs on thyroid disease, but the thyroglobulin results may indicate an increased susceptibility to autoimmune thyroiditis in PBDE-exposed persons.  a Adjusted for age, BMI, and serum lipids. Unless otherwise noted, significance of adjusted models did not change with further adjustment for the following covariates (added individually to model): smoking, alcohol use, antilipid medications, beta blocker medications, furosamide medication, Great Lakes sport fish meals in the past year, sport fish meals in the past year, ln ΣPCBs, ln DDE, years consuming sport fish meals, years consuming Great Lakes sport fish meals, and HA1c level. Unless otherwise noted, significance of adjusted models for thyroid hormones did not change with further adjustment for testosterone level, ln SHBG level, and SHBG-bound testosterone level. b Significant with further adjustment for Great Lakes fish meals or sport fish meals in the last year (r = -0.12, p = 0.04). c Borderline significant (0.05 < p < 0.10) with further adjustment for HA1c, ln SHBG, SHBG-bound testosterone, or ln DDE. d Borderline significant (0.05 < p < 0.10) with further adjustment for alcohol consumption, Great Lakes fish meals in the last year, ln SHBG, or SHBG-bound testosterone. e Adjusted for urinary creatinine. f Significant with further adjustment for testosterone (r = -0.12, p = 0.05). g Significant after exclusion of two extreme outliers for T 4 -bound albumin (r = 0.13, p = 0.03).

Figure 1. β-Coefficients and 95%
CIs from regression models for associations of individual ΣPBDE quartiles with hormone levels. p-Values are from regression models for associations of ordinal ΣPBDE quartile variables with hormone levels. All models adjusted for age, BMI, and serum lipids. Urinary total T 4 was also adjusted for urinary creatinine. To our knowledge, epidemiologic data on the effects of PBDEs on thyroid hormones in adults is limited to four published studies. First, a longitudinal study of 11 electronic recycling employees found no significant associations of BDE congeners with TSH, total T 3 , or free T 4 , but did note nonsignificant trends for increasing free T 4 with BDEs 28, 153, and 183 (Julander et al. 2005). Second, free T 4 and TSH were not significantly associated with PBDEs in 36 New York anglers, although the associations of BDE congeners with free T 4 were consistently positive (Bloom et al. 2008), and the authors estimated that 318 persons would be required to reach significance for the association of ΣPBDEs with free T 4 . Third, Hagmar et al. (2001) found a significant negative association of BDE-47 with TSH but no significant association with free and total T 3 and T 4 in 110 men with high consumption of fish from the Baltic Sea. Our results are consistent with the decreased TSH in Hagmar et al.'s study and with the positive direction of the free T 4 associations of Bloom et al. (2008) and Julander et al. (2005). Fourth, Yuan et al. (2008) found higher TSH levels in electronic waste workers compared with unexposed persons, but PBDE exposures levels were substantially higher in that study than in our fish consumer cohort.
Our findings of a positive association of PBDEs with T 4 and free T 4 are not, however, consistent with results of laboratory animal studies. In rats and mice, PBDE mixtures and BDE-47 have been shown to decrease T 4 and free T 4 (Hallgren et al. 2001;Hallgren and Darnerud 2002;Stoker et al. 2004;van der Ven et al. 2008;Zhou et al. 2001Zhou et al. , 2002. T 3 was also decreased in some studies, but to a lesser extent than total T 4 (Zhou et al. 2001), and TSH was not affected, except in a 31-day exposure in male rats that had decreased TSH (Stoker et al. 2004). It is not clear why our results are inconsistent with decreased T 4 found in PBDE-exposed laboratory animals. Thyroid hormone regulation is similar in vertebrates, but some functions differ by species. For example, more T 3 is produced by the thyroid in rats than in that of humans (40% vs. 20%), increasing the importance of deiodinases in controlling T 3 levels in humans. In addition, TTR is the dominant binding protein in rats, whereas most thyroid hormone circulates bound to TBG in humans. Rats are more sensitive to effects of PBDEs on thyroid hormones than are mice (Hallgren et al. 2001). Inconsistencies could also be related to generally higher exposure levels in animals, younger life stage at exposure, and congener-specific effects. Mice exposed to BDE-209 had decreased T 3 but not T 4 (Tseng et al. 2008). Finally, there may be substantial differences in the effects of acute versus chronic exposure.
A major strength of our study is the measurement of specific hormones and BDE congeners, which may offer insights into potential biological pathways. The analysis of the independence of associations between thyroid hormones and PBDEs regarding other measured thyroid hormones suggests independent pathways for PBDE effects on urinary T 4 levels and T 4 serum protein binding proportions, whereas changes in rT 3 , total T 4 , free T 4 , total T 3 , and TSH were interrelated. BDE-congener-specific relationships also support different pathways, with associations of BDEs 47, 99, 100, and 153 with urinary T 4 , BDEs 100 and 153 with T 4 serum protein binding proportions, and BDEs 99 and 153 with rT 3 , total T 4 , and free T 4 .
The association of PBDEs with rT 3 suggests that PBDEs may affect thyroid hormone deiodinases. Deiodinases play a key role in control of cellular levels of T 3 (Bianco and Kim 2006). D 2 deiodinase removes iodide from outer ring of thyroid hormones (meta position), converting T 4 to T 3 , whereas D 3 deiodinase removes an iodide from the inner ring (ortho position) converting T 4 to rT 3 . D 1 can remove iodide from the inner and outer Unless otherwise noted, significance of adjusted models did not change with further adjustment for the following covariates (added individually to model): smoking, alcohol use, antilipid medications, beta blocker medications, furosamide medication, Great Lakes sport fish meals in the past year, sport fish meals in the past year, ln ΣPCBs, ln DDE, years consuming sport fish meals, years consuming Great Lakes sport fish meals, and HA1c level. Unless otherwise noted, significance of adjusted models for thyroid hormones did not change with further adjustment for testosterone level, ln SHBG level, and SHBG-bound testosterone level. b p-Value from linear regression model for ordinal BDE tertile variable indicating trend over BDE tertiles. c Borderline significant (0.05 < p < 0.10) with further adjustment for years consuming sport fish or years consuming Great Lakes sport fish. d Borderline significant (0.05 < p < 0.10) with further adjustment for smoking, alcohol use, or years consuming Great Lakes sport fish. e Significant with exclusion of extreme hormone outliers or with further adjustment for smoking. f Borderline significant (0.05 < p < 0.10) with further adjustment for HA1c, ln SHBG, or SHBGbound testosterone. g Significant with further adjustment for years consuming sport fish or Great Lakes sport fish. h Borderline significant (0.05 < p < 0.10) with further adjustment for HA1c, ln SHBG, SHBG-bound testosterone, smoking, alcohol use, or Great Lake sport fish meals in last year. i Adjusted for urinary creatinine. j Significant with further adjustment for years consuming sport fish or beta blocker use. k Significant with further adjustment for testosterone levels. l Borderline significant (0.05 < p < 0.10) with further adjustment for HA1c or alcohol use. *Individual tertile 0.05 < p < 0.10. **Individual tertile p < 0.05.
rings of thyroid hormones. Changes in deiodinase activity can affect circulating hormone levels, as demonstrated by studies of mice carrying deletion mutations. For example, mice carrying deletion mutations in D 2 have elevated T 4 and TSH but no changes in T 3 (Schneider et al. 2001), those with D 1 mutations have elevated T 4 and rT 3 but no change in T 3 and TSH (Schneider et al. 2006), whereas those with D 3 mutations are hypothyroid with decreased T 4 and T 3 but no change in TSH (Hernandez et al. 2006). These studies suggest that inhibition of outer ring deiodinases, most likely D 1 , by PBDEs could account for the increased T 4 and rT 3 in our participants with higher exposures. A possible mechanism is competitive inhibition of outer ring deiodinase by BDEs. Evidence that outer ring deiodinases in fish may debrominate BDE-99 to BDE-47 by removal of a bromine from the meta position (Benedict et al. 2007), as well as our finding that BDEs 99 and 153, both of which have a bromine in the meta position, were positively associated with rT 3 , free T 4 , and total T 4 , supports this hypothesis. The negative relationship of PBDEs with TSH might be a normal feedback response to elevated T 4 levels. Decreased production of total T 3 could also be a consequence of decreased outer ring deiodinase activity, although mice with outer ring deiodinase deletion mutations did not have abnormal T 3 levels.
The strongest PBDE association we observed was related to urinary total T 4 levels. Urinary total T 4 levels are not routinely assessed clinically. The increase we found, however, is consistent with the noted increases in serum free T 4 and albumin-bound T 4 , although the association of urinary T 4 with PBDEs was independent of serum free T 4 and albumin-bound T 4 . Future studies could assess effects of PBDEs on urinary thyroid hormone metabolites as a potential mechanism.
The associations of PBDEs with T 4 serum protein binding proportions suggest that PBDEs could be displacing T 4 from TBG. Hydroxylated BDE metabolites were able to bind to TTR in vitro (Hamers et al. 2006), and TTR in serum from BDE-47-treated rats showed decreased binding to 125 I-T 4 serum compared with serum from untreated rats (Hallgren and Darnerud 2002). However, to our knowledge, the potential for BDE congeners and metabolites to compete with T 4 binding to TBG has not been tested.
Steroid hormones can affect thyroid hormones through changes in TBG production. We did not find that testosterone levels modified the effects of PBDEs on thyroid hormones, but testosterone and SHBG did confound several associations of PBDEs and thyroid hormones. In addition, we found a positive association of testosterone with BDE-47. Hagmar et al. (2001) did not find associations of BDE-47 with free testosterone, follicle-stimulating hormone, luteinizing hormone, or prolactin in men. In male rats, the onset of prenuptial separation was delayed and ventral prostate and seminal vesicle weights were decreased, but luteinizing hormone and testosterone were not changed by PBDE exposure (Stoker et al. 2004). However, Stoker et al. (2005) found increased luteinizing hormone and a trend for increased steroid concentrations in PBDE-exposed adult male rats, and Lilienthal et al. (2006) observed that testosterone was decreased in male pups prenatally exposed to BDE-99. BDE congeners, in particular BDE-100, are androgen antagonists in vitro (Hamers et al. 2006;Stoker et al. 2005).
Although we excluded persons with diabetes from the hormone analyses, our data suggest that the effects of PBDEs on rT 3 , free T 4 , and albumin-and TBG-bound T 4 are stronger in persons with higher HA1c levels, which could place persons with moderately increased blood glucose at higher risk of thyroid hormone disruption by PBDEs. rT 3 is increased by fasting, malnutrition, and poorly controlled diabetes. Alternatively, changes in thyroid hormones may affect blood glucose (Chidakel et al. 2005).
Our results also suggest that fish consumption may modify the effect of PBDEs on thyroid function. We saw a stronger effect of PBDEs on rT 3 and albumin-bound T 4 among infrequent consumers and stronger effects on TSH among frequent sport fish consumers. Furthermore, some associations of PBDEs with hormones were modified by consumption of sports fish, but not by PCB or DDE body burdens. These findings are consistent with an interaction between PBDEs and other contaminants in fish on thyroid hormones.
The strengths of the present study include the use of a large, well-defined cohort; assessment of multiple hormones; and consideration of other environmental exposures that can affect thyroid hormones. Our conclusions are limited by those of any cross-sectional investigation. Although our results are inconsistent with animal studies, they are consistent with several human studies. The associations we found were relatively weak, and the highest proportion of hormone variation explained by All linear regression were adjusted for age, BMI, and serum lipid; urinary T 4 was also adjusted for creatinine; and TSH models were also adjusted for Great Lakes fish meals. Extreme outliers for albumin-bound T 4 were excluded for albuminbound T 4 models. Ordinal variables for ΣPBDE quartiles were used in total T 3 models. b β-Coefficient and p-value for ΣPBDEs from linear regression model predicting hormone levels. *β-Coefficient change of > 20% with control for second hormone. PBDE was approximately 6% for urinary T 4 (r = 0.25). There were some inconsistencies between results of models with continuous and ordinal exposure variables, with effects seen only in highest exposure category for some hormones. This pattern might be related to the extremely skewed distribution of PBDEs in the study cohort (Table 1). In addition, some hormone parameters show inconsistencies between models of ΣPBDEs and individual BDE congeners, which may be explained by congener-specific effects, as supported by animal and in vitro data. Although PBDE levels are lower than PCB or DDE levels, PBDE body burdens are increasing (Schecter et al. 2005). Older adults, who have a high risk of thyroid disease, are more likely to have BDE-47 levels above the 95th percentile level of 291 ng/g lipid (Sjodin et al. 2008). In the present study, exposures were similar to those of the U.S. population (Anderson et al. 2008). With increasing PBDE body burdens, we found increases in T 4 , but decreases in T 3 and TSH. In addition, thyroglobulin antibodies were higher in men with the highest PBDE body burdens. This is the first large study to link PBDE exposure with changes in thyroid antibodies and thyroid hormone homeostasis in men.