Thyroid Hormones in Relation to Lead, Mercury, and Cadmium Exposure in the National Health and Nutrition Examination Survey, 2007–2008

Background: Heavy metals, such as lead (Pb), mercury (Hg), and cadmium (Cd), are known toxicants, but their associations with the thyroid axis have not been well quantified at U.S. background levels. Objectives: We investigated the relationships between thyroid hormones (total and free thyroxine [TT4 and FT4], total and free triiodothyronine [TT3 and FT3], thyroid-stimulating hormone [TSH], and thyroglobulin [Tg]) and levels of Pb, Hg, and Cd in blood and Cd in urine. Methods: We separately analyzed a sample of 1,109 adolescents (12–19 years of age) and a sample of 4,409 adults from the U.S. National Health and Nutrition Examination Survey (NHANES) 2007–2008. We estimated associations after adjusting for age, sex, race, urinary iodine, body mass index, and serum cotinine. Results: The geometric mean (GM) levels of blood Pb (BPb), total Hg, and Cd were 0.81 µg/dL, 0.47 µg/L, and 0.21 µg/L in adolescents and 1.43 µg/dL, 0.96 µg/L, and 0.38 µg/L in adults, respectively. The GMs of urinary Cd were 0.07 and 0.25 µg/g creatinine in adolescents and adults, respectively. No consistent pattern of metal and thyroid hormone associations was observed in adolescents. In adults, blood Hg was inversely related to TT4, TT3, and FT3 and urinary Cd was positively associated with TT4, TT3, FT3, and Tg, but there were no associations with Pb. Associations were relatively weak at an individual level, with about 1–4% change in thyroid hormones per interquartile range increase in Hg or Cd. Conclusions: Our analysis suggests an inverse association between Hg exposure and thyroid hormones, and a positive association between Cd exposure and thyroid hormones in adults.

Thyroid hormones (THs) play a critical role in the functions of nervous, reproductive, and cardiovascular systems in both children and adults (Danzi and Klein 2012;Williams 2008;Yazbeck and Sullivan 2012). The hypothalamus-pituitary-thyroid (HPT) axis regulates thyroid function through thyro tropin releasing hormone, thyroidstimulating hormone (TSH), and the THs [thyroxine (T 4 ) and triiodothyronine (T 3 )]. Circulating T 4 and T 3 are mostly bound to thyroxine binding globulin, transthyretin, and albumin; < 1% is unbound and biologically active. In peripheral tissues, T 4 is converted to T 3 by type 1 and type 2 deiodinases; T 3 in turn binds thyroid receptors α and β and initi ates target gene expression (Stathatos 2012). Disruption of TH synthesis, transport, deio dination, and metabolism can result in clinical or subclinical thyroid diseases (Cooper and Biondi 2012). Circulating TSH and THs, even at levels within reference ranges, are significantly associated with effects in neuro development (Ghassabian et al. 2011;Pop et al. 2003), blood pressure (Asvold et al. 2007), cholesterol, triglycerides, and insulin resistance (Roos et al. 2007).
Environmental chemicals might alter TH levels via several mechanisms, including dis ruption of iodine (I) transport, thyroid peroxi dase, TH binding proteins, hepatic catabolism, deiodinases, and receptor binding (Miller et al. 2009). Studies of human popu la tions have focused primarily on chemicals that are struc turally similar to T 4 , such as polychlorinated biphenyls (PCBs), polybrominated diphenyl ethers, and bisphenol A (BPA), with little attention on heavy metals (Boas et al. 2006;Pearce and Braverman 2009). Lead (Pb), mer cury (Hg), and cadmium (Cd) are known environmental toxicants, but only a few studies have examined associations with total and free T 4 (TT 4, FT 4 ), total and free T 3 (TT 3, FT 3 ), or TSH (Dundar et al. 2006;Jin et al. 2006;Lamb et al. 2008;Pearce and Braverman 2009;Robins et al. 1983;Schell et al. 2008).
Pb is known to have adverse neuro logical, hemato logical, renal, and gastro intestinal effects (Bellinger 2004;GurerOrhan et al. 2004); however, associations with THs have been inconsistent (Meeker et al. 2009). Pb exposure [mean 15 µg/dL blood Pb (BPb) level] was negatively correlated with trans thyretin levels in cerebro spinal fluid samples from human patients (Zheng et al. 2001). Previous studies of popu la tions with high exposure to Pb (indicated by BPb levels of > 20 µg/dL) suggested negative associations with circulating T 4 , FT 4 , or T 3 (López et al. 2000;Robins et al. 1983;Singh et al. 2000;Tuppurainen et al. 1988); however, associations were not evident in other similar study populations (Erfurth et al. 2001;Schumacher et al. 1998;Siegel et al. 1989). Fewer studies have investigated associations of BPb levels of < 10 µg/dL with THs. Dundar and colleagues reported a negative association between BPb and FT 4 levels in adolescents with mean BPb of 7 µg/dL (Dundar et al. 2006). A recent study (Meeker et al. 2009) has suggested an inverse association between BPb (median, 1.5 µg/dL) and TSH levels in men of the couples presenting at infertility clinics. Another study, in the lakeside communities of Quebec, Canada, found no association between BPb (median, 3.1 µg/dL) and THs in men, but identified a positive association with T 3 and an inverse association with TSH in females with median BPb of 1.7 µg/dL (Abdelouahab et al. 2008).
Hg has adverse effects on a variety of sys tems that vary with the level, length of expo sure, and time window of exposure (Tan et al. 2009). Proposed mechanisms of Hgrelated TH disruption involve selective binding to sulfhydryl (SH)containing ligands in the thy roid, reduced TSH production, and inhibition of deiodination (Soldin et al. 2008;Tan et al. 2009). FT 3 levels were reduced in associa tion with occupational exposure to Hg vapor among chloralkali plant workers (Barregard et al. 1994;Ellingsen et al. 2000). Studies of populations with environmental exposure, for example, from fish consumption and from dental amalgams, have had mixed findings (Abdelouahab et al. 2008;Meeker et al. 2009;Schell et al. 2008;Takser et al. 2005). A study in a Canadian lakeside community with exposure levels slightly higher than reported for the U.S. National Health and Nutrition Examination Survey (NHANES) [median total Hg, 2.25 µg/L in men and 1.50 µg/L in women compared with median total Hg, 0.8 µg/L in both men and women for NHANES 2003(Caldwell et al. 2009 suggested a positive association between Hg Background: Heavy metals, such as lead (Pb), mercury (Hg), and cadmium (Cd), are known toxicants, but their associations with the thyroid axis have not been well quantified at U.S. background levels. oBjectives: We investigated the relationships between thyroid hormones (total and free thyroxine [TT 4 and FT 4 ], total and free triiodothyronine [TT 3 and FT 3 ], thyroid-stimulating hormone [TSH], and thyroglobulin [Tg]) and levels of Pb, Hg, and Cd in blood and Cd in urine. Methods: We separately analyzed a sample of 1,109 adolescents (12-19 years of age) and a sample of 4,409 adults from the U.S. National Health and Nutrition Examination Survey (NHANES) 2007-2008. We estimated associations after adjusting for age, sex, race, urinary iodine, body mass index, and serum cotinine. results: The geometric mean (GM) levels of blood Pb (BPb), total Hg, and Cd were 0.81 µg/dL, 0.47 µg/L, and 0.21 µg/L in adolescents and 1.43 µg/dL, 0.96 µg/L, and 0.38 µg/L in adults, respectively. The GMs of urinary Cd were 0.07 and 0.25 µg/g creatinine in adolescents and adults, respectively. No consistent pattern of metal and thyroid hormone associations was observed in adolescents. In adults, blood Hg was inversely related to TT 4 , TT 3 , and FT 3 and urinary Cd was posi tively associated with TT 4 , TT 3 , FT 3 , and Tg, but there were no associations with Pb. Associations were relatively weak at an individual level, with about 1-4% change in thyroid hormones per interquartile range increase in Hg or Cd. conclusions: Our analysis suggests an inverse association between Hg exposure and thyroid hormones, and a positive association between Cd exposure and thyroid hormones in adults.  (Abdelouahab et al. 2008). Hg was not associated with TSH or FT 4 in 232 Akwesasne Mohawk adolescents with a geometric mean (GM) Hg level of 1 µg/L (Schell et al. 2008).
Cd affects the renal, skeletal, and respira tory systems and is classified as a Group 1 carcinogen by the International Agency for Research on Cancer (Järup and Åkesson 2009). Cd exposure in animal studies has been related to decreased serum TT 4 levels, and interference with deiodination has been sug gested as a possible mechanism (Hammouda et al. 2008;Mori et al. 2006). A Japanese study comparing residents of the Cdpolluted Kakehashi River basin with residents of a nonpolluted area reported lower FT 4 levels in exposed females but higher TT 3 levels in both sexes (Nishijo et al. 1994). Studies of neo nates and children with environmental expo sures have reported inconsistent results (Iijima et al. 2007;Maervoet et al. 2007;Osius et al. 1999). Blood Cd (median, 0.2 µg/L) was not associated with TSH in male infertility clinic patients (Meeker et al. 2009).
Many previous studies have had fairly small sample sizes or have been based on popu lations with occupational exposures that may not be relevant to the general public. In addi tion, many studies have measured blood Cd, which is a good biomarker for recent exposure, but urinary Cd is a better indicator of long term exposure (Järup and Åkesson 2009). In this study, we analyzed NHANES data from 2007-2008 to estimate associations of Pb, Hg, and Cd with TH levels in a large U.S. popula tion with background levels of exposure.

Methods
We used study subjects' data from NHANES 2007-2008 [National Center for Health Statistics (NCHS) 2009a] to examine the asso ciation between heavy metals and TH levels in the general population with environ mental exposure levels. NHANES is conducted in a nationally representative sample of the U.S. civilian population by the Centers for Disease Control and Prevention (CDC; Atlanta, GA). In 2007-2008, a sample of 10,149 subjects was included in this complex multistage, strati fied cluster survey. Of those participants, TH levels were measured in 6,260 subjects ≥ 12 years of age. We excluded subjects with no BPb, Hg, or Cd measures (n = 5), those who had been told by a doctor or health pro fessional that they have thyroid problems or were currently taking thyroid medications (n = 520) (Belin et al. 2004), and those cur rently pregnant or taking steroid hormones (i.e, estrogen, androgen) that might alter TH or thyroxine binding globulin levels (n = 317). The analytical sample for this analy sis was 5,418, including 1,009 adolescents (12-19 years of age) and 4,409 adults (20-80 years of age). After consideration of sam pling weights, this analytic sample represents 26,770,162 adolescents and 159,282,838 adults in the general U.S. population who had no reported thyroid diseases, thyroid medi cations, pregnancy, and sex steroid medica tions. The analysis was exempt from review by the University of Cincinnati Institutional Review Board, but each subject had provided written informed consent to participate in the NHANES study.
Heavy metals. In the NHANES 2007-2008 cycle, metal assays were conducted in whole blood or urine samples at the Division of Laboratory Sciences, National Center for Environmental Health of the CDC. BPb, total Hg, and Cd levels were determined by inductively coupled plasma mass spectrometry (ICPMS; CDC method no. ITB0001A) with modification from a published method (Nixon et al. 1999) with limits of detection (LOD) of 0.25 µg/dL for Pb, 0.33 µg/L for total Hg, and 0.2 µg/L for Cd (NCHS 2009d). Inorganic Hg in whole blood was measured using Flow Injection Mercury System Cold Vapor Atomic Absorption (PerkinElmer, Norwalk, CT), with an LOD of 0.35 µg/L. In the data set provided by the CDC (2009), levels < LOD were imputed as being the metalspecific LOD divided by the square root of 2 (Hornung and Reed 1990).
Only 6 participants had BPb levels < LOD. In the United States, organic Hg accounts for the majority of total blood Hg (Mahaffey et al. 2004). Therefore, if the total Hg level was < LOD (n = 884), we assumed that organic Hg was equal to the imputed total Hg level (0.2 µg/L). If the total Hg level was > LOD, we calculated organic Hg as the difference between total and inorganic Hg. In this data set, 4,062 subjects (75%) had inorganic Hg levels < LOD; therefore, we did not test for associations between inorganic Hg and TH levels. Blood Cd levels were < LOD in 1,282 subjects. In addition to whole blood samples, a onethirdsample subset of par ticipants in the NHANES 2007-2008 had urine samples tested for Cd (n = 1,767) using ICPMS. Among them, 106 had urine Cd levels < LOD (< 0.042 µg/L). We calcu lated creatinineadjusted urinary Cd levels to account for urine dilution.
TH levels. Serum TH and thyroid antibody levels were determined in the Department of Laboratory Medicine at the University of Washington (Seattle, WA) (NCHS 2009b). Access HYPERsensitive hTSH assay (Beckman Coulter Inc., Brea, CA) was used to assay TSH. Competitive binding immuno enzymatic assay was used to deter mine TT 4 , FT 4 , TT 3 , and FT 3 . In addition, NHANES 2007-2008 samples were assayed for thyro globulin (Tg), Tg antibody (TgAb), and thyroid peroxidase antibody (TPOAb) using immunoenzymatic assays.
Statistical analyses. In this study, we per formed separate analyses for adolescents and adults. In samples from both adolescent (12-19 years of age) and adult subjects (≥ 20 years of age), we first examined the association between heavy metals and TH levels using linear regres sion models. Because both the exposure and outcome variables were not normally distri bu ted, we used natural log transformation to analyze the data. We examined associations of BPb, blood total Hg, blood organic Hg, blood Cd, and urinary Cd with TT 4 , FT 4 , TT 3 , FT 3 , TSH, and Tg, using separate regression mod els for each exposure-outcome association. Second, we categorized exposures into quintiles and estimated differences in mean values for the 2nd, 3rd, 4th, and 5th quintiles compared with the first quintile. Third, we examined the proportion of subjects with high levels of TgAb (> 4 IU/mL) or TPOAb (> 9 IU/mL), an indicator of immunologic disturbance of thyroid tissue functions, based on NHANES laboratory method references (NCHS 2009c). Logistic regression was used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for metal exposure. Because this ana lytical sample was thyroid disease-free and medicationfree (based on selfreport) and the percentage of subjects with clinical and sub clinical hyper thyroidism or hypo thyroidism was < 2% according to NHANES reference values (TSH 0.34-5.6 µIU/mL) (NCHS 2009c), we did not examine hyper thyroidism or hypo thyroidism as dichotomous outcomes. Fourth, we did a subset analysis restricted to women of reproductive age (15-44 years of age, n = 1,095) to examine whether metal exposure levels in nonpregnant U.S. women had a discernible association with TH levels during reproductive age. Fifth, we summarized significant findings in adolescents and adults by calculating percentage change in TH lev els with an interquartile range (IQR) increase in metal exposure levels. Sixth and last, we examined associations between TH levels and multiple metal exposures by categorizing adult BPb, total Hg, and Cd by their corresponding medians (1.39 µg/dL, 0.88 µg/L, 0.33 µg/dL) and modeling the eight possible comparison groups, using the group with levels of all three metals < median as the reference group.
In the regression models, we adjusted for a priori covariates (Caldwell et al. 2009;Hollowell et al. 2002;Muntner et al. 2005;TellezPlaza et al. 2012): age (continuous), sex (male, female), race/ethnicity (white, black, Hispanic/other), natural log transformed cre atinine-adjusted urinary iodine (measured by ICPMS at the CDC), body mass index (BMI; age and sexspecific zscore in adolescent mod els, original value in adult models), and serum cotinine levels [measured by high performance liquid chromatography tandem mass spectrom etry at the CDC; < 1 ng/mL as non smoking, 1-9.9 ng/mL as environmental tobacco smoke (ETS) exposure, ≥ 10 ng/mL as active smoking, dummy variables used] (CDC 2009). Because NHANES is a complex multi stage sampling survey, we used PROC SURVEYREG and PROC SURVEYLOGISTIC in SAS version 9.2 (SAS Institute Inc., Cary, NC) to calcu late regression parameters and 95% CIs after accounting for sampling weights and survey methods. The significance level was set at α = 0.05 for twosided tests.

Results
In the adolescent subjects sample, the mean age was 15.5 years, with 55% male, 60% white, 15% black, and 25% Hispanic and other eth nicity. Twelve percent were exposed to ETS, and 15% were active smokers. The mean BMI zscore was 0.54. The GM of urinary iodine was 140 µg/g creatinine. For the adult sub jects sample, the mean age was 46.4 years, with 55% male, 68% white, 11% black, and 22% Hispanic and other ethnicity. Adults had 28% active smoking percentage, with 5% exposed to ETS. The mean adult BMI was 28.5 kg/m 2 .
The GM of adult urinary iodine was 156 µg/g creatinine. The covariates used in adjusted regression models, including age, sex, race and ethnicity, smoking status, BMI, urinary iodine, were associated with metal exposures and TH levels in most models (detailed data not shown). We also observed increased Pb and Hg levels with age, higher Cd levels in smokers versus nonsmokers and in females versus males, and lower TSH levels in smokers overall. Table 1 displays the means, ranges, and GMs for Pb, Hg, Cd, and TH levels in both adolescents and adults. Adults had statistically significant higher levels of metal exposures than adolescents.
Statistically significant negative associa tions between blood total Hg and TT 4 and FT 3 were observed in adolescents (Table 2). Blood Cd was positively associated with FT 3 and urinary Cd was positively associated with FT 4 . Table 3 shows that in adults, BPb expo sure was not associated with any TH levels. Both total and organic Hg had significant negative associations with TT 4 , TT 3 , and FT 3 . Blood Cd was positively associated with FT 4 and Tg; urinary Cd was positively associated with TT 4 , TT 3 , FT 3 , and Tg. In the analyses of quintiles of metal exposures and TH levels in adolescents, the dose responses were not evident [see Supplemental Material, Table S1 (http://dx.doi.org/10.1289/ehp.1205239)]. However, in adults, the dose-response pat terns consistent with the modeling of con tinuous exposure were evident for total Hg, organic Hg, and urinary Cd (see Supplemental Material, Table S2). Figure 1 shows that total blood Hg levels were associated with lower TT 4 and TT 3 , with the 5th quintile of expo sure (≥ 2.16 µg/dL) showing the strongest associations. Urinary Cd levels were positively associated with TT 4 and TT 3 . In contrast, no consistent patterns were found for BPb levels (see Supplemental Material, Table S2).
The percentages of high thyroid antibody levels were slightly higher in adults than adolescents (5.77% vs. 4.79% for TgAb; 8.83% vs. 6.04% for TPOAb, Table 4). However, in neither adolescents nor adults were blood or urinary Pb, Hg, Cd levels significantly associated with high TgAb or TPOAb levels.
Subset analyses of continuous exposures among women of reproductive age were gener ally consistent with associations in the adult population as a whole [see Supplemental   Adjusted for age, sex, race/ethnicity, creatinine-adjusted urinary iodine, BMI z-score, and serum cotinine level. *p < 0.05. Table 3. Adjusted regression coefficients (95% CIs) of blood and urinary Pb, Hg, and Cd in relation to THs in adults.  . Urinary Cd levels were posi tively related to TT 4 , but associations with other TH levels did not reach statistical significance. Table 5 gives the estimated percentage difference in TH levels per IQR increase in exposures that were significantly associated with TH levels when modeled as continuous variables. Overall estimated differences in mean levels were small, at 1-4%. However, we estimated a 12% increase in Tg associated with blood Cd at 0.61 vs. 0.21 µg/L, and an 18% increase in Tg associated with urinary Cd of 0.41 vs. 0.14 µg/g creatinine in adults.

Metal
In the threemetal analysis in adults, the nega tive association between total Hg and TT 4 and TT 3 was evident with and without exposures to Cd or Pb above median levels, and the positive association between blood Cd and Tg was evident for all combinations with exposure to Cd above the median, regardless of exposure to Hg or Pb [see Supplemental Material,

Discussion
In adults, Hg exposure was negatively associ ated with TH levels, whereas Cd exposure was positively associated with TH levels and the prehormone Tg. TSH levels were not con sistently associated with Hg or Cd exposure, suggesting that these exposures may not affect pituitary function. In women of reproduc tive age, the inverse associations between Hg and TH levels persisted, whereas the associa tions between Cd and TH levels were mostly positive but not statistically significant.
The lack of association between BPb and TH levels suggests current exposure levels experienced by the U.S. population do not adversely affect TH synthesis and regulation, though effects at higher environmental expo sure levels cannot be ruled out. Occupational Pb exposure has been associated with signifi cant reductions in TH levels (Robins et al. 1983), and a recent animal study also noted an effect of Pb on TH levels in rats (Wu et al. 2011).     Abbreviations: P 75 , 75th percentile; P 25 , 25th percentile. a Instead of first quartile, the interval starts from < LOD (value replaced with LOD divided by the square root of 2) to reflect > 25% participants with exposure < LOD.
Negative associations observed between Hg and TH levels are consistent with pro posed mechanisms for Hg toxicity in which Hg accumulates in the thyroid and reduces iodide uptake at the sodium/iodide symporter by binding to iodide (Nishida et al. 1986) and inhibits TH deiodinase function in peripheral tissues (Soldin et al. 2008;Tan et al. 2009). In the Abdelouahab et al. (2008) study, TSH was positively associated with hair and blood Hgs, but we did not find an association between TSH and blood Hg in our population. A recent analysis of NHANES 2007-2008 data suggested an increase in the prevalence of TgAb in women with blood Hg > 1.8 µg/L versus ≤ 0.4 µg/L (Gallagher and Meliker 2012), but that analysis did not exclude subjects that had thyroid disease or were taking medications to treat thyroid disease. We did not evaluate inorganic Hg, which has been associated with a higher FT 4 /FT 3 ratio in two occupational studies (Barregard et al. 1994;Ellingsen et al. 2000). Prior research has suggested that PCBs may influence TH levels, and effects of PCB could therefore confound associations with Hg because both may be consumed in fish (Hagmar 2003). However, we did not have data on PCB exposures.
Studies in experimental animals suggest lower TH levels in Cdexposed mice and rats Kar 1997, 1998;Hammouda et al. 2008;Yoshizuka et al. 1991), in contrast with our finding of positive associations between Cd and TH levels. This discrepancy could be due to species differences or to higher exposure doses being used in animal studies, although we cannot rule out the possibility that the associations we observed were due to chance or bias. TT 3 was increased in residents of a Cdpolluted region (GM urinary Cd levels, 6.6 and 9.2 µg/g creatinine in males and females, respectively) compared with TT 3 levels in res idents of a control region (GMs of 2.6 and 4.4 µg/g creatinine, respectively) (Nishijo et al. 1994). In the present study population, GM urinary Cd levels were an order of magnitude lower, but we observed higher TH levels with increased urinary Cd. Additional research is needed to clarify the potential effects of Cd exposure on thyroid function in humans.
The observed associations between Hg and Cd exposures and TH levels were rela tively weak on the individual level, with an IQR increase in exposure associated with a 1-4% change in hormone levels. The HPT axis is precisely regulated and it is plausible that the low environmental exposures expe rienced by the U.S. population do not sub stantially influence individual thyroid profiles. However, exposures at the very high end (e.g., ≥ 5th quintile compared with the 1st quintile) may be associated with a TH level change of > 5%, such as total Hg at ≥ 2.16 µg/L ver sus < 0.42 µg/L and TT 3 in adults as shown in Supplemental Material, Table S2 (http:// dx.doi.org/10.1289/ehp.1205239). This may be related to significant health effects in indi viduals who already have compromised thyroid functions. More research is needed for Hg and Cd exposure at the higher end of exposures, for example, in specific populations such as people who eat a large amount of fish or live close to Cdcontaminated areas.
This analysis has limitations related to the crosssectional design of the NHANES. The associations cannot be interpreted as cau sation, and the results are more relevant to background exposure in the general popula tion. Although we did analyze levels of BPb, Hg, and Cd as well as urinary Cd, the latter is only available in a subset of onethird of the total sample population. The research was only assessing one time point and we lacked longi tudinal data. We performed multiple compari sons in the analysis and may have encountered the problem of false positive findings. Instead of adopting a strict Bonferroni correction, we mainly focused the patterns (relation to more than one TH), dose response (significance for more than one quintile), and consistency between exposures (total and organic Hg, blood and urinary Cd). In spite of these limi tations, we were able to test the associations between metals and FT 4 and FT 3 , which often were not measured in prior studies. We strati fied the analysis by adolescents and adults, and replicated results in women of reproductive age. We completed the analysis using con tinuous exposure variable and exposure quin tiles, and we summarized percentage change in hormone levels by IQR of exposure.

Conclusions
In the general adult U.S. population, we observed inverse associations between Hg and TH levels and positive associations between Cd and TH levels. Research is needed to quan tify the associations at higher levels of exposure and to examine potential mechanisms of Hg and Cd thyroid toxicity.