The Effects of Perchlorate, Nitrate, and Thiocyanate on Free Thyroxine for Potentially Sensitive Subpopulations of the 2001

Home , Perchlorate

ORIGINAL ARTICLE The effects of perchlorate, nitrate, and thiocyanate on free thyroxine for potentially sensitive subpopulations of the 2001–2002 and 2007–2008 National Health and Nutrition Examination Surveys

Mina Suh1, Liz Abraham2, J. Gregory Hixon3 and Deborah M. Proctor1

Among women with urinary iodine concentration o100 mg/l in the 2001–2002 National Health and Nutrition Examination Survey (NHANES), urinary perchlorate was associated with significant changes in thyroid stimulating hormone and total thyroxine (T4). Although perchlorate, nitrate, and thiocyanate all potentially act to inhibit iodide uptake, free T4 was not found to be associated with exposure to these chemicals in the same data. Fetuses of pregnant mothers with iodine deficiency are thought to be a sensitive subpopulation for perchlorate exposure, but the potential associations between free T4 and exposure to these chemicals among pregnant mothers in NHANES 2001–2002 and 2007–2008 have not been specifically evaluated to date. This study investigates the potential associations between urinary perchlorate, nitrate, and thiocyanate and serum free T4 in individuals with low urinary iodine levels and pregnant women. Multivariate regression models of free T4 were conducted and included urinary perchlorate, nitrate, thiocyanate, and covariates known to have an impact on the thyroid (anti-thyroid peroxidase (TPO) antibodies, age, race/ethnicity, body mass index, and hours of fasting). Meta-analyses were also conducted on non-pregnant and on pregnant women from the two survey cycles. Urinary nitrate was associated with serum free T4 in non-pregnant women of NHANES 2001–2002 who had urinary iodine Z100 mg/l. In the meta-analysis, urinary perchlorate, nitrate, and thiocyanate were significant predictors of serum free T4 in non-pregnant women. No association was found in men and pregnant women. TPO antibodies were significant predictors of free T4 among non-pregnant women only when the models included urinary perchlorate, nitrate, or thiocyanate. Risk assessment for perchlorate exposure should consider co-exposure to nitrate and thiocyanate.

Journal of Exposure Science and Environmental Epidemiology (2014) 24, 579–587; doi:10.1038/jes.2013.67; published online 23 October 2013 Keywords: perchlorate; thiocyanate; nitrate; thyroid; risk assessment

INTRODUCTION NIS.11,12 All three chemicals are ubiquitous in the environment, Iodide is an essential element required for thyroid hormone and exposures can occur through diet (for example, milk, 13,14 synthesis, and deficiency in iodide uptake is a major cause for cruciferous, and leafy vegetables) and drinking water. disturbance in hypothalamus-pituitary-thyroid (HPT) axis that can Exposure to these chemicals may also occur because of lead to hypothyroidism.1,2 Perchlorate, nitrate, and thiocyanate are anthropogenic sources. Used for rockets, missiles, and fireworks, chemicals that can have an impact on thyroid hormone levels by perchlorate is a naturally occurring anion that is also manmade; causing iodide uptake inhibition.3–7 The potential mode of action thiocyanate is a metabolic component of cigarette smoke, and (MOA) for hypothyroidism from exposures to these chemicals, at nitrate is a preservative that can be added to food such as fish and sufficient dose, is thought to involve iodide uptake inhibition by meat, and is used as fertilizer.4,13 Both perchlorate and nitrate are sodium-iodide symporter (NIS) and subsequent perturbations also found in drinking water. Hence, assessing the effects of of thyroid hormone production (Figure 1).4,8,9 It is thought that cumulative exposure to these chemicals that can cause NIS perchlorate binds to NIS on the surface of thyroid follicles to inhibition is of public health interest. disrupt iodide uptake and, potentially also, thyroid hormone As the human body has compensation mechanisms to regulate productions including thyroxine (T4) and triiodothyronine (T3).4 T3 thyroid hormone and iodide levels, short-term perchlorate is the biologically active thyroid hormone and is needed for exposures will not induce prolonged inhibition of thyroidal iodide normal development of the central nervous system in fetuses and uptake and subsequent adverse health effects.4 In fact, it is infants.1,10 T4 is primarily a precursor hormone converted to T3, expected that iodide uptake would have to be reduced at least and its production is under tight biological control to protect 75% for months or longer period for adverse health effects to against hypo- and hyperthyroidism.1 Although the effect of occur from abnormal thyroid hormone production.4 However, perchlorate on iodide uptake inhibition cannot be distinguished neurodevelopmental decrements in newborns of pregnant from the effects of nitrate or thiocyanate, and perchlorate occurs mothers who have decreased thyroid hormone production at much lower levels in urine, it is a far more effective inhibitor of (low T4) is thought to be the most sensitive adverse health

1ToxStrategies, Inc., Mission Viejo, California, USA; 2ToxStrategies, Inc., Katy, Texas, USA and 3ToxStrategies, Inc., Austin, Texas, USA. Correspondence to: Mina Suh, ToxStrategies, Inc., 20532 El Toro Road, Suite 206, Mission Viejo, CA 92692, USA. Tel.: þ 949 459 5491. E-mail: [email protected] Received 19 June 2013; accepted 10 September 2013; published online 23 October 2013 NIS inhibitors, free T4, NHANES Suh et al 580 Exposure to NIS inhibitors Metabolic Sequelae at any age

Inhibition of Decrease in Thyroid lodide Uptake T3 and T4 hypertrophy or Hypothyroidism Thyroid Increase TSH hyperplasia

Abnormal Fetal and Child Growth Adverse Effect and Development Biochemical Changes Not Adverse Effects

Figure 1. The mode of action (MOA) for the sodium-iodide symporter (NIS) inhibitors (that is, perchlorate, nitrate, and thiocyanate) based on the MOA for perchlorate.4

effect potentially associated with perchlorate exposure.4 Thus, the mothers, an important subpopulation for these chemicals, were fetuses of pregnant women who have iodine deficiency have excluded from assessment in the aforementioned studies.20–22 been identified as the most sensitive subpopulation susceptible We assessed the potential relationships of urinary perchlorate, to the effects of perchlorate.4 Newborns have been identified as nitrate, and thiocyanate with serum free T4 collected during the a subpopulation requiring further evaluation based on updated 2001–2002 and 2007–2008 examination cycles. As iodine defi- physiologically based pharmacokinetic (PBPK) models of ciency and pregnancy status are known risk factors for negative perchlorate inhibition of radioiodide uptake.15 Further, indivi- thyroid function, individuals with low urinary iodine levels and duals with less than sufficient iodide intake, infants, children, and pregnant women were primarily assessed. The objective of this those with iodide deficiency or compromised iodide uptake may study was to determine whether elevated urinary perchlorate, be at greater risk of thyroid dysfunction or abnormality from thiocyanate, and nitrate are significant predictors of free T4 in elevated and prolonged exposures to chemicals that can cause serum among potentially sensitive subpopulations and character- NIS inhibition. ize which of the three chemicals have the most significant effect Previously, urinary perchlorate levels were analyzed for on free T4. Multivariate regression models of free T4 were associations with thyroid stimulating hormone (TSH) and total conducted with the inclusion of urinary perchlorate, nitrate, and T4 among men and women of the 2001–2002 National Health and thiocyanate and covariates associated with thyroid hormone status Nutrition Examination Survey (NHANES).16 Blount et al.16 also including age, body mass index (BMI), race/ethnicity, and hour of assessed urinary nitrate and thiocyanate for potential impact on fasting. As TPO antibodies have been linked to hypothyroxinemia, total T4 or TSH. Among women with urinary iodine concentration they were also evaluated in the models as a covariate. o100 mg/l, increased urinary perchlorate was associated with increased TSH and decreased total T4.16 Urinary nitrate and thiocyanate were not found to be meaningful predictors of TSH METHODS and total T4 levels.16 One limitation noted in the study was the Study Participants lack of assessment on free T4 and anti-thyroid peroxidase (TPO) Initiated in the early 1960s, NHANES assesses the health and nutrition antibodies.16 These data were not evaluated because they were status of adults and children across the United States and includes data on measurements of demographic, socioeconomic, dietary, clinical examina- only recently available. TPO antibodies have been linked to 23 subclinical hypothyroidism and hypothyroxinemia and are a tion, and laboratory variables. Findings from these surveys are used to 17,18 evaluate nutritional status, and also serve as the basis for national covariate that should be explored. Free T4 is thought to be standards such as ideal weight, height, and blood pressure.23 For the a better indicator of thyroid status than total T4, which consists of 2001–2002 survey, 13,156 persons were selected as a representative bound and unbound (free) T4 because abnormal total T4 levels sample across all age groups, and from these individuals, 11,039 (83.9%) can be due to protein-bound T4 that is elevated or decreased and 10,477 (79.6%) underwent survey interviews and medical examina- based on concentrations of circulating proteins.19 tions, respectively.23 Spot laboratory sampling was conducted on a To better assess the association, if any, between iodide uptake representative subset of the participants, and 2,013 individuals with inhibition induced by NIS inhibitors and thyroid hormone laboratory results from NHANES 2001–2002 were initially considered in this disturbance, Bruce et al.20 conducted additional evaluations of study. the NHANES 2001–2002 data set; in this study, perchlorate, nitrate, Neonates, infants, and young children were identified as potentially sensitive subpopulations because they have minimal stores of thyroid and thiocyanate urinary concentrations were evaluated as total 20 hormones and iodide, and as such, are less able to maintain normal serum perchlorate equivalent concentrations (PEC). PEC was weakly thyroid hormone levels than older children or adults.3,24,25 Fetuses of and negatively associated with total T4, whereas no relationship pregnant women with thyroid deficiency were identified to be susceptible 20 was observed for total T3, TSH, and free T3 and T4. The authors because maternal hypothyroidism, which can result from maternal iodine concluded that the association between PEC and total T4 was deficiency, can be detrimental to fetal neurodevelopment.4 However, serum dominated by nitrate and thiocyanate.20 Conversely, Mendez and levels of free T4 and TPO antibodies were collected only from individuals at Eftim21 evaluated biomarkers of perchlorate exposure and serum age 12 and older in NHANES 2001–2002 and 2007–2008. As such, only men thyroid hormone levels in 2007–2008 NHANES participants and and women older than age 12 were evaluated herein. found that urinary perchlorate was a significant predictor of free Seven NHANES population groups were assessed from the 2001 to 2002 cycle: all participants (all men and women regardless of iodine and T4. In a study by Steinmaus et al.,22 NHANES 2007–2008 subjects pregnancy status), men with urinary iodine o100 mg/l, men with urinary exposed to higher levels of perchlorate and thiocyanate, and with iodine Z100 mg/l, non-pregnant women with urinary iodine Z100 mg/l, low urinary iodine, had lower total and free T4 levels suggesting non-pregnant women with urinary iodine o100 mg/l, pregnant women that elevated concomitant exposures to these two NIS inhibitors regardless of iodine status, and pregnant women with urinary iodine 22 and low iodine affect T4 production. However, pregnant o150 mg/l. From the 2007 to 2008 cycle, only women were assessed based

Journal of Exposure Science and Environmental Epidemiology (2014), 579 – 587 & 2014 Nature America, Inc. NIS inhibitors, free T4, NHANES Suh et al 581 on results of the NHANES 2001–2002 data analysis in this study. Pregnancy urinary iodine.34 As individuals with TPO antibodies may have underlying status, for both surveys, was determined using reproductive health thyroid conditions including hypothyroxinemia and are less likely to maintain questionnaire response and positive pregnancy test result based on human normal thyroid function, TPO antibody was considered as a covariate and chorionic gonadotropin levels. In general, the median urinary iodine included in the regression models. Urinary creatinine was also included in the concentrations among children, men, and non-pregnant women range regression models to account for variable water excretion. Gestational age from 100 to 299 mg/l.26 As low iodine levels may potentially have an impact was also included as a covariate in the regression models of pregnant on the relationships between the NIS inhibitor chemicals and serum free T4, women. Distributions for each of the covariates were evaluated using men and non-pregnant women with urinary iodine o100 mg/l were analy- histograms. Log transformations of the following covariates were used in the zed as well as those with urinary iodineZ100 mg/l. Pregnant women have regression analysis, as these variables exhibited skewed distributions: BMI, higher iodide intake requirements than the general population;26 hence, serum concentrations of cotinine and C-reactive protein, and urinary nitrate, pregnant women with urinary iodineo150 mg/l were also assessed. thiocyanate, perchlorate, and creatinine. Stepwise regression analyses using Individuals using medicines containing thyroid hormones such as backward elimination procedures were used on population-weighted data. levothyroxine, liotrix, liothyronine, natural thyroid drugs, for example, desic- Multicollinearity was evaluated using the variance inflation factor. A retaining cated thyroid or anti-thyroid drugs (methimazole or propylthiouracil), were two-tailed P-value of 0.05 was used to identify significant predictor variables excluded from study analysis. Individuals with missing data on serum levels for the final regression models. of free T4 and TPO antibodies, urinary perchlorate, nitrate, thiocyanate, and Meta-analysis was conducted using the data of non-pregnant women creatinine, and pregnancy status were also excluded from this study. with o100 and Z100 mg/l for urinary iodine, and pregnant women, regardless of urinary iodine status, from the two surveys. Meta-analysis Laboratory Methods could not be conducted for pregnant women with urinary iodine 27 r150 mg/l because of collinearity pertaining to small sample size. The Detailed descriptions of laboratory methods are provided elsewhere. final regression models from the stepwise regression analyses of NHANES Briefly, spot urine and blood samples were collected from a subset of 2001–2002 data were applied to the 2007–2008 data. Unweighted participants during their medical examinations. Perchlorate, nitrate, Stouffer’s z-test and Fisher’s w2-test were used35,36 to evaluate potential thiocyanate, and iodine content were measured in all available urine 28,29 relationships between the NIS inhibitors and serum free T4. One-tailed samples from participants at age 6 and older. The chemicals were P-value of 0.05 was used to identify significant associations. measured using ion chromatography tandem mass spectrometry, and the limits of detection for perchlorate, nitrate, and thiocyanate have been determined to be 0.05, 700, and 20 mg/l, respectively.28,30 Urinary iodine levels were measured using inductively coupled plasma-mass spectro- RESULTS metry; serum samples were analyzed for free T4 and TPO antibodies using Study Population and Urinary Chemicals and Sera Thyroid enzyme immunoassays.27,28 Hormone Measures Seventy-two 2001–2002 NHANES survey participants taking Statistical Analyses medications containing thyroid hormones were excluded from All data were analyzed using R software (http://www.R-project. the study. Additional participants were excluded for whom data org). Descriptive statistics were obtained using the Survey package within on serum thyroid hormone levels and/or urinary chemical R, incorporating sample weights and masked variance units for strata and concentrations were missing (n ¼ 66). Of the 1,875 individuals primary sampling units, to capture the NHANES survey design. The urinary from NHANES 2001–2002 that met the inclusion criteria, there concentrations of perchlorate, nitrate, and thiocyanate were adjusted by were 890 men and 985 women (Table 1). Among the men, one urinary creatinine, that is, urinary chemical concentration divided by individual did not have urinary iodine measurement and 22 urinary creatinine, and reported as geometric means in mg/g. Urinary creatinine is excreted from the body at a relatively constant rate. As urine women had unknown status for pregnancy; these individuals were samples in NHANES were spot samples collected at different times of day excluded from additional analyses. From NHANES 2007–2008, depending on the examination session, urine dilution will likely vary 2,568 non-pregnant women and 49 pregnant women were between individuals and within individuals during the course of a day. assessed (Tables 1 and 2). Hence, adjustment by urinary creatinine was necessary to correct for Ninety-two women of the 2001–2002 survey indicated that they variable dilutions among spot samples.31 Free T4 and TPO antibody in were pregnant and had positive pregnancy test results. Of these serum were reported as geometric means. It is thought that perchlorate, pregnant women, 48 had urinary iodine o150 mg/l (Table 2). Free nitrate, and thiocyanate can decrease free T4 levels. The normal reference T4 was decreased among pregnant women with urinary iodine range for free T4 levels in adults can vary from 0.7 to 2.0 ng/dl.32 Hence, o150 mg/l compared with the other groups. This group of individuals with serum free T4o0.7 ng/dl for each population group were also identified. Among pregnant women, serum free T4 is reported to be pregnant women also had the lowest adjusted geometric mean higher in the first trimester than in the second.33 However, standard US concentration for urinary perchlorate (2.73 mg/g creatinine) reference ranges for serum free T4 by trimester have not been reported compared with the other groups (Figure 2a). Urinary perchlorate consistently; in this study, geometric means of serum free T4 by trimester concentrations in pregnant women were significantly lower were reported for pregnant women. compared with males with urinary iodine o100 mg/l (P ¼ 0.03) NHANES has constructed various sample weights for each 2-year survey and men with urinary iodine Z100 mg/l (P ¼ 0.02). cycle to account for survey non-response, oversampling, poststratification, Non-pregnant women from the 2001 to 2002 survey with and sampling error. To assess for differences in urinary levels of urinary iodine o100 mg/l had the highest adjusted geometric perchlorate, nitrate, and thiocyanate and serum free T4 among the seven mean concentration for urinary nitrate (52,945 mg/g creatinine) different groups from the 2001 to 2002 survey analyzed in this study, independent groups t-testing was conducted. Each group (for example, (Figure 2b). Urinary nitrate concentrations in this group of non- men with urinary iodineo100 mg/l) was compared with the other six pregnant women were significantly elevated compared with groups one to one for statistical significance. Individual measurements of males with urinary iodine concentration Z100 mg/l (Po0.001) perchlorate, nitrate, thiocyanate, and free T4 along with the associated and non-pregnant women with urinary iodine concentration weights from the 2001 to 2002 NHANES database for each group were Z100 mg/l (Po0.001). However, urinary perchlorate concentra- used in the t-test comparisons. P-valueso0.05 in a two-tailed test were tions for this group of non-pregnant women were not significantly considered to be significant. different from the other groups with the exception of men Multivariate regression models were conducted to assess the associations with urinary iodine concentration 4100 mg/l (P ¼ 0.01). Men with between urinary perchlorate, nitrate, and thiocyanate and free T4. Sample urinary iodine concentration 100 mg/l had the highest adjusted weights and stratum were incorporated, and the models also included other o covariates known to have an impact on the thyroid, that is, age, race/ geometric mean concentration for urinary thiocyanate (2,320 mg/g ethnicity, BMI, serum cotinine, estimated total caloric intake, postmenopausal creatinine) (Figure 2c). Urinary thiocyanate concentrations in this status, premenarche status, estrogen use, serum C-reactive protein, and hours group of men were significantly elevated compared with the other offastingbeforesamplecollection.Thesecovariateswerealsoexploredby groups except non-pregnant women with urinary iodine concen- Blount et al.16 Race/ethnicity and urine creatinine were also associated with tration o100 mg/l. Thiocyanate is a metabolite of cyanide in

& 2014 Nature America, Inc. Journal of Exposure Science and Environmental Epidemiology (2014), 579 – 587 NIS inhibitors, free T4, NHANES Suh et al 582 Table 1. Geometric mean and 95% CIs of serum free T4 and TPO antibodies in NHANES 2001–2002 and 2007–2008 participants.

Group N Geometric mean (95% CIs) N (%) free T4o0.7 ng/dla Serum free T4 (ng/dl) Serum TPO antibodies (IU/ml)

NHANES 2001–2002 All participants 1875 0.76 (0.76–0.77) 0.98 (0.89–1.09) 254 (13.5%) Men (iodineo100 mg/l) 171 0.78 (0.76–0.79) 0.79 (0.65–0.96) 17 (9.9%) Men (iodine Z100 mg/l) 718 0.77 (0.76–0.78) 0.82 (0.72–0.93) 76 (10.6%) Women (non-pregnant, iodineo100 mg/l)b 307 0.77 (0.75–0.79) 1.44 (0.93–2.23) 34 (11.1%) Women (non-pregnant, iodineZ100 mg/l)b 564 0.75 (0.73–0.78) 1.05 (0.85–1.29) 69 (12.2%)

NHANES 2007–2008c Women (non-pregnant, iodineo100 mg/l)b 848 0.76 (0.74–0.77) 0.98 (0.85–1.12) 121 (14.3%) Women (non-pregnant, iodineZ100 mg/l)b 1720 0.76 (0.75–0.78) 1.25 (1.09–1.42) 260 (15.1%) Abbreviation: 95% CI, 95% conﬁdence interval. aNormal free T4 levels range from 0.7 to 2.0 ng/dl.29 As it is thought that perchlorate, nitrate, and thiocyanate can decrease free T4 levels, those with free T4o0.7 ng/dl were noted. bPregnancy status is indicated by four categories (i.e., pregnant, not pregnant, not sure, and menopause). Non-pregnant women include individuals that are not pregnant and menopausal and did not have positive laboratory results for pregnancy (based on human chorionic gonadotropin). cFor the 2007–2008 cycle, non-pregnant and pregnant women were evaluated; women were analyzed for this cycle based on results of NHANES 2001–2002 data analyses in this study and in Blount et al.16

Table 2. Geometric mean and 95% CIs of serum free T4 by trimester and TPO antibodies in NHANES 2001–2002 and 2007–2008 pregnant women.

Group N Geometric mean (95% CIs)

Serum free T4 (ng/dl) Serum TPO antibodies (IU/ml) Trimester 1 Trimester 2 Trimester 3

NHANES 2001–2002 Women (all pregnant)a 92 0.62 (0.55–0.71) 0.57 (0.53–0.61) 0.61 (0.55–0.67) 1.45 (0.40–5.28) Women (pregnant, iodineo150 mg/l)a 48 0.58 (0.49–0.7) 0.55 (0.5–0.6) 0.58 (0.55–0.62) 2.34 (0.26–20.75)

NHANES 2007–2008 Women (all pregnant)a 49 0.81 (0.71–0.92) 0.61 (0.57–0.65) 0.58 (0.48–0.71) 1.02 (0.5–2.09) Women (pregnant, iodineo150 mg/l)a 27 0.75 (0.71–0.79) 0.62 (0.59–0.65) 0.54 (0.41–0.69) 1.64 (0.5–5.38) Abbreviation: 95% CI: 95% conﬁdence interval aPregnancy status is indicated by survey response to questionnaires, designated in four categories (i.e., pregnant, not pregnant, not sure, and menopause) as well as positive laboratory results for pregnancy (based on human chorionic gonadotropin). Pregnant women include individuals indicating that they were pregnant and/or had a positive pregnancy test.

tobacco smoke and is used as a biomarker for validation of self- the perchlorate model (Table 3). For non-pregnant women with reported smoking behavior.11,37,38 Elevation in urinary concen- urinary iodine concentration Z100 mg/l, nitrate was a significant trations of thiocyanate may be partially attributed to tobacco use predictor of free T4 (P ¼ 0.003) with a coefficient for log nitrate of in NHANES men. Pregnant women had the lowest thiocyanate À 0.0311 (Table 4). In this group of non-pregnant women, nitrate concentrations compared with other groups likely indicating remained a significant predictor of free T4 (P ¼ 0.02) with a smoking cessation during pregnancy. coefficient for log nitrate of À 0.0259 in the perchlorate model In the 2007–2008 survey, 27 of 49 pregnant women had urinary (Table 4). In the regression model of thiocyanate, perchlorate was iodineo150 mg/l; among these women, first trimester free T4 also a significant predictor of serum free T4 (P ¼ 0.04) for non- levels were higher than the other trimesters, which is consistent pregnant women with urinary iodineZ100 mg/l; the coefficient for with the reported profile during pregnancy (Table 2).33 Non- log perchlorate was À 0.019 (Table 4). Among this group of non- pregnant women from this survey cycle had serum free T4 levels pregnant women, TPO antibody was a significant predictor of free that were similar to those of NHANES 2001–2002 non-pregnant T4 with coefficients for log TPO antibody of À 0.014 (P ¼ 0.02), women. Urinary concentrations of nitrate, perchlorate, and À 0.014 (P ¼ 0.02), or À 0.013 (P ¼ 0.03) in perchlorate, nitrate, or thiocyanate in non-pregnant and pregnant women were compar- thiocyanate models, respectively (Table 4). able to those of NHANES 2001–2002 women (Figure 3). For pregnant women of the 2001–2002 survey, nitrate and thiocyanate were not significant predictors of free T4 (Table 5). In addition, TPO antibody was not a significant predictor of free T4 Multivariate Regression Analyses among pregnant women. Urinary perchlorate was associated with In non-pregnant women of the 2001–2002 survey, nitrate and free T4 but not in the direction expected (regression perchlorate were found to be significant predictors of serum free coefficient ¼ 0.040) based on the mechanism of NIS inhibition T4 (Tables 3 and 4). Among non-pregnant women with urinary (Table 5). Gestational age and urinary creatinine were found iodine concentration o100 mg/l, nitrate was a significant predictor to be significant predictors of free T4 (Table 5). Among pregnant of free T4 (P ¼ 0.03) with a coefficient for log nitrate of À 0.0368 in women with urinary iodineo150 mg/l, nitrate, perchlorate, and

Journal of Exposure Science and Environmental Epidemiology (2014), 579 – 587 & 2014 Nature America, Inc. NIS inhibitors, free T4, NHANES Suh et al 583

3 g/ g creatinine) µ

2 g/ g creatinine) µ

1 Nitrate ( Perchlorate ( 0 s ) ) (All)

(Low iodine) All participant Men (Normal) Men (Low iodine men (Low iodine Pregnant women

Non-pregnant women (Normal)Pregnant women Non-pregnant wo

Figure 2. Geometric mean and 95% CIs of urinary chemical concentrations (in mg/g creatinine) of (a) perchlorate; (b) nitrate; and (c) thiocyanate in National Health and Nutrition Examination Survey (NHANES) 2001–2002 participants.

thiocyanate were not observed to be significant predictors of free DISCUSSION T4 (Table 6). The regression analyses for this group exhibited Across several NHANES cycles (2001–2008), Lau et al.39 identified multicollinearity where in addition to race/ethnicity (which had dietary sources of perchlorate using perchlorate measured in urine been eliminated from every regression analyses because of samples. Tap water contributed to adult urinary perchlorate, but collinearity), the predictor variables serum C-reactive protein, dietary sources were found to be more significant contributors. serum cotinine, and TPO antibody also presented as collinear. This The consumption of milk products was associated with statistically is likely attributable to the small group size for pregnant women significant contributions to urinary perchlorate for all age groups; with urinary o150 mg/l (n ¼ 48). for adults, vegetables, fruits, and eggs contributed significantly.39 When all men and women of the 2001–2002 survey, regardless However, because nitrate and thiocyanate are ubiquitous in food of iodine and pregnancy status, were combined for the regression and drinking water, and occur at higher levels compared with analyses, urinary perchlorate, nitrate, and thiocyanate were perchlorate, these two chemicals have been identified to account not found to be significant predictors of serum free T4 11 (Supplementary Table 1). For both men groups (regardless of for larger portions of iodine uptake inhibition in humans. In iodine status), none of the urinary chemicals were significant Chinese hamster ovary cells, the relative potencies of thiocyanate predictors of serum free T4 (Supplementary Tables 2 and 3). Race/ and nitrate were lower than that of perchlorate (15 and 240 times, respectively) for radioactive iodide uptake inhibition at NIS.12 Van ethnicity and serum cotinine were not observed to be significant 40 predictors of serum free T4 for all groups evaluated in this study. Sande et al. measured the affinity of various anions for For the meta-analysis, urinary nitrate and thiocyanate were NIS in COS-7 cells and thyroid cells and found that perchlorate significant predictors of serum free T4 among non-pregnant had the greatest affinity followed by iodide and thiocyanate and 40 women with urinary iodineo100 mg/l from the two survey cycles then bromide. Bromide is known to decrease thyroidal (Table 7). Among non-pregnant women with urinary iodineZ100 iodide uptake or NIS expression.41 Although not evaluated in mg/l, urinary perchlorate and nitrate were significant predictors of this study, it is likely that there are other anions that have serum free T4 in the meta-analysis (Table 7). No significant an impact on thyroid NIS status and possibly serum free T4 levels associations were observed among pregnant women for urinary among NHANES subjects. nitrate and thiocyanate. Urinary perchlorate was found to be In this study, nitrate was found to be a significant predictor of significantly associated with serum free T4 in pregnant women serum free T4, and the effect was observed in non-pregnant (Table 7); however, the coefficients in the regression models were women in NHANES 2001–2002, but not in pregnant women or noted to be positive and thus not in the expected direction based other groups. The effect of perchlorate on free T4 is also only on the MOA. observed among non-pregnant women who participated in the

& 2014 Nature America, Inc. Journal of Exposure Science and Environmental Epidemiology (2014), 579 – 587 NIS inhibitors, free T4, NHANES Suh et al 584

5 60000 4 50000 3 40000 g/ g creatinine)

µ 30000 g/ g creatinine)

2 µ 20000 1 10000 Nitrate ( 0 Perchlorate ( 0 ) e)

iodine) (Normal) (Low iodine) n (Low iodin

Pregnant women (All) Pregnant women (All

Non-pregnant women (Normal)Pregnant wome Non-pregnant women Pregnant women (Low Non-pregnant women (Low iodine) Non-pregnant women

3000 g/ g

µ 2000

1000 creatinine) Thiocyanate (

0 )

Pregnant women (All)

Non-pregnant women (Normal)Pregnant women (Low iodine) Non-pregnant women (Low iodine Figure 3. Geometric mean and 95% CIs of urinary chemical concentrations (in mg/g creatinine) of (a) perchlorate; (b) nitrate; and (c) thiocyanate in National Health and Nutrition Examination Survey (NHANES) 2007–2008 non-pregnant and pregnant women.

including those with low iodine excretion. As non-pregnant Table 3. Multiple regression models of urinary perchlorate, urinary women had high urinary nitrate and perchlorate concentrations nitrate, and urinary thiocyanate predicting serum free T4 in non- (Figure 2), this finding suggests that effects on free T4 levels by pregnant women (n ¼ 307) with urinary iodineo100 mg/l, NHANES urinary nitrate and perchlorate may occur among more highly 2001–2002. exposed individuals. Thiocyanate, in addition to perchlorate and nitrate, was found Independent variable Coefficient SE P-value to be a significant predictor of serum free T4 in non-pregnant Intercept 1.034889 0.102945 3.38e À 07* women when NHANES 2007–2008 data were assessed for the Log urine perchlorate 0.003223 0.014706 0.8302 meta-analysis. It is interesting that urinary nitrate and thiocyanate Log urine nitrate À 0.036830 0.015067 0.0309* were significant predictors of serum free T4 among non-pregnant Log urine creatinine 0.029802 0.022972 0.2189 women with urinary iodineo100 mg/l, whereas urinary perchlorate and nitrate were significant predictors among those with urinary Intercept 1.01933 0.11446 6.78e À 07* iodineZ100 mg/l. Previous studies of NHANES subjects have Log urine nitrate À 0.03551 0.01705 0.0577 observed associations between NIS inhibitors and thyroid Log urine creatinine 0.03084 0.02110 0.1675 hormones among those with urinary iodineo100 mg/l; in the 21 Intercept 0.826011 0.046467 1.67e À 10* study by Mendez and Eftim, significant relationships between Log urine thiocyanate À 0.018020 0.009599 0.0831 perchlorate excretion and thyroid hormones were detected in an 5,16,21 Log urine creatinine 0.019584 0.017296 0.2780 iodine-sufficient population. As such, additional investi- gations are needed to further elucidate the meta-analysis results *Statistically significant (Po0.05). observed herein. The effect on serum free T4 from urinary NIS inhibitors was not 2001–2002 examination cycle. The geometric mean concentra- observed in pregnant womens. Gibb et al.43 did not observe an tions of urinary nitrate and thiocyanate are B400-fold higher than association between perchlorate and free T4 among pregnant geometric mean urinary concentration of perchlorate among the women in Chile. A surge in free T4 levels is expected to occur in 2001–2002 survey participants.16 The 95th percentile urinary pregnant women at the end of first trimester in pregnancy to perchlorate and nitrate concentrations were observed to be ensure adequate supply of free T4 for the fetus.10 However, these highest in non-pregnant women of NHANES 2001–2002 (data not changes in free T4 levels that occur during pregnancy are not shown). This may partially explain why nitrate and perchlorate likely captured by spot samples collected at a single point in time. were significant predictors of free T4 in non-pregnant women, but Gestation age and urinary creatinine were significant predictors of not in other NHANES populations. Similarly, Pearce et al.42 serum free T4 in pregnant women; hence, potential associations observed that urinary perchlorate was not a significant predictor between exposures to perchlorate, nitrate, and thiocyanate of thyroid function among 241 first-trimester pregnant women and serum free T4 levels are expected to be more difficult to

Journal of Exposure Science and Environmental Epidemiology (2014), 579 – 587 & 2014 Nature America, Inc. NIS inhibitors, free T4, NHANES Suh et al 585 Table 4. Multiple regression models of urinary perchlorate, urinary Table 6. Multiple regression models of urinary perchlorate, urinary nitrate, and urinary thiocyanate predicting serum free T4 in non- nitrate, andurinary thiocyanate predicting serum free T4 in pregnant pregnant women (n ¼ 564) with urinary iodineZ100 mg/l, NHANES women with urinary iodineo150 mg/l (n ¼ 48), NHANES 2001–2002. 2001–2002. Independent variable Coefﬁcient SE P-value Independent variable Coefﬁcient SE P-value Intercept 0.805 0.075 0.0004* Intercept 0.756533 0.086337 1.06e-05* Fasting À 0.009 0.004 0.06 Log urine perchlorate À 0.014136 0.008697 0.13851 Log urine perchlorate 0.019 0.015 0.26 Log urine nitrate À 0.025860 0.009329 0.02168* Log urine creatinine À 0.024 0.019 0.27 Neither premenarche or 0.086993 0.032884 0.02667* Gestational age À 0.006 0.004 0.23 postmenopausal Postmenopausal 0.120608 0.032493 0.00483* Intercept 0.682 0.288 0.08 Log TPO antibody À 0.013686 0.004908 0.02111* Log urine nitrate 0.017 0.047 0.73 Log urine creatinine 0.045244 0.012574 0.00576* Log urine perchlorate 0.046 0.019 0.07 Log urine creatinine À 0.062 0.044 0.23 Intercept 0.806828 0.084809 2.51e-06* Gestational age À 0.011 0.008 0.23 Log urine nitrate À 0.031100 0.008131 0.00335* Neither premenarche or 0.089687 0.032789 0.02100* Intercept 0.734 0.105 0.002* postmenopausal Log urine thiocyanate 0.012 0.028 0.68 Postmenopausal 0.120666 0.032603 0.00410* Fasting À 0.011 0.004 0.06 Log TPO antibody À 0.013762 0.005001 0.02041* Log urine creatinine À 0.023493 0.027708 0.44 Log urine creatinine 0.041970 0.014314 0.01499* Gestational age À 0.003819 0.006373 0.58

Intercept 0.590891 0.079389 7.31e-05* Fasting, hours of fasting before sample collection *Statistically significant (P 0.05). Log urine thiocyanate À 0.006595 0.008549 0.46265 o Log urine perchlorate À 0.019109 0.007551 0.03521* Neither premenarche or 0.094783 0.034653 0.02564* postmenopausal whether pregnant women experience any effects of these Postmenopausal 0.139638 0.035232 0.00416* chemicals on free T4 levels. Estrogen use À 0.046019 0.019847 0.04902* It is important to note that demands for iodide are greater Log TPO antibody À 0.012849 0.004788 0.02777* among pregnant women compared with other populations, and Log urine creatinine 0.032541 0.012443 0.03088* recommended urinary iodine concentrations for pregnant 26 *Statistically significant (Po0.05). women are higher for thyroid hormone production. Among the 2001–2002 NHANES participants evaluated in this study, 35% of non-pregnant women had urinary iodineo100 mg/l, whereas 52% of pregnant women had urinary iodineo150 mg/l. On the basis Table 5. Multiple regression models of urinary perchlorate, urinary of this sampling, more than half of pregnant women in the United nitrate, and urinary thiocyanate predicting serum free T4 in pregnant States have low urinary iodine levels. Consistent with this study, women (n ¼ 92), NHANES 2001–2002. 56.9% of pregnant women (n ¼ 184) surveyed in NHANES 2005– 44 Independent variable Coefficient SE P-value 2006 and 2007–2008 cycles had urinary iodide levelso150 mg/l. As such, it is possible that many pregnant women are more Intercept 0.898 0.062 5.39e À 07* vulnerable to the potential goitrogenic effects of perchlorate, Fasting À 0.007 0.002 0.02* nitrate, and thiocyanate. Based on these data, hypothyroxinemia Log urine perchlorate 0.040 0.008 0.001*a during pregnancy is common, and mothers with this condition may Log urine creatinine À 0.051 0.013 0.004* have insufficient thyroid hormone supply for their fetus.45–48 Gestational age À 0.011 0.004 0.01* However, no association was observed in this study between free T4 and urinary chemical concentrations among pregnant women, Intercept 0.862 0.107 8.63e 05* À regardless of iodine status, suggesting that even with decreased Fasting À 0.006 0.002 0.02* Log urine nitrate 0.005 0.016 0.77 urinary iodide in the most potentially sensitive population, Log urine perchlorate 0.039 0.009 0.002*a population-level exposures to nitrate, perchlorate, and thiocyanate Log urine creatinine À 0.054 0.018 0.02* do not affect free T4 levels. Gestational age À 0.011 0.003 0.01* Although the observations for pregnant women in this study are limited to a relatively small sample size and spot urine data, Intercept 0.896 0.063 1.99e À 06* they are consistent with the biologically based dose-response Fasting À 0.006 0.002 0.03* (BBDR) model HPT predictions (BBDR-HPT) by Lumen et al.9 This Log urine thiocyanate 0.001 0.013 0.96 model predicts that it would require on the order of 32 mg/kg-day Log urine perchlorate 0.040 0.009 0.002*a of perchlorate to cause hypothyroxinemia in pregnant women Log urine creatinine À 0.051 0.020 0.04* Gestational age À 0.012 0.004 0.03* with low iodide intake. Although perchlorate dietary intake in the general population is only B0.1 mg/kg-day,14 the intake of nitrate Fasting, hours of fasting before sample collection. and thiocyanate is expected to be considerably greater and, thus, a *Statistically significant (Po0.05). Urinary perchlorate is significantly it is reasonable to consider the potential cumulative effect of associated with serum free T4 but not in the expected direction (positive exposure to multiple chemicals that act by similar MOA (that is, coefficients in the regression models) based on the MOA. NIS inhibition) in environmental risk assessment. In 2005, the National Research Council developed a reference dose (RfD) for perchlorate using data characterizing iodide uptake detect in pregnant women using spot samples, which is one inhibition in healthy adults following voluntary consumption of limitation of this study. Repeated urinary measures of perchlorate, perchlorate in drinking water.4 The RfD of 0.7 mg/kg-day for nitrate, and thiocyanate along with repeated serum measures perchlorate was developed using the no-observed-effect-level of free T4 during the pregnancy term would better inform from Greer et al.,49 which was the dose at which iodide uptake

& 2014 Nature America, Inc. Journal of Exposure Science and Environmental Epidemiology (2014), 579 – 587 NIS inhibitors, free T4, NHANES Suh et al 586 Table 7. Meta-analysis of NHANES 2001–2002 and 2007–2008 non-pregnant and pregnant women for potential associations between urinary chemicals and serum free T4.a

Group Urinary chemical Stouffer z-testb Fisher’s w2-testb

z-score P-value w2 P-value

Non-pregnant women (iodineo100 mg/l) Perchlorate À 0.5782 0.7184 2.0859 0.7200 Nitrate 2.3726 0.0088* 12.3396 0.0150* Thiocyanate 1.9257 0.0271* 10.0144 0.0402* Non-pregnant women (iodineZ100 mg/l) Perchlorate 2.1869 0.0144* 11.1942 0.0245* Nitrate 2.7449 0.0030* 16.3102 0.0026* Thiocyanate 1.6088 0.0538 8.4988 0.0749 Pregnant women Perchloratec 2.3456 0.0095*c 16.6311 0.0023*c Nitrate 0.0802 0.4680 3.0298 0.5529 Thiocyanate À 0.7617 0.7769 1.7460 0.7823

*Statistically significant (Po0.05). aNon-pregnant and pregnant women were evaluated based on analyses of NHANES 2001–2002 data in this study. Meta- analysis could not be conducted for pregnant women with urinary iodiner150 mg/l because of collinearity pertaining to small sample size. bUnweighted Stouffer z-test and Fisher’s w2-tests were used. cIn our regression models of pregnant women, urinary perchlorate was significantly associated with serum free T4 but not in the expected direction (positive coefficients) based on the MOA.

inhibition was not significantly affected. As thyroid hormones are Among sensitive individuals with mild or moderate hypothyr- required for normal physical and mental develop- oxinemia or who are also exposed to other goitrogenic chemicals, ment, elevated exposures to chemicals that can disturb proper it is challenging to determine whether subjects with normal thyroid function in pregnant women and in early infancy are thyroid status are shifted to slightly perturbed or abnormal status thought to potentially cause neurodevelopmental decrements in with environmental perchlorate exposures. The findings of our children. PBPK models suggest that the fetal and neonatal life study would suggest that, on a population basis, environmental stages have up to two times greater thyroidal iodide uptake perchlorate intake do not affect pregnant women; however, the inhibition for a given external dose of perchlorate as compared with NHANES sample of pregnant women has lower urinary perchlo- adults.50 It is notable that Blount et al.51 estimated perchlorate, nitrate, and thiocyanate levels than the general population rate doses corresponding to the reported urinary perchlorate and it may be that the pregnant women with lower exposures concentrations in NHANES 2001–2002.51 In this NHANES popula- were randomly oversampled. Our study indicates that perchlorate, tion, the 95th percentile of estimated perchlorate doses was less nitrate, and thiocyanate can affect free T4 levels, and although the than the RfD (0.7 mg/kg-day), and 11 adults were estimated to have effect was not observed in pregnant women, for ensuring perchlorate doses in excess of the RfD.51 In the current study, protection of public health, the potential for cumulative risk of urinary concentrations for perchlorate are observed to be lower in coexposures should be assessed. pregnant women than the general population, and lower than the RfD. Hence, an effect in populations exposed above the RfD cannot be ruled out. At perchlorate exposures experienced by pregnant CONFLICT OF INTEREST women in the NHANES 2001–2002 and 2007–2008 populations, The authors declare no conflict of interest. there appears to be no adverse effect on free T4 levels within the limits of these data to assess such. Thus, our findings support that the current RfD is health protective. ACKNOWLEDGEMENTS With the BBDR-HPT model, a maternal iodide intake of 75, 100, We thank Mr. Kenneth Burkhalter and Ms. Meng Ke for contributing to the regression 150, 200, and 250 mg per day (200 mg per day was the intake level analyses. for an euthyroid term pregnancy) with an environmentally relevant exposure of perchlorate (B0.1 mg/kg-day) was estimated to not result in hypothyroxinemia or hypothyroidism. However, REFERENCES the effect of perchlorate on fetal serum free T4 appeared to be 1 Yen PM. Physiological and molecular basis of thyroid hormone action. Physiol Rev more pronounced compared with that in pregnant women.9 2001; 81: 1097–1142. Lumen et al.9 noted that additional analysis was needed to 2 IOM. Dietary reference intakes for vitamin A, vitamin K, arsenic, boron, address variability in the human population and that the greatest chromium, copper, iodine, iron, manganese, molybdenum, nickel, silicon, vana- uncertainty in the BBDR-HPT model was predicting serum thyroid dium, and zinc 2001; 258–289. hormones in the fetus and alterations caused in this sensitive 3 Cao Y, Blount BC, Valentin-Blasini L, Bernbaum JC, Phillips TM, Rogan WJ. Goi- trogenic anions, thyroid-stimulating hormone, and thyroid hormone in infants. subpopulation by changes in dietary iodide and perchlorate Environ Health Perspect 2010; 118: 1332–1337. exposure. Evaluating the 2007–2008 NHANES data for correlation 4 NRC. Health Implications Of Perchlorate Ingestion. National Academies Press: between biomarkers of perchlorate exposure and circulating Washington, D.C., USA, 2005. 21 thyroid hormone levels, Mendez and Eftim noted that HPT axis 5 Steinmaus C, Miller MD, Howd R. Impact of smoking and thiocyanate on per- controls do not completely compensate for small changes in chlorate and thyroid hormone associations in the 2001-2002 national health and thyroid hormone levels associated with perchlorate exposures. nutrition examination survey. Environ Health Perspect 2007; 115: 1333–1338. Further, the BBDR-HPT model does not consider contributions to 6 Wolff J. Perchlorate and the thyroid gland. Pharmacol Rev 1998; 50: 89–105. NIS inhibition from thiocyanate and nitrate. As our findings 7 Ward MH, Kilfoy BA, Weyer PJ, Anderson KE, Folsom AR, Cerhan JR. Nitrate intake suggest that nitrate, thiocyanate, and perchlorate are significant and the risk of thyroid cancer and thyroid disease. Epidemiology 2010; 21: 389– 395. predictors of T4 in non-pregnant women, the cumulative effect of 8 Fisher J, Lumen A, Latendresse J, Mattie D. Extrapolation of hypothalamic-pitui- exposure to these chemicals that can cause NIS inhibition in tary-thyroid axis perturbations and associated toxicity in rodents to humans: case pregnant women and early life stages should be considered in study with perchlorate. J Environ Sci Health C Environ Carcinog Ecotoxicol Rev 2012; future modeling analyses and risk assessment applications. 30: 81–105.

Journal of Exposure Science and Environmental Epidemiology (2014), 579 – 587 & 2014 Nature America, Inc. NIS inhibitors, free T4, NHANES Suh et al 587 9 Lumen A, Mattie DR, Fisher JW. Evaluation of perturbations in serum thyroid 30 Valentin-Blasini L, Mauldin JP, Maple D, Blount BC. Analysis of perchlorate in hormones during human pregnancy due to dietary iodide and perchlorate expo- human urine using ion chromatography and electrospray tandem mass spec- sure using a biologically based dose response model. Toxicol Sci 2013; 133: trometry. Anal Chem 2005; 77: 2475–2481. 320–341. 31 Barr DB, Wilder LC, Caudill SP, Gonzalez AJ, Needham LL, Pirkle JL. Urinary crea- 10 Skeaff SA. Iodine deﬁciency in pregnancy: the effect on neurodevelopment in the tinine concentrations in the U.S. population: implications for urinary biologic child. Nutrients 2011; 3: 265–273. monitoring measurements. Environ Health Perspect 2005; 113: 192–200. 11 De Groef B, Decallonne BR, Van der Geyten S, Darras VM, Bouillon R. Perchlorate 32 Toft AD, Beckett GJ. Measuring serum thyrotropin and thyroid hormone and versus other environmental sodium/iodide symporter inhibitors: potential thyr- assessing thyroid hormone transport. In: Braverman LE, Utiger RD (edits). Werner & oid-related health effects. Eur J Endocrinol 2006; 155: 17–25. Ingbar’s the thyroid: A fundametnal and Clinical Text, 9th edn (Lippincott Williams 12 Tonacchera M, Pinchera A, Dimida A, Ferrarini E, Agretti P, Vitti P et al. Relative & Wilkins: Philadelphia, PA, USA, 2005, pp 319–344. potencies and additivity of perchlorate, thiocyanate, nitrate, and iodide on the 33 Lambert-Messerlian G, McClain M, Haddow JE, Palomaki GE, Canick JA, Cleary- inhibition of radioactive iodide uptake by the human sodium iodide symporter. Goldman J et al. First- and second-trimester thyroid hormone reference data in Thyroid 2004; 14: 1012–1019. pregnant women: a FaSTER (First- and Second-Trimester Evaluation of Risk for 13 Mervish N, Blount B, Valentin-Blasini L, Brenner B, Galvez MP, Wolff MS et al. aneuploidy) Research Consortium study. Am J Obstet Gynecol 2008; 199:62e1–62e6. Temporal variability in urinary concentrations of perchlorate, nitrate, thiocyanate 34 Pfeiffer CM, Sternberg MR, Caldwell KL, Pan Y. Race-ethnicity is related to bio- and iodide among children. J Expo Sci Environ Epidemiol 2012; 22: 212–218. markers of iron and iodine status after adjusting for sociodemographic and life- 14 Huber DR, Blount BC, Mage DT, Letkiewicz FJ, Kumar A, Allen RH. Estimating style variables in NHANES 2003-2006. J Nutr 2013; 143: 977S–985SS. perchlorate exposure from food and tap water based on US biomonitoring and 35 Mosteller F, Fisher RA. Questions and answers. Am Stat 1948; 2: 30–31. occurrence data. J Expo Sci Environ Epidemiol 2011; 21: 395–407. 36 Stouffer SA, Suchman EA, DeVinney LC, Williams Jr RM. The American Soldier Volume 15 McLanahan ED, White P, Flowers L, Schlosser PM. The use of PBPK models to 1: Adjustment during Army Life. Princeton University Press: Princeton, UK, 1949. inform human health risk assessment: case study on perchlorate and radioiodide 37 Buratti M, Xaiz D, Carevelland G, Colombi A. Validation of urinary thiocyanate as a human lifestage models. Risk Anal 2013, (doi:10.1111/risa.12101; e-pub ahead of biomarker of tobacco smoking. Biomarkers 1997; 2: 81–85. print). 38 Pre J, Vassy R. Urinary thiocyanate:creatinine ratio as reliable indicator of cigarette 16 Blount BC, Pirkle JL, Osterloh JD, Valentin-Blasini L, Caldwell KL. Urinary smoking. Clinica Chimica Acta 1991; 204: 87–94. perchlorate and thyroid hormone levels in adolescent and adult men and women 39 Lau FK, deCastro BR, Mills-Herring L, Tao L, Valentin-Blasini L, Alwis KU et al. living in the United States. Environ Health Perspect 2006; 114: 1865–1871. Urinary perchlorate as a measure of dietary and drinking water exposure in a 17 Kaczur V, Vereb G, Molnar I, Krajczar G, Kiss E, Farid NR et al. Effect of anti-thyroid representative sample of the United States population 2001-2008. J Expo Sci peroxidase (TPO) antibodies on TPO activity measured by chemiluminescence Environ Epidemiol 2013; 23: 207–214. assay. Clin Chem 1997; 43(8 Pt 1): 1392–1396. 40 Van Sande J, Massart C, Beauwens R, Schoutens A, Costagliola S, Dumont JE et al. 18 Zelaya AS, Stotts A, Nader S, Moreno CA. Antithyroid peroxidase antibodies in Anion selectivity by the sodium iodide symporter. Endocrinology 2003; 144: patients with high normal range thyroid stimulating hormone. Fam Med 2010; 42: 247–252. 111–115. 41 Shen DH, Kloos RT, Mazzaferri EL, Jhian SM. Sodium iodide symporter in health 19 ATA. Thyroid Function Tests. American Thyroid Association, ATA: Falls Church, VA, and disease. Thyroid 2001; 11: 415–425. 2012. 42 Pearce EN, Spencer CA, Mestman JH, Lee RH, Bergoglio LM, Mereshian P et al. 20 Bruce GM, Corey LM, Mandel JH, Pleus RC. Urinary nitrate, thiocyanate, and Effect of environmental perchlorate on thyroid function in pregnant women from perchlorate and serum thyroid endpoints based on NHANES 2001 to 2002. Cordoba, Argentina, and Los Angeles, California. Endocr Pract 2011; 17: 412–417. J Occup Environ Med 2013; 55: 52–58. 43 Gibbs JP, Van Landingham C. Urinary perchlorate excretion does not predict 21 Mendez Jr W, Eftim SE. Biomarkers of perchlorate exposure are correlated with thyroid function among pregnant women. Thyroid 2008; 18: 807–808. circulating thyroid hormone levels in the 2007-2008 NHANES. Environ Res 2012; 44 Caldwell KL, Makhmudov A, Ely E, Jones RL, Wang RY. Iodine status of the U.S. 118: 137–144. population, National Health and Nutrition Examination Survey, 2005-2006 and 22 Steinmaus C, Miller MD, Cushing L, Blount BC, Smith AH. Combined effects of 2007-2008. Thyroid 2011; 21: 419–427. perchlorate, thiocyanate, and iodine on thyroid function in the National Health 45 Moleti M, Trimarchi F, Vermiglio F. Doubts and concerns about isolated maternal and Nutrition Examination Survey 2007-08. Environ Res 2013; 123: 17–24. hypothyroxinemia. J Thyroid Res 2011; 2011: 463029. 23 CDC. National Health and Nutrition Examination Survey 2012, (cited 7 February 46 Henrichs J, Ghassabian A, Peeters RP, Tiemeier H. Maternal hypothyroxinemia and 2012); Available from http://www.cdc.gov/nchs/nhanes/about_nhanes.htm. effects on cognitive functioning in childhood: how and why? Clin Endocrinol (Oxf) 24 van den Hove MF, Beckers C, Devlieger H, de Zegher F, De Nayer P. Hormone 2013; 79: 152–162. synthesis and storage in the thyroid of human preterm and term newborns: effect 47 van Mil NH, Steegers-Theunissen RP, Bongers-Schokking JJ, El Marroun H, Ghas- of thyroxine treatment. Biochimie 1999; 81: 563–570. sabian A, Hofman A et al. Maternal hypothyroxinemia during pregnancy and 25 Lawrence JE, Lamm SH, Pino S, Richman K, Braverman LE. The effect of short-term growth of the fetal and infant head. Reprod Sci 2012; 19: 1315–1322. low-dose perchlorate on various aspects of thyroid function. Thyroid 2000; 10: 48 Suarez-Rodriguez M, Azcona-San Julian C, Alzina de Aguilar V. Hypothyroxinemia 659–663. during pregnancy: the effect on neurodevelopment in the child. Int J Dev Neurosci 26 WHO. Assessment of iodine deﬁciency disorders and monitoring their elimination. 2012; 30: 435–438. World Health Organization, 2007. 49 Greer MA, Goodman G, Pleus RC, Greer SE. Health effects assessment for envir- 27 CDC. 2001–2002 Laboratory Methods 2012, Available from http://www.cdc. onmental perchlorate contamination: the dose response for inhibition of thyr- gov/nchs/nhanes/nhanes2001-2002/lab_methods_01_02.htm. oidal radioiodine uptake in humans. Environ Health Perspect 2002; 110: 927–937. 28 CDC. 2001–2002 Data documentation, Codebook, and Frequencies: Nitrate, 50 Clewell RA, Merrill EA, Yu KO, Mahle DA, Sterner TR, Fisher JW et al. Predicting thiocyanate, perchlorate 2012, (cited 16 March 2012); Available from http:// neonatal perchlorate dose and inhibition of iodide uptake in the rat during lac- www.cdc.gov/nchs/nhanes/nhanes2001-2002/SSNO3P_B.htm. tation using physiologically-based pharmacokinetic modeling. Toxicol Sci 2003; 29 CDC. NHANES 2007–2008 Data Documentation, Codebook, and Frequencies: 74: 416–436. Urinary Perchlorate, Urinary Nitrate, and Urinary Thiocyanate. 2013 (cited 28 51 Blount BC, Valentin-Blasini L, Osterloh JD, Mauldin JP, Pirkle JL. Perchlorate August 2013); Available from http://www.cdc.gov/nchs/nhanes/nhanes2007- exposure of the US Population, 2001-2002. J Expo Sci Environ Epidemiol 2007; 17: 2008/PERNT_E.htm. 400–407.

Supplementary Information accompanies the paper on the Journal of Exposure Science and Environmental Epidemiology website (http:// www.nature.com/jes)

& 2014 Nature America, Inc. Journal of Exposure Science and Environmental Epidemiology (2014), 579 – 587