Associations between Prenatal Exposure to Perchlorate, Thiocyanate, and Nitrate and Child Neurodevelopment in the MARBLES Study

Abstract

Associations between prenatal exposure to perchlorate, thiocyanate, and nitrate and autism spectrum disorder (ASD) remain largely unexplored. We investigated these associations in the MARBLES cohort, a prospective cohort with an elevated familial likelihood of ASD. A total of 310 mothers provided 1916 urine samples during pregnancy, and their children were classified at age 3 years as typically developing (TD), ASD, or non-TD (developmental concerns not meeting ASD diagnostic criteria). We estimated relative risks for ASD and non-TD compared with TD using Poisson regression and assessed mixture effects using quantile G-computation and Bayesian kernel machine regression. Overall and mixture analyses showed no significant associations between prenatal exposure to these NIS inhibitors and the child neurodevelopmental outcomes. However, stratified analyses identified several context-specific associations according to child sex, trimester of exposure, maternal urinary iodine concentration, and passive smoking status, some of which remained significant after false-discovery-rate correction. These findings suggest that susceptibility to prenatal exposure to perchlorate, thiocyanate, and nitrate may vary across population subgroups and exposure windows. Although these subgroup-specific associations should be considered hypothesis generating, they highlight the need for future studies integrating thyroid biomarkers and mechanistic approaches to better understand thyroid-related pathways underlying child neurodevelopment.

1. Introduction

Perchlorate, thiocyanate, and nitrate are widespread environmental chemicals found in both the natural and living environments. Perchlorate is an inorganic compound that occurs naturally and is also produced industrially for military purposes, space exploration, and fireworks. Thiocyanate occurs naturally, functions as an endogenous metabolite, and is mainly derived from dietary sources such as broccoli, Brussels sprouts, and cabbage, as well as from cigarette smoke. Nitrate, composed of nitrogen and oxygen, is ubiquitous in air, soil, water, and food, with major sources including contaminated drinking water, cured meats, and green leafy vegetables. , These three compounds are ubiquitously detected in human populations, although exposure levels vary across studies according to dietary habits, drinking water sources, smoking-related exposures, and other population characteristics. ,

All three compounds inhibit the sodium iodide symporter (NIS), thereby reducing iodide transport into the thyroid and interfering with thyroid hormone (TH) synthesis. ,,, This mechanism is well established in animal studies, − and several epidemiological studies have reported corresponding effects on circulating TH levels. For example, a meta-analysis reported that exposure to perchlorate or thiocyanate was significantly associated with lower free thyroxine (FT4) and higher thyroid-stimulating hormone (TSH) levels. ,

Autism spectrum disorder (ASD) is a neurodevelopmental condition that affects social perception, communication, and behavior. − Over the past three decades, its prevalence has increased, , reaching one in 31 (3.2%) among U.S. children aged eight years. ASD is understood as a multistage, brain-wide developmental difference that originates during prenatal and early postnatal life. Among the various hypothesized risk factors, maternal thyroid dysfunction has received particular attention because the fetus depends on maternal THs for brain development before its own thyroid gland matures. − Epidemiological studies support this link: Danish and Californian cohort studies reported a higher likelihood of ASD among children born to mothers with hypothyroidism during pregnancy, , and other studies have associated maternal thyroid dysfunction during pregnancy with reduced intelligence quotient (IQ) and increased likelihood of attention-deficit/hyperactivity disorder (ADHD). ,

Given these findings, prenatal exposure to perchlorate, thiocyanate, and nitrate may affect child neurodevelopment, possibly via the disruption of maternal TH homeostasis. A randomized controlled trial in Italy reported that higher maternal perchlorate exposure was associated with lower child IQ at age three, and a longitudinal study in China found that higher thiocyanate and nitrate exposures during early pregnancy were linked to lower mental development index (MDI) scores, with thiocyanate showing additional effects in late pregnancy. Despite these indications, no prior study has examined prenatal exposure to all three NIS inhibitors in relation to the ASD in children.

In this study, we leveraged data and samples from MARBLES (Markers of Autism Risk in BabiesLearning Early Signs), a prospective cohort enriched for ASD due to elevated familial likelihood. MARBLES enrolls mothers during pregnancy and clinically assesses the child for ASD at 3 years. This design helps address the challenge of ASD’s low prevalence, which typically requires large population-based samples to detect differences. Multiple urine samples were collected per mother to minimize exposure misclassification, given the short biological half-lives of these compounds (perchlorate: 6–8 h, thiocyanate: 1–5 days, and nitrate: 5–8 h). ,, Using these unique samples, we examined (1) individual prenatal exposures to perchlorate, thiocyanate, and nitrate and (2) their combined mixture effects in relation to the likelihood of ASD.

2. Materials and Methods

2.1. Study Population

This study utilized data and samples from the MARBLES cohort, which began in 2006 in the United States. MARBLES is a prospective cohort study that follows pregnant women who already have a child diagnosed with ASD, placing them at an elevated likelihood (∼20%) of having another child who develops ASD. Most participants were recruited from a list of families receiving state-funded services for children with ASD. Eligible participants were women aged 18 years or older who had at least one child with ASD; could speak, read, and understand English; and lived within 2.5 h of the Davis/Sacramento region. Details regarding the study design, recruitment methods, and data collection have been described elsewhere. Between 2006 and 2020, 414 pregnant women provided urine samples at least once during pregnancy. For the present study, we excluded mother–child pairs whose (1) child lacked an ASD diagnostic classification at age three because they were lost to follow-up or did not complete the assessments required for algorithm-based outcome determination (N = 78) and (2) urine sample was not sent for quantification of perchlorate, thiocyanate, nitrate, and creatinine due to low volume (N = 26), resulting in 310 mother–child pairs. A flowchart detailing this study population selection from the full cohort is shown in Figure .

1.

Flowchart of the study population selection from the MARBLES full cohort.

All study procedures were approved by the Institutional Review Boards at the University of California Davis (UC Davis) and the state of California. Written informed consent was obtained from all of the participants.

2.2. Child Neurodevelopmental Assessment

Children were classified into three groupstypically developing (TD), ASD, or nontypically developing (non-TD)using an algorithm based on the Diagnostic and Statistical Manual of Mental Disorders, fifth Edition (DSM-5) criteria, the Autism Diagnostic Observation Schedule (ADOS), , and the Mullen Scales of Early Learning (MSEL). The ADOS, a standardized ASD diagnostic tool, was administered by trained psychologists at approximately 3 years of age. , The ADOS calibrated severity score (CSS), ranging from 1 to 10, provides a normalized measure of ASD trait severity, with higher scores indicating greater severity. The MSEL, which evaluates cognitive development from birth to 36 months, was used to assess developmental functioning across multiple domains. Children classified as ASD met DSM-5 criteria and had an ADOS CSS ≥4 (n = 74, 23.9%). Non-TD children (n = 45, 14.5%) did not meet DSM-5 criteria for ASD but had an ADOS CSS ≥3 and/or lower MSEL scores (>1.5 standard deviation [SD] below the mean in two or more MSEL subscales or 2 SD below the mean in at least one MSEL subscale). The non-TD category represents a clinically heterogeneous group of children exhibiting developmental difficulties that do not fulfill criteria for ASD. Similar classifications have been used in studies from the Baby Siblings Research Consortium (BSRC), which have shown that many high-risk siblings who do not meet ASD diagnostic criteria nevertheless exhibit elevated ASD symptom severity, lower developmental functioning, and broader autism phenotype characteristics. Nevertheless, ASD and non-TD represent distinct developmental outcomes and were analyzed as separate outcome categories relative to the TD reference group in the present study. TD children (n = 191, 61.6%) did not meet any of these criteria.

2.3. Urine Sample Collection

A total of 1916 maternal urine samples were collected from 310 mother–child pairs throughout pregnancy. Of these, 109 mothers provided samples during the first trimester, 211 during the second trimester, and 296 during the third trimester (Figure ). Each mother was instructed to collect three first-morning-void (FMV) urine samples over three consecutive weeks and one 24 h sample in each trimester. Samples were stored in home freezers until retrieved by study staff, transported to the UC Davis laboratory, aliquoted, and stored at −80 °C until analysis. Details of urine sample collection, transport, and storage procedures are described elsewhere. When mothers provided three or more urine samples within a trimester, the first FMV sample was designated as the individual sample, and the remaining samples were pooled to minimize analytical cost. Each pooled specimen was prepared by combining equal volume (5 mL) aliquots from each contributing urine sample into a sterile conical tube, vortex mixing for 10 s, and transferring 2 mL aliquots into sterile cryovials for storage and subsequent analysis. After pooling, 1092 urine samples were analyzed: 468 FMVs (43%), 195 24 h samples (18%), and 429 pooled samples (39%).

2.4. Quantification of Urinary Perchlorate, Thiocyanate, Nitrate, Creatinine, and Iodide

Urinary concentrations of perchlorate, thiocyanate, nitrate, creatinine, and iodide were quantified at the Division of Laboratory Sciences of the National Center for Environmental Health at the Centers for Disease Control and Prevention (CDC). Analyses for perchlorate, nitrate, thiocyanate, and iodide were performed using ion chromatography coupled with electrospray-ionization tandem mass spectrometry. The limits of detection (LOD) were 0.05 ng/mL for perchlorate, 10 ng/mL for thiocyanate, 500 ng/mL for nitrate, and 0.33 μg/L for iodide. Details of sample preparation and instrumental analysis are reported elsewhere. Creatinine concentrations were measured using an Enzymatic Roche Cobas 6000 system with an LOD of 3.17 mg/dL. For concentrations below the LOD, values were imputed as LOD divided by the square root of two. For quality assurance, 50 duplicate samples (4%) were analyzed alongside study samples. The median relative percent difference (interquartile range) was 3% (2–6%) for perchlorate, 4% (2–5%) for thiocyanate, 4% (2–7%) for nitrate, 4% (2–7%) for creatinine, and 4% (2–7%) for iodide.

2.5. Exposure Metrics and Urine Dilution

We computed a weighted average concentration (C _avg) of perchlorate, thiocyanate, nitrate, creatinine, and iodide across all trimesters, following the approach described in a previous study. The weighted average concentration was calculated using the following formula: C _avg=(C _ind × N _ind + C _pooled × N _pooled)/(N _ind + N _pooled), where C _ind is the concentration in an individual sample, C _pooled is the concentration in a pooled sample, N _ind is the number of individual samples, and N _pooled is the number of composites in a pooled sample. This measure was used to summarize repeated urine measurements collected across pregnancy while accounting for the unequal contribution of individual and pooled urine samples based on the number of original urine samples represented by each pooled specimen.

We also calculated trimester-specific weighted averages using the same formula to examine trimester-specific associations among prenatal exposure to perchlorate, thiocyanate, and nitrate and child neurodevelopmental outcomes. Additionally, we calculated the perchlorate equivalent concentration (PEC) to estimate the combined inhibitory effects of perchlorate, thiocyanate, and nitrate on NIS function. PEC was derived using relative potency factors (RPFs) for iodine uptake inhibition (IUI): $C_{\mathrm{P}\mathrm{E}\mathrm{C}}=C_{\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{c}\mathrm{h}\mathrm{l}\mathrm{o}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{e}}+\frac{C_{\mathrm{t}\mathrm{h}\mathrm{i}\mathrm{o}\mathrm{c}\mathrm{y}\mathrm{a}\mathrm{a}\mathrm{n}\mathrm{t}\mathrm{e}}}{17.6}+\frac{C_{\mathrm{n}\mathrm{i}\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{e}}}{150}$ . , This composite metric captures the combined inhibitory potential of the three NIS inhibitors; therefore, associations observed for PEC are not necessarily expected to mirror those observed for the individual component compounds or for statistical mixture analyses. To account for urinary dilution, urinary creatinine concentration was included as an independent covariate in statistical models, controlling for hydration status while preserving the statistical independence of exposure metrics from creatinine effects. , We used creatinine rather than specific gravity to adjust for urine dilution because creatinine measurements were available for all urine samples, whereas specific gravity values were missing for a subset of samples.

2.6. Covariates

We constructed a directed acyclic graph (DAG) to identify potential confounders (Figure S1), selecting variables a priori based on previous studies. , The final set of nine covariates included maternal age at delivery (<35 vs ≥ 35 years), maternal race/ethnicity (non-Hispanic White vs other, which included Hispanic, Asian, Black, and multiracial persons), maternal education (less than bachelor’s vs Bachelor’s or higher) as a proxy for socioeconomic status, child sex (male vs female), passive smoking during pregnancy (no, yes, or unknown) to minimize self-report bias, maternal prepregnancy body mass index (BMI) (<25 vs ≥ 25 kg/m²), parity (≤1 vs > 1), year of birth (2007-2011, 2012-2015, or 2016-2020) to account for temporal trends in ASD prevalence, based on results from a generalized additive model (GAM) (Figure S2), and urinary iodine concentration (UIC) group during pregnancy. Maternal UIC group was classified using each participant’s weighted-average uncorrected UIC according to the WHO-recommended cutoff for pregnant women (<150 μg/L vs ≥ 150 μg/L), which is based on uncorrected UIC (μg/L). Because thyroid-related clinical factors may lie on pathways related to thyroid-disrupting mechanisms, maternal thyroid disease history and thyroid medication use were not included in the primary models. As a sensitivity analysis, we additionally adjusted for maternal thyroid conditions, defined as a reported history of thyroid disorders and/or thyroid medication use during pregnancy based on questionnaires and medication records, to assess the robustness of the findings.

2.7. Statistical Analyses

We compared maternal demographic characteristics across the three child neurodevelopmental classification groups. Descriptive statistics were summarized using counts (n) and proportions (%) for categorical variables and means with standard deviations for continuous variables. Group differences were assessed using chi-squared tests for categorical variables and Kruskal–Wallis tests for continuous variables. To assess the potential for selection bias arising from differential urine sample availability across pregnancy, we additionally compared maternal demographic characteristics between participants with and without first-trimester urine samples. Geometric means (GM) and percentiles were calculated to describe the distributions of perchlorate, thiocyanate, nitrate, and PEC overall and by trimester. Intraclass correlation coefficients (ICCs) were computed using ICCest­() function in R to assess within-person variability across trimesters. Because the exposure variables were right-skewed, concentrations were log₂-transformed before analysis. Pearson’s correlation coefficients among the log₂-transformed concentrations were visualized in a heatmap. All analyses were performed in R version 4.3.1 (R Core Team, https://www.R-project.org/). Statistical significance was defined as p < 0.05.

We estimated relative risks (RRs) for ASD and non-TD, compared with TD, in association with prenatal average concentrations of perchlorate, thiocyanate, nitrate, and PEC using Poisson regression models with robust standard errors. False discovery rate (FDR) corrections were applied to account for multiple comparisons. As a sensitivity analysis, we compared estimates from two models: (1) the primary model including urinary creatinine as a covariate and (2) a model using raw (unadjusted) concentrations.

To evaluate the joint effects of perchlorate, thiocyanate, and nitrate, we applied quantile g-computation and Bayesian Kernel Machine Regression (BKMR) with multiple parallel chains. Unlike the PEC metric, which was used as a biologically informed composite exposure metric representing the cumulative inhibitory potential of the three NIS inhibitors based on their relative potencies, quantile g-computation was used to estimate the overall joint effect of the exposure mixture under an approximately linear and additive modeling framework, whereas BKMR was additionally used to evaluate potential nonlinear exposure–response relationships and interactions among mixture components. − Quantile g-computation was implemented using four quantiles and a Poisson distribution to estimate the overall linear mixture effect and to determine the relative weight of each compound’s contribution. BKMR models were run using five chains of 5000 iterations each, following a 500-iteration burn-in, under default parameter settings. From the BKMR model, we assessed the nonlinear overall mixture association by estimating changes in RRs for ASD and non-TD when all three exposure concentrations were simultaneously set at percentiles ranging from 10th to 95th (in 5-percentile increments) relative to the 50th percentile. We further estimated univariate exposure–response functions with 95% confidence intervals (CIs), while holding the other two compounds at their median values. In addition, we evaluated bivariate exposure–response surfaces for each compound pair, while fixing the second compound at the 25th, 50th, and 75th percentiles and the third compound at the 50th percentile. , All models were adjusted for nine covariates described above.

We conducted stratified analyses by trimester, child sex, maternal UIC group, and passive smoking during pregnancy to explore the potential effect modification. Previous studies suggest that these factors may influence associations between prenatal exposure to endocrine-disrupting chemicals (EDCs) and child neurodevelopment. Several studies have reported trimester-specific associations between prenatal exposure to perchlorate, thiocyanate, nitrate, phthalates, and child neurodevelopmental outcomes, , while a systematic review found limited evidence linking prenatal EDC exposure with autistic traits. Another review identified sex-specific differences in neurobehavioral development following prenatal EDC exposure. Because perchlorate, thiocyanate, and nitrate disrupt thyroid function, their neurodevelopmental impacts may vary depending on maternal iodine status and thyroid hormone balance, which are essential for fetal brain development. ,,, Passive smoking during pregnancy may also modify associations through its effects on thyroid hormone regulation and oxidative stress. To evaluate potential effect modification, interaction terms between each exposure and the stratifying variables were included in the regression models, and stratified analyses were conducted following the same procedures described above.

3. Results

3.1. Participant Characteristics

Among the 310 mother–child pairs, 136 mothers (43.9%) delivered female infants and 174 (56.1%) delivered male infants. Most mothers (71%) reported no exposure to passive smoking during pregnancy. A total of 195 mothers (62.9%) had UIC <150 μg/L, whereas 115 (37.1%) had UIC ≥150 μg/L. The proportion of male children was higher in the ASD group than in the TD and non-TD groups. In addition, the median urinary creatinine concentration was highest in the ASD group, followed by the non-TD and TD groups (Table ). Overall, concentrations of perchlorate, thiocyanate, nitrate, and PEC tended to be higher among mothers with UIC ≥150 μg/L. Mothers exposed to passive smoking during pregnancy had higher thiocyanate concentrations than those not exposed (Table S1). Participant characteristics were generally comparable between mothers with and without first-trimester urine samples, with no statistically significant differences observed for any of the characteristics evaluated (Table S2).

1. Participant Characteristics of the 310 Mother–Child Pairs Included in the Present Study .

	All	TD	ASD	non-TD	p-value
	(N = 310)	(N = 191)	(N = 74)	(N = 45)
maternal age at delivery (years), n (%)					0.621
20–34	163 (52.6)	101 (52.9)	36 (48.6)	26 (57.8)
35–50	147 (47.4)	90 (47.1)	38 (51.4)	19 (42.2)
parity, n (%)					0.035
≤1	129 (41.6)	87 (45.5)	23 (31.1)	19 (42.2)
>1	174 (56.1)	103 (53.9)	47 (63.5)	24 (53.3)
unknown	7 (2.3)	1 (0.5)	4 (5.4)	2 (4.4)
maternal race/ethnicity, n (%)					0.871
non-Hispanic white	162 (52.3)	102 (53.4)	37 (50.0)	23 (51.1)
Hispanic and others	148 (47.7)	89 (46.6)	37 (50.0)	22 (48.9)
maternal education, n (%)					0.032
less than bachelor’s degree	148 (47.7)	80 (41.9)	42 (56.8)	26 (57.8)
Bachelor’s degree or higher	162 (52.3)	111 (58.1)	32 (43.2)	19 (42.2)
child sex, n (%)					0.019
male	174 (56.1)	98 (51.3)	52 (70.3)	24 (53.3)
female	136 (43.9)	93 (48.7)	22 (29.7)	21 (46.7)
passive smoking during pregnancy, n (%)
no	220 (71.0)	142 (74.3)	46 (62.2)	32 (71.1)	0.202
yes	55 (17.7)	27 (14.1)	21 (28.4)	7 (15.6)
unknown	35 (11.3)	22 (11.5)	7 (9.5)	6 (13.3)
maternal prepregnancy BMI (kg/m2), n (%)
≤24.9	136 (43.9)	92 (48.2)	26 (35.1)	18 (40.0)	0.156
>24.9	174 (56.1)	99 (51.8)	48 (64.9)	27 (60.0)
year of birth, n (%)					0.192
2007-2011	141 (45.5)	97 (50.8)	27 (36.5)	17 (37.8)
2012-2015	125 (40.3)	69 (36.1)	34 (45.9)	22 (48.9)
2016-2020	44 (14.2)	25 (13.1)	13 (17.6)	6 (13.3)
Urinary iodine concentration (μg/L), n (%)					0.538
≥150	115 (37.1)	69 (36.1)	26 (35.1)	20 (44.4)
<150	195 (62.9)	122 (63.9)	48 (64.9)	25 (55.6)
creatinine concentration (mg/dl), median (IQR)	72.4 (50.9, 96.9)	67.8 (47.3, 91.0)	81.5 (58.7, 101.6)	78.5 (59.7, 101.5)	0.025
urinary concentrations (ng/mL), median (IQR)
Perchlorate	3.9 (2.6, 5.7)	3.8 (2.7, 5.7)	3.8 (2.5, 5.4)	4.3 (3.0, 5.8)	0.484
thiocyanate	969 (618, 1399)	901 (567, 1331)	1061 (620, 1360)	1223 (687, 1680)	0.058
nitrate	52022 (37255, 70400)	51088 (37550, 67704)	52138 (37063, 73625)	59400 (36300, 73688)	0.500
PEC	415 (310, 544)	412 (302, 528)	407 (310, 550)	467 (342, 641)	0.215

^aAbbreviations: TD, typical development; ASD, autism spectrum disorder; non-TD, nontypical development; BMI, body mass index; IQR, interquartile range; PEC, perchlorate equivalent concentration.

^b P-values were obtained using chi-squared tests for categorical variables and Kruskal–Wallis tests for continuous variables.

^cIncludes Hispanic persons, Black persons, Asian persons, and persons with other races.

3.2. Maternal Perchlorate, Thiocyanate, Nitrate Concentrations, and PEC Levels

Perchlorate, thiocyanate, and nitrate were detected in nearly all urine samples. Nitrate concentrations were substantially higher than those of the other two compounds (overall GM: nitrate = 45,885 ng/mL; thiocyanate = 893 ng/mL; and perchlorate = 3.5 ng/mL) (Table ). Compared with the National Health and Nutrition Examination Survey (NHANES) data, urinary perchlorate, thiocyanate, and nitrate concentrations in our study population were generally similar to those reported among pregnant women in the U.S. general population (Table S3). Across pregnancy, ICCs calculated for all trimesters combined were lower than trimester-specific ICCs. Perchlorate (1st = 0.39, second = 0.23, and third = 0.27) and nitrate (1st = 0.31, second = 0.32, and third = 0.30) showed greater within-person variability across trimesters, whereas thiocyanate displayed lower variability (1st = 0.72, second = 0.79, and third = 0.79) (Table ). The three compounds were moderately correlated (r = 0.23–0.35), and PEC was highly correlated with nitrate (r = 0.94) (Figure S3). No significant differences in log₂-transformed concentrations of perchlorate, thiocyanate, nitrate, or PEC were observed among the three child-neurodevelopment groups, either overall or by trimester (Figure S4).

2. Distribution of Perchlorate, Thiocyanate, Nitrate, and Perchlorate Equivalent Concentration (PEC) in Maternal Urine Samples Collected from 2007 to 2020 ^,

	LOD (ng/mL)	N	detection frequency		concentration (ng/mL)	ICC
				GM	percentile
					5th	25th	50th	75th	95th
perchlorate
total	0.05	1092	100	3.5	1.1	2.2	3.3	5.5	14.1	0.17 (0.11, 0.23)
first trimester		169	100	3.8	1.1	2.3	3.5	6.3	21.5	0.39 (0.15, 0.56)
second trimester		399	100	3.6	1.2	2.2	3.4	5.5	11.6	0.23 (0.10, 0.36)
third trimester		524	100	3.4	1.2	2.2	3.3	5.1	14.3	0.27 (0.15, 0.38)
thiocyanate
total	10	1092	99.9	893	262	550	933	1440	2875	0.63 (0.58, 0.68)
first trimester		169	100	856	234	545	907	1290	3194	0.72 (0.59, 0.81)
second trimester		399	100	928	265	554	983	1495	2956	0.79 (0.73, 0.83)
third trimester		524	99.8	878	272	552	893	1440	2698	0.79 (0.74, 0.83)
nitrate
total	500	1092	99.7	45885	13965	32875	50250	70200	125450	0.22 (0.16, 0.29)
first trimester		169	100	48267	15800	36000	51000	71800	125200	0.31 (0.06, 0.50)
second trimester		399	100	46728	14250	32650	51000	70300	122200	0.32 (0.19, 0.44)
third trimester		524	99.4	44521	12175	32000	48300	69525	127700	0.30 (0.18, 0.41)
PEC
total	-	1092	-	386	129	277	397	572	987	0.25 (0.19, 0.32)
first trimester		169	-	400	135	299	425	586	911	0.14 (<0.001, 0.37)
second trimester		399	-	396	135	280	403	581	1003	0.32 (0.19, 0.44)
third trimester		524	-	375	124	262	394	554	967	0.38 (0.27, 0.48)

^aNote: ICC was calculated using log₂-transformed values, and the confidence intervals (CIs) were estimated based on the exact confidence limit equation in Searle et al. to address unbalanced measurement. ,

^bAbbreviations: PEC, perchlorate equivalent concentration; LOD, limit of detection; GM, geometric mean; ICC, intraclass correlation coefficient.

3.3. Associations of Prenatal Perchlorate, Thiocyanate, Nitrate, and PEC with Child Neurodevelopmental Outcomes

We estimated the individual associations between prenatal exposure to perchlorate, thiocyanate, nitrate, and PEC and the child neurodevelopmental outcomes. Several significant associations were observed in the stratified analyses. Trimester-specific analyses indicated that higher prenatal perchlorate concentrations in the first trimester were associated with greater likelihood of non-TD compared with TD (RR = 1.61, 95% CI: 1.08, 2.40), whereas third-trimester perchlorate concentrations were inversely associated with non-TD likelihood (RR = 0.77, 95% CI: 0.60, 0.98) (Figure , Table S4).

2.

Relative risks (RRs) and 95% confidence intervals (CIs) of non-TD compared with TD in relation to prenatal exposure to (A) perchlorate, (B) thiocyanate, (C) nitrate, and (D) PEC, overall and stratified by trimester, child sex, maternal urinary iodine concentration (UIC), and passive smoking during pregnancy. All models were adjusted for nine covariates, with additional adjustment for urinary creatinine. Bolded RRs denote nominal statistical significance, and asterisks (*) indicate associations that remained statistically significant after Benjamini–Hochberg false discovery rate (FDR) adjustment. Abbreviations: ASD, autism spectrum disorder; PEC, perchlorate equivalent concentration; p-int, interaction p-value.

Among female children, higher perchlorate concentrations were associated with greater likelihood of ASD (per 2-fold increase: RR = 1.41, 95% CI: 1.04, 1.91), and higher thiocyanate, nitrate, and PEC were associated with greater likelihood of non-TD (RR = 1.50, 95% CI: 1.07, 2.10 for thiocyanate, RR = 2.04, 95% CI: 1.05, 3.96 for nitrate, and RR = 2.98, 95% CI: 1.50, 5.96 for PEC). Among mothers with UIC ≥150 μg/L, higher nitrate concentrations were associated with greater likelihood of ASD (RR = 1.84, 95% CI: 1.04, 3.25), and higher thiocyanate (RR = 1.37, 95% CI: 1.02, 1.85) and PEC (RR = 2.20, 95% CI: 1.15, 4.19) were associated with greater likelihood of non-TD (Figure , Table S4).

3.

Relative risks (RRs) and 95% confidence intervals (CIs) of ASD compared with TD in relation to prenatal exposure to (A) perchlorate, (B) thiocyanate, (C) nitrate, and (D) PEC, overall and stratified by trimester, child sex, maternal urinary iodine concentration (UIC), and passive smoking during pregnancy. All models were adjusted for nine covariates, with additional adjustment for urinary creatinine. Bolded RRs denote nominal statistical significance, and asterisks (*) indicate associations that remained statistically significant after Benjamini–Hochberg false discovery rate (FDR) adjustment. Abbreviations: ASD, autism spectrum disorder; PEC, perchlorate equivalent concentration; p-int, interaction p-value.

Additionally, among mothers not exposed to passive smoking, higher PEC was associated with greater likelihood of non-TD (RR = 1.92, 95% CI: 1.02, 3.63), whereas among mothers exposed to passive smoking, higher PEC was associated with greater likelihood of ASD (RR = 2.19, 95% CI: 1.09, 4.40). Overall, the direction and magnitude of the associations were consistent regardless of whether urinary creatinine adjustment was applied (Figures and , Table S4).

Additional adjustment for maternal thyroid disease history and thyroid medication use produced results that were generally similar in magnitude and direction to those observed in the primary analyses. However, some differences were observed in analyses stratified by the UIC group. Specifically, several associations among women with UIC ≥150 μg/L were attenuated and no longer statistically significant, whereas positive associations of thiocyanate and PEC with the likelihood of non-TD became statistically significant among women with UIC <150 μg/L. These findings suggest that thyroid-related clinical factors may contribute to heterogeneity across UIC subgroups; however, the overall interpretation of the findings remained unchanged. Further studies are needed to better understand the role of thyroid-related clinical factors and iodine status in these associations (Table S5).

3.4. Associations between Prenatal Perchlorate, Thiocyanate, and Nitrate Mixture and Child Neurodevelopmental Outcomes

On the basis of the quantile g-computation model, no significant overall mixture association of perchlorate, thiocyanate, and nitrate was observed on the likelihood of ASD and non-TD compared with TD. Although no overall effect was detected, the three compounds contributed in different directions to the likelihood of ASD and non-TD under the mixture exposure scenario (Figure ).

4.

Relative risks (RRs) and 95% confidence intervals (CIs) for autism spectrum disorder (ASD) and nontypical development (TD) compared with TD in relation to prenatal mixture exposure to perchlorate, thiocyanate, and nitrate during pregnancy, overall and stratified by trimester, child sex, maternal urinary iodine concentration (UIC), and passive smoking during pregnancy. All models were adjusted for nine covariates, with additional adjustment for urinary creatinine.

BKMR results indicated increasing trends in likelihood of non-TD relative to TD with higher overall mixture concentrations, although these trends were not statistically significant. The three compounds also exhibited U-shaped exposure–response relationships with the likelihood of ASD and non-TD relative to TD, but these associations were not statistically significant (Figure ).

5.

Associations of mixture exposures to perchlorate, thiocyanate, and nitrate with the risks of ASD (A) and non-TD (B) compared with TD, estimated using the Bayesian kernel machine regression (BKMR) model. Estimates are presented on the probit scale. (Left panel) Overall mixture effects (estimates and 95% confidence intervals [CIs]) represent changes in relative risks (RRs) for ASD and non-TD when all three compounds were simultaneously set at varying percentiles compared with their median values. (Middle panel) Univariate exposure–response functions and 95% CIs for each compound, with the other compounds fixed at their median values. (Right panel) Bivariate exposure–response functions for each pair of compounds, with the second compound fixed at the 25th, 50th, and 75th percentiles, and the third compound fixed at the median value.

4. Discussion

To our knowledge, this is the first study to examine associations between prenatal exposure to perchlorate, thiocyanate, and nitrate and both ASD and non-TD at three years of age. Leveraging multiple maternal urine samples from a cohort enriched for familial ASD likelihood, we observed that prenatal exposure to these NIS inhibitors was associated with greater likelihood of ASD and non-TD compared with TD in several subgroups. Among female children, perchlorate was positively associated with likelihood of ASD, and thiocyanate and nitrate were positively associated with likelihood of non-TD. Among mothers with UIC ≥150 μg/L, nitrate and thiocyanate showed positive associations with ASD and non-TD, respectively. Among mothers exposed to passive smoking, higher PEC was associated with greater likelihood of ASD, whereas among mothers not exposed to passive smoking, higher PEC was associated with greater likelihood of non-TD. No significant overall mixture effects of the three compounds were observed. Together, these findings suggest context-specific associations of prenatal NIS inhibitors with early neurodevelopment, potentially through thyroid-mediated pathways.

Our findings are consistent with those of two previous epidemiologic studies. An Italian randomized controlled trial (n = 487) reported that children born to mothers with urinary perchlorate levels in the highest decile during pregnancy had higher odds of scoring in the lowest decile for total or verbal IQ at age three. Similarly, we found that higher first-trimester perchlorate levels were associated with a greater likelihood of non-TD compared with TD. A longitudinal study in China (n = 1028) reported that higher maternal thiocyanate, nitrate, and PEC were associated with lower MDI scores in the first trimester and that higher thiocyanate was linked to lower MDI scores in the third trimester. Consistent with these findings, we observed the strongest associations for thiocyanate, nitrate, and PEC. Even though thiocyanate and nitrate have lower IUI potency than perchlorate, their substantially higher exposure levels may still pose meaningful neurodevelopmental concerns. Perchlorate is approximately 17 and 150 times more potent for IUI than thiocyanate and nitrate, respectively, yet maternal exposure concentrations were roughly 250 times higher for thiocyanate and 13,000 times higher for nitrate than for perchlorate based on geometric mean concentrations.

A biologically plausible mechanism linking prenatal exposure to these compounds with neurodevelopment alterations involves their inhibition of iodide uptake via the NIS, thereby reducing TH synthesis. THs are essential for fetal brain development, including neurogenesis, neural differentiation, synapse formation, and myelination. , Before maturation of the fetal thyroid gland, the developing brain depends primarily on maternal TH supply. Perchlorate, thiocyanate, and nitrate compete with iodide for NIS transport, potentially reducing iodide availability and impairing TH production. ,,, Although few animal studies have evaluated downstream neurodevelopmental effects, available findings support this mechanism. In pregnant rats exposed to perchlorate in drinking water, dose-dependent reductions in TH levels were observed in both the thyroid gland and serum, accompanied by reduced TH-responsive gene expression in the fetal cortex. Additional studies have reported reduced baseline synaptic transmission in hippocampal field potentials across exposure levels. Notably, we observed significant associations even among participants with UIC ≥150 μg/L, suggesting that these inhibitors may exert neurodevelopmental effects despite sufficient iodine concentrations in the bloodstream. Further studies are needed to elucidate the interplay between iodine status, thyroid regulation, and NIS inhibition.

We also found sex-specific associations between thiocyanate and nitrate exposure and the likelihood of ASD and non-TD among female children. The potential mechanisms underlying these sex differences remain to be clarified. One potential explanation involves the sex hormones. Prenatal testosterone plays an important role in fetal brain development and has been implicated in ASD. , Thyroid hormones and sex steroids are also known to interact during fetal development, and thyroid dysfunction can influence circulating sex steroid levels. Because perchlorate, thiocyanate, and nitrate may influence thyroid-related pathways, sex-specific susceptibility cannot be excluded. However, the present study was not designed to evaluate these mechanisms. Therefore, the biological basis of the observed sex-specific associations remains speculative and warrants further investigation.

Among mothers exposed to passive smoking, higher PECstrongly correlated with nitratewas associated with greater likelihood of ASD. Passive smoking during pregnancy has been linked to elevated likelihood of ASD in offspring, although findings remain inconsistent. − Several biological mechanisms have been proposed. Carbon monoxide in cigarette smoke forms carboxyhemoglobin, which crosses the placenta and reduces fetal oxygen delivery, potentially impairing brain development. Reactive oxygen species generated from smoking increase oxidative stress, leading to endothelial dysfunction and reduced nitric oxide (NO) bioavailability in the uteroplacental and fetal–placental circulation. Nitrate is recycled in blood and tissues to generate NO and other reactive nitrogen species. Therefore, maternal exposure to secondhand smoke may enhance nitrate-related effects by altering oxidative balance and NO metabolism, with downstream consequences for thyroid regulation and fetal neurodevelopment. Given the small sample size in stratified analyses, these interactions should be interpreted cautiously but warrant further investigation.

This study has several notable strengths. First, the prospective design allowed for the temporal assessment of prenatal exposures in relation to the clinically confirmed ASD classifications. Second, our study population consisted of families at elevated familial likelihood for ASD, a design that may capture heterogeneity in susceptibility to environmental agents within this high-likelihood populations. Our findings are therefore particularly informative for understanding exposure–outcome relationships in this population. Lastly, multiple urine samples collected within and across trimesters improved the reliability of exposure estimates and enabled trimester-specific analyses. ,

This study also has several limitations. First, the generalizability of our findings may be limited, as the study sample consisted of families with elevated familial likelihood of ASD. Genetic susceptibility, environmental exposures, and gene–environment interactions may differ between families with elevated familial likelihood of ASD and the broader population. Second, residual confounding from unmeasured variables, such as diet and pharmaceuticals use, may be present. Drinking water and dietary intake are major exposure sources for perchlorate, thiocyanate, and nitrate, − while prenatal dietary factors may also affect maternal thyroid function and child neurodevelopment, including ASD-related outcomes. , Because urinary biomarkers primarily reflect recent exposure and detailed dietary information was not available, we were unable to fully account for dietary factors that may influence both exposure levels and neurodevelopmental outcomes. Furthermore, several dietary sources of nitrate and thiocyanate may additionally confer nutritional or physiological benefits, whereas these compounds may contribute to thyroid-disrupting effects through NIS inhibition. Because the present study focused on thyroid-disrupting mechanisms related to NIS inhibition, the absence of detailed dietary measures limits our ability to disentangle exposure-related effects from potential dietary confounding and to distinguish potential beneficial effects associated with dietary sources from adverse effects related to NIS inhibition. Third, most biospecimens were collected during the second and third trimesters due to variability in enrollment timing, limiting assessment of early pregnancy exposures. Fourth, subgroup-specific associations warrant careful interpretation, particularly because significant findings were observed primarily in stratified analyses, whereas overall mixture analyses were largely null. Although perchlorate, thiocyanate, and nitrate share a common mechanism of NIS inhibition, susceptibility to thyroid disruption may vary according to factors such as child sex, maternal iodine status, passive smoking exposure, and timing of exposure. Consequently, associations present within susceptible subgroups may not be apparent in overall mixture analyses. In addition, reduced subgroup sample sizes may have limited statistical power to reliably detect stable mixture effects, potentially contributing to variability and statistical instability in subgroup-specific findings. Furthermore, the large number of subgroup comparisons increases the possibility that some statistically significant associations may represent chance findings despite correction for multiple testing. Accordingly, these findings should be considered hypothesis-generating rather than confirmatory and require replication in larger studies. Fifth, because active smoking during pregnancy was rare in this cohort, tobacco-related sources for thiocyanate exposure could not be fully evaluated. Although passive smoking was accounted for in the analyses, it may not fully capture variability in tobacco-related thiocyanate exposure. Accordingly, associations involving thiocyanate, particularly subgroup-specific findings, should be interpreted cautiously. Lastly, we did not evaluate maternal thyroid function, which constrains our ability to elucidate mechanistic pathways or potential synergistic interactions among these compounds.

In conclusion, although overall and mixture analyses were largely null, this study identified potential context-specific associations between prenatal exposure to perchlorate, thiocyanate, and nitrate and child neurodevelopmental outcomes by trimester, child sex, maternal UIC group, and passive smoking, some of which persisted after FDR adjustment. Because these findings were observed in a cohort with an elevated familial likelihood of ASD, they should be interpreted with caution and warrant replication in independent, more diverse populations, as well as further research integrating thyroid biomarkers and mechanistic studies to clarify these patterns and strengthen causal inference.

Glossary

Abbreviations

ADHD

attention deficit hyperactivity disorder

ASD

autism spectrum disorder

BKMR

Bayesian kernel machine regression

BMI

body mass index

CDC

Centers for Disease Control and Prevention

DAG

directed acyclic graph

DSM-5

diagnostic and statistical manual of mental disorders, fifth edition

EDCs

endocrine disrupting chemicals

FDR

false discovery rate

FT4

free thyroxine

GAM

generalized additive model

ICCs

intraclass correlation coefficients

IQ

intelligence quotient

LOD

limit of detection

MARBLES

markers of autism risk in babies–learning early signs

MDI

mental development index

MSEL

Mullen scales of early learning

NIS

sodium iodide symporter

NO

nitric oxide;

Non-TD

nontypical development

PEC

perchlorate equivalent concentrations

RR

relative risk

TD

typical development

TSH

thyroid-stimulating hormone

THs

thyroid hormones

WHO

World Health Organization.

Data will be made available on request.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.6c03054.

• Distribution of maternal perchlorate, thiocyanate, and nitrate concentrations; mother’s demographic characteristics by first-trimester sample availability; comparison of associations before and after creatinine adjustment; comparison of associations before and after additional adjustment for mother’s thyroid disease history or mediation use; directed acyclic graph (DAG) for covariate selection; cutoff of child birth year; correlation heatmap among exposure; and histogram of exposure concentrations by trimester and child neurodevelopment group (PDF)

HJ (Writingoriginal draft, methodology, formal analysis, and conceptualization), DHB (writingreview and editing), DJT (writingreview and editing and methodology), RJS (writingreview and editing), HMS (writingreview and editing, conceptualization, methodology, funding acquisition, and supervision).

This research was supported by grants from the National Institute of Environmental Health Sciences (R21-ES033389, R01-ES020392, U24ES028533, R01ES028089, P01ES011269) and was supported in part by the UC Davis MIND Institute Intellectual and Developmental Disabilities Research Center (U54HD079125); additional support was provided by the U.S. EPA STAR (#R829388, R833292, and RD835432) and the Simons Foundation (SFARI #863967, RJS). This work was also supported by the National Institutes of Health (U2CE026542).

The MARBLES study protocol and this study were approved by the institutional review boards for the State of California and the University of California-Davis (UC-Davis). Participants provided written informed consent before collection of any data and AI-assisted technologies in the writing process During the preparation of this work, the authors used ChatGPT to improve the language and readability of the manuscript. After using this tool, the authors reviewed and edited the content as needed and took full responsibility for the content of the published article.

The authors declare the following competing financial interest(s): RJS received travel reimbursement for invited talks, and consultancy fees from the Beasley Allen Law Firm and Linus Biotechnology, Inc. DHB received consultancy fees from Linus Biotechnology, Inc. RJS received support for MARBLES from the Simons Foundation. The other authors declare they have no actual or potential competing financial interests.