Prenatal fluoride exposure and autistic behaviors in preschool-aged children: the Maternal-Infant Research on Environmental Chemicals (MIREC) cohort study

Abstract

Background

The prevalence of autism spectrum disorder has risen in recent decades. Given the growing evidence that prenatal fluoride exposure may be neurotoxic, we examined associations between prenatal fluoride exposure and parent-reported autistic behaviors in preschool-aged children.

Methods

We studied 453 mother-child pairs using data from the Maternal-Infant Research on Environmental Chemicals (MIREC) study, a prospective Canadian birth cohort. Autistic behaviors were assessed in children at 3 to 4 years using the Social Responsiveness Scale-Second Edition (SRS-2) Preschool Form, where a higher score indicates more autistic behaviors. We estimated prenatal fluoride exposure using three methods: (i) maternal urinary fluoride adjusted for specific gravity (MUF_SG), from spot urine samples collected at each trimester and the mean calculated across samples, (ii) daily maternal fluoride intake, based on self-reported consumption of tap water, coffee, and tea during the first and third trimesters, and (iii) water fluoride concentration in tap water. We used multivariable linear regression models to estimate associations between the SRS-2 scale T-scores and each fluoride exposure separately. We used multivariable logistic regression to estimate the association between each prenatal fluoride exposure and an elevated SRS-2 total T-score (i.e., 90th percentile or higher). Potential effect modification of MUF_SG was examined by child sex, daily folic acid supplementation, and plasma total folate in pregnancy.

Results

The mean SRS-2 total T-score for children aged 3 to 4 years was 45.3 (SD = 6.1, range = 34 to 85). The median MUF_SG concentration was 0.43 mg/L (interquartile range = 0.33 mg/L). MUF_SG was not significantly associated with the SRS-2 total T-score in multivariable linear regression (β = -0.16; 95% CI, -1.70, 1.39) or logistic regression (OR = 0.76; 95% CI, 0.29, 1.96). Similarly, estimated daily fluoride intake and water fluoride concentration were not associated with the SRS-2 total T-score. No effect modification was observed.

Conclusions

There was no evidence of an association between prenatal fluoride exposure and autistic behaviors in preschool-aged children, in contrast to previous MIREC research findings on lead and phthalates. Given that this cohort has relatively few children with high SRS-2 scores, further research is needed in other groups of children to more fully explore this association.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12940-025-01237-9.

Background

Autism Spectrum Disorder (ASD) is a neurodevelopmental condition characterized by impairments in social communication and interactions, as well as repetitive and stereotyped behaviors [1]. The prevalence of ASD has steadily increased over the past three decades. Approximately 1 in 36 children in the United States (U.S.) [2] and 1 in 63 in Canada [3] are diagnosed with ASD. The development of autism is influenced by a combination of genetic and environmental factors [4]. Epidemiologic studies have linked increased risk of autism-related symptoms with exposure to air pollution [5], bisphenols [6], lead [7], and phthalates [8].

Fluoride has been linked with oxidative stress, inflammation, and immuno-excitotoxicity, which are cellular and molecular mechanisms associated with ASD [9, 10]. Fluoride is a mineral that is released from rocks into water and soil, as well as the environment through industrial processes. It is found in foods, beverages, and dental products and is also added to some public water supplies or salt as a public health strategy for preventing dental caries [11]. Approximately 39% of the Canadian population has access to community water fluoridation [12] compared to about 72% in the U.S [13]. In both Canada and the U.S., 0.7 mg/L is the recommended water fluoride concentration that is considered effective for preventing dental caries [12, 14, 15]. The maximum acceptable fluoride concentration in drinking water recommended by the World Health Organization [11] and Health Canada [16] is 1.5 mg/L.

Over the past decade, growing evidence has suggested that prenatal exposure to fluoride may be neurotoxic, which has caused increased scrutiny of the safety of fluoridation. A recent systematic review and meta-analysis conducted by the U.S. National Toxicology Program reported inverse associations between fluoride exposure and children’s intelligence quotient (IQ) scores [17]. These associations remained significant when analyses were restricted to studies with low risk of bias and urinary fluoride concentrations below 1.5 mg/L. Given that urinary fluoride captures all sources of fluoride, the review could not determine whether the 0.7 mg/L recommended level of fluoride in drinking water negatively affects children’s IQ. Fluoride exposure has also been associated with behavioral problems in children, including symptoms of attention-deficit hyperactivity disorder [18]. These findings raise concerns about fluoride’s potential for neurotoxicity, but the evidence remains limited with few studies investigating the association between fluoride exposure and autistic behaviors.

Two studies have investigated the relationship between fluoride exposure and autistic behaviors [19, 20]. In a U.S. prospective cohort study of 229 mother-child pairs, higher maternal urinary fluoride (MUF) concentrations (median = 0.76 mg/L) were associated with more parent-reported autistic behaviors in children at 3 years of age [19]. In contrast, a prospective cohort study in northern Sweden found no association between prenatal MUF (median = 0.72 mg/L) or postnatal child urinary fluoride (median = 0.86 mg/L) and autistic behaviors at 4 years of age in over 300 mother-child pairs [20]. Another study from Canada found that among preschool-aged girls, higher prenatal exposure to fluoride in drinking water was linked with poorer inhibitory control and decreased cognitive flexibility [21]. These executive function deficits are also commonly observed in ASD [22]. Further, human and animal studies have shown that fluoride exposure can reduce the expression of nicotinic acetylcholine receptor subunits that have been implicated in neurodevelopmental disorders, including ASD [23, 24].

Given the potential neurotoxicity of early-life exposure to fluoride and the rising prevalence of ASD, further evaluation of fluoride as a risk factor for ASD is needed. Specifically, prospective cohort studies with fluoride biomarkers and validated questionnaires designed to assess autistic traits are needed to understand whether there is a link between prenatal exposure to fluoride and autistic behaviors. This study aimed to examine associations between prenatal fluoride exposure and autistic behaviors among Canadian preschool-aged children living in fluoridated and non-fluoridated communities. We also investigated the potential for sex-specific effects given previous findings of sex differences associated with fluoride exposure [25, 26] and the higher prevalence of ASD in males [27]. Finally, we evaluated prenatal folic acid supplementation and plasma total folate concentration as potential effect modifiers given previous findings showing that folic acid may mitigate adverse effects of neurotoxicants [7, 8, 28].

Methods

Study participants

We used data from the Maternal-Infant Research on Environmental Chemicals (MIREC) study, a prospective Canadian birth cohort that enrolled 2001 pregnant women from 10 cities (11 sites) between 2008 and 2011 during their first trimester of pregnancy. The aim of the MIREC study was to examine potential effects of prenatal exposure to environmental chemicals on maternal and child health. Women were eligible for recruitment if they could communicate in English or French, were < 14 weeks gestation, ≥ 18 years old, planning to deliver at a local hospital and agreed to provide a cord blood sample. The study included 1983 women who consented to participate. Further information about the study design and recruitment process are detailed in a cohort profile [29].

A neurodevelopmental follow-up study called MIREC-Child Development (MIREC-CD) Plus was conducted between 2013 and 2015 when children were between 15 months and 5 years of age. Due to resource limitations, MIREC families from six (seven sites) of the 10 recruitment cities were invited to participate; these cities represented the largest sample sizes from the original MIREC study and included Vancouver, Toronto, Hamilton, Halifax, Kingston, and Montreal, as previously described [30]. Among 1459 eligible families, 1207 were contacted and 808 enrolled to participate in the follow-up study. Of these, 610 singleton-born children underwent cognitive and behavioral assessment at 3 to 4 years of age; 604 consented to participate in the Biomonitoring visit, and 595 completed an assessment of autistic behaviors. Our study sample was further restricted to 453 mother-child pairs with complete data on prenatal fluoride exposure and covariates, including prenatal folic acid supplementation and plasma total folate concentration (Fig. 1). A detailed comparison between the full MIREC cohort (n = 1983), the sample eligible to participate in the MIREC-CD Plus (n = 1630), and the mother-child pairs who consented to the Neurodevelopment visit (n = 610) was published by Fisher et al. [30]. Sociodemographic characteristics were generally similar with a few exceptions: the Neurodevelopment sample (relative to the full MIREC cohort and the eligible CD-Plus cohort) had a lower proportion of younger, less educated, lower SES, first trimester smokers, and foreign-born mothers, and a higher proportion of White participants.

Fig. 1.

Study sample flow chart of participants from the MIREC study

All participants in the MIREC study gave informed consent for themselves and their children. This study obtained approval from Research Ethics Boards at Health Canada, Centre hospitalier universitaire (CHU) Sainte-Justine Research Center, ethics committees at participating hospitals, York University, and the University of Toronto for the analysis of secondary data.

Autistic behavioral outcome

During an in-person assessment, trained research personnel asked mothers to complete the Social Responsiveness Scale-Second Edition (SRS-2) Preschool Form, a 65-item questionnaire designed to assess symptoms associated with ASD [31]. Responses were recorded on a 4-point Likert scale ranging from “not true” to “almost always true”. Parent responses are summed to produce an age-standardized total T-score with a mean of 50 and standard deviation (SD) of 10, with a higher score indicating more autistic behaviors. A total T-score ≥ 60 indicates elevated autistic behaviors and a need for further evaluation. The SRS-2 includes five scales: Social Awareness, Social Cognition, Social Communication, Social Motivation, and Restricted Interests and Repetitive Behavior, and two Diagnostic and Statistical Manual of Mental Disorders-Fifth Edition compatible scales: Social Communication and Interaction, and Restricted Interests and Repetitive Behavior. Previous research has reported the instrument to be valid and reliable in preschool-aged children [31–34].

Fluoride exposures

Maternal urinary fluoride (MUF)

Maternal urine spot samples were collected as part of MIREC at each trimester of pregnancy and shipped to the Indiana University School of Dentistry for analysis, as described previously [35]. The fluoride concentration of the samples was assessed using a modified version of the hexamethyldisiloxane (HMDS; Sigma Chemical Co.) microdiffusion method, originally developed by Taves [36] and adapted by Martínez-Mier et al. [37]. This method has been shown to yield the highest recoveries of fluoride for undiluted samples. The method’s limit of detection (LoD) for fluoride concentration was 0.02 mg/L with an intra-day precision >97% at a concentration of >0.2 ppm fluoride, where the standard reference loses linearity. Reference standard solutions were monitored daily by a quality assurance officer for stability; technicians reanalyzed, on a rotating basis, one of three standards daily. In addition, urine-based certified reference materials (PCU-F1305; Institut National de Santé Publique Québec, Laboratoire de Toxicologie) were analyzed every 200 to 300 samples. Among all available spot urine samples for fluoride analysis, two first trimester samples (0.002%) were excluded because their fluoride concentrations exceeded the 5 mg/L upper limit of quantification and thus may not accurately represent the true exposure value. We included only women who had urinary fluoride measured in each trimester due to the variability of fluoride levels throughout pregnancy [38].

Each trimester MUF value was adjusted for specific gravity (SG) to account for variability in urine dilution using Eq. 1:

1

where MUF_SG is the SG-adjusted urinary fluoride concentration (mg/L), MUF_i is the observed fluoride concentration, SG_i is the specific gravity of the urine sample, and SG_M is the median SG for the cohort [39]. The mean MUF_SG concentration was then calculated by taking the mean across all three trimesters for each woman.

MUF was also adjusted for urinary creatinine (MUF_CRE) to account for dilution, using a method adapted by Thomas et al. [40] and described in detail elsewhere [35]. Our prior work has shown that correcting MUF for urinary dilution using SG and creatinine (CRE) methods are interchangeable [35]. We chose to use MUF_SG for our primary analyses due to the larger sample size compared to MUF_CRE.

Estimated daily maternal fluoride intake

We collected information on fluoride intake using self-report questionnaires administered to mothers during their first and third trimesters. The questionnaires inquired about the consumption of tap water, coffee, and tea; black tea can be a particularly important source of fluoride [35, 41–43]. For women who reported drinking tap water during pregnancy, we estimated daily maternal fluoride intake (mg/L) independently for the first and third trimesters by multiplying women’s individual drinking water fluoride concentration (mg/L) by total volume (L) of water, coffee, and tea consumed. Next, we added the expected fluoride intake from tea, using 0.326 mg and 0.260 mg for a 200 mL cup of black tea and green tea if prepared with deionized water, respectively [43]. The final estimate was obtained by taking the mean of the fluoride intake across the two trimesters. See Additional Fig. 1 for the formula used to calculate estimated daily maternal fluoride intake.

Water fluoride concentration

For each of the 10 cities included in the MIREC study, municipal water treatment plant (WTP) drinking water reports were requested from 2008 to 2012. These reports included daily fluoride concentration for WTPs that added fluoride to their public drinking water, and weekly or monthly data for those that did not. In this study, a non-fluoridated site was defined as having a water fluoride level below 0.3 mg/L. This is lower than the recommended optimal fluoride concentration of 0.7 mg/L in drinking water for dental health by Health Canada, recognizing that actual water fluoride levels can vary more widely in practice [12]. For each city, WTP boundaries were identified and matched to the first three letters of the postal codes for women who reported drinking tap water during pregnancy. We then estimated water fluoride concentration for each woman by calculating their geometric mean (mg/L) across pregnancy; more details are described elsewhere [35].

Covariates and potential confounders

We identified predictors of autistic behaviors and potential confounding variables in the relationship between prenatal fluoride exposure and autistic behaviors a priori through a literature review [2, 44–46], which informed the development of a conceptual model (Additional Fig. 2).

Data on sociodemographic characteristics were collected by trained research staff during pregnancy in the first and third trimesters and at delivery by administering questionnaires, as well as through medical chart abstraction. We considered the following maternal characteristics: age at enrollment (years), race (White, other), born in Canada (yes, no), education level (high school or less, more than high school), annual household income (≤ $70,000 CAD, >$70,000 CAD), monthly alcohol consumption (none, >0 drinks), smoked in trimester 1 (yes, no), second-hand smoke in trimester 1 (yes, no), marital status (married, other), first birth (yes, no), and study city (Vancouver, Toronto, Hamilton, Kingston, Montreal, Halifax). Child characteristics included biological sex (male, female), and age at SRS-2 testing (months). The level of stimulation and support in the children’s home environment was examined using the Home Observation for Measurement of the Environment (HOME) Inventory-Revised Edition, where a higher score indicates a more enriched environment [47, 48]. A trained interviewer completed the inventory by conducting in-home observations and interviews with caregiver(s).

Prenatal folic acid supplementation

Data on daily total folic acid intake from supplements was collected by surveying study participants after the first trimester study visit. Mothers were asked to provide the name and description of the supplement they were taking, the product identification number, amount taken each time, and frequency of intake in the last 30 days. As described previously [49], intake levels were categorized as < 400 µg/day (below the recommended intake), 400 to 1000 µg/day (meeting the recommended intake), and >1000 µg/day (exceeding the recommended intake) [50–52].

Plasma total folate concentration

Non-fasting peripheral blood samples were collected at the first and third trimester clinic visits and subsequently shipped to Health Canada Food Laboratory for analysis of plasma folate. Using liquid chromatography-tandem mass spectrometry, the contributions of five individual folate vitamers - tetrahydrofolate (THF), 5,10-methenylTHF, 5-formylTHF, 5-methylTHF, and unmetabolized folic acid - to plasma total folate were calculated, as previously described [49]. Plasma total folate concentration for the first and third trimesters was then determined by summing the five vitamers for each trimester. We categorized plasma total folate concentrations into three groups: < 10th percentile, ≥ 10th and < 80th percentiles, and ≥ 80th percentile, in line with previously reported methodological cutoffs [7].

Statistical analysis

We compared the sociodemographic characteristics of participants who consented to the Neurodevelopment visit from the MIREC-CD Plus with complete SRS-2 outcome, MUF_SG exposure, and covariate data to those excluded from the study due to incomplete data. We then explored the central tendency and distribution of maternal fluoride exposures, as well as SRS-2 scale T-scores for our study sample. For independent data, we used t-tests or Satterthwaite t-tests for unequal variances to compare the means between two groups, while ANOVA or Welch’s test for non-homogeneity was used to compare the means among more than two groups. For categorical covariates, chi-square tests examined associations between groups, while Fisher’s exact test was used when expected frequencies were less than five. When data were non-independent, repeated measures ANOVA was used to compare means across multiple timepoints. Spearman rank correlation (r) examined the relationship between prenatal fluoride exposure measurements, interpreted using Cohen’s conventions: r < 0.3 for small, 0.3 ≤ r < 0.5 for moderate, and r ≥ 0.5 for strong correlations [53]. In addition, we calculated the intraclass correlation coefficient (ICC) to assess the reliability of MUF_SG measurements across the three pregnancy trimesters. The ICC ranges from 0 to 1, with a higher value indicating greater consistency among measurements.

We used linear regression models to estimate the association between SRS-2 total T-score and subscale scores and each fluoride exposure separately. Guided by our conceptual model (Additional Fig. 2), a covariate was included in the model if its p-value was < 0.2 or if its inclusion changed the MUF_SG beta coefficient by more than 10%. Regardless of statistical significance, models were adjusted for child sex, and study city. The final covariate-adjusted models included the following variables: annual household income, maternal second-hand smoke in trimester 1, marital status, first birth, HOME score, child sex, and study city. Residual plots were used to assess model assumptions of linearity and homoscedasticity, along with the Breusch-Pagan test for homoscedasticity. Additionally, Q-Q plots were used to evaluate normality. We checked for multicollinearity using the variance inflation factor and assessed potential outliers and influential observations with Cook’s D and studentized residuals, and leverage using leverage plots. Four observations were identified as outliers and influential points, two of which had high leverage; all four observations had the highest SRS-2 total T-scores in our sample, ranging from 62 to 85. Given the plausibility of these T-scores and the fact that their removal did not alter our findings (Additional Table 1), these observations were retained in our models. Additionally, because the distribution of MUF_SG is right skewed (Additional Fig. 3), we separately modeled the natural log (ln) and quadratic; these transformations of MUF_SG did not improve the model (results not presented), thus we present linear regression results. To address potential selection bias, we applied inverse probability weighting (IPW) in adjusted linear regression models assessing the effect of MUF_SG on SRS-2 scores (Additional Table 2). Since IPW did not alter the model findings, we present the unweighted linear regression results.

We examined potential effect modification by testing the interaction between MUF_SG and the following covariates: (i) child sex, (ii) daily folic acid supplementation (below, at, or above recommended intake levels), (iii) first trimester plasma total folate (< 10th percentile, ≥ 10th and < 80th percentiles, ≥ 80th percentile), and (iv) third trimester plasma total folate (< 10th percentile, ≥ 10th and < 80th percentiles, ≥ 80th percentile), in SRS-2 total T-score models. Child sex was examined due to the higher global prevalence of ASD among males than females [27], and prenatal folate measures were assessed because maternal folate status has been linked to the odds of ASD in offspring such that adequate folic acid intake is associated with lower odds of having a child with ASD [54]. As a secondary aim, logistic regression models were used to examine whether each of our fluoride measures is associated with an increased odds of having an SRS-2 total T-score in the 90th percentile or higher (T-score ≥ 53) compared to a score below the 90th percentile. We selected the 90th percentile as the cut-point for dichotomizing total T-scores, as opposed to clinically elevated scores (i.e., T-score ≥ 60), due to the limited number of children in our sample with elevated scores (n = 8). Further, we modeled MUF_SG using linear splines with one knot, placed at 0.7 mg/L (optimal fluoride concentration) [12], to examine how the association between MUF_SG and SRS-2 total T-score changes across different levels of MUF_SG exposure.

In sensitivity analyses, we examined whether first trimester maternal blood concentrations of lead (ln), manganese, mercury (ln) [55], and urinary arsenic (ln) [56] independently influenced the relationship between MUF_SG and SRS-2 total T-score in multivariable linear regression models. These heavy metals were examined as they are established toxicants [57], however, studies show mixed evidence as to whether each is associated with autism or autism-like behaviors [58–60]. We also assessed postnatal maternal stress and depressive symptoms, measured during the Biomonitoring visit using the Parenting Stress Index (PSI) [61] and the Center for Epidemiological Studies-Depression (CES-D) Scale 10 [62], both linked to ASD risk in offspring [63, 64]. Given missing data for maternal pre-pregnancy body mass index (BMI), which may influence fluoride exposure concentration [65], we evaluated its potential as a confounder in the relationship between MUF_SG and SRS-2 total T-score. Based on prior research from our group demonstrating child IQ is associated with the SRS-2 and MUF_SG [25, 66], we tested whether child IQ, measured using the Full Scale IQ (FSIQ) from the Wechsler Preschool and Primary Scale of Intelligence-Third Edition [67], confounded the relationship. We also examined first, second, and third trimester MUF_SG as independent predictors, based on our previous work examining critical windows of prenatal fluoride exposure [68]. Finally, we assessed whether adjusting MUF for CRE rather than SG altered our model findings.

Analyses were adjusted for the same set of covariates, except in the case of analyses involving water fluoride, which did not adjust for study city due to collinearity between the two variables. Significance tests were 2-sided and evaluated at p-value < 0.05. All statistical analyses were completed using SAS version 9.4 Software (SAS Institute, Inc., Cary, North Carolina).

Results

Compared to participants who consented to the Neurodevelopment visit from the MIREC-CD Plus that were excluded due to incomplete data (n = 142), the sample included in our analyses (n = 453) had a significantly higher proportion of mothers who were White, experiencing their first birth, and had a child that was younger in age at SRS-2 testing (Table 1); all other sociodemographic characteristics were similar. The overall median MUF_SG was 0.43 mg/L (interquartile range = 0.33 mg/L), with values ranging from 0.10 to 2.48 mg/L (Table 2). The mean MUF_SG was significantly higher in fluoridated regions (mean = 0.72 mg/L, SD = 0.45) compared to non-fluoridated regions (mean = 0.42 mg/L, SD = 0.30); all samples were above the LoD. Mean MUF_SG progressively increased from trimester 1 to trimester 3 (Table 2). Overall, trimester-specific MUF_SG measurements were moderately consistent (ICC = 0.36) (Additional Table 3). We observed a strong positive association between each prenatal fluoride exposure measure (r ≥ 0.5) and a moderate positive association between MUF_SG and tap water fluoride concentration (r = 0.47) (Additional Table 3).

Table 1.

Comparison of sociodemographic characteristics among MIREC-CD plus neurodevelopment participants with complete SRS-2 data who were included^a and excluded^b from analyses

	Included (n = 453)			Excluded (n = 142)
Characteristic	n	Mean (SD) or %		nc	Mean (SD) or %	P Valued
Maternal
Age, y	453	32.6 (4.4)		137	32.4 (5.2)	0.72
Race						0.05
White	413	91.2		116	85.3
Other	40	8.8		20	14.7
Born in Canada						0.11
Yes	380	83.9		106	77.9
No	73	16.1		30	22.1
Education level						0.53
High school or less	21	4.6		8	6.0
More than high school	432	95.4		126	94.0
Annual household income						0.55
≤ $70,000 CAD	124	27.4		35	30.2
>$70,000 CAD	329	72.6		81	69.8
Alcohol consumption, drink/mo						0.16
None	371	81.9		98	87.5
>0	82	18.1		14	12.5
Smoked in trimester 1						0.55
Yes	11	2.4		5	3.6
No	442	97.6		132	96.4
Second-hand smoke in trimester 1						0.80
Yes	17	3.8		4	2.9
No	436	96.2		132	97.1
Marital status						0.50
Married	323	71.3		101	74.3
Other	130	28.7		35	25.7
First birth						< 0.01
Yes	211	46.6		45	32.9
No	242	53.4		92	67.1
HOME score	453	47.4 (4.3)		124	47.4 (4.1)	0.85
Folic acid supplementation, µg/day						0.50
< 400	25	5.5		8	5.6
400–1000	325	71.8		95	66.9
>1000	103	22.7		39	27.5
Plasma total folate, trimester 1						0.30
< 10th percentile (66.02 nmol/L)	46	10.1		17	14.9
≥ 10th & < 80th percentiles	316	69.8		78	68.4
≥ 80th percentile (126.26 nmol/L)	91	20.1		19	16.7
Plasma total folate, trimester 3						0.55
< 10th percentile (49.14 nmol/L)	45	9.9		7	9.5
≥ 10th & < 80th percentiles	317	70.0		48	64.8
≥ 80th percentile (151.53 nmol/L)	91	20.1		19	25.7
Child
Age at SRS-2 testing, mo	453	40.5 (3.8)		142	41.3 (3.6)	0.03
Sex						0.32
Male	224	49.4		62	44.6
Female	229	50.6		77	55.4
SRS-2 total T-score
Overall	453	45.3 (6.1)		142	45.4 (6.2)	0.86
< 60	445	98.2		139	97.9	0.73
≥ 60	8	1.8		3	2.1

CD child development, SRS Social Responsiveness Scale, MUF_SG maternal urinary fluoride adjusted for specific gravity, SD standard deviation, y year, mo month, HOME Home Observation for the Measurement of the Environment.

^aComplete MUF_SG exposure and covariate data

^bIncomplete MUF_SG exposure and/or covariate data

^cCounts not adding to n = 142 are due to missing covariate data

^dIndependent sample t-test, interpreted Welch-Satterthwaite for unequal variances; or chi-square test, interpreted Fisher’s exact test when expected cell count < 5

Table 2.

Distribution of maternal fluoride exposures by water fluoridation status

Percentile
Fluoride exposurea	Mean	SD	Min	25	50	75	Max	IQR	P Valueb
MUFSG (mg/L)
Overall (n = 453)	0.52	0.38	0.10	0.28	0.43	0.61	2.48	0.33	< 0.01
Fluoridated (n = 151)	0.72	0.45	0.10	0.46	0.60	0.82	2.48	0.36
Non-fluoridated (n = 216)	0.42	0.30	0.11	0.23	0.36	0.48	2.06	0.25
Trimester 1	0.45	0.49	0.01	0.17	0.31	0.56	4.29	0.39	< 0.01
Trimester 2	0.50	0.49	0.03	0.23	0.36	0.59	5.28	0.37
Trimester 3	0.62	0.52	0.08	0.31	0.47	0.79	5.56	0.48
Fluoride intake (mg/day)c
Overall (n = 369)	0.52	0.42	0.01	0.19	0.37	0.77	1.92	0.58	< 0.01
Fluoridated (n = 151)	0.89	0.38	0.19	0.60	0.81	1.10	1.92	0.49
Non-fluoridated (n = 216)	0.26	0.20	0.01	0.15	0.21	0.32	1.80	0.17
Water fluoride (mg/L)c
Overall (n = 369)	0.32	0.23	0.04	0.13	0.20	0.56	0.76	0.43	< 0.01
Fluoridated (n = 151)	0.59	0.08	0.40	0.53	0.56	0.65	0.76	0.12
Non-fluoridated (n = 216)	0.13	0.06	0.04	0.13	0.13	0.18	0.20	0.05

SD standard deviation, min minimum, max maximum, IQR interquartile range, MUF_SG maternal urinary fluoride adjusted for specific gravity

^aOverall counts not adding to n = 453 are due to missing data

^bIndependent sample t-test, interpreted Welch-Satterthwaite for unequal variances; or repeated measures ANOVA

^cTwo subjects are missing data on community water fluoridation status (fluoridated or non-fluoridated)

The overall mean SRS-2 total T-score for children 3 to 4 years old was 45.3 (SD = 6.1, range = 34 to 85), with males having a significantly higher score (mean = 46.4, SD = 6.5) compared to females (mean = 44.2, SD = 5.4) (Additional Table 4), as previously reported by our group [66]. Most scores fell within the ‘normal range’ with a total T-score < 60 (Additional Fig. 4). For all other SRS-2 scales, except the Motivation scale, males had a significantly higher SRS-2 score compared to females. Children had a significantly higher SRS-2 total T-score among mothers who were ≤ 30 years old at enrollment, non-White, had a high school education or less, annual household income ≤ $70,000 CAD, exposed to second-hand smoke in their first trimester, not married, and had a HOME score ≤ 48 (median score) (Additional Table 5).

The bivariate regression analysis between MUF_SG and SRS-2 total T-score revealed no association (Fig. 2). In multivariable linear regression, a 1 mg/L increase in MUF_SG concentration was not associated with the SRS-2 total T-score (β = −0.16; 95% CI, −1.70, 1.39) (Table 3). Similarly, estimated daily fluoride intake and water fluoride concentration were not associated with the SRS-2 total T-score in multivariable models. Additionally, no associations were observed between any fluoride exposure variable, and SRS-2 scale T-scores.

Fig. 2.

Bivariate association between SRS-2 total T-score and MUF_SG (n = 453)

Table 3.

Adjusted^a linear regression models for the effect of fluoride exposures on SRS-2 scores

	MUFSG (mg/L) (n = 453)b					Fluoride intake (mg/day) (n = 369)					Water fluoride (mg/L) (n = 369)c
	95% CI					95% CI					95% CI
Model	β	Lower	Upper	P Value		β	Lower	Upper	P Value		β	Lower	Upper	P Value
Total	−0.16	−1.70	1.39	0.84		−0.93	−2.87	1.00	0.34		−0.66	−3.12	1.80	0.60
Boysd	−0.06	−2.16	2.04	0.96		−1.71	−4.54	1.12	0.24		0.63	−3.24	4.50	0.75
Girlse	0.16	−2.31	2.62	0.90		0.22	−2.54	2.98	0.88		−1.89	−5.00	1.22	0.23
Subscales
Awareness	−0.88	−2.92	1.17	0.40		−0.39	−3.11	2.33	0.78		−0.02	−3.48	3.44	0.99
Cognition	−0.18	−1.81	1.46	0.83		−1.39	−3.44	0.66	0.18		0.23	−2.38	2.83	0.87
Communication	−0.13	−1.72	1.46	0.88		−0.76	−2.75	1.24	0.46		−0.34	−2.85	2.17	0.79
Motivation	0.28	−1.63	2.19	0.77		−0.96	−3.43	1.51	0.44		−1.82	−4.96	1.33	0.26
Restricted and repetitive	0.12	−1.61	1.86	0.89		−0.77	−2.92	1.38	0.48		−0.98	−3.67	1.72	0.48
DSM social	−0.08	−1.63	1.47	0.92		−0.89	−2.84	1.06	0.37		−0.49	−2.97	1.99	0.70
DSM restricted	0.12	−1.62	1.86	0.89		−0.77	−2.92	1.38	0.48		−0.98	−3.67	1.72	0.48

MUF_SG maternal urinary fluoride adjusted for specific gravity, SRS Social Responsiveness Scale, DSM Diagnostic and Statistical Manual of Mental Disorders

^aAdjusted for: maternal - income, second-hand smoke, married, parity, HOME score; child - sex, city

^bInteraction term not statistically significant between MUF_SG and (i) child sex, (ii) folic acid supplementation, (iii) trimester 1 plasma total folate, and (iv) trimester 3 plasma total folate, in SRS-2 total T-score models

^cAdjustment does not include city as it is collinear with water fluoride concentration

^dMUFSG (n = 224), fluoride intake and water fluoride (n = 186)

^eMUF_SG (n = 229), fluoride intake and water fluoride (n = 183)

We found no evidence of effect modification in the SRS-2 total T-score models between MUF_SG and the following variables: child sex, daily folic acid supplementation, first and third trimester plasma total folate (Additional Table 6). In secondary analyses, no association was observed between each prenatal fluoride exposure and the odds of having an SRS-2 total T-score ≥ 90th percentile compared to a total T-score < 90th percentile (Table 4). Further, when MUF_SG was modeled using linear splines, the association with the SRS-2 total T-score remained non-significant (Additional Table 7).

Table 4.

Adjusted^a logistic regression models for the association between each fluoride exposure and increased odds of SRS-2 total T-score ≥ 90th percentile

95% CI
Model	n	OR	Lower	Upper	P Value
MUFSG (mg/L)	453	0.76	0.29	1.96	0.56
Fluoride intake (mg/day)	369	0.36	0.09	1.53	0.17
Water fluoride (mg/L)b	369	0.49	0.10	2.42	0.38

MUF_SG maternal urinary fluoride adjusted for specific gravity, SRS Social Responsiveness Scale, OR odds ratio

^aAdjusted for: maternal - income, second-hand smoke, married, parity, HOME score; child - sex, city

^bAdjustment does not include city as it is collinear with water fluoride concentration

In sensitivity analyses, adjusting for first trimester maternal blood lead, manganese, mercury, and urinary arsenic, the multivariable regression results for the effect of MUF_SG on SRS-2 total T-score remained non-significant (Additional Table 8). Similarly, no significant results were observed after adjusting for maternal stress, depressive symptoms, pre-pregnancy BMI, as well as child FSIQ. Lastly, after independently adjusting for various fluoride exposures - including first, second and third trimester MUF_SG, and MUF_CRE - no associations with SRS-2 scale T-scores were observed (Additional Table 9).

Discussion

We assessed the association between prenatal fluoride exposure and autistic behaviors measured with the SRS-2 in preschool-aged children from a prospective Canadian birth cohort. We found no evidence of an association between prenatal fluoride exposure, including MUF_SG, estimated daily maternal fluoride intake or water fluoride concentration with parent-reported autistic behaviors. These findings contrast with previous research in this cohort of preschool-aged children, which found that lead and phthalate exposure is associated with more autistic behaviors [7, 8]. Notably, our cohort has relatively few children with high SRS-2 scores; this limited variability may make it difficult to detect an effect if it exists, as may be possible in other populations that have different exposure and outcome profiles. The mean SRS-2 total T-score in our sample was approximately half a SD lower than the U.S. normative sample used to standardize the SRS-2 (45.3 vs. 50.0, respectively) [31]. Sociodemographic factors may be contributing to the low scores on the SRS-2 in the MIREC cohort [66].

Our null findings align with those of a prospective cohort study (NICE) conducted in northern Sweden where fluoride is not added to drinking water, but concentrations can vary from natural sources. The NICE study found no association between prenatal MUF_SG and SRS-2 total T-score in 4-year-old children [20]. In contrast, a U.S. prospective cohort study (MADRES) reported an association between higher MUF_SG concentrations and more parent-reported autistic behaviors in 3-year-old children using the Child Behavior Checklist [19]. The MADRES cohort primarily consisted of Hispanic women living in Los Angeles, California, where most of the county’s community water is at least partially fluoridated [69]. Moreover, the women were of lower socioeconomic status, a risk factor for ASD [2], which may have contributed to higher risk of autistic behaviors in that study sample. The median MUF_SG concentration was slightly lower for the MIREC cohort (0.60 mg/L) relative to the NICE (0.72 mg/L) and MADRES (0.76 mg/L) cohorts.

We previously showed that prenatal fluoride exposure is significantly inversely associated with performance IQ (PIQ), but not with verbal IQ (VIQ), in children [25, 68]. VIQ pertains to abilities related to language, communication and verbal reasoning, while PIQ reflects non-verbal reasoning, spatial skills and problem solving. Additionally, we found that SRS-2 total T-scores are more strongly negatively correlated with VIQ (r = −0.37) than PIQ (r = −0.21) [66], consistent with other research [70]. Taken together, the pattern of findings observed in the current study showing no association between fluoride and social communication is consistent with past findings in the MIREC cohort [68] showing no association between fluoride and verbal reasoning.

In the present study, males had a higher prevalence of autistic behaviors than females, which is consistent with global estimates [27]. However, no effect modification by sex was observed in the association between prenatal MUF_SG and SRS-2 total T-score. Similarly, no effect modification by folic acid was observed, though only 5.5% of pregnant women in our sample did not meet daily recommendations for folic acid intake. Health officials recommend that individuals who could become pregnant, are pregnant or lactating take a daily multivitamin containing folic acid to reduce the risk of neural tube defects [50]. Future studies can extend the scope of the analyses to include other dietary factors, such as vitamin D, B₁₂ or B₆.

Limited variability of clinically elevated SRS-2 scores (total T-score ≥ 60) in the MIREC cohort can affect the precision of covariate effect estimates in regression models (as we observed with some wide estimate confidence intervals), and reduce the statistical power to detect small or subtle associations, even if a true relationship exists [71]. Considering the low variability of SRS-2 scores in our sample, with over 98% of total T-scores falling within the ‘normal range’ (T-score < 60), we conducted a secondary analysis to explore the association between prenatal fluoride exposure and autistic behaviors using multivariable logistic regression. We used the 90th percentile as the cutoff to dichotomize the SRS-2 total T-score (which corresponded to a T-score of 53 in our sample), and results remained non-significant. In the MIREC cohort, when children were between 8 and 11 years old, parents were asked whether their child had ever been diagnosed with ASD. Among our study sample, only 0.8% of participants who responded reported an ASD diagnosis for their child, which is considerably lower than current estimates of ASD in the general population [2, 3] and further aligns with the limited variability of elevated SRS-2 scores in our study.

Previous studies of MIREC children found significant associations between other neurotoxicants (i.e., lead, phthalates) and higher SRS-2 scores with effect modification by folic acid and sex [7, 8]. If fluoride increases the risk of autistic behaviors, it does so with less potency (or at higher population levels of exposure than observed in this study) than other neurotoxicants; alternatively, the biological mechanism underlying the adverse impacts of fluoride on intelligence and behavior that have been seen in MIREC and other longitudinal cohorts [17] may not be as relevant to the pathogenesis of autism and may be different to those of other neurotoxicants.

The main strengths of our study include its prospective, longitudinal design, which enabled the collection of prenatal and postnatal mother-child data to assess the impact of environmental exposures on child neurodevelopment. A key strength is the measurement of three independent prenatal fluoride exposure measures, enabling a comprehensive assessment of the association with autistic behaviors in preschool-aged children and facilitating comparisons across exposure measures. Additionally, a valid and reliable standardized questionnaire was used for assessing autistic behaviors, and robust covariate data collection allowed for the evaluation of potential confounders [72] and risk factors for ASD, including folate supplementation and other well-known neurotoxicants, including lead exposure.

The study also has several limitations. First, prenatal MUF_SG was measured using spot urine samples, rather than 24-hour fasting or first-morning void samples. Given fluoride’s short half-life of approximately 5 h [73], this method may capture recent exposure, but not usual total exposure, and has relatively high intra-individual variation in urinary fluoride concentrations. Variability in MUF_SG can lead to non-differential exposure misclassification, underestimating true MUF_SG concentrations across all levels of the autistic behavioral outcome, potentially attenuating observed effects [74]. We attempted to mitigate this by calculating the mean of urinary fluoride samples across the three trimesters, although fluoride levels may fluctuate throughout pregnancy, contributing to variability. Second, our data does not include information on day-to-day fluoride use (i.e., toothpaste, mouthwash, etc.), which may influence prenatal exposure levels. Third, the mechanism underlying ASD is not fully understood, and unknown/unmeasured factors may lead to residual confounding. Fourth, while there is evidence suggesting that the synergistic interaction of fluoride and aluminum could increase potential neurotoxic effects of fluoride [10], our cohort does not include aluminum biomarkers, so we were unable to examine this interaction. Further research should explore the potential role of aluminofluoride in ASD risk. Fifth, given the prevalence of ASD is approximately 1 in 63 among Canadian children [3] and considering our study sample size (n = 453), statistical power may be limited in our study, particularly for assessing effect modification, potentially contributing to the null findings we observed. Lastly, our results may be limited in their generalizability due to the sample’s predominance of participants that are White and have high socioeconomic status.

Conclusion

In this Canadian cohort, we found no evidence of an association between prenatal fluoride exposure and autistic behaviors in children aged 3 to 4 years. Given the limited and contradictory nature of existing research, we recommend further epidemiologic studies to explore the association between prenatal fluoride exposure and autistic behaviors in other cohorts, including children older than preschool age.

Abbreviations

ASD

Autism Spectrum Disorder

BMI

Body mass index

CES-D

Center for Epidemiological Studies-Depression

CRE

Creatinine

FSIQ

Full Scale IQ

HOME

Home Observation for Measurement of the Environment

ICC

Intraclass correlation coefficient

IPW

Inverse probability weighting

IQ

Intelligence quotient

In

Natural log

LoD

Limit of detection

MIREC

Maternal-Infant Research on Environmental Chemicals

MIREC-CD

Maternal-Infant Research on Environmental Chemicals-Child Development

MUF

Maternal urinary fluoride

MUF_CRE

Maternal urinary fluoride adjusted for urinary creatinine

MUF_SG

Maternal urinary fluoride adjusted for specific gravity

PIQ

Performance IQ

PSI

Parenting Stress Index

SD

standard deviation

SG

Specific gravity

SRS-2

Social Responsiveness Scale-Second Edition

THF

Tetrahydrofolate

US

United States

VIQ

Verbal IQ

WTP

Water treatment plant

Authors’ contributions

AC and CT conceptualized and formulated the study. BL and GM designed and conducted the MIREC-Child Development (MIREC-CD) Plus study. CT and EAM acquired fluoride data. AJM acquired folate data. AC, RG and CVG prepared data for analysis. AC performed data analysis, and all authors synthesized results. AC drafted the manuscript and was critically reviewed by all authors. All authors reviewed and approved the final manuscript.

Funding

This study was funded by the National Institute of Environmental Health Science (grants R21ES027044 and R01ES030365). The Maternal-Infant Research on Environmental Chemicals (MIREC) study was funded by the Canadian Institutes of Health Research (grant MOP-81285), and the Chemicals Management Plan at Health Canada.

Data availability

The source data were and can be accessed through an application and review process as required by the MIREC Biobank and described here: https://www.mirec-canada.ca/.

Declarations

Ethics approval and consent to participate

This study obtained approval from Research Ethics Boards at Health Canada, Centre hospitalier universitaire (CHU) Sainte-Justine Research Center, ethics committees at participating hospitals, York University, and the University of Toronto for the analysis of secondary data. All participants completed informed consent for themselves and children.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.