Abstract
Background
Previous research on the role of maternal diet in relation to autism has focused on examining individual nutrient associations. Few studies have examined associations with multiple nutrients using mixtures approaches, which may better reflect true exposure scenarios.
Objectives
This study aims to examine associations of nutrient mixtures with children’s autism diagnosis and trait scores within a large, diverse population.
Methods
Participants were drawn from the United States Environmental influences on Child Health Outcomes (ECHO) consortium. Maternal prenatal diet was reported via validated food frequency questionnaires. Children’s autism-related traits were measured using the Social Responsiveness Scale (SRS) and autism diagnoses were from parent reports of physician diagnosis. Bayesian kernel machine regression was used to examine the overall mixture effect and interactions between a set of 5 primary nutrients (folate, vitamin D, omega 3 and omega 6 fatty acids, and iron), adjusted for potential confounders, in relationship to child outcomes. Secondary analyses were conducted in a subset of cohorts with an expanded set of 14 nutrients. Traditional linear and logistic regression models were also analyzed for comparison of results to mixture models.
Results
A total of 2614 participants drawn from 7 ECHO cohorts were included in primary analysis. Mixture analyses suggested that increasing the overall 5-nutrient mixture was associated with lower SRS scores. Individual U-shaped associations and bivariate interactions between folate and omega 3 fatty acids were suggested. In the subset included in the secondary analyses of the 14-nutrient mixture, a modest inverse trend remained, but individual nutrient associations were altered, with vitamin D demonstrating higher relative importance than other nutrients. Strong associations with autism diagnosis were not observed.
Conclusions
In this large sample, we found evidence for combined nutrient effects with broader autism-related traits. Because results for individual nutrients were sensitive to mixture components, replication of combined associations between nutrients and autism-related outcomes is needed.
Keywords: neurodevelopment, nutrients, statistical mixtures methods, epidemiology, autism
Introduction
Prenatal nutrition is critical for children’s neurodevelopment. Nutrients such as iron, fatty acids, and folic acid have been associated with neurodevelopmental outcomes [1] and could potentially influence the development of autism, but the evidence is inconclusive [2]. Autism is a neurodevelopmental condition that significantly affects communication and social interaction. In the United States, the prevalence of autism has increased over time, with the most recent estimate suggesting 1 in 36 children have an autism diagnosis by age 8 [3]. The etiology of autism is thought to be multifactorial, including both genetic and environmental contributions [4,5], with the perinatal period potentially a crucial window for autism development [6,7]. Additionally, evidence suggests traits related to autism follow a continuous distribution in the general population and may be useful to consider for etiologic investigations [8,9].
Higher prenatal intakes of folic acid, polyunsaturated fatty acids (PUFAs), and vitamin D have each been associated with lower odds of autism diagnosis and autism traits scores [2]. These nutrients are thought to influence neurodevelopment via effects on gene expression, oxidation, and inflammation, among other pathways [10,11]. For example, folic acid intake during early pregnancy has been associated with lower odds of autism diagnosis in several studies [12,13]. Folic acid influences 1-carbon metabolism and gene methylation, a pathway implicated in autism development, given evidence for altered methylation patterns among autistic individuals [[12], [13], [14], [15]]. PUFAs and vitamin D have also been associated with autism and general neurodevelopment [[16], [17], [18], [19], [20]]. These nutrients influence multiple pathways, including those related to inflammation and oxidation, which have in turn been associated with autism development [21]. Other nutrients that influence these pathways (such as choline and betaine, which influence 1-carbon metabolism) may also have the potential to influence autism outcomes but have been less studied to date.
Although previous work has focused on the role of individual nutrients (that is, folic acid) in relationship to autism, nutrients occur together within foods and dietary patterns, and intakes are highly correlated. Nutrients may also interact with each other biologically, underscoring the importance of considering their coexposures. Traditional statistical models may not adequately capture these biological interactions and correlations. Moreover, statistical methods designed to assess mixtures are particularly useful for examining multiple, interrelated exposures. One such method is Bayesian kernel machine regression (BKMR), which uses a kernel function to flexibly estimate exposure–response associations for a mixture of potentially correlated exposures with adjustment for covariates [22]. One strength of BKMR is that it can accommodate nonlinear, nonadditive relationships, and the exposure–response relationship does not have to be in the same direction (that is, positive or negative) for all exposures in the mixture. Additionally, BKMR allows for hierarchical grouping of exposures based on correlation or hypothesized relationship. Full descriptions of BKMR are available elsewhere [22,23]. Similar mixtures methods have been used to examine associations of environmental chemical mixtures with neurodevelopmental outcomes [24,25], but the application of these methods for diet-related analyses is far less common. In prior work, members of our team used BKMR to examine the association of prenatal intake of 10 nutrients with autism outcomes in 2 prospective cohorts: the Early Autism Risk Longitudinal Investigation (EARLI) study, a familial autism cohort study following younger siblings of autistic children, and the Nurses Health Study II (NHSII), a cohort of registered United States nurses drawn from the general population [26]. Although we did not find strong individual or combined associations of nutrients with autism traits scores in the EARLI familial autism cohort, there were moderate inverse associations of the overall mixture and of individual nutrients (omega 6 fatty acids, iron, vitamin B12) with Social Responsiveness Scale (SRS) score and moderate positive associations with omega 3 fatty acids in the NHSII cohort. However, this prior analysis included relatively small samples (n = 127 in EARLI and 713 in NHSII) with limited racial/ethnic diversity, potentially limiting the ability to detect associations.
Therefore, to address the role of combined effects of a suite of nutrients and clarify associations with autism-related outcomes and specific key nutrients in a large diverse sample, we employed mixture methods using data from the United States Environmental influences on Child Health Outcomes (ECHO) cohort consortium [27]. We hypothesized, given prior signals with individual nutrients, that overall intake of a mixture of neurodevelopmentally relevant nutrients would be inversely associated with autism related-outcomes, and that novel interactions between nutrients may be identified using mixture modeling approaches.
Methods
Participants
The ECHO Program includes over 38,000 pregnancies and 60,000 children with an overall goal of understanding the role of early-life exposures on children’s health, including neurodevelopment [27]. As of 2023, 69 prenatal and pediatric cohort studies across the United States had joined the ECHO Program, representing births spanning from the 1980s to the present [28], contributing previously collected data and agreeing to further data collection under a common ECHO Protocol. The ECHO Cohort includes pregnancies at enriched risk of autism along with those drawn from the general population [28].
All dyads with available maternal prenatal food frequency data and child autism outcomes data as of 1 March, 2023 were included in the current study, as long as 1) both mother and child had consented to participation in ECHO, and 2) children were the result of a singleton pregnancy. For mothers with multiple eligible children, we randomly selected 1 child. Descriptions of the 7 ECHO cohort study sites (corresponding to 2614 unique mother–child dyads) that contributed data for this analysis are provided in Supplemental Table 1.
Prenatal nutrient intake measures
ECHO cohort study sites have collected several types of dietary data, including 24-h recalls and food frequency questionnaires (FFQs) [29]. Differences in nutrient intake by the assessment method have been documented in ECHO [30] and elsewhere [31,32]. For this analysis, only prenatal FFQ data were included, as FFQs better capture usual daily intake and dietary patterns [33].
Three different FFQs were collected by the cohorts included in this analysis: the Block FFQ [34], the Harvard Willett FFQ [35], and the Diet History Questionnaire (DHQ) [36]. Each of these validated [37] questionnaires asks respondents to describe how often they consumed a list of >100 foods and beverages over a specified time frame. Participants in a subset of cohorts were also asked about supplement use. All questionnaires were completed by participants during pregnancy. If mothers completed >1 FFQ during pregnancy, the first FFQ was chosen to align with the early pregnancy period, a timepoint in which nutritional exposures have been linked to autism outcomes in previous studies [2].
Cohorts used standardized nutrient databases appropriate for their FFQ to convert raw food frequency data into usual nutrient intake and then shared the data with ECHO analysts, who harmonized nutrients to standard units. Because of differences in the underlying nutrient databases and original cohort study aims, the nutrients available for each cohort varied slightly. All cohorts provided data on dietary intake of our 5 priority nutrients: folate, vitamin D, omega 3 fatty acids, omega 6 fatty acids, and iron. These nutrients were a priori selected for analyses based on their associations with autism outcomes in previous literature [2]. Of the 7 cohorts, 4 also provided data on 9 additional nutrients: vitamin B12, vitamin B6, choline, betaine, methionine, zinc, vitamin E (alpha-tocopherol), vitamin A, and vitamin C. These nutrients are involved in related biological pathways (including antioxidant, inflammatory, and epigenetic) that may influence neurodevelopmental outcomes but have been less studied in prior literature. Because dietary supplement use was not consistently captured across cohorts, only intake from dietary sources is included in the primary analysis.
Children’s autism outcomes
Two autism outcomes were considered in this analysis: autism diagnosis (yes/no) and scores from the SRS-2, a quantitative measure of autism-related traits [38,39]. Parents were asked whether a doctor or other health care provider ever told them that their child had autism, autism spectrum disorder (ASD), Asperger’s disorder, or pervasive developmental disorder. In some cohorts, diagnoses were confirmed according to clinical assessments or medical record abstraction; however, parent-reported diagnosis was used here because it was harmonized across cohorts. The SRS is a validated and widely used informant-report questionnaire that quantitatively measures children’s autism-related traits, including social communication, social cognition, restricted repetitive behaviors, autistic mannerisms, attention, and social motivation [40]. The SRS has well-established psychometric properties, with high internal validity, reliability, and reproducibility [40] and can be applied to both clinical populations and the general population [8]. Higher scores from general population samples have also been shown to correlate well with clinical autism diagnosis [40]. There are several forms of the SRS based on age at administration and length: a preschool form (for children aged 2.5–4.5 y), a school-age form (4–18 y), full-length forms (65 items), and short forms (16 items). All versions of the SRS have been validated and yield comparable total raw scores, with higher scores reflecting greater deficits in social communication. Short-form scores are scaled to yield the same total score range as full forms. We included all versions of the SRS in this analysis, as they have been found to have strong agreement in their distributional properties, risk factor associations, predictions of autism diagnosis, and familiality, including in the ECHO population [[41], [42], [43], [44], [45]]. T-scores can also be calculated; in this case, raw scores were chosen for primary analysis, rather than T-scores, as recommended by the publisher for use in studies of population-level effects. Forms were excluded if there was >15% missingness or the age of administration was <2.5 or >18 y.
Although all cohorts included in this analysis collected information on both autism diagnosis and SRS scores, the number of individuals with data varied across autism outcomes (that is, some participants only had information on autism diagnosis, whereas some only had SRS scores). Because our previous work with BKMR has focused on continuous outcomes [26] and because the evidence supports the use of quantitative traits scores in autism research [9], SRS was the primary, continuous outcome examined in this analysis. Autism diagnosis was considered a secondary, binary outcome given the modest number of individuals with an autism diagnosis (n = 137) in our analytic sample.
Statistical analysis
Descriptive statistics were calculated for nutrient intakes, autism outcomes, and covariates of interest. Dietary nutrient intakes were compared with United States recommended dietary allowances or adequate intake levels for pregnancy where available. United States recommended intakes during pregnancy were not available for total omega 3 or total omega 6 fatty acids, betaine, or methionine. Because some of the dietary intake variables had skewed distributions, Spearman’s correlations were calculated to examine monotonic relationships among nutrients.
Primary analysis: BKMR analysis of 5 nutrients with SRS score
We used BKMR to examine the combined, individual, and joint associations of maternal dietary nutrient intakes with SRS total raw scores, adjusting for potential confounders.
We used the R package bkmr (version 0.2.2) for analysis. First, we ran the Markov chain Monte Carlo (MCMC) sampler, which randomly samples from the distribution to fit the model. To improve computing performance while maximizing the number of iterations, parallel chains were utilized, with 5 chains and 5000 iterations per chain. Default prior parameters were used. The R package bkmrhat (version 1.1.3) was used to assess model fit. Prior work has shown that BKMR is able to detect modest associations in sample sizes as small as 100, suggesting our sample size was more than adequate for estimating associations under this approach [22].
BKMR models were used to estimate exposure–outcome relationships in different ways, each illustrated in a separate figure. First, we estimated the overall mixture effect, or the combined association of the nutrient mixture with SRS scores. Second, we estimated individual nutrient associations with SRS scores while holding the remaining nutrients constant at their median value. Third, we examined how associations of single nutrients with SRS scores changed when the remaining nutrients were set to their 25th, 50th, or 75th percentiles. Finally, we used BKMR to estimate bivariate exposure–response functions, which show the association of a single nutrient with SRS at varying levels of a second nutrient.
In addition, BKMR produces posterior inclusion probabilities (PIPs) for the inclusion of a nutrient in the MCMC sampler, representing the relative importance of each exposure in the model. PIPs range from 0 to 1, with higher values representing higher importance. In the same way, group PIPs describe the relative importance of a group of nutrients. Conditional PIPs describe the relative strength of each nutrient within a group, as given by the percentage of iterations in which the MCMC sampler identifies each nutrient as the most important when the group is included in the model [22].
For the primary model, we included 5 nutrients in our exposure mixture: folate, vitamin D, omega 3 fatty acids, omega 6 fatty acids, and iron. Before inclusion in the models, all nutrients were energy-adjusted and standardized using a z-score. In addition to treating each nutrient independently within the model, we used a hierarchical grouping of nutrients, with vitamin D, iron, and omega 3, omega 6 fatty acids representing an “inflammatory pathways” group, and folate in a separate “1-carbon metabolism” group.
Models were adjusted by covariates with previously reported associations with autism outcomes or maternal prenatal diet: maternal age (continuous), prepregnancy BMI (continuous), cohort type (familial autism cohort or general population), maternal race and ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, and other), maternal education (less than high school, high school degree/General Educational Development test (GED), some college/associate’s/trade school, bachelor’s degree, master’s/professional/doctorate degree), maternal smoking (yes/no), child sex (male, female), and child year of birth (1998–2004, 2005–2009, 2010–2014, 2015+). We include race and ethnicity as a social construct which may serve as a proxy for structural inequities associated with maternal diet and children’s autism-related outcomes [46]. Missingness was low (0%–3%) for variables included in primary analyses. Complete information on missingness is included in Table 1. Missing values were filled using single imputation estimated from linear regression using values from all other covariates.
TABLE 1.
Characteristics of ECHO participants included in primary analysis of nutrient intake with SRS scores (n = 2614).
| Characteristic | n (%) |
|---|---|
| Cohort type | |
| Autism (familial) enriched risk | 222 (8.49) |
| Nonenriched/population-based | 2392 (91.51) |
| Maternal race/ethnicity | |
| Non-Hispanic White | 1719 (65.76) |
| Non-Hispanic Black | 536 (20.50) |
| Hispanic | 178 (6.81) |
| Other | 176 (6.73) |
| Missing | 5 (0.19) |
| Maternal smoking | |
| Yes | 145 (5.55) |
| No | 2448 (93.65) |
| Missing | 21 (0.8) |
| Prenatal supplement use (ever) | |
| Yes | 2102 (80.41) |
| No | 134 (5.13) |
| Missing | 378 (14.46) |
| Maternal education | |
| <High school | 102 (3.9) |
| High school degree, GED | 427 (16.34) |
| Some college/associate’s/trade school | 653 (24.98) |
| Bachelor’s degree | 648 (24.79) |
| Master’s, professional or doctorate degree | 704 (26.93) |
| Missing | 80 (3.06) |
| Household income | |
| <$30,000 | 225 (8.61) |
| $30,000–$49,999 | 142 (5.43) |
| $50,000–$74,999 | 188 (7.19) |
| $75,000–$99,999 | 214 (8.19) |
| $100,000 or more | 767 (29.34) |
| Missing | 1078 (41.24) |
| Child sex | |
| Male | 1313 (50.23) |
| Female | 1301 (49.77) |
| Preterm birth | |
| Term (≥37 wk gestation) | 2451 (93.76) |
| Preterm (<37 gestation) | <1651 |
| Missing | <51 |
| Child birth year | |
| 1998–2004 | 574 (21.96) |
| 2005–2009 | 453 (17.333) |
| 2010–2014 | 1261 (48.24) |
| 2015+ | 326 (12.47) |
| Breastfed (ever) | |
| Yes | 1546 (59.14) |
| No | 24 (0.92) |
| Missing | 1044 (39.94) |
| Child autism diagnosis | |
| Autism | 107 (4.09) |
| No autism | 2285 (87.41) |
| Missing | 222 (8.49) |
| SRS-2 instrument | |
| 16 items | 380 (14.54) |
| 65 items | 2234 (85.46) |
| SRS-2 form type | |
| School age | 2123 (81.22) |
| Preschool | 491 (18.78) |
| Diet form type | |
| Block | 952 (36.42) |
| DHQ2 | 91 (3.48) |
| Harvard | 1571 (60.10) |
| Mean (SD) | |
| Maternal age | 31.30 (5.42) |
| Birthweight | 3393.05 (541.70) |
| Prepregnancy BMI | 26.08 (6.37) |
| Parity | 0.95 (1.03) |
| SRS total raw score | 25.17 (20.62) |
| SRS total T score | 46.60 (8.03) |
Abbreviations: DHQ, Diet History Questionnaire; ECHO, Environmental influences on Child Health Outcomes; GED, General Educational Development Test; SRS, Social Responsiveness Scale.
To preserve confidentiality, we have suppressed exact values in bins with small n’s (<5), along with the value for ≥1 other bin in that variable.
Secondary analyses
We performed 3 secondary analyses. First, we included an expanded suite of 14 nutrients as our exposure mixture in the BKMR model. The additional nutrients were as follows: vitamin B12, vitamin B6, choline, betaine, and methionine (included with folate in the “1-carbon metabolism” group); vitamin E and zinc (included in the “inflammatory pathways” group); and vitamins A and C, which were grouped together as the “antioxidants” group. These analyses were secondary because only a subset of participants had complete data on all 14 nutrients. Second, we used BKMR models with a probit link function to examine the association of nutrient intakes with dichotomous autism diagnosis, using the same methods as described previously, in the sample with diagnostic data. Third, to compare with previous work assessing individual nutrients, we also conducted analyses using traditional regression models. Nutrients were individually examined in separate quantile regression (for SRS, modeled at the 50th percentile) and logistic regression (for autism diagnosis) models. Bivariate interactions suggested in BKMR analysis were then also explored in standard adjusted regression models using interaction terms to obtain a P value for the interaction.
Sensitivity analyses
To test the robustness of results in our primary analysis, we performed several sensitivity analyses. First, BKMR analysis of the 5 primary nutrients was examined in the analytic subsample of participants who had data for all 14 nutrients to assess differences between the subsample and the full study group. Second, BKMR analysis was performed for total intake from diet and supplements of the 5 primary nutrients in the analytic subsample with data on supplement intake. Although supplemental intake of omega 6 fatty acids was not measured in any cohort, dietary intake of this nutrient was included in the model because we considered omega 6 intake from supplements to be uncommon [47]. Because Bayesian kernal machine regression (BKMR) is computationally intensive, the remaining sensitivity analyses were conducted in traditional regression models; if differences from the primary analysis were identified, the sensitivity analysis was also performed using BKMR models. To assess the role of potential cohort-specific effects, we conducted covariate-adjusted quantile regression analyses of nutrient-SRS associations leaving out individual cohorts 1 at a time. Additionally, we tested primary models for further adjustment of several additional variables of interest, which were not included in primary models because of high missingness or limited variability within the sample, including prenatal vitamin use (ever during pregnancy), pregnancy complications, preterm birth, breastfeeding (ever), parity, and income. For variables with <25% missingness, imputation was used to assign missing values as in the primary analysis, whereas for variables with higher missingness (25%–50%), a “missing” category was created. Finally, we included adjustment for FFQ form type (Block compared with Willett compared with DHQ), which was not included in primary models because of the high correlation with cohort type. Because the addition of this variable altered regression results, we performed regression models stratified by FFQ type and, separately, tested FFQ type as a covariate in primary BKMR models.
Results
Participant characteristics
A total of 2614 maternal-child dyads from 7 ECHO cohort study sites were included in our primary analysis of nutrient intake and SRS scores (Supplemental Figure 1). Participants in this analysis were broadly comparable to the general ECHO population [28], with most maternal participants being non-Hispanic White, nonsmokers, with more than a high school education, and enrolled in cohorts that sampled from the general population (Table 1). Most children were born at term (≥37 wk gestational age). The mean (SD) SRS total raw score was 25.2 (20.6). The majority of dietary data included in our primary sample was collected using the Harvard/Willett FFQ (60%), with most of the remaining data collected via the Block FFQ. Average intakes of folate, zinc, and vitamins B6, B12, A, and C approached or exceeded United States recommended intakes for pregnancy, whereas average intakes of vitamins D, E, iron, and choline were below recommendations (Supplemental Table 2). A Spearman correlation matrix revealed associations between several nutrient pairs, including a strong positive correlation between omega 3 and omega 6 fatty acids for both the 5-nutrient and 14-nutrient samples (0.78), as well as iron and folate (0.82) and vitamin B12 and zinc (0.72) for the 14-nutrient sample (Supplemental Figure 2). A comparison of nutrient intake values by cohort is shown in Supplemental Table 3.
Primary analysis: BKMR analysis of 5 nutrients with SRS score
We ran BKMR models to estimate combined, individual, and bivariate associations of a suite of nutrients with autism outcomes. Convergence criteria were met for all models. Because results were similar with and without hierarchical grouping of exposures, we have only presented results from hierarchical models here. In covariate-adjusted models, we observed an inverse association between the overall 5-nutrient mixture and SRS scores. A dose–response trend across the mixture was suggested, with increases of ∼3 points in SRS scores for the lowest quantile of nutrient intake (with credible intervals narrowly excluding the null) to reductions of ∼1.5–2 points at higher levels of the mixture, compared with median intakes (Figure 1A). Plots of the associations of individual nutrients with SRS revealed U-shaped relationships with total omega 3 fatty acids and folate, with credible intervals excluding the null at certain points in the distribution (Figure 1B). Inverse associations of folate and omega 3 fatty acids with SRS were somewhat stronger when other nutrients were fixed at the 75th quantile (compared with 50th or 25th) (Figure 1C), but credible intervals for these differences included the null (Figure 1D).
FIGURE 1.
Adjusted associations between prenatal intake of 5 nutrients and child SRS scores using hierarchical BKMR (n = 2614). Results of Bayesian kernel machine regression (BKMR) analyses in ECHO including energy-adjusted nutrient intake for omega 3 and omega 6 fatty acids, iron, vitamin D, and folate, adjusted for maternal age, maternal prepregnancy BMI, child sex (male and female), cohort type (enriched familial autism probability and general population), maternal ethnicity/race (non-Hispanic White, non-Hispanic Black, Hispanic, and other), maternal education (less than high school, high school/GED, some college/associates degree/trade school, bachelor’s degree, and graduate degree), maternal smoking (yes and no), and child year of birth (1998–2004, 2005–2009, 2010–2014, and 2015+). Plots show: (A) the overall mixture effect, or the relationship between the nutrient mixture and child SRS scores. (B) Individual nutrient associations with SRS scores, holding all other nutrients at their 50th percentile. (C) Single-exposure effects; plot shows the impact on a child SRS score when each exposure increases from its 25th to 75th percentile (whereas other exposures are fixed at their 25th, 50th, or 75th percentiles). (D) Interaction parameters; plot compares the single-exposure health risk of each exposure (associated with a change from its 25th to 75th percentile) when other exposures are fixed to their 75th percentile compared with their 25th percentile. The primary set of nutrients included here was selected as described in the text based on a priori interest for neurodevelopmental relevance and grouped by shared pathway. ECHO, Environmental influences on Child Health Outcomes; SRS, Social Responsiveness Scale.
When assessing bivariate nutrient interactions, there was suggestive evidence of a complex interaction between folate and omega 3 fatty acids, as evidenced by intersecting slopes in Figure 2. At lower folate intakes, there was a slightly stronger inverse association with SRS scores if omega 3 fatty acid intake was in the highest quantile, compared with lower quantiles. However, at higher folate intakes, there were stronger positive associations with SRS scores if omega 3 fatty acid intake was in the highest quantile. This suggests that omega 3 fatty acid intake may enhance associations of folate with SRS score (both in a positive and negative direction).
FIGURE 2.
Adjusted bivariate associations between prenatal intake of 5 nutrients and child SRS scores using hierarchical Bayesian kernel machine regression (BKMR) (n = 2614). Results of BKMR analyses in ECHO including energy-adjusted nutrient intake for omega 3 and omega 6 fatty acids, iron, vitamin D, and folate, adjusted for maternal age, maternal prepregnancy BMI, child sex (male and female), cohort type (enriched familial autism probability and general population), maternal ethnicity/race (non-Hispanic White, non-Hispanic Black, Hispanic, and other), maternal education (less than high school, high school/GED, some college/associates degree/trade school, bachelor’s degree, and graduate degree), maternal smoking (yes and no), and child year of birth (1998–2004, 2005–2009, 2010–2014, and 2015+). Plots show bivariate exposure–response function for 2 predictors, where the second predictor is fixed at various percentiles (10th, 50th, and 90th), and all other predictors are fixed at the 50th percentile. The magnified plot shows the association of folate with SRS score at different quantiles of omega 3 fatty acids. Intersecting lines suggest an interaction between folate and omega 3 fatty acids. ECHO, Environmental influences on Child Health Outcomes; SRS, Social Responsiveness Scale.
When examining the relative contribution of each nutrient and group of nutrients, results suggested folate played the strongest role in relationships with SRS scores, as evidenced by a higher group PIP compared with the “inflammatory pathways” group (Table 2) [22]. Within the “inflammatory pathway” group, omega 3 fatty acids appeared to play the strongest role, with the highest conditional PIP.
TABLE 2.
Posterior inclusion probabilities (PIPs) from adjusted, hierarchical BKMR with child SRS scores in 5-nutrient and 14-nutrient analyses1.
| 5 nutrient analysis (n = 2614) | 14 nutrient analysis (n = 952) | |||||
|---|---|---|---|---|---|---|
| Nutrient | Group | Group PIP | Conditional PIP | Group | Group PIP | Conditional PIP |
| Folate | 1-C metab | 0.86 | 1 | 1-C metab | 0.21 | 0.06 |
| Omega 3 fatty acids | Inflammatory | 0.47 | 0.51 | Inflammatory | 0.54 | 0.04 |
| Iron | Inflammatory | 0.47 | 0.37 | Inflammatory | 0.54 | 0.04 |
| Omega 6 fatty acids | Inflammatory | 0.47 | 0.09 | Inflammatory | 0.54 | 0.01 |
| Vitamin D | Inflammatory | 0.47 | 0.03 | Inflammatory | 0.54 | 0.84 |
| Zinc | Inflammatory | 0.54 | 0.04 | |||
| Vitamin E | Inflammatory | 0.54 | 0.02 | |||
| Vitamin A | Antioxidant | 0.27 | 0.91 | |||
| Vitamin C | Antioxidant | 0.27 | 0.09 | |||
| Betaine | 1-C metab | 0.21 | 0.44 | |||
| Methionine | 1-C metab | 0.21 | 0.24 | |||
| Vitamin B6 | 1-C metab | 0.21 | 0.12 | |||
| Choline | 1-C metab | 0.21 | 0.08 | |||
| Vitamin B12 | 1-C metab | 0.21 | 0.07 | |||
Abbreviations: 1-C metab, 1-carbon metabolism pathway; BKMR, Bayesian kernel machine regression; SRS, Social Responsiveness Scale.
PIPs are from BKMR models of the association of prenatal nutrient intake with child SRS total raw score, adjusted for maternal age, maternal prepregnancy BMI, child sex, cohort type, maternal ethnicity/race, maternal education, maternal smoking, and child year of birth. PIPs are the probabilities for the inclusion of a nutrient in the Markov Chain Monte Carlo (MCMC) sampler, representing the relative importance of each exposure in the model. PIPs range from 0 to 1, with higher values representing higher importance. In the same way, group PIPs describe the relative importance of a group of nutrients. Conditional PIPs describe the relative strength of each nutrient within a group, as given by the percentage of iterations in which the MCMC sampler identifies each nutrient as the most important when the group is included in the model [22].
Secondary analyses
BKMR analysis of 14 nutrients with SRS score
Compared with participants included in the primary analysis, maternal participants in the subsample with complete data on intake of 14 nutrients (4 cohorts, n = 952) were more likely to be non-Hispanic Black, have only a high school education, and to be drawn from an autism familial enriched risk cohort (Supplemental Table 4). Nutrient intakes were similar to those in the primary analysis (Supplemental Table 2). The association of the overall mixture with SRS scores remained similar in the expanded 14-nutrient mixture (Figure 3A). However, individual nutrient associations with SRS were altered. Previously suggested relationships with folate and omega 3 fatty acids were null in this nutrient mixture, and betaine was positively associated with SRS score, though credible intervals were wide (Figure 3B). Inverse associations of vitamin D with SRS were somewhat stronger when other nutrients were fixed at the 75th quantile (compared with 50th or 25th) (Figure 3C), but this difference had credible intervals that crossed the null (Figure 3D). There was no clear evidence of bivariate nutrient interactions (Figure 4). Assessing relative contributions in this nutrient mixture, the “inflammatory pathways” nutrient group had the highest group PIP, with vitamin D having the highest conditional PIP within the group, suggesting the strongest relationship with SRS score (Table 2). Within the “1 carbon metabolism” and “antioxidant” groups, betaine and vitamin A had the highest conditional PIPs, respectively.
FIGURE 3.
Adjusted associations between prenatal intake of 14 nutrients and child SRS scores using hierarchical Bayesian kernel machine regression (BKMR) (n = 952). Results of BKMR analyses in ECHO including energy-adjusted nutrient intake for omega 3 and omega 6 fatty acids, iron, vitamin D, folate, zinc, vitamin B12, vitamin B6, choline, betaine, vitamin A, vitamin C, vitamin E, and methionine, adjusted for maternal age, maternal prepregnancy BMI, child sex (male, female), cohort type (enriched familial autism probability, general population), maternal ethnicity/race (non-Hispanic White, non-Hispanic Black, Hispanic, and other), maternal education (less than high school, high school/GED, some college/associates degree/trade school, bachelor’s degree, and graduate degree), maternal smoking (yes, no), and child year of birth (1998–2004, 2005–2009, 2010–2014, and 2015+). Plots show: (A) the overall mixture effect, or the relationship between the nutrient mixture and child SRS scores. (B) Individual nutrient associations with SRS scores, holding all other nutrients at their 50th percentile. (C) Single-exposure effects; plot shows the impact on child SRS score when each exposure increases from its 25th to 75th percentile (whereas other exposures are fixed at their 25th, 50th, or 75th percentiles). (D) Interaction parameters; plot compares the single-exposure health risk of each exposure (associated with a change from its 25th to 75th percentile) when other exposures are fixed to their 75th percentile compared with their 25th percentile. The set of nutrients included here was selected as described in the text based on a priori interest for neurodevelopmental relevance and grouped by shared pathway. ECHO, Environmental influences on Child Health Outcomes; SRS, Social Responsiveness Scale.
FIGURE 4.
Adjusted bivariate associations between prenatal intake of 14 nutrients and child SRS scores using hierarchical Bayesian kernel machine regression (BKMR) (n = 952). Results of BKMR analyses in ECHO including energy-adjusted nutrient intake for omega 3 and omega 6 fatty acids, iron, vitamin D, folate, zinc, vitamin B12, vitamin B6, choline, betaine, vitamin A, vitamin C, vitamin E, and methionine, adjusted for maternal age, maternal prepregnancy BMI, child sex (male, female), cohort type (enriched familial autism probability, and general population), maternal ethnicity/race (non-Hispanic White, non-Hispanic Black, Hispanic, and other), maternal education (less than high school, high school/GED, some college/associates degree/trade school, bachelor’s degree, and graduate degree), maternal smoking (yes and no), and child year of birth (1998–2004, 2005–2009, 2010–2014 and 2015+). Plots show bivariate exposure–response function for 2 predictors, where the second predictor is fixed at various percentiles (10th, 50th, and 90th) and all other predictors are fixed at the 50th percentile. The set of nutrients included here was selected as described in the text based on a priori interest for neurodevelopmental relevance and grouped by shared pathway. ECHO, Environmental influences on Child Health Outcomes.
BKMR analysis of nutrients with autism diagnosis
Most (91%) of the participants in the primary analyses also provided information on children’s autism diagnosis and were included in this secondary analysis. A total of 3280 dyads provided diagnostic information, with 137 children (4.2%) identified as autism cases. Participant characteristics were similar to those in the primary analysis (Supplemental Table 5). As in primary analyses, we conducted BKMR with a 5-nutrient mixture (7 cohorts, n = 3280) and a 14-nutrient mixture (4 cohorts, n = 1596); nutrient intakes across these groups were broadly comparable (Supplemental Table 6).
In covariate-adjusted probit BKMR models, credible intervals included the null for associations of the overall mixture and individual nutrients with an autism diagnosis, either when examining 5 nutrients (Supplemental Figure 3) or 14 nutrients (Supplemental Figure 4). When assessing relative contributions in these nutrient mixtures, the inflammatory group and omega 6 fatty acids had the highest group and conditional PIPs in the 5-nutrient analysis, respectively, and the inflammatory group and vitamin E and omega 6 fatty acids had the highest group and conditional PIPs in the 14-nutrient analysis (Supplemental Table 7).
Comparison analyses using traditional regression approaches
To compare with previous work assessing individual nutrients, we conducted regression analyses of single nutrients with autism outcomes. Using adjusted quantile regression models to examine the association of individual nutrients with SRS scores, statistically significant associations with SRS scores were identified for omega 3 and omega 6 fatty acid intake. Compared with the lowest quartile of intake, higher intake was consistently inversely associated with SRS score, with beta coefficients ranging from –1.63 to –3.91 (Supplemental Table 8). Folate intake above the first quartile was also consistently, inversely associated with SRS score; however, the results were not statistically significant. Given the suggestive evidence for an interaction of folate and omega 3 fatty acids in the primary BKMR model, we examined these 2 nutrients in an adjusted regression model together with an interaction term, but the term was not statistically significant (P-interaction = 0.98). When the expanded suite of 14 nutrients was examined in the subsample with complete data, vitamins A, B6, B12, D, and zinc had statistically significant inverse associations with SRS score in ≥1 quartile, with betas ranging from –2.18 to –5.40 (Supplemental Table 9). In adjusted logistic regression models, no nutrients were consistently significantly associated with autism diagnosis across intake quartiles, whether considering the primary or expanded sets of nutrients (Supplemental Tables 10 and 11).
Sensitivity analyses
In sensitivity analyses, the association of the overall mixture remained similar, but associations with individual nutrients changed. For example, when the primary 5-nutrient BKMR analysis was performed in the subsample with complete data on 14 nutrients (4 cohorts, n = 952), vitamin D and iron were inversely associated with SRS score, and folate was positively associated (Supplemental Figure 5). In analyses examining total intake of the 5 primary nutrients from diet and supplements (4 cohorts, n = 952; Supplemental Table 12), vitamin D was inversely associated with SRS score, and folate was positively associated with SRS score at higher folate intakes (Supplemental Figure 6). In both sensitivity analyses, there was no clear association with omega 3 fatty acids.
Given that these subsamples included only 4 ECHO cohorts (compared with 7), we examined whether associations changed with the exclusion of individual cohorts using sensitivity analyses excluding 1 cohort at a time. In quantile regression models, the removal of the large New Hampshire Birth Cohort Study (NHBCS) led to changes in estimates for associations of nutrients with SRS score (Supplemental Figure 7). In particular, associations with vitamin D became inverse rather than positive, and associations with folate and omega 6 fatty acids became null. Sensitivity analyses examining additional adjustment for covariates in regression models did not substantially alter results, with the exception of adjustment for FFQ type, which attenuated the significant inverse associations of omega 3 and omega 6 fatty acids with SRS score (Supplemental Table 13). In regression models stratified by FFQ type (not conducted for DHQ given small sample size), although there was some variability for certain nutrients across FFQs, overall patterns of associations with SRS and ASD were similar across stratified and full sample results (Supplemental Tables 14 and 15). When FFQ type was added as a covariate to the primary BKMR analysis, the overall mixture effect was attenuated (Supplemental Figure 8), although FFQ type differed by cohort (Supplemental Table 1).
Discussion
In this large and geographically and demographically diverse sample from the ECHO cohort consortium, we found some evidence for associations of prenatal dietary nutrient intake with children’s autism outcomes. By employing a Bayesian mixtures approach, we examined complex relationships among a suite of nutrients while accounting for multicollinearity. In our primary models, we found evidence for U-shaped relationships of folate and omega 3 fatty acids with children’s autism-related traits, as well as a trend for lower SRS scores as the overall nutrient mixture—which included several nutrients important for neurodevelopment—increased. These results align with prior work suggesting the importance of prenatal intake of these nutrients for neurodevelopment and involvement in autism development. After expanding our models to include a larger suite of nutrients, associations with specific nutrients were altered, with vitamin D gaining importance in the model, suggesting the need to carefully consider the relative contributions of components of the diet and their potential combined effects.
To our knowledge, this is among the largest analyses of dietary exposures and neurodevelopmental outcomes using mixture models [24,48,49]. Previously, we performed BKMR analysis using data from the NHSII (n = 713) and the EARLI study (n = 127) [26]. In the EARLI familial autism cohort, there was no evidence for associations of maternal nutrient intake with children’s SRS scores; however, in the NHSII general population cohort, there was suggestive evidence for a positive association of omega 3 fatty acids and negative associations of omega 6 fatty acids, iron, and vitamin B12 with SRS score, as well as a trend for lower SRS score with increasing values of the overall nutrient mixture. In the current study, we did not find evidence for associations with the same nutrients; however, we did identify a similar trend in overall mixtures effects, underscoring the importance of taking a comprehensive look at diet and dietary patterns for neurodevelopment. Of note, the EARLI study was also included in this analysis, as it is an ECHO cohort study site. However, it contributed a relatively small sample and did not strongly influence results in leave-one-out sensitivity analysis.
There are several possible explanations for differences in the individual nutrient associations identified with BKMR, across studies and between the primary (5 nutrient) and secondary (14 nutrient) analyses of the current study. First, the differences may be because of the inclusion of different combinations of nutrients in BKMR models. The results of these models may be sensitive to the complex interrelationships characteristic of nutrient intake, including true biological interactions. This is a strength of mixtures modeling; however, researchers should be thoughtful when selecting nutrients to include in future analyses. For example, in this analysis, the relative importance of folate was reduced after the inclusion of other nutrients on the methylation pathway, especially betaine. In fact, betaine was identified as being individually associated with SRS score. Given that betaine and folate are complementary sources of 1-carbon units used in methylation pathways [50], researchers should consider the complex interplay of these nutrients in gene methylation and autism development.
Differences in associations identified by BKMR models could also be because of differences in sample characteristics, such as nutrient intake or SRS scores. In the current analysis, the exclusion of the NHBCS influenced associations with vitamin D, folate, and omega 6 fatty acids in traditional regression models. NHBCS is a largely rural cohort, with different access to food systems as well as health care. Participants also tended to be non-Hispanic White, with a relatively high level of education, in part due to the selection of pregnancies that were using private, unregulated water systems at their place of residence [51]. Despite these differences, nutrient intake distributions and SRS scores were similar between this cohort and others included in this analysis. Furthermore, we included adjustments for sample characteristics including race, ethnicity, and education. It is possible that these results are because of the reduced sample size of models excluding NHBCS, as this was also the largest cohort included in the analysis, contributing nearly 40% of participants.
Differing results could also be because of variations in assessment methodology across cohorts. For example, this analysis used dietary data collected from 3 different validated FFQs. The 14-nutrient sample data were all drawn from the Block FFQ, whereas the 5-nutrient sample had greater contributions from the Harvard/Willett FFQ. Though prior work supports the comparability of nutrient data from these sources [37], minor measurement differences may have contributed to variability. We also observed differences in the overall mixture effect when further adjusting for diet form type, although this was also highly related to cohort and may have inflated variance in coadjustment, particularly in standard models.
Most previous work examining links between prenatal diet and children’s autism development has used traditional single-exposure regression models, rather than mixtures models. A description of this work, focusing on folate, fatty acids, and vitamin D, is included next. In our analysis, associations of certain nutrients with autism outcomes were present in traditional regression models but not in mixtures models. This suggests that concerns of coexposure confounding across nutrients may impact traditional analyses of nutrient-autism associations. Mixtures models may be better able to handle complex associations between nutrients and within metabolic pathways.
Despite the challenges of traditional regression models, studies using these models have also identified similar associations between folate, omega 3 fatty acids, and vitamin D with autism outcomes as those identified by mixtures models here. Folate, in particular, has been studied for its role in autism development [52]. Previous work has generally reported negative associations between supplemental folic acid intake and autism diagnosis or traits [12,53], although there is also evidence for positive associations at the highest intake level [13]. In this study, we found evidence for a U-shaped relationship between dietary folate and SRS score, supporting previous research, although credible intervals were wide at higher folate intakes. Folate is thought to influence neurodevelopment primarily via its involvement in the 1-carbon metabolism pathway, which impacts gene methylation. Other nutrients on the 1-carbon metabolism pathway have been studied less. Prenatal betaine, and its precursor choline, have been positively associated with children’s neurodevelopment [54,55]. Evidence from animal models suggests that these nutrients may also influence autism development specifically [[56], [57], [58]]. Although there are no United States recommendations for betaine intake during pregnancy, insufficient intake of choline during pregnancy is common [59,60]. Given that betaine was associated with SRS score in this analysis, further research including all nutrients on the methylation pathway is recommended.
Omega 3 and 6 fatty acids were also associated with autism outcomes in this study and in traditional analyses [16,17]. These fatty acids are crucial for membrane formation, cell signaling, and inflammatory processes. In our analysis, we found a U-shaped association of omega 3 fatty acids with autism trait scores and a possible interaction with folate in the 5-nutrient analysis. Our data suggest that higher prenatal intake of omega 3 fatty acids may modestly increase both the inverse association (at moderate intake levels) and positive association (at extreme intake levels) of folate with SRS score. Other studies have noted a link between folate and omega 3 fatty acids (especially DHA). For example, maternal plasma levels of folate and DHA are positively associated during pregnancy [61]. Additionally, animal models have noted improvements in cognitive performance after prenatal co-supplementation of B vitamins (including folate) and fatty acids, above supplementation with a single nutrient [62]. However, to our knowledge, no other studies have identified this interaction in relation to autism development.
In this study, there was suggestive evidence for a negative association of dietary vitamin D with SRS score in mixtures models, similar to previous literature using traditional regression models [[18], [19], [20]], but only when a wider suite of nutrients was also included in the model. Perhaps this could be because of correlated food sources of certain nutrients (that is, fish as a source of both omega 3s and vitamin D), although the diet is not generally the major contributor to vitamin D status. Mechanistic studies of the role of vitamin D in the context of other nutrients may be needed to explain this finding more clearly.
Although this work represents the largest and most detailed BKMR mixtures analysis of nutrient intake and autism outcomes to date, several limitations apply. First, cohorts differed in the timing and type of FFQ used for dietary assessment. Although pooling of data is a major strength of the ECHO consortium, these differences may have contributed to our varying results across primary and secondary analyses, as noted above. Differences across dietary assessment methods also meant that we were not able to assess potential differences by trimester of pregnancy and could examine nutrient intake from supplements in only a subset of participants. Nutrient supplements are a major source of some of our primary nutrients of interest [47], and prenatal vitamin use has been consistently associated with autism development [12,53]. In sensitivity analyses with supplemental intake, differences in individual nutrient associations were noted, suggesting future analyses should further explore the role of nutrient supplements in complex nutrient-SRS associations. Because data collection continues in the ECHO study under a standardized protocol, harmonized information on dietary intake, supplement use, and nutrient biomarkers will become more widely available [29]. Additionally, all dietary intake assessments are subject to error and bias, and FFQs tend to overestimate intake [32]; however, we reduced the potential for these errors and biases by using only prospectively collected, validated FFQs and adjusting for energy intake.
Another limitation is that several cohorts only provided parent reports of physician autism diagnoses, without clinical confirmation. This may partly explain why we did not find similar associations of nutrients with autism diagnosis as with SRS score. However, 49% of our sample also conducted clinical assessments or medical record abstraction to confirm autism diagnosis, and results were similar in these cohorts. Alternatively, it may be that quantitative autism trait scores are more sensitive to associations with nutrients across the distribution.
Finally, without mechanistic studies, we cannot be certain that the differences in nutrient associations seen in the 5-nutrient primary analysis and 14-nutrient secondary analysis were because of true biological interactions rather than statistical modeling effects or differences in sample characteristics. As models evolve and improve, and as animal and human evidence increases, we may have a better understanding of the complex interrelationships of nutrients within and across pathways that influence neurodevelopment.
In conclusion, this study highlights the use of mixtures models to explore the interrelated associations of prenatal nutrient intakes with autism development. Using these models, we found evidence for individual nutrient associations previously documented using traditional regression analysis, as well as novel nutrient interactions. Continued examination of the complex associations between nutrients and autism development using mixtures models, complemented by mechanistic studies, can help refine our understanding of nutrition’s impact on neurodevelopment.
Author contributions
The authors’ responsibilities were as follows – KNC, SME, MRK, P-IL, RJS, KL: directed cohort-level data collection; KL: designed the analysis; JR: performed statistical analysis; MGB: aided in study oversight, wrote the paper, and had primary responsibility for final content; and all authors: read and approved the final manuscript.
Data availability
Select deidentified data from the ECHO Program are available through NICHD’s Data and Specimen Hub (DASH). Information on study data not available on DASH, such as some Indigenous datasets, can be found on the ECHO study DASH webpage.
Funding
Research reported in this publication was supported by the Environmental influences on Child Health Outcomes (ECHO) Program, Office of the Director, National Institutes of Health, under Award Numbers U2COD023375 (Coordinating Center), U24OD023382 (Data Analysis Center), U24OD023319 with cofunding from the Office of Behavioral and Social Science Research (Measurement Core), U24OD035523 (Lab Core), ES0266542 (HHEAR), U24ES026539 (HHEAR Barbara O’Brien), U2CES026533 (HHEAR Lisa Peterson), U2CES026542 (HHEAR Patrick Parsons, Kannan Kurunthacalam), U2CES030859 (HHEAR Manish Arora), U2CES030857 (HHEAR Timothy R. Fennell, Susan J. Sumner, Xiuxia Du), U2CES026555 (HHEAR Susan L. Teitelbaum), U2CES026561 (HHEAR Robert O. Wright), U2CES030851 (HHEAR Heather M. Stapleton, P. Lee Ferguson), UG3/UH3OD023251 (Akram Alshawabkeh), UH3OD023320 and UG3OD035546 (Judy Aschner), UH3OD023332 (Clancy Blair, Leonardo Trasande), UG3/UH3OD023253 (Carlos Camargo), UG3/UH3OD023248 and UG3OD035526 (Dana Dabelea), UG3/UH3OD023313 (Daphne Koinis Mitchell), UH3OD023328 (Cristiane Duarte), UH3OD023318 (Anne Dunlop), UG3/UH3OD023279 (Amy Elliott), UG3/UH3OD023289 (Assiamira Ferrara), UG3/UH3OD023282 (James Gern), UH3OD023287 (Carrie Breton), UG3/UH3OD023365 (Irva Hertz-Picciotto), UG3/UH3OD023244 (Alison Hipwell), UG3/UH3OD023275 (Margaret Karagas), UH3OD023271 and UG3OD035528 (Catherine Karr), UH3OD023347 (Barry Lester), UG3/UH3OD023389 (Leslie Leve), UG3/UH3OD023344 (Debra MacKenzie), UH3OD023268 (Scott Weiss), UG3/UH3OD023288 (Cynthia McEvoy), UG3/UH3OD023342 (Kristen Lyall), UG3/UH3OD023349 (Thomas O’Connor), UH3OD023286 and UG3OD035533 (Emily Oken), UG3/UH3OD023348 (Mike O’Shea), UG3/UH3OD023285 (Jean Kerver), UG3/UH3OD023290 (Julie Herbstman), UG3/UH3OD023272 (Susan Schantz), UG3/UH3OD023249 (Joseph Stanford), UG3/UH3OD023305 (Leonardo Trasande), UG3/UH3OD023337 (Rosalind Wright), UG3OD035508 (Sheela Sathyanarayana), UG3OD035509 (Anne Marie Singh), UG3OD035513 and UG3OD035532 (Annemarie Stroustrup), UG3OD035516 and UG3OD035517 (Tina Hartert), UG3OD035518 (Jennifer Straughen), UG3OD035519 (Qi Zhao), UG3OD035521 (Katherine Rivera-Spoljaric), UG3OD035527 (Emily S Barrett), UG3OD035540 (Monique Marie Hedderson), UG3OD035543 (Kelly J Hunt), UG3OD035537 (Sunni L Mumford), UG3OD035529 (Hong-Ngoc Nguyen), UG3OD035542 (Hudson Santos), UG3OD035550 (Rebecca Schmidt), UG3OD035536 (Jonathan Slaughter), UG3OD035544 (Kristina Whitworth).
Conflict of interest
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The authors report no conflicts of interest.
Acknowledgments
We wish to thank our ECHO Colleagues; the medical, nursing, and program staff; and the children and families participating in the ECHO cohort.
Footnotes
A version of this work was presented at the ASN 2024 Conference 29 June–2 July as a Late Breaking Poster.
Supplementary data to this article can be found online at https://doi.org/10.1016/j.tjnut.2025.02.025.
Contributor Information
Kristen Lyall, Email: kld98@drexel.edu.
ECHO Cohort Consortium:
P Brian Smith, L Kristin Newby, Linda Adair, Lisa P Jacobson, Diane Catellier, Monica McGrath, Christian Douglas, Priya Duggal, Emily Knapp, Amii Kress, Courtney K Blackwell, Maxwell A Mansolf, Jin-Shei Lai, Emily Ho, David Cella, Richard Gershon, Michelle L Macy, Suman R Das, Jane E Freedman, Simon A Mallal, John A McLean, Ravi V Shah, Meghan H Shilts, Akram N Alshawabkeh, Jose F Cordero, John Meeker, Leonardo Trasande, Carlos A Camargo, Zhaozhong Zhu, Ashley F Sullivan, Dana Dabelea, Wei Perng, Traci A Bekelman, Greta Wilkening, Sheryl Magzamen, Brianna F Moore, Anne P Starling, Deborah J Rinehart, Daphne Koinis Mitchell, Viren D'Sa, Sean CL Deoni, Hans-Georg Mueller, Cristiane S Duarte, Catherine Monk, Glorisa Canino, Jonathan Posner, Tenneill Murray, Claudia Lugo-Candelas, Anne L Dunlop, Patricia A Brennan, Christine Hockett, Amy Elliott, Assiamira Ferrara, Lisa A Croen, Monique M Hedderson, John Ainsworth, Leonard B Bacharier, Casper G Bendixsen, James E Gern, Diane R Gold, Tina V Hartert, Daniel J Jackson, Christine C Johnson, Christine LM Joseph, Meyer Kattan, Gurjit K Khurana Hershey, Robert F Lemanske, Susan V Lynch, Rachel L Miller, George T O’Connor, Carole Ober, Dennis Ownby, Katherine Rivera-Spoljaric, Patrick H Ryan, Christine M Seroogy, Anne Marie Singh, Robert A Wood, Edward M Zoratti, Rima Habre, Shohreh Farzan, Frank D Gilliland, Irva Hertz-Picciotto, Deborah H Bennett, Julie B Schweitzer, Rebecca J Schmidt, Janine M LaSalle, Alison E Hipwell, Kate E Keenan, Catherine J Karr, Nicole R Bush, Kaja Z LeWinn, Sheela Sathyanarayana, Qi Zhao, Frances Tylavsky, Kecia N Carroll, Christine T Loftus, Leslie D Leve, Jody M Ganiban, Jenae M Neiderhiser, Scott T Weiss, Augusto A Litonjua, Cindy T McEvoy, Eliot R Spindel, Robert S Tepper, Craig J Newschaffer, Kristen Lyall, Heather E Volk, Rebecca Landa, Sally Ozonoff, Joseph Piven, Heather Hazlett, Juhi Pandey, Robert Schultz, Steven Dager, Kelly Botteron, Daniel Messinger, Wendy Stone, Jennifer Ames, Thomas G O'Connor, Richard K Miller, Emily Oken, Michele R Hacker, Tamarra James-Todd, T Michael O'Shea, Rebecca C Fry, Jean A Frazier, Rachana Singh, Caitlin Rollins, Angela Montgomery, Ruben Vaidya, Robert M Joseph, Lisa K Washburn, Semsa Gogcu, Kelly Bear, Julie V Rollins, Stephen R Hooper, Genevieve Taylor, Wesley Jackson, Amanda Thompson, Julie Daniels, Michelle Hernandez, Kun Lu, Michael Msall, Madeleine Lenski, Rawad Obeid, Steven L Pastyrnak, Elizabeth Jensen, Christina Sakai, Hudson Santos, Jean M Kerver, Nigel Paneth, Charles J Barone, Michael R Elliott, Douglas M Ruden, Chris Fussman, Julie B Herbstman, Amy Margolis, Susan L Schantz, Sarah Dee Geiger, Andrea Aguiar, Karen Tabb, Rita Strakovsky, Tracey Woodruff, Rachel Morello-Frosch, Amy Padula, Joseph B Stanford, Christina A Porucznik, Angelo P Giardino, Rosalind J Wright, Robert O Wright, Brent Collett, Nicole Baumann-Blackmore, Ronald Gangnon, Daniel J Jackson, Chris G McKennan, Jo Wilson, Matt Altman, Judy L Aschner, Annemarie Stroustrup, Stephanie L Merhar, Paul E Moore, Gloria S Pryhuber, Mark Hudak, Ann Marie Reynolds Lyndaker, Andrea L Lampland, Burton Rochelson, Sophia Jan, Matthew J Blitz, Michelle W Katzow, Zenobia Brown, Codruta Chiuzan, Timothy Rafael, Dawnette Lewis, Natalie Meirowitz, Brenda Poindexter, Tebeb Gebretsadik, Sarah Osmundson, Jennifer K Straughen, Amy Eapen, Andrea Cassidy-Bushrow, Ganesa Wegienka, Alex Sitarik, Kim Woodcroft, Audrey Urquhart, Albert Levin, Tisa Johnson-Hooper, Brent Davidson, Tengfei Ma, Emily S Barrett, Martin J Blaser, Maria Gloria Dominguez-Bello, Daniel B Horton, Manuel Jimenez, Todd Rosen, Kristy Palomares, Lyndsay A Avalos, Yeyi Zhu, Kelly J Hunt, Roger B Newman, Michael S Bloom, Mallory H Alkis, James R Roberts, Sunni L Mumford, Heather H Burris, Sara B DeMauro, Lynn M Yee, Aaron Hamvas, Antonia F Olidipo, Andrew S Haddad, Lisa R Eiland, Nicole T Spillane, Kirin N Suri, Stephanie A Fisher, Jeffrey A Goldstein, Leena B Mithal, Raye-Ann O DeRegnier, Nathalie L Maitre, Ruby HN Nguyen, Meghan M JaKa, Abbey C Sidebottom, Michael J Paidas, JoNell E Potter, Natale Ruby, Lunthita Duthely, Arumugam Jayakumar, Karen Young, Isabel Maldonado, Meghan Miller, Jonathan L Slaughter, Sarah A Keim, Courtney D Lynch, Kartik K Venkatesh, Kristina W Whitworth, Elaine Symanski, Thomas F Northrup, Hector Mendez-Figueroa, Ricardo A Mosquera, Margaret R Karagas, Juliette C Madan, Debra M MacKenzie, Johnnye L Lewis, Brandon J Rennie, Bennett L Leventhal, Young Shin Kim, Somer Bishop, Sara S Nozadi, Li Luo, Barry M Lester, Carmen J Marsit, Todd Everson, Cynthia M Loncar, Elisabeth C McGowan, Stephen J Sheinkopf, Brian S Carter, Jennifer Check, Jennifer B Helderman, Charles R Neal, and Lynne M Smith
Appendix A. Supplementary data
The following is the Supplementary data to this article:
References
- 1.Prado E.L., Dewey K.G. Nutrition and brain development in early life. Nutr. Rev. 2014;72(4):267–284. doi: 10.1111/nure.12102. [DOI] [PubMed] [Google Scholar]
- 2.Zhong C., Tessing J., Lee B.K., Lyall K. Maternal dietary factors and the risk of autism spectrum disorders: a systematic review of existing evidence. Autism. Res. 2020;13(10):1634–1658. doi: 10.1002/aur.2402. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Maenner M.J., Warren Z., Williams A., Amoakohene E., Bakian A.V., Bilder D.A., et al. Prevalence and characteristics of autism spectrum disorder among children aged 8 years — autism and developmental disabilities monitoring network, 11 sites, United States, 2020, Morb. Mortal. Wkly Rep. MMWR. 2023;72(2):1–14. doi: 10.15585/mmwr.ss7202a1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Bölte S., Girdler S., Marschik P.B. The contribution of environmental exposure to the etiology of autism spectrum disorder. Cell. Mol. Life Sci. 2019;76(7):1275–1297. doi: 10.1007/s00018-018-2988-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Hallmayer J., Cleveland S., Torres A., Phillips J., Cohen B., Torigoe T., et al. Genetic heritability and shared environmental factors among twin pairs with autism. Arch. Gen. Psychiatry. 2011;68(11):1095–1102. doi: 10.1001/archgenpsychiatry.2011.76. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Rahman M.M., Shu Y.H., Chow T., Lurmann F.W., Yu X., Martinez M.P., et al. Prenatal exposure to air pollution and autism spectrum disorder: sensitive windows of exposure and sex differences. Environ. Health Perspect. 2022;130(1) doi: 10.1289/EHP9509. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Christensen J., Grønborg T.K., Sørensen M.J., Schendel D., Parner E.T., Pedersen L.H., et al. Prenatal valproate exposure and risk of autism spectrum disorders and childhood autism. JAMA. 2013;309(16):1696–1703. doi: 10.1001/jama.2013.2270. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Constantino J.N., Todd R.D. Autistic traits in the general population: a twin study. Arch. Gen. Psychiatry. 2003;60:524–530. doi: 10.1001/archpsyc.60.5.524. [DOI] [PubMed] [Google Scholar]
- 9.Lyall K. What are quantitative traits and how can they be used in autism research? Autism Res. 2023;16(7):1289–1298. doi: 10.1002/aur.2937. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Dou J.F., Middleton L.Y.M., Zhu Y., Benke K.S., Feinberg J.I., Croen L.A., et al. Prenatal vitamin intake in first month of pregnancy and DNA methylation in cord blood and placenta in two prospective cohorts. Epigenetics Chromatin. 2022;15(1):28. doi: 10.1186/s13072-022-00460-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Bhutta Z.A., Guerrant R.L., Nelson C.A. Neurodevelopment, nutrition, and inflammation: the evolving global child health landscape. Pediatrics. 2017;139:S12–S22. doi: 10.1542/peds.2016-2828D. [DOI] [PubMed] [Google Scholar]
- 12.Schmidt R.J., Iosif A.-M., Angel E.G., Ozonoff S. Association of maternal prenatal vitamin use with risk for autism spectrum disorder recurrence in young siblings. JAMA Psychiatry. 2019;76(4):391–398. doi: 10.1001/jamapsychiatry.2018.3901. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Wiens D., DeSoto M.C. Is high folic acid intake a risk factor for autism? —A Review. Brain Sci. 2017;7(11):149. doi: 10.3390/brainsci7110149. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Tremblay M.W., Jiang Y.H. DNA methylation and susceptibility to autism spectrum disorder. Annu. Rev. Med. 2019;70:151–166. doi: 10.1146/annurev-med-120417-091431. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Keil K.P., Lein P.J. DNA methylation: a mechanism linking environmental chemical exposures to risk of autism spectrum disorders? Environ. Epigenetics. 2016;2 doi: 10.1093/eep/dvv012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Huang Y., Iosif A.M., Hansen R.L., Schmidt R.J. Maternal polyunsaturated fatty acids and risk for autism spectrum disorder in the MARBLES high-risk study. Autism. 2020;24(5):1191–1200. doi: 10.1177/1362361319877792. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Lyall K., Munger K.L., O'Reilly É.J., Santangelo S.L., Ascherio A. Maternal dietary fat intake in association with autism spectrum disorders. Am. J. Epidemiol. 2013;178(2):209–220. doi: 10.1093/aje/kws433. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Whitehouse A.J., Holt B.J., Serralha M., Holt P.G., Hart P.H., Kusel M.M. Maternal vitamin D levels and the autism phenotype among offspring. J. Autism Dev. Disord. 2013;43(7):1495–1504. doi: 10.1007/s10803-012-1676-8. [DOI] [PubMed] [Google Scholar]
- 19.Chen J., Xin K., Wei J., Zhang K., Xiao H. Lower maternal serum 25(OH) D in first trimester associated with higher autism risk in Chinese offspring. J. Psychosom. Res. 2016;89:98–101. doi: 10.1016/j.jpsychores.2016.08.013. [DOI] [PubMed] [Google Scholar]
- 20.Vinkhuyzen A.A.E., Eyles D.W., Burne T.H.J., Blanken L.M.E., Kruithof C.J., Verhulst F., et al. Gestational vitamin D deficiency and autism-related traits: the Generation R Study. Mol. Psychiatry. 2018;23(2):240–246. doi: 10.1038/mp.2016.213. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Zawadzka A., Cieślik M., Adamczyk A. The role of maternal immune activation in the pathogenesis of autism: a review of the evidence, proposed mechanisms and implications for treatment. Int. J. Mol. Sci. 2021;22(21) doi: 10.3390/ijms222111516. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Bobb J.F., Valeri L., Claus Henn B., Christiani D.C., Wright R.O., Mazumdar M., et al. Bayesian kernel machine regression for estimating the health effects of multi-pollutant mixtures. Biostatistics. 2015;16(3):493–508. doi: 10.1093/biostatistics/kxu058. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Bobb J.F., Claus Henn B., Valeri L., Coull B.A. Statistical software for analyzing the health effects of multiple concurrent exposures via Bayesian kernel machine regression. Environ. Health. 2018;17(1):67. doi: 10.1186/s12940-018-0413-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Goodrich A.J., Kleeman M.J., Tancredi D.J., Ludeña Y.J., Bennett D.H., Hertz-Picciotto I., et al. Ultrafine particulate matter exposure during second year of life, but not before, associated with increased risk of autism spectrum disorder in BKMR mixtures model of multiple air pollutants. Environ. Res. 2024;242 doi: 10.1016/j.envres.2023.117624. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Liu S.H., Bobb J.F., Claus Henn B., Gennings C., Schnaas L., Tellez-Rojo M., et al. Bayesian varying coefficient kernel machine regression to assess neurodevelopmental trajectories associated with exposure to complex mixtures. Stat. Med. 2018;37(30):4680–4694. doi: 10.1002/sim.7947. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Lyall K., Rando J., Wang S., Hamra G.B., Chavarro J., Weisskopf M.G., et al. Examining prenatal dietary factors in association with child autism-related traits using a Bayesian mixture approach: results from 2 United States Cohorts. Curr. Dev. Nutr. 2023;7(8) doi: 10.1016/j.cdnut.2023.101978. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Gillman M.W., Blaisdell C.J. Environmental influences on child health outcomes, a research program of the National Institutes of Health. Curr. Opin Pediatr. 2018;30(2):260–262. doi: 10.1097/MOP.0000000000000600. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Knapp E.A., Kress A.M., Parker C.B., Grier P., McArthur K., Gachigi K.K., et al. Cohort profile: the Environmental influences on Child Health Outcomes (ECHO)-wide cohort. Int. J. Epidemiol. 2023;192(8):1249–1263. doi: 10.1093/aje/kwad071. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Bragg M.G., Westlake M., Alshawabkeh A.N., Bekelman T.A., Camargo C.A., Catellier D.J., et al. Opportunities for examining child health impacts of early-life nutrition in the ECHO program: maternal and child dietary intake data from pregnancy to adolescence. Curr. Dev. Nutr. 2023;7(11) doi: 10.1016/j.cdnut.2023.102019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Sauder K.A., Harte R.N., Ringham B.M., Guenther P.M., Bailey R.L., Alshawabkeh A., et al. Disparities in risks of inadequate and excessive intake of micronutrients during pregnancy. J. Nutr. 2021;151(11):3555–3569. doi: 10.1093/jn/nxab273. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Freedman L.S., Commins J.M., Moler J.E., Arab L., Baer D.J., Kipnis V., et al. Pooled results from 5 validation studies of dietary self-report instruments using recovery biomarkers for energy and protein intake. Am. J. Epidemiol. 2014;180(2):172–188. doi: 10.1093/aje/kwu116. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Kipnis V., Midthune D., Freedman L., Bingham S., Day N.E., Riboli E., et al. Bias in dietary-report instruments and its implications for nutritional epidemiology. Public Health Nutr. 2002;5(6a):915–923. doi: 10.1079/phn2002383. [DOI] [PubMed] [Google Scholar]
- 33.Gibson R.S. Principles of nutritional assessment [Internet] 2020. https://nutritionalassessment.org/ [date updated March 2025; date cited April 10, 2025]
- 34.Block G., Woods M., Potosky A., Clifford C. Validation of a self-administered diet history questionnaire using multiple diet records. J. Clin. Epidemiol. 1990;43(12):1327–1335. doi: 10.1016/0895-4356(90)90099-b. [DOI] [PubMed] [Google Scholar]
- 35.Willett W.C., Sampson L., Stampfer M.J., Rosner B., Bain C., Witschi J., et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am. J. Epidemiol. 1985;122(1):51–65. doi: 10.1093/oxfordjournals.aje.a114086. [DOI] [PubMed] [Google Scholar]
- 36.Diet History Questionnaire, Version 2.0 . National Cancer Institute; 2010. National Institutes of Health, Epidemiology and Genomics Research Program.https://epi.grants.cancer.gov/dhq2/about/citations.html [Google Scholar]
- 37.Subar A.F., Thompson F.E., Kipnis V., Midthune D., Hurwitz P., McNutt S., et al. Comparative validation of the Block, Willett, and National Cancer Institute food frequency questionnaires: the Eating at America’s Table Study. Am. J. Epidemiol. 2001;154(12):1089–1099. doi: 10.1093/aje/154.12.1089. [DOI] [PubMed] [Google Scholar]
- 38.Constantino J., Gruber J. Western Psychological Services; Los Angeles: 2005. Social Responsiveness Scale (SRS) Manual. [Google Scholar]
- 39.Sturm A., Kuhfeld M., Kasari C., McCracken J.T. Development and validation of an item response theory-based Social Responsiveness Scale short form. J. Child Psychol. Psychiatry. 2017;58(9):1053–1061. doi: 10.1111/jcpp.12731. [DOI] [PubMed] [Google Scholar]
- 40.Constantino J.N., Davis S.A., Todd R.D., Schindler M.K., Gross M.M., Brophy S.L., et al. Validation of a brief quantitative measure of autistic traits: comparison of the social responsiveness scale with the autism diagnostic interview-revised. J. Autism Dev. Disord. 2003;33(4):427–433. doi: 10.1023/a:1025014929212. [DOI] [PubMed] [Google Scholar]
- 41.Patti M.A., Ning X., Hosseini M., Croen L.A., Joseph R.M., Karagas M.R., et al. A comparative analysis of the full and short versions of the Social Responsiveness Scale in estimating an established autism risk factor association in ECHO: do we get the same estimates? J. Autism. Dev. Disord. 2023 doi: 10.1007/s10803-023-06020-8. Online ahead of print. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Patti M.A., Croen L.A., Dickerson A.S., Joseph R.M., Ames J.L., Ladd-Acosta C., et al. Reproducibility between preschool and school-age Social Responsiveness Scale forms in the Environmental Influences on Child Health Outcomes Program. Autism Res. 2024;17(6):1187–1204. doi: 10.1002/aur.3147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Lyall K., Rando J., Toroni B., Ezeh T., Constantino J.N., Croen L.A., et al. Examining shortened versions of the Social Responsiveness Scale for use in autism spectrum disorder prediction and as a quantitative trait measure: results from a validation study of 3-5 year old children. JCPP Adv. 2022;2(4) doi: 10.1002/jcv2.12106. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Lyall K., Hosseini M., Ladd-Acosta C., Ning X., Catellier D., Constantino J.N., et al. Distributional properties and criterion validity of a shortened version of the Social Responsiveness Scale: results from the ECHO Program and Implications for Social Communication Research. J. Autism Dev. Disord. 2021;51(7):2241–2253. doi: 10.1007/s10803-020-04667-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Oppenheimer A.V., Weisskopf M.G., Lyall K. An examination of family transmission of traits measured by the Social Responsiveness Scale-Short Form. J. Autism Dev. Disord. 2024;54(11):4034–4044. doi: 10.1007/s10803-023-06115-2. [DOI] [PubMed] [Google Scholar]
- 46.Duncan A.F., Montoya-Williams D. Recommendations for reporting research about racial disparities in medical and scientific journals. JAMA Pediatr. 2024;178(3):221–224. doi: 10.1001/jamapediatrics.2023.5718. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Bailey R.L., Pac S.G., Fulgoni V.L., III, Reidy K.C., Catalano P.M. Estimation of total usual dietary intakes of pregnant women in the United States, JAMA Netw. Open. 2019;2(6) doi: 10.1001/jamanetworkopen.2019.5967. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Liu C., Huang L., Huang S., Wei L., Cao D., Zan G., et al. Association of both prenatal and early childhood multiple metals exposure with neurodevelopment in infant: a prospective cohort study. Environ. Res. 2022;205 doi: 10.1016/j.envres.2021.112450. [DOI] [PubMed] [Google Scholar]
- 49.Chiu Y.M., Wilson A., Hsu H.L., Jamal H., Mathews N., Kloog I., et al. Prenatal ambient air pollutant mixture exposure and neurodevelopment in urban children in the Northeastern United States. Environ. Res. 2023;233 doi: 10.1016/j.envres.2023.116394. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Niculescu M.D., Zeisel S.H. Diet, methyl donors and DNA methylation: interactions between dietary folate, methionine and choline. J. Nutr. 2002;132(8 Suppl):2333s–2335s. doi: 10.1093/jn/132.8.2333S. [DOI] [PubMed] [Google Scholar]
- 51.Karagas M.R., Punshon T., Sayarath V., Jackson B.P., Folt C.L., Cottingham K.L. Association of rice and rice-product consumption with arsenic exposure early in life. JAMA Pediatr. 2016;170(6):609–616. doi: 10.1001/jamapediatrics.2016.0120. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Hoxha B., Hoxha M., Domi E., Gervasoni J., Persichilli S., Malaj V., et al. Folic acid and autism: a systematic review of the current state of knowledge. Cells. 2021;10(8):1976. doi: 10.3390/cells10081976. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Levine S.Z., Kodesh A., Viktorin A., Smith L., Uher R., Reichenberg A., et al. Association of maternal use of folic acid and multivitamin supplements in the periods before and during pregnancy with the risk of autism spectrum disorder in offspring. JAMA Psychiatry. 2018;75(2):176–184. doi: 10.1001/jamapsychiatry.2017.4050. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Wu B.T.F., Dyer R.A., King D.J., Richardson K.J., Innis S.M. Early second trimester maternal plasma choline and betaine are related to measures of early cognitive development in term infants. PLOS ONE. 2012;7(8) doi: 10.1371/journal.pone.0043448. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Caudill M.A., Strupp B.J., Muscalu L., Nevins J.E.H., Canfield R.L. Maternal choline supplementation during the third trimester of pregnancy improves infant information processing speed: a randomized, double-blind, controlled feeding study. FASEB J. 2018;32(4):2172–2180. doi: 10.1096/fj.201700692RR. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56.Huang F., Chen X., Jiang X., Niu J., Cui C., Chen Z., et al. Betaine ameliorates prenatal valproic-acid-induced autism-like behavioral abnormalities in mice by promoting homocysteine metabolism. Psychiatry Clin. Neurosci. 2019;73(6):317–322. doi: 10.1111/pcn.12833. [DOI] [PubMed] [Google Scholar]
- 57.Langley E.A., Krykbaeva M., Blusztajn J.K., Mellott T.J. High maternal choline consumption during pregnancy and nursing alleviates deficits in social interaction and improves anxiety-like behaviors in the BTBR T+Itpr3tf/J mouse model of autism. Behav. Brain Res. 2015;278:210–220. doi: 10.1016/j.bbr.2014.09.043. [DOI] [PubMed] [Google Scholar]
- 58.Agam G., Taylor Z., Vainer E., Golan H.M. The influence of choline treatment on behavioral and neurochemical autistic-like phenotype in Mthfr-deficient mice. Transl. Psychiatry. 2020;10(1):316. doi: 10.1038/s41398-020-01002-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59.Wiedeman A.M., Barr S.I., Green T.J., Xu Z., Innis S.M., Kitts D.D. Dietary choline intake: current state of knowledge across the life cycle. Nutrients. 2018;10(10):1513. doi: 10.3390/nu10101513. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60.Wallace T.C., Fulgoni V.L., III Assessment of total choline intakes in the United States. J. Am. College Nutr. 2016;35(2):108–112. doi: 10.1080/07315724.2015.1080127. [DOI] [PubMed] [Google Scholar]
- 61.Wadhwani N.S., Pisal H.R., Mehendale S.S., Joshi S.R. A prospective study of maternal fatty acids, micronutrients and homocysteine and their association with birth outcome, Matern. Child. Nutr. 2015;11(4):559–573. doi: 10.1111/mcn.12062. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62.Kemse N., Kale A., Chavan-Gautam P., Joshi S. Increased intake of vitamin B12, folate, and omega-3 fatty acids to improve cognitive performance in offspring born to rats with induced hypertension during pregnancy. Food Funct. 2018;9(7):3872–3883. doi: 10.1039/c8fo00467f. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Select deidentified data from the ECHO Program are available through NICHD’s Data and Specimen Hub (DASH). Information on study data not available on DASH, such as some Indigenous datasets, can be found on the ECHO study DASH webpage.




