Prenatal diet as a modifier of environmental risk factors for autism and related neurodevelopmental outcomes

Abstract

Purpose of review:

Environmental chemicals and toxins have been associated with increased risk of impaired neurodevelopment and specific conditions like autism spectrum disorder (ASD). Prenatal diet is an individually modifiable factor that may alter associations with such environmental factors. The purpose of this review is to summarize studies examining prenatal dietary factors as potential modifiers of the relationship between environmental exposures and ASD or related neurodevelopmental outcomes.

Recent findings:

Twelve studies were identified; five examined ASD diagnosis or ASD-related traits as the outcome (age at assessment range: 2–5 years) while the remainder addressed associations with neurodevelopmental scores (age at assessment range: 6 months to 6 years). Most studies focused on folic acid, prenatal vitamins, or omega-3 fatty acids as potentially beneficial effect modifiers. Environmental risk factors examined included air pollutants, endocrine disrupting chemicals, pesticides, and heavy metals. Most studies took place in North America. In 10/12 studies, the prenatal dietary factor under study was identified as a significant modifier, generally attenuating the association between the environmental exposure and ASD or neurodevelopment.

Summary:

Prenatal diet may be a promising target to mitigate adverse effects of environmental exposures on neurodevelopmental outcomes. Further research focused on joint effects is needed that encompasses a broader variety of dietary factors, guided by our understanding of mechanisms linking environmental exposures with neurodevelopment. Future studies should also aim to include diverse populations, utilize advanced methods to optimize detection of novel joint effects, incorporate consideration of timing, and consider both synergistic and antagonistic potential of diet.

Introduction

Autism spectrum disorder (ASD) is a neurodevelopmental disorder often diagnosed in early childhood that is characterized by deficits in social communication and restricted, repetitive behaviors¹. In 2018, 1 in 44 (23.0 per 1000) children age 8 years were diagnosed with ASD across 11 sites in the United States (US)². More broadly, prevalence of any developmental disability among US children age 3–17 was 15% in 2006–2008³. Development of ASD, and neurodevelopmental disorders in general, is influenced by a myriad of genetic and environmental factors^{4, 5}.

Prenatal exposure to several classes of environmental chemicals and toxicants has been linked in ASD^{6, 7}. Exposure to air pollutants has been examined in a large number of studies of ASD, with work suggesting increased risk of ASD with prenatal exposure to particulate matter (especially particulate matter less than 2.5 microns in diameter (PM_2.5)), nitrogen dioxide (NO₂), and ozone specifically^8–11. Prenatal exposure to endocrine disrupting chemicals (EDCs), including polychlorinated biphenyls (PCBs), bisphenol A (BPA), polybrominated diphenyl ethers (PBDEs), and phthalates, have also been linked to impaired neurodevelopment and ASD^{12 13}. Certain pesticides have also been associated with ASD¹⁴, and dietary intake is a major route of exposure¹⁵. Finally, heavy metals like mercury and lead, as well as metal mixtures¹⁶, have also been associated with developmental delays. Various broad mechanisms have been hypothesized for how these environmental factors impact neurodevelopment, including increased oxidative stress and inflammation¹⁷, disrupted hormonal signaling¹⁸, and altered DNA methylation¹⁹.

Prenatal diet may be an important modifier of the link between environmental risk factors and risk for ASD or impaired neurodevelopment, given known roles in the same critical pathways affected by above-noted environmental risk factors^20–22 (see Figure 1). Modification via dietary factors may occur via protection against or mitigation of the adverse effects of other environmental exposures as a result of adequate or optimal intake, or via synergistic effects due to inadequate intake or presence of adverse components of the diet. Several prenatal dietary factors have been independently associated with ASD and neurodevelopment^{7, 20, 23, 24} as reviewed elsewhere²⁰, most notably folate, with evidence for an inverse association with neurodevelopmental delays. However, concerns have also been raised related to excess nutrient intake via supplement use, which raise additional questions about how associations may differ across nutrient levels²⁵. In addition, certain foods and food packaging are also a known direct source of exposure to metals and chemicals like phthalates, BPA, and pesticides^{15, 26}, and thus there is also the potential for inherent combined effects within foods containing both potentially beneficial nutrients, as well as potentially harmful chemicals. Given inadequate intake of several neuroprotective nutrients, such as iron, folate, and choline, is common among pregnant women in the United States²⁷, as is suboptimal intake of other foods and nutrients influencing key pathways, clarifying potential modifying effects of these factors is of high public health relevance. Furthermore, because individuals may be able to change their dietary habits more easily than their broad environmental exposures (which often require policy to enact meaningful changes in exposure), the role of diet as a potential modifier is a promising public health target.

Figure 1.

Key pathways that may link environmental exposures, nutrients, and neurodevelopmental outcomes.

The aim of this narrative review is to summarize studies examining prenatal diet as a modifier of environmental exposures- specifically, chemicals and toxicants- associated with ASD, in order to guide future work in this area. Given limited prior work on this topic, we also secondarily considered research focused on related neurodevelopmental outcomes.

Methods

A search on PubMed, conducted in the Fall of 2021, was performed to identify research articles examining dietary exposure during pregnancy as a modifier of the risk for ASD or neurodevelopmental outcomes. Developmental outcomes included behavioral and physiological assessments and could be measured at any postnatal age. The primary risk factor must have been prenatal exposure to an environmental chemical or toxicant. Search terms included: pregnan*, maternal, prenatal; environmental, pollutant, endocrine disrupting chemical, pesticide, heavy metal, toxicant; diet*, nutrition, nutrient, food, intake; attenuate*, synergy, moderat*, interaction; autis*, neurodevelopment, developmental, brain development. A complete list of search terms is provided in the Appendix. Only human studies were included. Finally, all studies must have had a comparison group; case studies were not included. Additional papers were identified via a review of reference lists from key articles. Initially, the search was restricted publications to within the past 5 years. However, as only two additional studies were published outside of that range, all studies meeting criteria and published in English before October 27, 2021 were considered.

Current Findings

Autism Spectrum Disorder

We identified 5 studies (all conducted in North America) investigating the role of prenatal diet as a modifier of the relationship between environmental factors and risk for ASD (Table 1). Three studies examined folic acid^28–30 and two examined prenatal supplements^{31, 32}. Environmental exposures in these studies included pesticides²⁸, air pollutants²⁹, phthalates^{30, 31}, and per- and polyfluoroalkyl substances (PFASs)³². Four of the 5 studies used data from California, USA^{28, 29, 31, 32}. Most studies relied on gold standard research measures and/or clinical assessment of diagnosis at 2–5 years of age (median: 3 years). Covariates considered in adjusted models varied across studies (Table 1); all included final adjustment for some proxy for income level; most included child birth year or sex; 2 included adjustment for two other nutrients; 2 for sociodemographic factors; and 2 for pre-pregnancy BMI.

Table 1.

Studies investigating the role of prenatal diet as a modifier of the relationship between environmental factors and ASD.

Reference	Study design and location	Sample size	Measure of ASD or autistic traits	Prenatal environmental exposure	Prenatal dietary exposure	Covariates	Findings
Schmidt et al, 2017	CHARGE study: case control study of children age 2–5y in California, USA	516 mother-child pairs (296 ASD, 220 typically developing controls)	ASD diagnosis: Autism Diagnostic Interview, Revised (ADI-R) and Autism Diagnostic Observation Schedule-Generic (ADOS-G)	Pesticide exposure: household (none vs any: indoor or outdoor sprays/foggers, pet flea & tick products, any); agricultural (none vs any: chlorpyrifos, organophosphates, pyrethroids, carbamates); and occupational (none vs any).	Reported maternal intake of FA (<800 μg vs ≥800 μg), vitamin B12 (<8 μg vs ≥ 8 μg), and vitamin B6 (<2.83 mg vs ≥ 2.83 mg) from foods or supplements in the first month of pregnancy	Child’s birth year; maternal vitamin B6 and vitamin D intake during 1st month of pregnancy; homeownership	Maternal FA intake attenuated association between ASD and household and agricultural pesticide exposure. Among women who consumed <800 μg FA, the OR (95% CI) for ASD related to any pesticide exposure was 2.1 (1.1, 4.1); among women who consumed ≥800 μg FA, the OR decreased to 1.7 (1.0, 2.9).
Goodrich et al, 2018	CHARGE study	760 mother-child pairs (434 ASD, 326 typically developing controls)	ASD diagnosis: ADI-R and ADOS-G	Air pollution exposure (near roadway air pollution, PM2.5, PM10, ozone, NO2) using geo-coded addresses, traffic and atmospheric data, and Air Quality System data	Reported maternal intake of FA (≤800 μg vs >800 μg) from foods or supplements in the first month of pregnancy	Child’s birth year; maternal vitamin A and zinc intake during 1st month of pregnancy; self-reported financial hardship	Maternal FA intake attenuated association between ASD and air pollutant exposure. p for interaction significant for NO2 (OR among ≤800 μg FA: 1.53 (0.91, 2.56); OR among >800 μg FA: 0.74 (0.46, 1.19); p: 0.039).
Oulhote et al, 2020	MIREC study: longitudinal cohort study of pregnant women across Canada	556 mother-child pairs	Autistic traits measured at 3y using the Social Responsiveness Scale-2 (SRS-2)	Phthalate exposure: phthalate monoester metabolites (MBP, MBzP, MEP, MCPP, sum of DEHP) measured in 1st trimester urine samples	Reported maternal intake of FA (<400 μg vs ≥400 μg) from supplements in the first trimester	Child sex; maternal age, education, parity, marital status, race/ethnicity; year of enrollment; study city; household income	Maternal FA intake attenuated association between ASD and phthalate exposure during pregnancy. p for interaction significant for all phthalates except MEP (p=0.17).
Shin et al, 2018	MARBLES study: longitudinal cohort study of pregnant women at high risk for delivering a child with ASD in California, USA	186 pregnant women and their 201 children (46 ASD, 55 non-typically developing, 100 typically developing)	ASD diagnosis: ADOS	Phthalate exposure: weighted averages of 14 phthalate metabolites measured in 2nd and 3rd trimester urine samples	Reported maternal use of prenatal vitamins (yes vs no) in the first month of pregnancy	Child’s birth year; maternal pre-pregnancy BMI; homeownership	Among prenatal vitamin users, ASD risk inversely associated with 3 phthalate metabolites: MiBP (RRR [95% CI]: 0.44 [0.21, 0.88]), MCPP (0.41 [0.20, 0.83]), and MCOP (0.49 [0.27, 0.88]). No association between metabolites and ASD among non-users.
Shin et al, 2020	CHARGE study	453 mother-child pairs (239 ASD, 214 typically developing controls)	ASD diagnosis: ADI-R and ADOS-G	Per- and polyfluoroalkyl substance (PFAS) exposure: 9 PFASs modeled based on postnatal maternal serum samples	Reported maternal use of prenatal vitamins (yes vs no) in the 1 month before and after conception	Child sex, age, birth year, gestational age at delivery; maternal age at delivery, parity, race/ethnicity, birthplace, pre-pregnancy BMI; breastfeeding duration; homeownership; study center	Maternal prenatal vitamin use did not modify association between prenatal exposure to PFAs and ASD.

Abbreviations: EPA - Environmental Protection Agency; MiBP - mono-iso- butyl phthalate; MBP - mono-n-butyl phthalate; MBzP - mono-benzyl phthalate; MEP - mono-ethyl phthalate; MCOP - mono-carboxyisooctyl phthalate; MCPP - mono-3- carboxypropyl phthalate; DEHP - di-(2-ethylhexyl) phthalate; FA - Folic acid.

Each of the three studies examining maternal folic acid found attenuation of the effects of environmental exposures. Both Schmidt et al and Goodrich et al used data from the Childhood Autism Risks from Genetics and the Environment (CHARGE) study, a population-based case-control study in California that enrolled children with diagnosed ASD (age 2–5 years) and age and sex matched typically developing controls. Both used self-reported, retrospective data to estimate maternal folic acid intake from foods and supplements during the first month of pregnancy, and both dichotomized folic acid intake as above or below 800 μg. Schmidt et al examined associations with household, agriculture, or occupational pesticide use, while Goodrich et al investigated the role of air pollutants (NO₂, PM_2.5, PM₁₀, and ozone). Schmidt et al found that associations with ASD were highest among offspring of women with household and agricultural pesticide exposure and folic acid intake <800 μg/d, compared to those with no pesticide exposure or higher folic acid intake²⁸. Goodrich et al similarly found that offspring of women exposed to higher levels of air pollution along with folic acid intake <800 μg/d had the highest risk of ASD, compared to lower levels of air pollution or higher folic acid intake²⁹. Although this pattern was true for most pollutants, the interaction reached statistical significance only for NO₂. Oulhote et al reported similar results for potential modification by folic acid for phthalate exposure using data from the Maternal-Infant Research on Environmental Chemicals (MIREC) Study, a longitudinal cohort study of pregnancy in Canada. The authors found a positive association between maternal urinary phthalate metabolites measured in first trimester urine samples and children’s Social Responsiveness Scale (SRS-2) scores, a measure of autism-related traits collected at 36–48 months in this study. This association was attenuated among women who consumed >400 μg/d of folic acid supplements in the first trimester according to prospective report³⁰. Of note, although this study used a lower folic acid cutoff compared to the others, it only included intake from supplements.

Two studies examined prenatal supplements as modifiers of the link between environmental exposures and ASD. Shin et al, 2018, used data from the Markers of Autism Risk in Babies – Learning Early Signs (MARBLES) study, a cohort study that enrolled pregnant women at increased familial likelihood for delivering a child with ASD³¹. Urinary phthalate metabolites were measured during the 2^nd and 3^rd trimesters and ASD diagnosis was determined on clinical best estimate at 3 years. Reported prenatal vitamin use during the first month of pregnancy (yes/no) was tested as a modifier of associations between 14 phthalate metabolites and ASD, with 4 highly correlated metabolites of di(2-ethylhexyl) phthalate (DEHP) summed together. Among prenatal supplement users only, 3 metabolites (mono-isobutyl phthalate [MiBP], mono(3-carboxypropyl) phthalate [MCPP], and mono-carboxyisooctyl phthalate [MCOP]) were significantly inversely associated with ASD, although the interaction between prenatal supplement use and phthalates was statistically significant only for MCPP. In Shin et al, 2020³², data from the CHARGE study was used, with prenatal exposure to PFAS modeled based on maternal serum PFAS at 2–5 years postpartum. Retrospectively reported prenatal supplement use during the months before and after conception (yes/no) was tested as a modifier of the association between 7 PFAS, considered both as natural-log transformed and untransformed variables, and ASD. PFASs were tested as independent predictors and as multiple predictors in a single model. No significant interactions between prenatal supplement use and any of the PFASs were observed at the p<0.05 level³², nor were stratified odds ratios meaningfully different between prenatal supplement users and non-users. Together, in contrast to findings for folic acid, these studies provide limited evidence for the role of prenatal supplement use as a modifier of associations between these environmental exposures and ASD.

Other neurodevelopmental outcomes

We identified 7 studies (conducted in North America, Africa, South Korea, and Europe) that examined the role of prenatal diet as a modifier of associations between environmental factors and other neurodevelopmental outcomes (Table 2). Two studies examined folic acid^{33, 34}, 2 examined fruit and vegetable intake^{34, 35}, and 3 examined fatty acids^36–38. Environmental exposures included heavy metals^{33, 38, 39}, EDCs^{36, 37, 39}, and air pollutants^{34, 35}. Though age at assessment and outcome measures used were more variable than for the studies focused on ASD, five studies measured neurodevelopment within the first 3 postnatal years using the Bayley Scales of Infant Development (BSID)^{33, 35, 36, 38, 39}. As for the ASD studies, covariate adjustment varied across studies (Table 2), with few considering the role of other dietary factors, and variability in sociodemographic factors included.

Table 2.

Studies investigating the role of prenatal diet as a modifier of the relationship between environmental factors and neurodevelopment

Reference	Study design and location	Sample size	Measure of neurodevelopment	Prenatal environmental exposure	Prenatal dietary exposure	Covariates	Findings
Liu et al, 2021	APrON study: longitudinal cohort study of pregnant women in Alberta, Canada	394 mother-child pairs	Bayley Scales of Infant Development-III (BSID-III) at 2y	Bisphenol (BPA and BPS) and cadmium exposure: measured in 2nd trimester maternal urine and red blood cells, respectively	Maternal red blood cell Se and Cu levels during the 2nd trimester	Child sex, ethnicity; household income	Among those with <median Se, BPS negatively associated with BSID-III motor scores. Among those with >median Se, there was a positive association.
Kim et al, 2020	MOCEH study: longitudinal cohort study of pregnant women in Seoul, Cheonan, and Ulsan, South Korea	451 mother-child pairs, categorized as average/slow vs rapid growers	BSID-II at 6, 12, 24, and 36 months	Mercury exposure: measured in maternal blood during late pregnancy (28–42 weeks gestation) and in cord blood samples	Serum folate (above vs below median), measured in early pregnancy (12–20 weeks gestation)	Child sex, gestational age; maternal age, education, pre-pregnancy BMI, drinking history, occupation; duration of breast feeding; household income	Maternal folate modified association between mercury and neurodevelopment for rapid growers. Among those with <median folate, maternal and cord blood mercury was negatively associated with BSID-II scores. No association among those with >median serum folate.
Loftus et al, 2019	CANDLES study: longitudinal cohort study of pregnant women in Shelby County, Tennessee, USA	1005 mother-child pairs	Stanford Binet Intelligence Scales, edition 5 (SB-5) at 4–6y	Air pollution exposure (PM10, NO2) using geo-coded addresses and Air Quality System data	Reported maternal intake of fruits and vegetables (below vs above median, measured by food frequency questionnaire) and maternal plasma folate quartile during the 2nd trimester	Child sex, age, date of birth, birth order; maternal age, race, education, IQ; prenatal depression, smoking; breastfeeding duration; child sleep; insurance status; Childhood Opportunity Index	PM10 exposure inversely associated with Full Scale IQ for those in lowest quartile of maternal plasma folate (−6.8 points [95% CI: 1.4, 12.3] per 5 unit increase in PM10). No association among those in the highest quartiles.
Merida-Ortega et al, 2019	Longitudinal cohort study of pregnant women in Mexico	276 mother-child pairs	BSID-II at 1–30 mos	DDT exposure: measured during 1st trimester as DDE in maternal serum	Dietary intake of fatty acids (ALA, EPA, DPA, DHA, LA, ARA; above vs below median) during 1st and 3rd trimesters, as measured by a food frequency questionnaire	Child sex, age; maternal IQ, energy intake; duration of breastfeeding; Home Observation for Measurement of the Environment scale	No significant interactions of any fatty acid and maternal serum DDE on BSID-II scores.
Ogaz-Gonzalez et al, 2018	Longitudinal cohort study of pregnant women in Mexico	142 mother-child pairs	McCarthy Scales of Children’s Ability at 42–60 months	DDT exposure: measured during 3rd trimester as DDE in maternal serum	Dietary intake of fatty acids (ALA, EPA, DPA, DHA, LA, ARA; above vs below median) during 1st and 3rd trimesters, as measured by a food frequency questionnaire	Child sex, age; maternal IQ, energy intake; duration of breastfeeding; HOME scale	DHA intake in 1st trimester and ARA intake in 3rd trimester were modifiers of association of DDE with motor and memory development, respectively. Among women with <median intake for DHA and ARA, serum DDE inversely associated with neurodevelopment.
Strain et al, 2015	Seychelles Child Development Study: longitudinal cohort study of pregnant women in the Republic of Seychelles	1265 mother-child pairs	BSID-II, Infant Behavior Record-Revised, MacArthur-Bates Communicative Development Inventories (CDI) at 20 months	Mercury exposure: measured in maternal hair samples at delivery	Serum fatty acids (ALA, EPA, DHA, LA, ARA) collected at 28 weeks gestation, categorized into tertiles of total n3, n6, and their ratio	Child sex, age; maternal age; socioeconomic status; number of parents living with the child	Total n3s and n6:n3 ratio modified the association between mercury and psychomotor development (p for interaction <0.05). Among those in the highest tertile of n3s, mercury was positively associated with motor development. Among those in the highest tertile of the n6:n3 ratio, mercury was negatively associated with motor development. No associations among those in lower tertiles.
Guxens et al, 2012	INMA study: longitudinal cohort study of pregnant women in Spain	1889 mother-child pairs	BSID at 14 months	Air pollution exposure (NO2, benzene) using passive samplers and land use regression models	Reported maternal intake of fruits and vegetables (low tertile vs medium/high tertile, as measured by FFQ) during the 1st trimester and maternal plasma vitamin D (season-specific tertiles)	Child’s sex, age, number of siblings; maternal age, education, height, pre-pregnancy BMI, prenatal alcohol use, consumption of fish in 1st trimester; use of gas stove; and evaluating psychologist.	Maternal intake of fruits and vegetables modified the association between NO2 and benzene and mental development (p for interaction = 0.073 and 0.047, respectively). Among those with low intake, air pollutants negatively associated with mental development. Results similar but not statistically significant for maternal vitamin D status.

Abbreviations: Se- selenium; Cu- copper; FFQ- Food Frequency Questionnaire; HOME- Home Observation for Measurement of the Environment.

A study by Liu et al in Canada examined several prenatal nutrients as potential modifiers of associations between bisphenols A and S (BPA and BPS) and neurodevelopment using data from the Alberta Pregnancy Outcomes and Nutrition (APrON) cohort study³⁹. The Bayley III scales were used to assess child neurodevelopment at age 2 years. Maternal levels of BPA and BPS were measured in 2^nd trimester urine samples, and arsenic, cadmium, lead, mercury, and several nutrients (serum ferritin, plasma vitamin B12, red blood cell folate, dietary choline, serum phospholipid docosahexaenoic acid (DHA) and arachidonic acid, and plasma copper, zinc, manganese, and selenium) were measured in 2^nd trimester blood samples. A statistically significant interaction between BPS and selenium in association with motor development was identified (p for interaction 0.026). Among offspring of women with selenium levels below the median, higher BPS was associated with poorer motor development; among those with selenium above the median, higher BPS was associated with improved motor development³⁹. Modification by copper and cadmium was also examined, but not identified; other nutrients and metals were not assessed as modifiers.

Two studies found that maternal serum or plasma folate significantly modified associations between environmental factors and neurodevelopmental outcomes. Kim et al used data from a longitudinal study in South Korea, Mothers and Children’s Environmental Health (MOCEH), to examine serum folate as a modifier of mercury exposure³³. Maternal serum folate was measured in early pregnancy (12–20 weeks gestation), and blood mercury in late pregnancy (28–42 weeks gestation) and in cord blood. Children’s neurodevelopment was assessed with the BSID-II at 6–36 months. Additionally, children were categorized as average/slow growers and rapid growers based on a change in weight z-score >0.67 from birth to 36 months. Among rapid-growing offspring of women with serum folate below the median, blood mercury was negatively associated with neurodevelopment, whereas there was no association among women with higher serum folate. Although this pattern was observed in stratified models of maternal and cord blood mercury with mental and psychomotor development scores, the interaction reached statistical significance only for folate and cord blood mercury in association with psychomotor development³³. In Loftus et al, maternal plasma folate during the 2^nd trimester was tested as a modifier of the relationship between prenatal air pollutant exposure and child IQ³⁴. Using data from the Conditions Affecting Neurocognitive Development and Learning in Early Childhood (CANDLE) cohort, they found a borderline significant interaction (p=0.07) as well as a significant negative association between PM₁₀ and Full Scale IQ only among those in the lowest quartile of plasma folate.

Two studies (one of which also examined folate, as noted above) found conflicting results when examining maternal intake of fruits and vegetables as a modifier of environmental exposures. In Loftus et al, maternal intake of fruits and vegetables (above or below the median) in early pregnancy was not a significant modifier of air pollutants (PM₁₀ or NO₂) in association with intelligence scores as assessed by the Stanford Binet scales³⁴. However, Guxens et al reported a significant interaction between fruit and vegetable intake and air pollution exposure using data from the Spanish Infancia y Medio Ambiente (INMA) Project³⁵. Among offspring of women in the lowest tertile of fruit and vegetable intake during pregnancy, both NO₂ and benzene were inversely associated with the BSID mental development score, while no association was observed among offspring of women in the mid or high tertiles.

Three studies examined the role of maternal long chain polyunsaturated fatty acids (PUFAs) as modifiers of environmental exposures. PUFAs include omega 3 fatty acids (α-linoleic acid, eicosapentanoic acid, docosapentanoic acid, and DHA) and the omega 6 fatty acids (linoleic acid and arachidonic acid). Strain et al used data from the longitudinal Seychelles Child Development Study to examine effect modification related to mercury exposure³⁸. Total omega 3, total omega 6, and the ratio of total omega 6:omega 3 fatty acids measured in mid-pregnancy maternal serum were tested as modifiers of the relationship between prenatal mercury exposure and neurodevelopment at 20 months. Although there was no overall association between mercury exposure and child cognitive scores, PUFAs were identified as significant modifiers. In the highest tertile of omega 3 levels, mercury was positively associated with psychomotor development, whereas in the highest tertile of the omega 6:omega 3 ratio, mercury was negatively associated with psychomotor development. There were no associations in the lower tertiles³⁸. Two studies examined maternal fatty acid intake as a modifier of prenatal exposure to the pesticide 1,1,1- trichloro-2,2-bis(p-chlorophenyl)ethane (DDT) and child neurodevelopment in the same longitudinal cohort study of Mexican women^{36, 37}. Fatty acid intake and DDT exposure (as its main metabolite, 1,1-dichloro-2,2-bis(p-chlorophenyl) ethylene (DDE)) were each measured during the 1^st and 3^rd trimesters. Child development was measured using the BSID-II at 1–30 months and the McCarthy Scales of Children’s Ability at 42–60 months. Using DDE levels during the 1^st trimester and BSID-II scores, Merida-Ortega et al did not find significant modification by any dietary fatty acid³⁶. However, using DDE levels during the 3^rd trimester and McCarthy scores, Ogaz-Gonzalez et al found evidence for modification by two PUFAs³⁷. Among those with 1^st trimester DHA intake below the median, DDE levels were inversely associated with motor development; the association was null for those with DHA intake above the median. A similar relationship was found between arachidonic acid intake during the 3^rd trimester and memory development (p for interaction with DDE for both fatty acids <0.05)³⁷.

Discussion

In this review, we found evidence supporting a role for several prenatal dietary factors as modifiers of relationships between environmental chemicals and ASD and broader neurodevelopment. Folic acid and long chain omega-3 fatty acids, in particular, were found to attenuate associations with environmental exposures and ASD or neurodevelopmental scores in multiple studies. Less evidence was observed for a modifying role of other dietary factors examined, including prenatal supplements and maternal fruit and vegetable intake. Although the number of studies identified that focus on this topic is limited, available data suggest interactions between dietary factors and environmental exposures is a promising target for future work. To our knowledge, this is the first review focused on the joint effects of diet and environmental exposures in relationship to ASD and neurodevelopment, and we believe this highlights a need in the field for growing work in this area given the likelihood of convergence of effects in shared pathways.

Folic acid is a form of the essential nutrient folate that is used in fortified foods and supplements. In the US, women of reproductive age are recommended to consume 400 μg of folic acid per day to prevent birth defects⁴⁰; to meet this recommendation, folic acid is added to refined grains such as bread and cereal. Folic acid intake during the periconceptional period has been directly associated with decreased risk of ASD^{7, 20}; in this review, folic acid was also identified as a modifier of the relationship between several environmental exposures (including air pollutants, metals, pesticides and pthathates) and ASD and neurodevelopment in several studies. There are two main hypotheses for how folic acid may attenuate the effects of environmental exposures. First, folate’s role as a methyl donor may counteract the effects of environmental toxins on DNA methylation. Folate works with vitamin B12 to donate a methyl group to homocysteine, forming methionine and then S-adenosyl-methionine, a major methyl donor. In rodent models, perinatal BPA provision induces changes in DNA methylation patterns and behavioral deficits among offspring, and these changes are reversed with perinatal provision of a combination of methyl donors (folic acid, vitamin B12, choline, and betaine)⁴¹. Second, folate may act as an antioxidant to mitigate toxin-associated oxidative stress and inflammation. Folic acid may have direct antioxidant properties⁴². Additionally, folate’s involvement in the methionine cycle is linked to the creation of cysteine (a reaction which requires vitamin B6) and then glutathione, a major antioxidant in the body. In rodent models, gestational exposure to air pollutants promotes neuroinflammation and oxidative stress, which can be attenuated with prenatal provision of folic acid, vitamin B12, and vitamin B6⁴³. Direct evidence for the mechanisms by which folic acid attenuates the effects of environmental toxins is lacking in human studies.

In the identified studies, folic acid exposure was measured in two ways: maternal self-report of folic acid supplementation during the periconceptional period^28–30 and maternal plasma/serum folate in the second trimester^{33, 34}. Both measures were associated with reduced risk for ASD or impaired neurodevelopment according to environmental exposures. However, folic acid and folate are distinct compounds and are metabolized differently in the body. Some concern has been raised that high intake of folic acid from supplements throughout pregnancy may lead to accumulation of unmetabolized folic acid in the blood, and that this may be linked to increased likelihood of ASD⁴⁴; thus, future studies further addressing potential differences in dose-response across pregnancy may be warranted.

Compared to folic acid, the evidence for modification by prenatal supplement use or intake of fruits and vegetables was relatively weaker, though fewer studies addressed these factors. Prenatal supplements and fruits and vegetables are rich sources of folic acid and natural folate, respectively, and also contain other neuroprotective compounds. It is possible that the combination of compounds may alter the specific effects as a modifier, or that differences in social or demographic factors influencing intake of these, play into these findings. The contrasting findings observed across the two studies addressing fruit and vegetable intake as a modifier of air pollution effects on neurodevelopment specifically may also be related to differences in populations, measurement of air pollutants, definition of intake categories, or outcome measures (IQ at 4–6 years in Loftus et al vs BSID score at 14 months in Guxens et al). In addition, it is worth noting considerations regarding timing; periconceptional or first month of pregnancy supplement use may also be a marker for pregnancy intent, and thereby serve as a proxy for a host of health-related factors that could also influence neurodevelopment. While most studies of folic acid and supplement use have focused on this periconceptional period, the limited studies on fruit and vegetable intake have generally not done so.

Other dietary factors also showed limited evidence for modification of exposure associations with neurodevelopmental outcomes. PUFAs are important for cell membrane structure, cell signaling, and neuronal development⁴⁵ and have been suggested as protective factors in some studies for ASD and neurodevelopment^{23, 46}. In this review, certain PUFAs were identified as modifiers of the effects of environmental exposures, though results varied across the three studies that examined maternal fatty acids. Whereas Ogaz-Gonzalez et al and Strain et al reported significant moderation by fatty acids, the specific PUFAs and the direction of the associations differed; on the other hand, Merida-Ortega et al found no evidence for modification by any PUFA. Differences across these studies could relate to different environmental exposures and/or differences in PUFA intake across study populations (in comparing Strain et al to the others), or differences in outcome timing and assessment (in comparing the two Mexican cohort studies). In Ogaz-Gonzalez et al, intake of omega 3 fatty acids was relatively low (~30 mg/day of DHA during the 3^rd trimester)³⁷, far below the recommended intake of 200 mg/day during pregnancy⁴⁷. Conversely, on the island nation of the Seychelles, women consumed fish (a rich source of omega 3 fatty acids) at 8.5 meals per week on average³⁸. It is possible that various PUFAs are able to act as modifiers only at certain intake levels.

Mechanisms that may underly PUFAs as modifiers of environmental exposures may be related to their roles in gene expression, antioxidation and inflammation, and/or direct incorporation into brain cell membranes. PUFAs may regulate gene expression in the brain by interacting with transcription factors⁴⁸ or by influencing gene methylation. For example, prenatal provision of DHA influences methylation of genes related to infant growth and immune development^{49, 50}. PUFAs may also be metabolized into bioactive lipids which influence inflammation and oxidation; for example, provision of omega 3 fatty acids during pregnancy has been shown to reduce markers of oxidation in mothers and newborns⁵¹. Further, the balance of DHA and ARA in early life is thought to be important for optimal neurodevelopment⁵².

Finally, vitamin D and selenium were examined as potential modifiers in only one study each. Although vitamin D was not a significant modifier of air pollution associations in Guxens et al³⁵, maternal vitamin D deficiency was shown to alter DNA methylation across two generations in rodent models⁵³, and has been associated with increased ASD in several studies^{20, 54}. Vitamin D is also a known immune moderator⁵⁵, and should thus be further considered in future work for potential joint effects in this pathway in particular. Selenium is a trace element that is required by a range of proteins involved in antioxidation, thyroid function, and DNA synthesis. Liu et al found that maternal selenium levels modified the association between the bisphenol BPS and motor development³⁹. Given connections to gene methylation and antioxidation⁵⁶, further study of Se and related nutrients is also warranted.

Challenges, Opportunities, and Future Directions

Although promising, the field is currently limited by a fairly narrow scope of dietary factors considered and, particularly for studies focused on ASD, relatively geographically homogeneous samples. All of the reports identified were observational studies. In the case control studies, there may be concerns regarding recall bias, or measurement error in dietary factors and supplements reported several years after pregnancy. Studies also face concerns regarding potential residual confounding by health-related behaviors that can be challenging to capture, or by other dietary factors not considered but involved in similar pathways. As summarized in our Tables, varying degrees of adjustment for demographic and sociodemographic factors, often tied to both diet and environmental exposures, have been considered across studies and represent an area for improved attention in future work. All but three studies used data from North America, and most studies took place in high-income countries. Populations in lower income countries are at higher risk of developmental delays⁵⁷, in part related to poor maternal nutrition status and harmful environmental exposures^{58 59}, and therefore conducting this work in ethnically and geographically diverse study populations is needed to better understand joint effects. The role of gene-environment and gene-nutrient interactions⁶⁰, several of which have already been linked to autism risk^{61, 62}, in these joint effect studies also represents an area of need, though these will require large samples. Likewise, while the focus of this review was on potential modification of the effects of environmental exposures in the narrow sense, a broader scope of exposures, such as gestational diabetes⁶³ and other conditions that act on these common pathways, should also be considered for potential joint effects.

As noted, this field would benefit from the investigation of a wider variety of dietary factors, guided by advancements in our understanding of mechanisms. For example, maternal iodine status, with its connection to thyroid function, may be a likely candidate for effect modification of EDCs, which disrupt thyroid signaling pathways¹⁸. Choline, a methyl donor that can (as an alternative to folate) convert homocysteine to methionine via its metabolite betaine, should be examined for interactions with exposures demonstrating modification by folate. High doses of choline have been shown to reverse oxidative stress markers among chick embryos exposed to dibutyl phthalate⁶⁴. Maternal choline consumption has also been shown to improve behavioral deficits in mouse models of autism⁶⁵. Vitamin E may modify exposures in oxidative and inflammatory pathways; it is a fat soluble antioxidant, and has been shown to be a modifier of neurologic outcomes associated with the phthalate DEHP in adult mice⁶⁶. In addition, pre- and probiotics may influence the composition of the gut microbiota, which has been linked to neuroinflammation and development of ASD⁶⁷, and flavonoids can act as neuroprotective phytochemicals⁶⁸. These factors have not been examined as modifiers of environmental associations with ASD and neurodevelopment to our knowledge. Relatedly, certain nutrients may also act as mediators of associations; it has been suggested that the inflammation caused by environmental toxins leads to prenatal deficiencies of nutrients affected by the acute phase response (iron, zinc, copper), and that these deficiencies contribute to ASD risk^{24, 69}.

We identified no studies that investigated potential antagonistic or negative effects of the diet in association with these exposures and outcomes; thus, consideration of pro-inflammatory dietary factors, dietary patterns, and diet as a source of exposures, should be considered. Similarly, while not included in this review, investigation of single foods as sources of both protective and risk factors merits discussion. Maternal intake of fish, for example, serves as a source of both environmental toxins (mercury) and dietary protective factors (fatty acids, selenium, etc), and has been examined in studies of cognitive scores⁷⁰ and linked to ASD-related traits^{71, 72}. Fruits and vegetables are also sources of environmental toxins (ie, pesticides¹⁵) as well as beneficial nutrients. When canned or packaged, these nutritious foods may also expose consumers to BPA or phthalates. The study of whole foods may help clarify how these positive and negative factors relate to neurodevelopment and ASD when combined.

The studies reviewed here employed standard methods for identification of joint effects (interaction terms and stratified models). Use of mixture modeling approaches^{73, 74} may be useful in advancing the field in identification of novel individual associations and interactions in the face of many correlated exposures, while approaches designed to examine time window effects^{73, 75, 76} may improve detection for nutrients with more specific critical windows. A number of studies of combined effects of metals (often including lead, mercury, arsenic, cadmium, etc)^77–80 have implemented such techniques and may be useful models. Several studies have identified novel interactions using these methods for other exposures and outcomes^{79, 81}. One study found suggestive evidence for an interaction but not independent effects, between selenium and arsenic in relationship to autistic traits as measured by the SRS⁸². Thus, examination of combined effects of nutrients and foods using mixtures approaches may reveal associations and interactions with ASD-related outcomes, even where interrogation of individual factors has failed to do so.

Finally, given similarities in risk factors for ASD and other neurodevelopmental outcomes, studying a range of outcomes may help to more fully understand relationships between nutrition, environment, and neurodevelopment, and determine specificity vs generalizability of effects across neurodevelopmental diagnoses. In addition, incorporation of quantitative traits, may also help to inform on associations with severity of symptoms and add to understanding of the nature of relationships, particularly if quantitative traits that are strong markers of heritability can be used.

Conclusion

Although currently limited in number, the available studies suggest a role for prenatal dietary factors (especially folic acid and fatty acids) as protective modifiers of the link between environmental chemicals and ASD and broader neurodevelopment. Replication of findings in diverse contexts and with powerful study designs is required. Investigation of other nutrients and dietary patterns, guided by understanding of mechanisms¹⁰, will help advance the field. Alteration of maternal diet is a promising public health target for reducing risk of adverse effects related to environmental exposures.