A systematic review of maternal exposures and the risk of autism spectrum disorders in their infants

INTRODUCTION

It is currently estimated that approximately one in thirty-six children in the United States are diagnosed with autism spectrum disorder (ASD) (Maenner et al. 2023). The etiology of ASD is not clearly understood and is speculated to arise from some single-gene disorders (Jin et al. 2020) and various complex interactions of genetic and environmental factors which result in changes in biochemistry and physiology, and ultimately neurologic function. Some of these changes may be present at birth and influenced by the intrauterine life.

The intrauterine environment and maternal exposures are important in predisposing or protecting the developing fetus from multiple health conditions such as diabetes, obesity, neurologic disorders, and hypertension, to name a few (Simkova, Veleminsky, and Sram 2020; Gómez-Roig et al. 2021). Epigenetics may play a critical role in the impact of exposures on fetal outcomes (Nye et al. 2014; Traglia et al. 2017). The ability to understand the impact and potential associated exposures during pregnancy related to the development of ASD could be crucial in assisting with early diagnoses and providing resources for parents to begin therapies to optimize the child’s development and function. The full scope of the impact of maternal exposures on ASD in the developing offspring is poorly understood.

The objective of this study was to systematically review and chronicle current literature evaluating maternal exposures associated with ASD diagnosis in their offspring.

MATERIALS AND METHODS

We performed a systematic review of the literature, using Ovid MEDLINE databases and reference lists of retrieved articles. Our search included articles published from January 2006 – December 2021 based on the keywords “autistic disorder,” “autism,” “maternal exposure,” “biomarkers” and “pregnancy.” We included articles that were published in English that studied pregnant individuals and the risk of ASD development. We excluded duplicate articles, those that included rodent and invertebrate analyses and all reviews and hypotheses manuscripts. Two authors independently reviewed abstracts found in the initial literature search and obtained full-text manuscripts for abstracts that met inclusion criteria. The full-text manuscripts were then evaluated for inclusion. Any discrepancy of opinion was discussed and settled via a third independent reviewer.

From the included articles, we collected information regarding the study population, gestational period at which the exposure was tested, exposure of interest, methodology for measurement of the exposure and outcome, study results, and study conclusions. We organized the exposures post-hoc into the following categories: biomarkers, environmental exposures, medication exposures, genetic variability, and maternal illnesses and conditions.

Descriptive summaries were presented. Quantitative meta-analyses were not possible due to the nature of the reports and heterogeneity of the measures and outcomes. The project was exempt from IRB approval as not being human subjects research. PRISMA guidelines for systematic review reporting were followed. As we had begun our search and study evaluation before we knew about PROSPERO registration, this study was not eligible for inclusion in PROSPERO per their guidance.

RESULTS

Of 433 reports identified, 47 studies published between 2006 and 2021 fulfilled the inclusion criteria (Figure 1).

Figure 1.Summary of search results and selection of studies for review

This figure summarizes the search outcomes, exclusions, and inclusion

In total, the studies reported on 6,172,363 pregnancies and children ages one to eighteen years with and without ASD. The studies were conducted globally, including 30 in the United States (Traglia et al. 2017; Braun et al. 2014; Granillo et al. 2019; J. H. Kim et al. 2021; Lyall et al. 2017, 2018; McKean et al. 2015; Millenson et al. 2017; Philippat et al. 2018; Ritz et al. 2020; Shin et al. 2018; Geier, Kern, and Geier 2009; Goodrich et al. 2017; Jo et al. 2019; Kalkbrenner et al. 2015; McCanlies et al. 2019; McGuinn et al. 2020; Patti, Li, et al. 2021; Raz et al. 2015; Schmidt et al. 2017; Singer et al. 2016; Vecchione et al. 2020; von Ehrenstein et al. 2019, 2020; Walton and Monte 2015; Windham et al. 2006; Hamad et al. 2019; Hollowood-Jones et al. 2020; Zhu et al. 2019) as well three in Sweden (Arora et al. 2017; Gong et al. 2017; Sujan et al. 2017), two studies in Canada (Bernardo et al. 2019; Oulhote et al. 2020), two in China (Gao, Xi, Wu, et al. 2015; Gao et al. 2016), two in Denmark (Liew et al. 2015; Singer et al. 2017), two in the Netherlands (Steenweg-de Graaff et al. 2016; van den Dries et al. 2019), two in Norway (Skogheim et al. 2021; Hornig et al. 2017), two in Finland (Malm et al. 2016; Kong et al. 2020), one in Taiwan (Chen et al. 2020) and one in Israel (Rotem et al. 2020). Of the 47 studies, 34 (72.3%) used validated tools for diagnosis of ASD including the Social Responsiveness Scale (SRS) (Maenner et al. 2023; Braun et al. 2014; Goodrich et al. 2017; McGuinn et al. 2020; Hollowood-Jones et al. 2020; Zhu et al. 2019; Oulhote et al. 2020; Gao, Xi, Wu, et al. 2015), Autism Diagnostic Observation Schedule (ADOS) (Jin et al. 2020; Traglia et al. 2017; Granillo et al. 2019; Lyall et al. 2017; Millenson et al. 2017; Shin et al. 2018; Geier, Kern, and Geier 2009; Kalkbrenner et al. 2015; McCanlies et al. 2019; von Ehrenstein et al. 2020; Walton and Monte 2015), Autism Diagnosis Interview-Revised (ADI-R) (Traglia et al. 2017; Granillo et al. 2019; Millenson et al. 2017; Shin et al. 2018; Geier, Kern, and Geier 2009; Jo et al. 2019; Kalkbrenner et al. 2015; McCanlies et al. 2019; von Ehrenstein et al. 2019; Walton and Monte 2015), Childhood Autism Rating Scale (CARS), and the Diagnostic and Statistical Manual criteria for ASD diagnosis (DSM-IV/V) (Jin et al. 2020; Simkova, Veleminsky, and Sram 2020; Gómez-Roig et al. 2021; Nye et al. 2014; J. H. Kim et al. 2021; Ritz et al. 2020; Patti, Li, et al. 2021; Raz et al. 2015; Singer et al. 2016; von Ehrenstein et al. 2020; Walton and Monte 2015; Windham et al. 2006; Gong et al. 2017; Liew et al. 2015; Skogheim et al. 2021; Volk et al. 2014). The tool for diagnosis of ASD in the remaining 13 studies was either unclear or not specified.

Detailed study characteristics and results are summarized in Table 1, organized by exposure categories and alphabetically by study author. This table also details the gestational age at exposure/measurement and analyses method of each study.

The concise summary of exposures and findings of associations with increased and decreased rates of ASD are presented in Table 2, organized again by exposure categories.

Biomarkers

Endocrine-disrupting chemicals

One study (Braun et al. 2014) analyzed the effects of endocrine-disrupting chemicals (EDCs) on 175 pregnant individuals, 22 of which had children diagnosed with ASD utilizing semi-Bayesian (β) regression models. Evidence from this study suggests an increased risk of developing ASD when exposed to trans­-nonachlor (β 4.1, 95% Confidence Interval [CI]: 0.8, 7.3). A significant decreased risk association was found in children born to individuals with detectable concentrations of β-hexachlorocyclohexane (β -3.3, 95% CI: -6.1, -0.5) or PBDE-85 (β -3.2, 95% CI: -5.9, -0.5).

Maternal metabolites

One study (Ritz et al. 2020) analyzed the risk associations between specific maternal metabolites in the serum of 114 pregnant mothers, 52 of whom had children diagnosed with ASD. The study found decreased risk associations between the metabolite quinoline and ASD. The study utilized partial least square discriminant analysis and found increased risk associations between the metabolites ornithine, 1-methylhistidine, methyl jasmonate, benzoate, nonanioic acid and 10-hydroxydecanoate with ASD diagnosis (statistical quantification not specified). The study also found significant risk associations with seven enriched metabolic pathways utilizing Mummichog pathway analysis, including: glycosphingolipid biosynthesis (p <0.001), phosphatidylinositol phosphate metabolism (p < 0.012), bile acid biosynthesis (p < 0.017), N-Glycan biosynthesis (p < 0.022), glycosphingolipid metabolism (p <0.036), pyrimidine metabolism (p <0.014), and C21-steroid hormone biosynthesis and metabolism (p <0.040).

Methylmercury

One study (McKean et al. 2015) analyzed 257 pregnant individuals exposed to methylmercury during pregnancy by analyzing newborn mercury serum concentrations and creating a toxicokinetic model with information from maternal fish questionnaire. Of these individuals, 164 children were diagnosed with ASD. The article reports that a cumulative exposure to methylmercury does not increase the risk for ASD diagnosis.

Organochlorine pesticides

One study (Lyall et al. 2017) analyzed 1,144 pregnant individuals exposed to organochlorine pesticides during pregnancy of which 545 offspring developed ASD. The study found no increased risk between organochlorine pesticides exposure and ASD.

Organohalogens

One study (Traglia et al. 2017) evaluated the exposure of various organohalogens in 366 children with an ASD diagnosis from a group of 735 total pregnant individuals evaluated. Three polybrominated compounds (PBDE-100, 153, and SUM PBDE) were associated with an increased risk of ASD in the ancestry-adjusted analysis, P < 0.05. Additionally, results supported the concept of genetic control of midgestational biomarkers for environmental exposures by nonoverlapping maternal and fetal genetic determinants, suggesting that the impact of environmental exposures may differ by genetic variation in mothers and/or fetus. Fetal genotypes expressed in placenta can influence maternal physiology and the transplacental transfer of organohalogens.

Organophosphates

Three studies (Millenson et al. 2017; Philippat et al. 2018; Ritz et al. 2020) analyzed exposure of 1,049 individuals to organophosphates during pregnancy of which 64 children were diagnosed with ASD. There was no association found between organophosphate exposure and development of ASD.

Perfluoroalkyls

Two studies analyzed the exposure of 2,165 pregnant individuals to perfluoroalkyls (PFAs) and 773 subsequent ASD diagnoses. Lyall et al. 2018 found a decreased risk association between PFOA (adjusted Odds Ratio [aOR] 0.62, 95% CI: 0.41, 0.93), PFOS (aOR 0.64, 95% CI: 0.43, 0.97) and ASD diagnosis. Liew et al. 2015 found no association between maternal PFA plasma level and autism (Liew et al. 2015). Other studies found increased risk associations between PFOAs and other PFAs (Oh et al. 2021).

Phthalates

Two studies (Oulhote et al. 2020; Shin et al. 2018) evaluated phthalate exposure in 711 pregnant individuals, 48 of whom had children who developed ASD. Oulhoute et al. found significant correlations between increased urinary MBP and MCPP concentrations and SRS score increases of 0.6 (95% CI: 0.1, 1.0) and 0.5 points (95% CI: 0.1, 0.8) respectively, indicating greater social behavior deficiencies. The authors found a decreased risk association in female children between two-fold increases in MEP concentrations and SRS score decreases of -0.4 points (95% CI: -0.7, 0.0). Finally, the authors found significant attenuation of the association between phthalate concentrations and higher SRS scores for DEHP, MCPP and MBP when folic acid was supplemented at ≥400µg/day (p<0.1) (Oulhote et al. 2020). Shin et al. found a significant decreased risk association between the MCPP phthalate in the second trimester (Relative Risk Reduction [RRR] 0.53, 95% CI: 0.32, 0.87) and the development of ASD (Shin et al. 2018). The authors also found a significant decreased risk association for ASD in mothers who took prenatal vitamins and were exposed to MiBP (RRR 0.44, 95% CI: 0.21, 0.88), MCPP (RRR 0.41, 95% CI: 0.20, 0.83), and MCOP (RRR 0.49, 95% CI: 0.27, 0.88) phthalates (Shin et al. 2018). Like Oulhoute et al., other studies also support the correlation between increased phthalates and elevated SRS scores (Patti, Newschaffer, et al. 2021).

Polychlorinated biphenyls

Three studies (Granillo et al. 2019; Lyall et al. 2017; Bernardo et al. 2019) analyzed a total of 2,444 individuals exposed to polychlorinated biphenyls (PCB) during pregnancy of which 574 children were diagnosed with ASD. Only Lyall et al. (2017) found a significant association between PCB exposure and ASD diagnosis (comparing highest versus lowest quartile of PCB 138/158 aOR 1.79, 95% CI: 1.10, 2.71 and PCB 153 aOR 1.82, 95% CI: 1.10, 3.02).

Polyunsaturated fatty acids

One study (Steenweg-de Graaff et al. 2016) reported the exposure of 4,624 pregnancies to polyunsaturated fatty acids and found an increased risk in ASD diagnosis when children were exposed to a lower maternal omega-3: omega-6 ratio (β = 0.009, 95% CI: -0.017, -0.001) as well as higher total omega-6 levels (β~total omega-6~ = 0.11, 95% CI: 0.002, 0.020).

Environmental Exposures

Agricultural pesticides

One study (von Ehrenstein et al. 2019) analyzed the ASD diagnosis of 2,961 children from 38,331 pregnant individuals exposed to agricultural pesticides and found an increased risk associated with pesticide exposure, including chlorpyrifos (Odds Ratio [OR] 1.15, 95% CI: 1.20, 1.29), diazinon (OR 1.14, 95% CI: 1.03, 1.26) and avermectin (OR 1.14, 95% CI: 1.03, 1.26) (von Ehrenstein et al. 2019).

Air conditioning

One study (Gao, Xi, Wu, et al. 2015) analyzed maternal air conditioning use during the pregnancy of 926 individuals, from whom 193 offspring were diagnosed with ASD. Use of air conditioning was found to have a decreased risk association with ASD development (OR 0.32, 95% CI: 0.22 – 0.46) (Gao, Xi, Wu, et al. 2015). The authors hypothesized this could be due to air conditioning correlating with family economic status, season of pregnancy and childbirth as air conditioners are primarily utilized in summer months, or indoor air pollution exposures decreased by air conditioner use.

Asthmagens

Two studies by Singer et al. (2016, 2017) examined the effect of immune-triggering exposures (“asthmagens”) during pregnancy on ASD diagnosis in two separate study populations. A total of 12,332 children from 61,781 pregnant mothers exposed to asthmagens were diagnosed with ASD. The 2016 study (including 463 cases with ASD) found no association between asthmagen exposure and ASD diagnosis. The 2017 more comprehensive study (11,869 cases with ASD) found a decreased risk association between maternal asthmagen exposure and ASD diagnosis (aOR 0.92, 95% CI: 0.86, 0.99).

Caffeine

One study (Patti, Li, et al. 2021) analyzed the intake of caffeine during the two halves of pregnancy in mothers from two different cohorts; 120 mothers from the EARLI birth cohort of which 20 had children diagnosed with ASD and 269 mothers from the HOME cohort of which 43 had children diagnosed with ASD. The study found no significant risk associations between caffeine intake in the EARLI cohort (β per Interquartile Range [IQR] increase [57mg]: 2.0, 95% CI: -0.1, 4.0) or the HOME cohort (β per IQR increase [0.43mg]: 0.6, 95% CI: -0.5, 1.6) and ASD diagnosis by SRS scores. The study did find a significant increased risk association between average caffeine intake and ASD diagnosis in the pooled data analysis (β per IQR increase: 1.2, 95% CI: 0.3, 2.1).

Commercial Pesticides

One study (Schmidt et al. 2017) examined the influence of commercial pesticide exposure, like carbamate and pyrethroid, in 806 pregnancies of which 466 developed ASD. The study found no significant association between commercial pesticide exposure and ASD diagnosis (OR 2.0, 95% CI: 0.9, 4.2).

Heavy Metals

One study (Windham et al. 2006) analyzed the exposure of 941 pregnant individuals to heavy metals. A total of 284 of the children from the cohort were diagnosed with ASD. The authors found a significant association between ASD diagnosis and heavy metal concentrations in ambient air, with increased risk at higher concentrations (fourth quartile aOR 1.7, 95% CI: 1.0, 3.0 and third quartile aOR 1.95, 95% CI: 1.2, 3.1).

One study (Skogheim et al. 2021) analyzed exposure to arsenic in 1,431 mothers, 397 of which had children diagnosed with ASD. Significant increased risk associations were found between the second quartile of arsenic (OR 1.77, 95% CI: 1.26, 2.49) and the fourth quartiles of cadmium (OR 1.57, 95% CI: 1.07, 2.31) and manganese (OR 1.84, 95% CI: 1.30, 2.59) and ASD diagnosis. The study found significant decreased risk associations between ASD diagnosis and the second quartile of copper (OR 0.69, 95% CI: 0.49, 0.98), fourth quartiles of cesium (OR 0.63, 95% CI: 0.44, 0.91) and zinc (OR 0.63, 95% CI: 0.45, 0.90) and the second, third and fourth quartiles of mercury (OR 0.43-0.56, 95% CI 0.30, 0.80).

High levels of folic acid intake

Two studies (Goodrich et al. 2017; Schmidt et al. 2017) examined the effects of high folic acid intake in 1,412 pregnant individuals. Other studies suggest a decreased risk association between maternal vitamin intake and ASD diagnosis (DeVilbiss et al. 2017). Goodrich et al. examined the influence of high versus low folic acid intake when mothers were exposed to high NO₂ levels and found a significant difference in the two populations (NO₂ and low FA intake OR 1.53, 95% CI: 0.91, 2.56 versus NO₂ and high FA intake OR 0.74, 95% CI: 0.46, 1.19, p-interaction = 0.04) (Goodrich et al. 2017). Schmidt et al. examined the effects of high FA intake combined with exposure to agricultural pesticides and found no significant association with ASD diagnosis (Schmidt et al. 2017). Thus, it appears that high intake of folic acid may be able to attenuate ASD risk from other substances, although the literature reports mixed results.

Indoor pesticide exposure

One study analyzed indoor household pesticide exposure during the pregnancy of 806 individuals (Schmidt et al. 2017). The study compared all groups to a cohort of individuals with low folic acid intake during the first month of pregnancy and no known pesticide exposure. They concluded that the combination of low folic acid intake (≤800 µg) and indoor pesticide exposure led to an increased risk for ASD (aOR 2.5, 95% CI: 1.3, 4.7). There was also a lesser, but still significantly increased risk for ASD among high folic acid intake (≥800 µg) and indoor pesticide use (aOR 1.7, 95% CI: 1.1, 2.8).

Maternal Dental Amalgams

One study (Geier, Kern, and Geier 2009) analyzed the effect of maternal dental amalgams on the development of severe versus mild autism in 100 children. Amalgams leak mercury vapor which poses neurotoxic consequences. However, there is also a possibility of amalgam quantity/exposure correlating with socioeconomic status, dental hygiene, or nutritional intake. These other risks could be considered as controls in future studies. The study found no significant association between the mean number of amalgams and patients with ASD (mild) versus autism (severe) (Geier, Kern, and Geier 2009). However, the study results showed children of mothers with ≥6 amalgams had 3.2 times greater risk association of autism diagnosis compared to mothers with ≤5 (χ² =6.2, df=1, p=0.0127).

Maternal Fish Consumption

Three studies analyzed maternal fish consumption in 5,345 pregnancies. Gao et al. (2016) reported a significant decreased risk association between fish consumption, particularly grass carp fish, and ASD diagnosis (OR 0.279, 95% CI: 0.095, 0.799). Steenweg-de Gaaf et al. (2016) found no significant association between maternal fish intake and child autistic traits. Vecchione et al. (2020) found an increased risk association between higher fish intake during 21-36 weeks of pregnancy and ASD diagnosis (OR 5.60, 95% CI: 1.03, 17.86).

Maternal Fruit Consumption

One study (Gao et al. 2016) surveyed mothers regarding fruit consumption during pregnancy and found a significant decreased risk association between fruit consumption and ASD diagnosis (OR 0.413, 95% CI: 0.216, 0.804).

Methanol

One study (Walton and Monte 2015) analyzed the effects of maternal methanol consumption in 711 pregnancies and the development of ASD in 161 of the subsequent offspring. There was a significant increased risk association between children who developed ASD having a higher level of maternal methanol exposure during pregnancy (p<0.001).

Air Pollution and Particulate Matter

Two studies analyzed the effects of air pollution on the serum of 214 pregnant mothers, 116 of which had children diagnosed with ASD. Kim et al. (2021) found increased risk associations between the metabolites hypotaurine and urate. The study also found decreased risk associations between the metabolites phenylalanine and 3-hydroxybutanic acid and ASD diagnosis. Goodrich et al. (2017) compared near roadway air pollution of maternal addresses during pregnancy and found no significant association with ASD diagnosis.

Particulate matter (PM) is a type of air pollution composed of various solid particles and liquid components. We identified six studies that analyzed exposure to PM during pregnancy and subsequent ASD diagnosis. A total of 288,721 individuals were exposed to PM during pregnancy with 9,766 children diagnosed with ASD. Three studies found an increased risk of ASD when exposed to PM with diameters ≤2.5µm (PM_2.5). Jo et al. (2019) reported a significant increased risk association with exposure during the entire pregnancy (HR 1.17, 95% CI: 1.04, 1.33) and ASD. Raz et al. (2015) reported an increased risk when exposure occurred in the third trimester (OR 1.42 per interquartile range increase in PM_2.5, 95% CI: 1.09, 1.86) compared to exposure occurring in the first or second trimester. McGuinn et al. (2020) found significant risk associations between PM_2.5 exposure during the third trimester and ASD diagnosis (OR 1.3, 95% CI: 1.0, 1.6). Two other studies found no association between PM exposure and ASD development (Goodrich et al. 2017; Gong et al. 2017). A fifth study (Kalkbrenner et al. 2015) found a significant increased risk association with exposure during the third trimester (aOR 1.36, 95% CI: 1.13, 1.63), but a significant decreased risk association with exposure during the first trimester (aOR 0.86, 95% CI: 0.74, 0.99). Other studies showed an increased risk association between PM exposure in mothers with specific genotypes (Volk et al. 2014).

Nitrogen oxides

Three studies evaluated individuals exposed to nitrogen oxides in traffic-related air pollution during pregnancy. Gong et al. (2017) analyzed 23,373 individuals of whom 5,136 had children with ASD diagnoses. There was no significant association found between nitric oxide exposure and ASD development. Goodrich et al. (2017) analyzed the exposure of 606 individuals to nitrous oxide (N₂O) of which 346 developed ASD. Goodrich et al. also concluded there is a significantly increased risk between exposure to N₂O in individuals who have low folic acid intake levels (OR 1.53, 95% CI: 0.91, 2.56) vs high folic acid intake levels (OR 0.74, 95% CI: 0.46, 1.19) (p-interaction = 0.04). Jo et al. (2019) studied 246,420 children with 2,471 cases of ASD and did not find any significant associations between nitrogen dioxide and ASD.

Occupational Exposures

One study (Windham et al. 2013) examined 941 pregnant individuals with various occupational exposures, of which 284 children were diagnosed with ASD. Categories of occupational exposure were based on employment listed on birth certificates. This study reported that mothers of children with ASD were twice as likely to have an occupational exposure as those with children without ASD (aOR 2.3, 95% CI: 1.3, 4.2). The occupational exposure categories with elevated odds were among mothers with engine exhaust/combustion products (aOR 12.0, 95% CI 1.5, 108.6) and disinfectants (aOR 4.0; 95% CI 1.4, 12.0).

Outdoor Household Pesticides

One study (Schmidt et al. 2017) analyzed the exposure of 806 pregnant individuals to outdoor household pesticides concurrent with either low or high folic acid intake. They found no association between outdoor household pesticide exposure and ASD diagnosis in either group.

Ozone

Three studies (Goodrich et al. 2018; Jo et al. 2019; McGuinn et al. 2020] analyzed the exposure of 248,555 pregnant individuals to ozone (O₃) of which 3,491 had children who developed ASD. Goodrich et al. (2017) and Jo et al. (2019) found no relationship between O₃ and ASD diagnosis (aOR 1.14; 95% CI 0.71, 1.82 and adjusted Hazard Ratio [aHR] 1.10, 95% CI 0.95, 1.26, respectively). McGuinn et al. (2020) found a significant increased risk association between ozone exposure during T3 and ASD diagnosis per IQR_Width of 6.6 parts per billion (OR 1.2, 95% CI: 1.1, 1.4).

Solvents

Two studies evaluated the risk of exposure to solvents, such as aromatic and chlorinated solvents, on the risk of ASD diagnosis (McCanlies et al. 2019; Windham et al. 2006). Of the 1,892 pregnant individuals included in these studies, 821 had children were diagnosed with ASD. Windham et al. (2006) found a significant increased risk association between solvent exposure and ASD diagnosis in two different exposure quartiles (third quartile aOR 1.95, 95% CI: 1.2-3.1 and fourth quartile aOR 1.7, 95% CI: 1.0-3.0). McCanlies et al. (2019) also found a significant increased risk association between solvent exposure and ASD diagnosis (OR 1.5, 95% CI: 1.01, 2.23).

Medication Exposures

Antibiotics

One study (Hamad et al. 2019) examined 214,834 pregnant individuals exposed to antibiotics, defined as filling an antibiotic prescription during pregnancy. Of the children born to these individuals, 2,965 were later diagnosed with ASD. The authors reported an increased risk of ASD with antibiotic exposure (aHR 1.10, 95% CI 1.01, 1.19).

Antidepressants

Two studies examined the effects of antidepressant exposure in 1,645,383 pregnancies with 14,924 ASD diagnoses. Malm et al. (2016) found a significant increased risk association between antidepressant exposure, specifically SSRIs, with ASD diagnosis (aHR 1.40, 95% CI: 1.02, 1.92). Sujan et al. (2017) found no association between antidepressant exposure during the first trimester and ASD diagnosis (aHR 0.83, 95% CI: 0.62, 1.13).

Genetic Variability

Metabolites

One study (Hollowood-Jones et al. 2020) analyzed maternal and urine blood samples from 59 mother-child pairs of which 30 children had diagnoses of ASD. The various methylenetetrahydrofolate reductase variants were examined in addition to several biomarkers and metabolites. The study did not find a significant risk association between MTHFR mutation A1298C (p=0.15) or C677T (p=0.90) and ASD. They did find significant increased risk associations between lower vitamin B12 levels (Ratio 0.75, p<0.00), SAM/SAH ratios (Ratio 0.93, p=0.01), 4-vinylphenol sulfate (Ratio 0.31, p<0.00), NAD+ (Ratio 0.41, p=0.03), gamma-glutamylglycine (Ratio 0.27, p<0.00), cinnamoylglycine (Ratio 0.46, p=0.01), propionylglycine (Ratio 0.48, p=0.01), carnitine-conjugated metabolites (Ratios 0.63-0.87, p≤0.03) and ASD diagnoses. The authors also found significant increased risk associations between higher Glu-Cyst (Ratio 1.10, p=0.01), fCysteine (Ratio 1.06, p=0.01), fCystine (Ratio 1.07, p=0.02), dimethyl sulfone (Ratio 18.7, p=0.01) and ASD diagnoses.

Placental Differently Methylated Regions (DMRs)

One study analyzed placental DMRs of 41 pregnant individuals who had 20 children diagnosed with ASD (Zhu et al. 2019). Of 400 potential DMRs, methylation of specific genes differed by up to 10% between the cases with ASD and typically developing controls. Two placental DMRs (CYP2E1 and IRS2) were found to be potentially early epigenetic markers in the placenta for ASD. DMRs were also enriched around placental H3K4me3 and brain H3K4me3 which regulate gene functions (von Ehrenstein et al. 2020).

Paraoxonase [PON1] genotype

One study examined 224 mother-child pairs of which 6 children were diagnosed with ASD (Millenson et al. 2017). PON1 genotypes in cord blood were analyzed to determine their genetic impact on the association between second trimester urine organophosphate concentrations and ASD diagnoses. The PON genotype did not modify the association between prenatal urinary organophosphate concentrations and the social behaviors of children.

Maternal Illnesses and Conditions

Androgen-related conditions

One study (Rotem et al. 2020) included 437,222 mother-child pairs of which 4,022 children were diagnosed with ASD. The authors report significant increased risk associations between individuals diagnosed preconceptionally with PCOS and subsequent ASD diagnosis (OR 1.42, 95% CI: 1.24, 1.64). The authors found a significant increased risk association between ASD diagnosis and mothers with any condition of hyperandrogenemia (OR 1.41, 95% CI: 1.10, 1.79), hirsutism (OR 1.30, 95% CI: 1.10, 1.53), acne (OR 1.16, 95% CI: 1.08, 1.25) and infertility (OR 1.19, 95% CI: 1.10, 1.28).

Diabetes

Two studies included 893,519 mother-child pairs, of whom 4,817 children were diagnosed with ASD. Jo et al. (2019) reported an increased risk of ASD diagnosis in children of those mothers with pre-existing type 2 diabetes (HR 1.60, 95% CI 1.26, 2.16). Kong et al. (2020) reported a significant risk association between DM2 in severely obese individuals and subsequent diagnosis of ASD (HR 2.28, 95% CI: 1.18, 4.41). Kong et al. also reported a significant increased risk association between insulin-treated pregestational diabetes in obese individuals (HR 3.61, 95% CI: 1.61, 8.07) severely obese individuals (HR 5.93, 95% CI: 2.81, 12.52) and ASD diagnosis. Children of mothers with gestational diabetes mellitus at fewer than 24 weeks gestational age were found to be at an increased risk of ASD when exposed to higher levels of residential ozone O₃ throughout pregnancy (HR 1.50, 95% CI 1.08, 2.09). Kong et al. found a significant increased risk association between gestational diabetes in overweight individuals (HR 1.28, 95% CI: 1.07, 1.53) and obese individuals (HR 1.57, 95% CI: 1.26, 1.95) and subsequent ASD diagnosis. Other studies also support the findings of an increased risk between gestational diabetes and ASD diagnosis (Li et al. 2021).

Fever

One study (Hornig et al. 2017) examined 95,754 pregnant individuals with reports on fevers during pregnancy. 583 children from this cohort developed ASD. The study reports maternal fever in the second trimester, regardless of antipyretic use, was associated with an increased risk of ASD diagnosis (aOR 1.40, 95% CI 1.11, 1.77).

Maternal depression

Two studies analyzed the presence of maternal depression in 709,441 pregnant individuals of whom 4,699 had children that developed ASD (Gao, Xi, Wu, et al. 2015; Chen et al. 2020). Gao et al. and Chen et al. both found an increased risk in ASD diagnosis in children born to mothers with maternal depression during pregnancy (OR 4.82, 95% CI: 3.08, 7.63 and HR 1.58, 95% CI: 1.11, 2.25, respectively). Neither study examined whether the depression was treated. Chen et al. also found a significant risk association between maternal depression diagnosed before pregnancy (HR 2.01, 95% CI: 1.70, 2.37) and subsequent diagnoses of ASD.

Psychiatric disorder

One study (Malm et al. 2016) examined the diagnosis of a psychiatric disorder in 64,754 pregnant individuals of whom 307 children developed ASD. These diagnoses included depression, anxiety, bipolar disorder, and undefined psychosis. The study found an increased risk of ASD development in children of mothers with psychiatric disorder diagnoses who were not on antidepressants during pregnancy (aHR 1.59, 95% CI: 1.16, 2.18). There was also an increased risk of ASD in pregnancies exposed to SSRIs, as detailed in medication exposure section above.

COMMENT

This descriptive systematic review found that several pregnancy exposures and biomarkers may be associated with ASD in offspring. In this systematic review, we assessed 46 exposures across 47 articles. There were exposures with significant, but mostly small, increased or decreased risk associations across all our classification categories [biomarkers, environmental exposures, medication exposures, genetic variability, and maternal illnesses and conditions]. The risk and protective associations are concisely depicted in Table 2.

Our work complements the work of meta-analyses published in 2009 and 2017, an umbrella review published in 2019, and a systematic review of biomarkers published in 2019 (Gardener, Spiegelman, and Buka 2009; Frye et al. 2019; Wang et al. 2017). These studies investigate the prenatal, perinatal, and postnatal risk factors for child autism and report inconsistent conclusions regarding exposures associated with autism. Our results complement some findings but update and are different from some other findings from these studies. The 2017 meta-analyses collated data from 17 studies published between 2002 and 2016 (Wang et al. 2017). The 2019 umbrella reviewed studies published between 2012-2018 that evaluated biomarkers and environmental risk factors related to autism and completed credibility assessments on these studies to quantitatively evaluate their levels of significant (J. Y. Kim et al. 2019). The biomarker review categorized biomarkers into the natural historical timeline of ASD including prenatal, pre-symptomatic, diagnostic, subgrouping, and treatment. Consistent with our study findings, these studies reported that factors associated with increased risk of autism include maternal metabolic syndromes like diabetes and fetal distress, and a factor that was not associated with an increased risk of autism, the presence of a nuchal cord. Using different inclusion criteria from our study results, others included labor characteristics and reported an increased risk of autism with breech presentations, gestational age ≤36 weeks, cesarean delivery, fetal distress, and pre-pregnancy antidepressant use, which our study found to be not significantly related with autism (Gardener, Spiegelman, and Buka 2009; Frye et al. 2019; Wang et al. 2017; J. Y. Kim et al. 2019).

Throughout our analysis of environmental exposures related to ASD, there tended to be some contradictory results about the risk associated with several environmental exposures including particulate matter, nitrous oxide, solvents, and maternal fish consumption. There are several possible explanations for the varied results found in these factors. Some environmental exposures have unmeasurable variations in exposure levels and also are complicated by confounding exposures. These potential confounders may not have been measured in the same way across different studies. Assessing the association of maternal fish consumption and ASD, for example, is complicated by the varied levels of healthy nutrients and harmful contaminates present in each fish. Fish can contain varying quantities of protein, long-chain polyunsaturated fatty acids, iodine, selenium, and vitamin D (healthy nutrients in moderation), in addition to mercury, arsenide, dioxin, and polychlorinated biphenyls (potential toxins). Without controlling the specific quantity of each component in the fish, the precise risk associated with fish consumption is not able to be precisely evaluated. The association between fish consumption and ASD is further complicated by epigenetic variability in prenatal exposure to developmental toxins like mercury and arsenic (Liew et al. 2015).

Epigenetic modifications may modify in-utero response to environmental toxins and nutrients.

In our review, several studies examined potential genetic components of ASD risks. Millenson et al. (2017) evaluated cord blood PON genotypes related to organophosphate concentrations and ASD. They hypothesized that because enzyme polymorphisms effect metabolism of exposures, ASD risk associated with specific exposures metabolized by these enzymes could be impacted by various genotypes of enzyme functioning. Zhu et al. (2019) measured placental DMRs and found 400 differently methylated regions accounting for 596 genes between placentas of children with ASD and children who were typically developing. These findings suggest that epigenetics could have a significant role in ASD outcomes, and that a compounded effect of genetic components combined with exposures could be related to these outcomes. Two of the studies in our review reported evidence in support of this theory. Traglia et al. (2017) and Lyall et al. (2017, 2018) reported more significant exposure risk associations in the ancestry-adjusted analyses of their exposures than in the analyses that did not account for ancestry genetics. There is also some literature suggesting that maternal smoking status, BMI, and antibiotic use during pregnancy can also impact gene regulation in newborns. This suggests that not only inherent genetic variations, but also other acquired epigenetic variations, may play a role in the development of ASD (Nye et al. 2014; Kile et al. 2012). These findings support the need for further work that examines compounded risks related to ASD, especially accounting for epigenetic variations.

According to the developmental origins of disease hypothesis, fetal development, physiology, metabolism, and programming may be altered by maternal nutritional and endocrine status. This embryonic developmental period is theorized to predispose fetuses to diseases even through adult life, including cardiovascular, metabolic, neurologic, and endocrine-related diseases (Nye et al. 2014). The pathophysiology behind this process may be through altered epigenetic programming that occurs because of embryonic exposures. Genetic inheritance of gene regulatory mechanisms is theorized to contribute to differences in circulating levels and processing of embryonic exposures (Traglia et al. 2017). As genetic contributions predispose fetuses to varying levels of circulating exposures during the embryonic developmental period, fetuses may then experience differential risks for developmental outcomes related to these exposures.

To minimize confounding variables in analyzing compounded exposure and factors related to autism, biomarkers are an objectively useful indicator. As defined by the World Health Organization, a biomarker is a chemical, its metabolite, or the product of an interaction between a chemical and target that is measured in the human body (Katherine and Shea 2011). Because biomarkers can be objective measurements, they provide insight into physiologic and pathophysiological processes and pharmacologic responses. Their objective specificity also supports their potential use as predictive and prognostic indicators. In our review, we analyzed biomarkers related to environmental exposures and nutritional intake. In the literature, other suggested potential biomarkers related to ASD include maternal IgG antibodies, amino acid and acyl-carnitine metabolites, and protein levels (Frye et al. 2019).

Most of the significant associations we encountered were mild, with odds ratios just above or below 1.0. A few biomarkers had stronger associations, such as: endocrine-disrupting chemicals (ORs 4.1, -3.3, and -3.2), some phthalates (ORs 0.44, 0.41, and 0.49), polyunsaturated fats (βs 0.009 and 0.11), using air conditioning (OR 0.32), indoor pesticide exposure in low folate intake women (OR 2.5), having ≥6 dental amalgams (OR 3.2), maternal fish and fruit consumption (OR 0.28 and 0.41, respectively), having newborn complications (OR 4.7), occupational exposures (OR 2.3), and maternal depression (OR 4.82). These exposures in particular call for further investigation and replication. Many of them may have roots in social determinants of health and the environment. Further research examining both maternal and newborn biomarkers in relation to ASD could progress the effort to identify predictive or prognostic biomarkers.

Developing a predictive model for ASD could be useful clinically. As Hollowood-Jones et al. demonstrated, models with clusters of multiple metabolites could create more specificity and sensitivity in predicting ASD (Hollowood-Jones et al. 2020). Understanding that a newborn may be at high risk given pregnancy exposures and potentially measuring predictive markers in cord blood could allow for new parents to be educated regarding resources and therapies that could optimize their newborn’s development. Additionally, requests for optimizing prenatal supplements to improve newborn outcomes have been published (Adams et al. 2021, 2022).

Our study is limited in that the array of different exposure biomarkers did not allow for a quantitative synthesis. The studies had valid findings internally, but the diversity of measured exposures and covariates in models precluded any meta-analysis. We attempted in Table 2 to systematically synthesize the data reviewed in a graphic way to guide other researchers and those analyzing exposures. Utilizing these various exposures and measures could be combined using machine learning techniques to build a predictive algorithm that could be independently tested in a cohort. This could then be coupled to trials of early childhood interventions aimed at optimizing childhood development. Additionally, educating pregnant individuals and families on pregnancy exposures that could reduce the risk of ASD might inspire other healthy behavior changes.

Our study has several strengths that contribute to the current published literature surrounding this subject. First, it includes articles published through 2021, adding new studies which were not present in other reviews. Additionally, we attempted to be more comprehensive in our approach to predictive exposures, including a wide array of exposure categories to assimilate a comprehensive overview of risk factors. As such, we were able to characterize several factors that were not included in other reviews. Organizing these exposures by category could allow for researchers to identify and potentially build models for upcoming analyses and trials.

In conclusion, several pregnancy exposures and biomarkers are associated with the development of ASD in children. Continued work to build combined models to allow for determination of the strongest associated modifiable factors is important to unlocking mechanisms and potentially protective measures for families. As ASD can have lifelong impact on an individual and family, better understanding its risks and being able to provide resources for families is crucial to optimal health.

HIGHLIGHTS

• Ovid MEDLINE systematic review 2006 – 2021 of maternal factors associated with autism

• 47 studies with exposures grouped as biomarkers, exposures, and genetic variability

• Some exposures with persistently increased risk of autism; others with decreased risk

• Genetic mutations may play role in impact of exposures’ risk of autism

ACKNOWLEDGEMENTS

Thanks to Dr. Joanne K Daggy^a for assistance with interpretation of statistical analyses.

^aDepartment of Biostatistics and Health Data Science, Indiana University School of Medicine, 550 N. University Blvd., Suite 2301, Indianapolis, IN, USA 46202

CONFLICTS OF INTEREST

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: James Adams reports a relationship with HealthyBaby that includes: equity or stocks. The other authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

SOURCE OF FUNDING

This report did not receive any financial support and is not affiliated with any organization.

PAPER PRESENTATION INFORMATION

Preliminary results were reported via poster presentation at ACOG 2021 Annual Clinical and Scientific Meeting virtually April 30 – May 2, 2021.

CORRESPONDING AUTHOR

Shae N. Jansen; Department of Obstetrics and Gynecology, Indiana University School of Medicine, 550 N. University Blvd., Suite 2301, Indianapolis, IN, USA 46202; Cell phone – (317) 694-8210; email address: shjansen@iu.edu.