Background

Autism spectrum disorder (ASD) is a lifelong neurodevelopmental condition that impacts communication, social interaction, and behavior. Its prevalence has been on the rise globally. In a recent systematic review, it was estimated that 1 in 100 children worldwide are affected by ASD [1]. The Centers for Disease Control and Prevention (CDC), 1 in 36 children aged eight years will be diagnosed with ASD [2]. In the United States, the economic burden of ASD was estimated to be $11.5 billion in 2011 [3]. A complex interplay of biological and environmental factors has been linked to autism spectrum disorder.

Gene-environment interactions are critical factors in ASD development. Environmental pollutants, including toxic metals, are linked to epigenetic modifications and de novo mutations, potentially contributing to ASD onset [4]. These pollutants, particularly during gestation and postnatal periods, pose health risks and are associated with ASD [5]. Toxic heavy metals can disrupt enzymatic functions, interfere with cell signaling, and trigger oxidative stress, potentially leading to cell death pathways. Elevated levels of cadmium and mercury are frequently found in children with ASD [6]. However, more research is needed to fully understand how metal-induced neurotoxicity might play a role in ASD.

Recent systematic reviews and meta-analyses have evaluated the link between environmental pollutants and the development of ASD, but these reviews exhibit notable limitations. Among the 21 identified studies, 18 relied on four or fewer databases [5, 7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23], 11 imposed language restrictions (English, French, or Chinese)) [7, 8, 12, 13, 15, 18,19,20, 22,23,24], and six confined their searches to brief periods [9, 13, 15, 19, 20, 22]. This approach potentially overlooks some available evidence. Additionally, most reviews concentrated on air pollutants [7, 8, 10, 12, 13, 15, 18, 20, 23,24,25,26], with fewer addressing metals [5, 8, 14, 15, 22, 27], pesticides [9, 11, 16, 17, 27], polychlorinated biphenyls [19], or perfluoroalkyl substances [21]. The evidence primarily stems from cross-sectional, case–control, ecological, and cohort studies, and some reviews failed to stratify results by study type, blending cohort and case–control data [5, 8, 10, 12, 14]. Only one review exclusively considered cohort studies [13], However, it was limited to children under five and focused solely on-air pollution, not accounting for prolonged exposures or older children. This study aims to analyze the association between various environmental pollutants and ASD incidence through cohort studies, evaluating different pollutants and their effects on subgroups.

Methods

Protocol and registration

This study was performed according to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses – PRISMA [28] and registered in the International Prospective Register of Systematic Reviews (PROSPERO) under the number CRD42018093510.

Eligibility criteria

The PECOS strategy was defined as follows: Population: children and adolescents from 0 to 18 years old; Exposure: higher levels of environmental pollutants during the prenatal and postnatal period; Comparison: lower levels of environmental pollutants; Outcome: incidence of ASD, Studies: cohort studies.

The exclusion criteria were as follows: (a) studies that only included participants older than 19 years; (b) tobacco exposure; and (c) response letters, reviews, editorials, animal, and duplicate studies. Duplicate studies were considered when they had the same author, title and year. Additionally, when the studies were updating previous versions, the most current version with the largest sample size was chosen.

Environmental pollutant exposure included air pollution; PM; inorganic carbon compounds; lead; sulfur oxides; nitrogen oxides; soot; polychlorinated biphenyls (PCBs), inorganic chemicals; pesticides; volatile organic compounds (VOC); hydrocarbons; endocrine disruptors; plasticizers, and plastics.

The air pollutants were classified according to American international guidelines [29]. The groups were ground-level ozone, PM, carbon monoxide (CO), sulfur oxides (SOx), and nitrogen oxides (NOx). The toxic substances included in this study were classified according to the International Guidelines on Toxic Substances [30]. The categories of these pollutants were coal ash; dioxins, furans, PCBs; benzidines/aromatic amines; inorganic substances; nitrosamines/ethers/alcohols; pesticides; phenols/phenoxy acids; organophosphates and carbamates; phthalates; halogenated pesticides and related compounds; volatile organic compounds; radionuclides (radioactive materials) and warfare and terrorism agents.

Sources of information and search strategy

We searched the Cochrane Central Register of Controlled Trials (COCHRANE CENTRAL), MEDLINE (via PUBMED), Cumulative Index to Nursing and Allied Health Literature (CINAHL), Scientific Electronic Library Online (SciELO), Latin American Caribbean Health Sciences Literature (LILACS), Excerpta Medica Database (EMBASE), American Psychological Association database (PsycINFO), Web of Science (WoS), and gray literature from inception to January 2023. We checked the references of the included studies and reviews (Additional file 1: Supplementary Chart 1. Databases search strategy). Searches were not limited by date or language.

Study selection

Three independent review authors independently (GZ and SK; MDBD and SE; FKN and EM) inspected all titles and abstracts identified. The second review stage consisted of reading the articles selected in the previous step in full text.

When a difference in opinion was found at each stage, the article selection was decided independently by the other two review authors (MDBD and EM).

Data extraction

Two authors (TDC and MDBD) extracted the following study characteristics: a) first author's name, b) publication year, c) population, d) the number of subjects in the studye) study location, f) pollution measurement method, g) types of contaminants, h) ASD diagnostic assessment, i) control group, j) participant age, k) follow-up time and l) results.

Summary measures and data analysis

Table 1 qualitatively summarizes the main characteristics of the included studies. Effect sizes using the beta coefficient (β) or relative risk (RR). The relative risk consolidates various metrics from individual studies, including the incidence rate ratio, odds ratio, hazard ratio, adjusted hazard ratio, cumulative hazard ratio, and Bayesian predictive odds ratio, each accompanied by its corresponding 95% confidence intervals (95% CI). If both the beta coefficient and relative risk were accessible for a specific outcome, both measures were described for comprehensive reporting. The meta-analyses were calculated with MetaXL 5.3 [31] software. The "ContCI" type for studies that report a β value and the "RRCI" type for studies that report an RR value were used for the meta-analysis. The method selected was inverse variance heterogeneity (IVhet) [32]. The meta-analyses were conducted exclusively with studies of high methodological quality (Additional file 1: Table S1). Furthermore, a sensitivity analysis was performed, considering the type of instrument used for ASD detection (either diagnosis or screening).

Table 1 Characteristics of included studies
Full size table

Cochran's Q, tau-squared (tau2) tests and I2 statistics were used to test heterogeneity. The I2 statistic interpretation was 0% to 40% might not be significant; 30% to 60% may represent moderate heterogeneity; 50% to 90% may represent substantial heterogeneity; 75% to 100% means considerable heterogeneity [60].

For studies that did not outline the mean and standard deviation, these values were computed using the Hozo et al. method specified in the research conducted by Wan X [61] and colleague.

Risk of bias in studies and certainty of evidence

Two reviewers independently judged the methodological quality of the individual studies (S.K. and E.M.) following the Quality Assessment Tool for Observational Cohort Studies from the National Institute of Health [62]. The studies were classified as good, fair, or poor (Additional file 1: Table S1). The GradePro tool was used to conduct the certainty analysis. The degree of evidence uncertainty was rated as high, moderate, low, and very low.

Results

A total of 5,780 studies were identified, of which 2,723 were duplicates. The remaining 3,057 studies of these 3,019 articles were excluded: 2,501 did not evaluate the association between environmental pollutants and ASD; 513 were studies with another type of design (cross-sectional, case studies, case series, experimental models, reviews, response letters or editorials); and five included participants older than 19 years. Thirty-eight articles were selected to read the full text, of which 11 were excluded: eight did not evaluate the association between air pollutants and ASD, and three had another design. Finally, 27 articles were included in the systematic review, and 22 were included in the meta-analysis (Fig. 1).

Fig. 1
figure 1

Flowchart of the study selection process

Full size image

The 27 articles included 1,289,183 individuals aged ranging from childhood to adolescence. Twenty-four studies were conducted on children [33,34,35,36,37,38,39,40,41,42,43,44, 46,47,48, 50,51,52,53,54,55, 57, 58] (n = 1,225,715), and three studies were conducted in adolescents [49, 56, 59] (n = 63,468). The average exposure duration was 6,9 years, with follow-up times ranging from 2 years [52] to 17 years [49, 59] across the studies (Table 1).

The studies reported one hundred twenty-nine pollutants, which are air pollutants and toxic substances. Seven studies reported nitrogen dioxide [40, 43, 45, 51, 52, 54, 59]; six studies reported PM 2.5 [37, 40, 43, 51, 52, 54]; five studies reported PM 10 [38, 40, 43, 45, 59]; four studies reported ozone [43, 45, 54, 59] and mono-n-butyl phthalate [33, 36, 41, 46, 50]. diethyl phosphite [33, 47, 53, 56], mono-ethyl phthalate [33, 36, 41, 50], PCB 118 [33, 35, 36, 39], and PCB 153 [33, 35, 36, 39]. Three studies reported bisphenol A [33, 36, 42], dialkyl phosphates [47, 56, 57], dimethyl phosphate [33, 47, 56], manganese [33, 55, 58], mono-(2-ethyl-5-hydroxyhexyl) phthalate [33, 36, 46], mono-3-carboxy propyl phthalate, mono-benzyl phthalate [33, 36, 50], nitrogen oxides [38, 40, 49], PCB 138 and PCB 180 [33, 35, 39]. 3,5,6-Trichloro-2-pyridinol [47, 53], carbon monoxide [45, 59], copper [55, 58], diazinon [47, 56], dimethylthiophosphate [33, 53], lead [33, 58], mono-(2-ethyl-5-oxohexyl) phthalate [33, 46], mono-2-ethyl-hexyl phthalate, trans-nonachlor, β-hexachlorocyclohexane [33, 36], mono-isobutyl phthalate [36, 41], oxychlordane, p,p-dichlorodiphenyldichloroethylene [33, 36], PCB 101 [36, 39], PCB 187 [35, 36] and sulfur dioxide [45, 59] were described in 2 studies each (Additional file 1: Table S2).

And finally, the following pollutants were described in a single study: 1,3-butadiene, acetaldehyde, benzene, chloroform, chromium, ethyl benzene, formaldehyde, hexavalent chromium, meta/para-xylen, methylene chloride, molybdenum, nickel, ortho-dichlorobenzene, ortho-xylene, paradichlorobenzene, perchloroethylene, polycyclic aromatic hydrocarbon, selenium, toluene, trichloroethylene, vanadium [58], 3-phenoxybenzoic acid [34], arsenic, cadmium, mercury, triclosan [33], brominated biphenyl 153, hexachlorobenzene, mono-2-ethyl-5-carboxypentyl phthalate, p,p-dichlorodiphenyltrichloroethane, perfluorohexane sulfonate, perfluorononanoate, perfluorooctane sulfate, perfluorooctanoate, polybrominated diphenyl ether (PBDE) 100, PBDE 153, PBDE 154, PBDE 183, PBDE 28, PBDE 47, PBDE 85, PBDE 99, PCB 172, PCB 105, PCB 138/158, PCB 146, PCB 156, PCB 157, PCB 167, PCB 170, PCB 177, PCB 178, PCB 183, PCB 194, PCB 195, PCB 196/203, PCB 199, PCB 206, PCB 209, PCB 28, PCB 66, PCB 74, PCB 99 [36], chlorpyrifos, malathion, oxydemeton-methyl [56], chlropyrifos-oxon, terbufos, [47], di-(2-ethylhexyl) phthalate [41], diethyl alkyl phosphates, dimethyl alkyl phosphates [57], elemental carbon, iron, organic carbon [55], nitric oxide [51], organochlorine, organophosphate [44], PM 2.5 absorbance, PM coarse [40], PCB 11, PCB 132, PCB 136, PCB 174, PCB 175, PCB 176, PCB 196, PCB 52, PCB 77, PCB 84, PCB 91, PCB 95 [39], polychlorinated biphenyls, polychlorinated dibenzo-p-dioxins and dibenzofurans [48] and three metabolites of di-(2-ethylhexyl) phthalate [50] (Additional file 1: Table S2).

Results of the association between environmental pollutants and ASD from individual studies

Individual studies reported a significant association with the following contaminants: cadmium, bisphenol A [33], PCB 138 [35], PBDE 28, PBDE 47, PBDE 99, PBDE 100, PBDE 154 [36], PM 2.5 [37, 43, 51, 52, 54], PCB 101 [39], mono-i-butyl phthalate [41], nitrogen dioxide [43, 45, 51, 59], carbon monoxide [45, 59], sulfur dioxide [45], chlropyrifos-oxon [47], nitrogen oxides [49], mono-n-butyl phthalate, mono-3-carboxypropyl phthalate [33, 50], nitric oxide [51], elemental carbon, organic carbon, iron, manganese [55], dialkylphosphates, dimethylphosphate [56], benzene, perchloroethylene, 1,3-butadiene, toluene, ortho-xylene, meta/para-xylen, ethyl benzene, lead, acetaldehyde, formaldehyde, trichloroethylene [58] and copper [55, 58] (Additional file 1: Table S3).

A significant association was also reported with PCB 74, PCB 146, PCB 153, PCB 156, PCB 157, PCB 170, PCB 172, PCB 177, PCB 178, PCB 183, PCB 187, PCB 194, PCB 195, PCB 196/203, PCB 199, PCB 209, β-hexachlorocyclohexane [36], brominated biphenyl 153, PCB 136, PCB 175, PCB 176 [39], ozone [45], polychlorinated dibenzo-p-dioxins and dibenzofurans [48], diethyl alkyl phosphates [57], vanadium [58] and PM 10 [59] (Additional file 1: Table S3). No significant associations were reported with the remaining pollutants.

Quality of studies

The evaluation of the quality of the studies is presented in Additional file 1: Table S1 and Figure S2. When carrying out this analysis, it was found that the majority of the included studies (n = 22) showed high quality [33,34,35, 37,38,39,40, 42, 43, 45, 46, 48,49,50,51,52,53,54,55, 57,58,59], a smaller proportion (n = 5) presented fair quality [36, 41, 44, 47, 56], and none presented poor quality.

Meta-analysis

The results of the individual and subgroup meta-analyses are presented below. It should be noted that, on some occasions, meta-analyses included the same study because they provide results for different environmental pollutants.

Meta-analysis of each pollutant and its association with ASD

The first meta-analysis was performed separately for each pollutant, and a significant association was found with nitrogen dioxide, copper, mono-3-carboxypropyl phthalate, monobutyl phthalate and PCB 138 (Additional file 1: Figure S2). Additionally, these meta-analyses suggest a potential association with PM 10 (Additional file 1: Figure S3). No associations were found with the other pollutants.

The association between nitrogen dioxide and ASD was significant RR 1.20 (95% CI: 1.03 to 1.38). However, this association showed high heterogeneity among studies (I2 = 91%) (Fig. 2). Conversely, copper exposure displayed a significant association with ASD, with an RR of 1.08 (95% CI: 1.03 to 1.13) and low heterogeneity (I2 = 0%) (Fig. 3A).

Fig. 2
figure 2

Meta-analysis association between nitrogen dioxide and ASD

Full size image
Fig. 3
figure 3

Meta-analysis association (A) copper; B mono-3-carboxypropyl phthalate; C monobutyl phthalate; D PCB 138 with ASD

Full size image

Similarly, exposure to mono-3-carboxy propyl phthalate was associated with ASD (β = 0.45, 95% CI: 0.20 to 0.70), with low heterogeneity (I2 = 0%) (Fig. 3B). Monobutyl phthalate also exhibited a positive coefficient (β = 0.43, 95% CI: 0.13 to 0.73) with low heterogeneity (I2 = 0%) (Fig. 3C). Lastly, PCB 138 showed an association with ASD, reflected in an RR of 1.84 (95% CI: 1.14 to 2.96) and low heterogeneity (I2 = 0%) (Fig. 3D).

Meta-analysis by subgroups of pollutants and their association with ASD

The pollutants were classified into 16 subgroups: ground-level ozone; PM; carbon monoxide; sulfur oxides; nitrogen oxides; volatile organic compounds; dioxins, furans, PCBs; hydrocarbons; inorganic substances; metals; organophosphates and carbamates; pesticides; phthalates; phenols/phenoxy acids; polyfluoroalkyl substances and plastics. Positive associations were found with carbon monoxide, nitrogen oxides, and metals (Fig. 4) (Additional file 1: Figure S2). A negative association with organophosphates and carbamates was observed (Fig. 5). Finally, potential associations with ozone (Additional file 1: Figure S4), PM (Additional file 1: Figure S5), inorganic substances (Additional file 1: Figure S6), pesticides (Additional file 1: Figure S7), dioxins, furans, and PCBs (Additional file 1: Figure S8).

Fig. 4
figure 4

Meta-analysis association of (A) carbon monoxide, B nitrogen oxides, and (C) metals with ASD

Full size image
Fig. 5
figure 5

Meta-analysis association between organophosphates and carbamates and ASD

Full size image

Also, the association between carbon monoxide and ASD was found to be significant, with an RR of 1.57 (95% CI: 1.25 to 1.97) and low heterogeneity (I2 = 0%), (Fig. 4A). Nitrogen oxides, including nitrogen dioxide and nitric oxides, were also associated with ASD, with an RR of 1.09 (95% CI: 1.04 to 1.15) and moderate heterogeneity (I2 = 34%) (Fig. 4B).

Metal elements such as iron and molybdenum were linked to ASD with an RR of 1.13 (95% CI: 1.01 to 1.27) and low heterogeneity (I2 = 24%) (Fig. 4C). Conversely, exposure to organophosphates and carbamates, which include compounds such as diethyl phosphate, dimethyl phosphate, dimethyl thiophosphate, dialkyl phosphates, diethyl alkyl phosphates, and dimethyl alkyl phosphates, showed a negative association with ASD (β = -0.49, 95% CI: -0.85 to -0.13) and high heterogeneity (I2 = 85%) (Fig. 5).

Sensitivity analysis

For the sensitivity analysis, we pooled the studies by pollutant and according to the instrument applied to determine autism, either diagnosis or screening. Once the meta-analyses were carried out, the results remained constant for nitrogen dioxide with diagnostic tools (Additional file 1: Figure S9), copper (Additional file 1: Figure S10), mono-3-carboxy propyl phthalate (Additional file 1: Figure S11), and mono-n-butyl phthalate (Additional file 1: Figure S12) with screening instruments. PCB 138 could not be meta-analyzed because there were not enough studies with either of the two types of tools.

In the case of the subgroups, the association between pollutants and ASD was maintained in carbon monoxide with diagnostic instruments (Additional file 1: Figure S13), nitrogen oxides with diagnostic and monitoring tools (Additional file 1: Figure S14-15), metals with diagnostic instruments (Additional file 1: Figure S16), and organophosphates and carbamates with monitoring instruments (Additional file 1: Figure S17). Finally, when meta-analyzed only with tracking instruments, the PM found as a possible association reported a significant association (Additional file 1: Figure S18).

Certainty of evidence

The analysis of evidence certainty using the GradePro tool consistently reveals a landscape characterized by low or very low certainty across all conducted analyses. The main factors contributing to this were heterogeneity between studies and the risk of publication bias (Additional file 1: Table S4-S7).

Discussion

This systematic review and meta-analysis investigated the association between environmental pollutants and the incidence ASD in children and adolescents. The results indicated that exposure to individual pollutants such nitrogen dioxide, copper, mono-3-carboxy propyl phthalate, monobutyl phthalate, and PCB 138 increases the risk of developing ASD. Subgroup analyses further linked carbon monoxide, nitrogen oxides, and metals to higher ASD risk. Additionally, trends suggested associations between ASD and exposure to particulate matter, inorganic substances, and pesticides. The associations found in this study can be explained according to the pollutant type, individually or by subgroup.

The associations with PCB 138 [19, 63], carbon monoxide [7], nitrogen oxides [10], and metals [64, 65] and risk of ASD were consistent with findings from other systematic reviews. However, there were discrepancies between those with nitrogen dioxide [10, 15], copper [66], mono-3-carboxypropyl phthalate and monobutyl phthalate [67, 68] and other reviews.

Differences can justify the possible differences between our findings from other reviews can be justified that some studies carried out subgroup analyses by exposure time [10, 15], did not only include a cohort study [15], considering the exposure window [15], the differences between pollutant concentrations and other methodologies for estimating associations [67, 68].

It is recognized that environmental pollutants disrupt cellular metabolism through mechanisms like breaching cell membranes, intracellular accumulation, and inhibition of critical metabolic pathways [69]. For instance, heavy metals can trigger oxidative stress by generating reactive oxygen species, which can harm lipids, proteins, and DNA and compromise mitochondrial function, potentially leading to cell death, tissue damage, or neurological disorders [70]. Particulate matter and polycyclic aromatic hydrocarbons can breach the blood–brain barrier, initiating brain inflammation that may disrupt neurotransmitter systems and synaptic function [69]. Persistent immune system activation by pollutants can induce chronic neuroinflammation, disrupting brain architecture connectivity and impeding normal brain development [71].

Moreover, pollutants can cause DNA damage, leading to epigenetic alterations like DNA methylation and histone modifications that influence gene expression tied to brain development and function [72, 73]. This, in turn, could potentially contribute to the pathophysiology of ASD. Compelling evidence suggests that environmental contaminants significantly impact cellular metabolism and neurological well-being, connecting these molecular changes to broader neurodevelopmental consequences [55, 74].

Exposure routes to environmental pollutants are crucial in ASD pathogenesis. Alterations in neuronal connectivity, occurring from prenatal to early adulthood, can result from genetic and epigenetic factors [76]. The ENVIRONAGE cohort study found that increased PM 2.5 exposure during pregnancy was associated with relationship and prosocial behavior problems in preschoolers [76]. These effects may be due to higher mutation rates and DNA repair alterations during fetal and neonatal stages [72]. Conversely, low-pollution maternal environments are associated with beneficial DNA methylation in neurodevelopmental genes, highlighting the importance of pollution levels and particulate matter composition in understanding ASD risk [73].

The characteristics of the population may impact the associations identified in this meta-analysis study and the timing of exposure. ASD symptoms typically manifest early in life, exerting significant developmental effects during the prenatal and early postnatal periods [76]. Both acute and chronic exposure to environmental pollutants during these critical phases can influence neurogenesis and neuronal maturation [76]. Evidence suggests that prenatal and postnatal exposure to contaminants can bring about developmental alterations in children, with the developing nervous system being especially vulnerable to environmental toxins, even at low exposure levels [76]. Accurate assessment of the timing of contaminant exposure is crucial for comprehending the underlying mechanisms and crafting effective interventions.

Recognizing that the absence of significant associations with specific contaminants, individually or in combination, or inconclusive findings does not imply their non-existence is crucial. Further research is imperative to pinpoint the risk factors contributing to our understanding of ASD and to inform the development of enhanced preventive measures.

Strengths and limitations of the systematic review

Our systematic review stands out for several key reasons. Firstly, it adopts a broad approach, incorporating a wide array of databases and gray literature sources. Unlike other reviews, our search was not constrained by time or language, ensuring inclusivity and breadth of scope. Additionally, we excluded observational studies, which often present limitations for causal inference, thereby enhancing the robustness of our findings.

Furthermore, our review maintained a stringent focus on studies of high methodological quality, ensuring the reliability of our results. Unlike comparable reviews, our analysis encompassed a broad range of contaminants, facilitating a deeper understanding of their impact on ASD incidence. Moreover, our study evaluated the effects of both individual and grouped contaminants, offering a novel perspective on the issue.

This study also has some limitations. First, the studies included a variability of exposure time, pollutant detection method, and the instrument used to determine ASD. However, to avoid overestimating the effect, sensitivity analyses were performed that supported the validity of the association with ASD. In addition, the instruments, although diverse, are all approved by the scientific community for the screening or diagnosis of ASD. Second, some meta-analyses had high heterogeneity.

In summary, our systematic review represents an original contribution to the field, distinguished by its meticulous methodology, broad inclusion, and comprehensive analysis of the effects of pollutants on the incidence of autism.

Conclusion and future directions

This systematic review and meta-analysis suggest that children and adolescents exposed to higher contamination levels by pollutants such as nitrogen dioxide, copper, mono-3-carboxy propyl phthalate, mono butyl phthalate, and PCB 138 have a higher risk of developing ASD. Likewise, those exposed to subgroups of environmental pollutants such as carbon monoxide, nitrogen oxides, and metals were associated with ASD. Therefore, it is important to identify the factors that underlie the susceptibility of children and adolescents to contribute effectively to ASD and identify prevention strategies. Future studies should standardize the exposure time to pollutants and the detection methods, allowing for more precise comparisons and better interpretation of the results.