Keywords

Neurodevelopmental disorders are a group of conditions characterized by impairments of social skills or intelligence due to perturbed growth and development of the brain, with onset in the developmental period. According to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V), they include intellectual disability (intellectual developmental disorder), communication disorders, autism spectrum disorder (ASD), attention deficit hyperactivity disorder (ADHD), specific learning disorders, and motor disorders [1]. They are among the most common chronic disorders in children worldwide. For instance, in 2016 9.4% of US children 2–17 years of age had at one time been diagnosed with ADHD [2]. The prevalence of children aged 3–17 years who had ever been diagnosed with a developmental disability increased from 5.8% to 7% between 2014 and 2016 [3], whereas the prevalence of ever receiving an ASD in children aged 3–17 years reached 2.5% in 2016 [4].

Although the conditions included among neurodevelopmental disorders are heterogeneous, these disorders are thought to be developmental, with early onset and involving a multitude of potential risk factors. The environment is now known to be an important determinant of child health, with increasing evidence that some chemicals are particularly toxic to the human brain. More than 140,000 new chemicals and pesticides have been synthesized since 1950, with about 5000 that are produced in high volume, and many have become ubiquitous in the environment [5]. The fetal period is a critical window of development, and in utero exposure to environmental chemicals has a significant influence on fetal growth and development, with consequences for birth outcomes, child development, and cognitive and behavioral functions [6, 7]. For instance, over 200 chemicals used in commerce or industry are known to be neurotoxic to humans [6]. Disruptions caused by exposure to these toxic chemicals may have far-reaching consequences with a considerable impact in regard to long-term complications including developmental delays, effects on socio-emotional adjustment, educational success, and quality of life [8,9,10]. In this chapter, we describe the accumulated evidence in regard to specific chemicals of interest, the potential mechanisms of action through which exposure to these chemicals affects child neurodevelopment, and the societal impacts of these neurotoxicants. We will finally discuss challenges arising when studying effects, and propose future research needs and directions.

1 Main Findings in Regard to Neurodevelopment and Developmental Chemical’s Exposure

1.1 Trace Elements: Lead, Manganese, Mercury, and Arsenic

There is a compelling evidence that trace metals adversely affect neurodevelopment and increase risk of neurodevelopmental disorders. Lead (Pb) is the most studied environmental toxicant, and its neurotoxic effects have been known and well described for centuries. For instance, Dioscorides, the well-known Greek physician and botanist once said: “Lead makes the mind give away”. Thus, Pb is considered a paradigm metal to study. Exposure to high concentrations of Pb has been linked to effects on the central nervous system, such as deficits in concentration, memory, cognition, and behavior [6, 10]. Epidemiological research has linked Pb exposure during childhood to deficits in cognitive skills and IQ, behavioral problems, and attention hyperactivity disorder (ADHD) [11,12,13,14]. Prenatal and early postnatal Pb poisoning and long-term exposure to Pb at low levels has also been associated to an increased risk of aggressive behaviors [15]. Recent studies also showed a link between early-life exposure to Pb and criminal behavior later in life [16]. Overall, Pb exposure unequivocally affects the developing brain, and its effects have been shown for a variety of cognitive and behavioral domains.

Although manganese (Mn) is an essential element, high exposure to Mn has been associated with behavioral problems, decreased cognitive function, and frontal lobe abnormalities [17]. The health implications for fetuses and infants are a concern given the high accumulation of Mn in tissue during development [18]. Indeed, maternal levels of Mn increase markedly during pregnancy, peaking in the third trimester [19]. Both prenatal and postnatal exposure to Mn has been linked to neurodevelopmental effects. In Bangladesh, greater exposure to manganese from drinking water was associated with reduced mathematics achievement scores in school children [20], whereas in Quebec, greater hair concentrations of Mn were associated with hyperactivity [21], reduced memory [22], and attention functions [23]. Other studies from Mexico and Brazil showed reduced intelligence quotient and impaired olfactory function [24, 25].

Arsenic (As) has also received much attention as a neurotoxicant, largely based on studies conducted in Bangladesh. Currently, the permitted concentration of arsenic in water is 10 μg/L (10 ppb); yet, an estimated 100 million people worldwide are exposed to excessive amounts of arsenic via drinking water [26]. As, including inorganic and methylated arsenicals, accumulates in many parts of the brain [27]. Studies of children from Bangladesh showed an inverse association between cognitive function and As levels in water [28, 29]. A recent meta-analysis on the associations between arsenic exposure and IQ concluded that a 50% increase in arsenic levels was associated with a 0.4% decrease in IQ in children [30]. Interestingly, studies on As exposure and neurodevelopment showed differential effects between boys and girls with a higher susceptibility among girls.

Mercury (Hg) is a trace metal of known toxicity. It naturally occurs in several physical and chemical forms, including metallic mercury, inorganic, and organic mercury (e.g., methylmercury, MeHg), with varying toxic effects [31]. Hg can have neurotoxic effects on the human central nervous system, particularly during fetal development. A number of prospective cohort studies assessed the neurodevelopmental effects of chronic low and moderate prenatal MeHg exposure from maternal fish consumption. The most influential studies were conducted in the Faroe Islands and were used by the EPA to establish guidelines for mercury toxicity [6, 32,33,34].

Although the individual toxicity of these trace elements is now well established, the neurotoxic effects of metals may occur in an interactive way, and several metals may have synergistic effects. However, to date few studies have investigated the independent and joint toxicity of metals within a mixture.

1.2 Polychlorinated Biphenyls (PCBs) and Brominated Flame Retardants (PBDEs)

Polychlorinated biphenyls (PCBs) are ubiquitous neurotoxicants used for various industrial applications as coolants and lubricants in electrical equipment because of their general chemical inertness and heat stability [35]. PCBs were manufactured as a mixture of congeners and banned from production and use in the late 1970s. The major exposure route for humans is through food, whereas inhalation and dermal routes are predominant in occupational settings [36]. There is a large body of scientific evidence showing neurodevelopmental effects of early exposure to PCBs. Recent systematic reviews showed that prenatal or early postnatal exposure to PCBs was associated with adverse cognitive and behavior outcomes in most studies, although null associations were also reported [37, 38]. Additionally, a few studies found associations between developmental exposure to PCBs and autism [39, 40].

Polybrominated diphenyl ethers (PBDEs) and their hydroxylated forms have been widely studied for their neurotoxicity. PBDE additives are not fixed in the polymer product through chemical binding, and can thus leak into the environment [41]. Like PCBs, they persist in the environment with relatively long half-lives. Human exposure is mainly due to ingestion of house dust, consumption of fish and other animal products, and breastfeeding for infants [42]. Higher maternal concentrations of PBDEs were associated with impaired cognitive and motor function, increased attention problems, anxious behavior, increased withdrawal, and altered response to frustration [43,44,45,46]. Additionally, early-life PBDE exposures were associated with decreased IQ and psychomotor development [47].

1.3 Pesticides

There is a strong scientific evidence that exposure to pesticides produces human health effects, including impairment of the central nervous system. Pesticides are widely used in both developed and developing countries, especially with the expansion of free-trade agreements and the focus of many developing countries on agricultural exports. This trend makes extensive pesticides use one of the major environmental health issues in low- and middle-income countries. The most prevalent chemical families of pesticides are organophosphates, organochlorines, and pyrethroids, with organophosphates being the most frequently used. Overall, research points to deleterious effects of pesticides on cognitive and behavioral development in children. Findings from prospective birth cohorts such as the Center for the Health Assessment of Mothers and Children of Salinas (CHAMACOS) point to a multitude of effects of organophosphates, including decreased IQ [48,49,50], impaired mental development, and pervasive developmental disorder at 24 months of age [51]. A study in New York City found a possible link between prenatal indoor pesticide use and abnormal neonatal reflexes [52]. Other studies from Brittany, France reported associations between prenatal pesticides exposure in an agricultural area and cranial growth in a prospective birth cohort of 3421 pregnant women [53]. Prenatal exposure to organochlorine pesticides (DDT and its metabolite DDE) was also associated with reduced psychomotor development index scores and significant reduction in the general cognitive index and neurodevelopmental delays [54,55,56]. However, many studies that investigated the associations between DDE exposure and cognitive and behavioral development reported null findings [57,58,59,60]. Unlike DDT, pyrethroids are rapidly metabolized and have short half-lives in humans (hours to days) [61], and very few studies investigated their associations with child neurodevelopment. Oulhote et al. found a positive association between concurrent exposure of some pyrethroid metabolites and child difficulties scores in a cross-sectional study in Canada [62]. Although a French study found no association between prenatal urinary pyrethroid metabolites and 6 year olds’ performance on WISC Verbal Comprehension or Working Memory [63], studies from China and Mexico reported that prenatal pyrethroid exposures were associated with lower cognitive scores in 1-year infants [64] and children 2–3 years of age [65]. Further, studies from France and the USA showed adverse effects of pyrethroids exposure on behavioral outcomes in children [66, 67].

1.4 Plasticizers

Studies investigating the associations between developmental exposure to plasticizers such as phthalates and bisphenol A (BPA) are still in their infancy but have shown some evidence of associations with both cognitive and behavioral outcomes. Developmental exposure to phthalates was shown to be associated with ADHD behaviors, autistic traits, reduced mental and psychomotor development, and reduced IQ [68,69,70,71]. BPA was also reported to be associated with more internalizing and externalizing behaviors, increased risk of ADHD, and executive function [72,73,74,75]. Most of these studies point to sexually dimorphic effects of these compounds on child neurodevelopment, with boys being more vulnerable [76]. Other studies showed no association between neither phthalates nor BPA with child neurodevelopment [71, 77].

1.5 Perchlorate

Perchlorate is used in a variety of industrial products including missile fuel, fireworks, and fertilizers, and industrial contamination of drinking water occurs in several areas [78]. Exposure to perchlorate is widespread and exposure from contaminated food and drinking water is ubiquitous [79]. Perchlorate can block iodide uptake into the thyroid gland, and result in decreased production and secretion of thyroid hormone, therefore potentially perturbing proper brain development that is dependent on thyroid function. One study investigated potential neurodevelopmental effects of gestational exposure to perchlorate in the UK and Italy. Findings from this study showed that higher maternal perchlorate levels in hypothyroid/hypothyroxinemic pregnant women were associated with adverse effects on offspring cognitive development [80].

2 Potential Mechanisms of Action

Traditionally, mechanisms of neurotoxicity have been identified as pathways leading to neuronal cell death, neuropathology, or neural injury. However, recent research highlights alternative mechanisms that result in more subtle but serious changes in cognition and behavior. These mechanisms include neuroendocrine and immune system pathways, in addition to inflammation and epigenetic mechanisms.

Early-life inflammation induced by exposure to chemicals, especially metals, has been suggested as a potential mechanism for the observed neurodevelopmental effects [81, 82]. For instance, chronic low-level lead exposure may trigger chronic inflammation and lead to adverse changes in inflammatory markers such as CRP [83]. Lead is also known to interfere with the N-methyl-D-aspartate receptor, which is essential for hippocampus-mediated learning and memory [84, 85], and to disrupt neurotransmission by inhibiting neuronal voltage-gated calcium channels and intracellular calcium dynamics [86]. Methylmercury also induces impairment in intracellular calcium homeostasis, in addition to alteration of glutamate homeostasis and oxidative stress [87]. Methylmercury perturbs cell proliferation and migration, producing widespread abnormalities including heterotopias, reduced cell densities, incomplete myelination, glial proliferation, and limited gyral differentiation [12]. The effects of Mn on behavior and cognitive abilities in children may be related to effects on the dopaminergic system during development. Mn accumulates in neurons, astrocytes, and oligodendrocytes, inhibits ATP synthesis in mitochondria, and interferes with the dopaminergic system [88, 89]. Arsenic has been shown to impact the synaptic activity of neurons localized to the hippocampus [26]. These effects may be attributable to alterations in synapse-related gene expression. Experimental studies showed that exposure to arsenic throughout gestation increased DNA methylation on two genes involved in neural plasticity in rat cortex and hippocampus [90]. Hypomethylation of these genes in both regions was observed later after 4 months of cumulative exposure to arsenic. Arsenic also interferes with glucocorticoid, cholinergic, and glutamatergic signaling, leading to hippocampal-related deficits [91,92,93].

Many pesticides, including organophosphates (OPs), carbamates, pyrethroids, and organochlorines (OCs), are designed to attack insects’ central nervous system by interfering with chemical neurotransmission or ion channels. They specifically work by targeting enzymes that regulate neurotransmitters, such as acetylcholine, which is released by motor neurons [94]. For instance, both OPs and carbamates have a common target of toxicity, the inhibition of acetylcholinesterase, whose physiological role is hydrolyzing acetylcholine, a major neurotransmitter in the central and peripheral nervous system [95, 96]. Pyrethroids bind to the sodium channel and slow its activation, as well as the rate of inactivation, leading to a stable hyperexcitable state, and inhibit Gamma aminobutyric acidA-gated chloride channels, leading to the choreoathetosis, salivation, and seizures seen in occupational exposures [96]. Finally, OCs such as DDT interfere with the sodium channels in the axonal membrane, by a mechanism like that of pyrethroids, prolonging the depolarizing (negative) afterpotential of the action potential, thus producing a period of increased neuronal excitability [97].

PCBs exert their neurotoxicity through a variety of mechanisms including thyroid disruption, interference with sex steroids, and aryl hydrocarbon receptor (AhR) activity, especially for dioxin-like PCBs [98,99,100]. Thyroid and sex steroid hormones are critical for proper brain development, and subtle perturbations of these systems have been shown to be associated with both cognitive and behavioral problems. PBDEs act on the developing brain mainly through a reduction in circulating thyroid hormones; however, direct effects on the developing brain have also been reported. Thyroid hormones are known to play a role in brain development, and a large body of experimental and human studies showed decreased levels of thyroxine (T4) and triiodothyronine (T3), and increased likelihood of hypothyroidism in relation to higher levels of PBDEs [101, 102]. These effects on the thyroid hormone levels are explained by an enhanced metabolism and excretion of T4 as a result of exposure to PBDEs and interactions of PBDEs (or their metabolites) with the thyroid hormone transport systems or with thyroid receptors [103]. In addition to their effect on the thyroid system, PBDEs exert effects by inducing oxidative stress in human neuroblastoma cells and hippocampal neurons, interfering with calcium homeostasis, and causing DNA damage and apoptotic cell death [104,105,106,107].

Plasticizers are endocrine-disrupting chemicals that can mimic the effects of endogenous hormones or disrupt the synthesis, metabolism, or uptake of endogenous hormones. For instance, phthalates may interfere with the action or metabolism of androgens and thyroid hormones, with subsequent anti-androgenic effects [108]. Increased concentrations of phthalates were inversely associated with total serum thyroid hormone levels in pregnant women and neonates and thyroid stimulating hormone in neonates [109]. Additionally, phthalate exposure may exert neurotoxic effects by increasing oxidative stress and via epigenetic re-programming of the fetus and placenta [76]. BPA may interfere with thyroid-specific gene expression, and affect androgen/estrogen concentrations by inhibiting key enzymes involved in gonadal hormone synthesis and metabolism [110]. These hormones play a substantial role in brain development [111].

Although perchlorate has not been extensively studied in relation to neurodevelopment, its effect on the thyroid gland through blocking the uptake of iodine is well studied. Low-level exposure to perchlorate has been shown to be positively associated with TSH and negatively associated with free T4 [112, 113]. Thyroid hormones are essential for proper brain development, and it is therefore hypothesized that the effects of perchlorate on thyroid function may lead to potential neurodevelopmental consequences [114, 115]. These effects may be stronger in children of women with low iodine status [116].

It is important to mention that these potential mechanisms of action are not mutually exclusive and many chemicals, such as metals, can exhibit endocrine-disrupting properties or increase oxidative stress and inflammation, mediated by epigenetic and neuroendocrine mechanisms.

3 Societal Impact of Environmental Chemicals

An argument frequently advanced by those skeptical about the importance of the neurodevelopmental impact of environmental chemicals is that the effect on an individual child is relatively modest, failing to reach the level of clinical significance. This argument fails to consider the issue in the context of population health. Effect estimates from epidemiologic data are population average effects and should be interpreted in the context of a population and not at the individual level. Some individuals will be resistant and some will be more sensitive. It is critical to view the issue of children’s exposures to environmental chemicals in the context of population health and not just the health of an individual child. Moreover, the impact of a factor at the population level depends not only on the magnitude of its impact on health, or its effect size, but also on the distribution of the factor or, in the case of a dichotomous factor, its incidence or prevalence. In a set of comparative analyses of pediatric disease and events, such as brain tumors, congenital heart disease, traumatic brain injury, iron deficiency, and lead exposure, Bellinger [97] estimated the total number of IQ points lost among US children younger than 5 years of age associated with each disease or event. The estimate for the loss associated with lead exposure was nearly 23 million IQ points, exceeded only by preterm birth. Among the reasons for this is the absence of a threshold for its inverse relationship with IQ and the fact that virtually every child has a quantifiable blood lead concentration. As a result, and in contrast to most other diseases and events, every child contributes to the total IQ loss in the population that is associated with lead exposure. In fact, the greatest contribution to the total loss is contributed by the very large proportion of children with blood lead concentrations at the lower end of the distribution. A similar calculation using the blood lead distribution of young US children from the late 1970s indicated that, at that time, the total loss of IQ points attributable to lead was approximately 125 million points, suggesting that the measures taken to reduce population lead exposure since that time produced savings of about 100 million IQ points in the current cohort of children. With approximately 25 million children in this age range, the average IQ benefit has been approximately 4 points. Analyzing temporal trends in the IQ scores of US adults over the period in which population blood lead concentrations declined, Kaufman et al. [117] estimated that the mean IQ has increased by 4–5 points.

An even more critical point with regard to the societal impact of environmental chemicals is that a reduction in a child’s IQ is only the “tip of the iceberg” in terms of their neurodevelopmental impacts. IQ is easily measured and widely recognized outcome, but it is merely a marker for a range of other neuropsychological and behavioral adversities that have a substantial impact on an individual’s future health and well-being. A full account of the burden of disease imposed by environmental chemicals must include these downstream impacts on mental health and economic success that can seriously impair quality of life [12].

This perspective is not captured by the method traditionally used to estimate global burden of diseases and risk factors. Current efforts to estimate the burden of disease associated with environmental chemicals consider only a clinically significant IQ deficit as the sequelae of chemicals such as lead [118], i.e., an IQ score <70, the criterion for identifying mild intellectual disability. As the research reviewed in this chapter illustrates, a reduction in IQ is only the “tip of the iceberg” in terms of the neurodevelopmental impacts of chemicals, which, as described, include deficits in specific neurodevelopmental domains such as executive functions, attention, language, and memory, and mental health disorders such as ADHD and, perhaps, autism spectrum disorders. IQ has the virtue of being easily measured, but an IQ deficit likely is only an indication of the likely presence of a range of other neuropsychological and behavioral adversities that subsequently unfold as a “developmental cascade” and have a substantial impact on an individual’s future health and well-being [119].

Another dimension of the societal impact of many environmental chemicals is that the non-random distribution of exposures perpetuates health disparities. In the case of lead, for example, poor and minority children tend to incur greater exposures due to the siting of point sources and the presence of greater lead hazards in the housing available to them. Compounding this problem is the fact that, at any given blood lead concentration, the adverse effects of lead are greater on children already at increased risk of poor outcomes due to the presence of other risk factors. In quantile regression analyses of the associations between blood lead concentration and reading scores, Miranda et al. [120] showed that lead exposure stretched out the left-hand tail of the performance distribution, such that the decrease per μg/dL increase in blood lead concentration was greater among poorer readers than among better readers. One effect would be to aggravate the disparities between the educational outcomes of poor children and their more advantaged peers and, as a study in New Zealand showed, reduce upward social mobility, limiting the socioeconomic success that is achieved in adulthood [121]. The origins of such environmental injustices can be located in the social and economic forces of modern society. One factor contributing to their persistence is the lack of voice affected communities have in influencing those with the power and the means to implement remedies. This was illustrated by the episode of lead contaminated water in Flint, Michigan. It took 18 months after citizens began to complain about the poor quality of their drinking water before government officials acknowledged the problem and began to address it.

4 Challenges and Future Directions in Studying Environmental Neurotoxicants

In the following section, we discuss some of the main challenges in investigating the potential neurotoxicity of chemicals. As summarized earlier, many chemicals have been shown to exert neurodevelopmental effects, and the body of literature keeps growing as new compounds are introduced into the market. The multiplicity of environmental neurotoxicants calls for a general framework that considers the cumulative effect of the mixtures, in addition to their potential interactive effects, while including other effect modifiers (i.e., Sociomics). It will also be important that the field adopt novel causal inference approaches that are widely used in clinical, pharmaco-epidemiology, social epidemiology, and econometrics. Finally, the complexity of child development calls for holistic approaches that consider all the facets of development and the multilayers of potential interactions, hence the need to consider broader definitions of the environment beyond just chemicals exposures.

4.1 The Issue of Chemical Mixtures and Cumulative Effects

Perhaps one of the most important issues for the field is the development of methods to account for the cumulative effects of chemical mixtures. Although the general population experiences exposure to multiple chemicals from many different sources at various doses, most studies in environmental epidemiology consider each chemical separately when assessing the adverse health effects of environmental exposures. This single pollutant approach suffers from several pitfalls including: (1) the risk of false positives in the case of multiple hypotheses testing; (2) confounding from correlated exposures; and (3) the lack of insights on the cumulative or synergistic effects of multiple exposures. These all weaken the inferences that can be drawn about the relationships between chemicals and neurodevelopmental outcomes. In recent years, several statistical methods have been proposed to address these issues. These methods include environmental-wide association studies (EWAS) [122], penalized regression methods (i.e., least angle selection and shrinkage operator [LASSO]) [123], Ridge regression [124], and elastic net regularization [125], dimension reduction methods, and exposure-response surface methodology such as generalized additive models and Bayesian kernel regression methods (i.e., BKMR) [126]. A recent study by Agier and colleagues used simulation to demonstrate feasibility of some established and emerging methods for handling multiple correlated exposures [127]. While promising, these methods underperformed when the goal was identification of individual exposures with an impact on the phenotype of interest. Many exhibited high false discovery proportions as the number of correlated exposures increased. Moreover, results from a recent NIEHS workshop showed that none of the tested approaches appeared to outperform the others [128], and most of these methods performed poorly in the context of complex mixtures [127]. More importantly, most of the methods (e.g., elastic net, LASSO) focus on the issues of correlated exposures while ignoring the threats arising from model misspecification (e.g., interactions, non-linearities). The most promising of these methods appears to be the BKMR. This method relaxes a priori parametric assumptions and allows for the investigation of cumulative effects and potential interactions between chemicals. However, one challenge is that such methods provide conditional estimates that are not applicable in simulating public health interventions since they do not provide marginal estimates of the exposures of interest. A new family of methods is emerging that takes advantage of the developments of machine learning, and specifically, ensemble learning techniques (i.e., stacked generalization). Machine-learning methods have shown great potential for quantifying the role of environmental exposures in regard to their effects on human health [129]. Typically, machine-learning approaches consist of algorithms that find variables (exposures) that are predictive of an outcome (phenotype) in two steps. In the first step, an algorithm “learns” the variables that are associated with the outcome. The algorithm is then tested in an independent dataset to estimate the predictive capability or generalizability of the algorithm [130]. Ensemble techniques reduce the variance and avoid overfitting by combining predictions from numerous “similar” algorithms. Although these methods are in their infancy, several of them have shown promising results in the context of chemical mixtures [131, 132], while others have been used in other contexts and could be leveraged to infer causal estimates of complex mixtures [133]. These methods allow for incorporation of multiple correlated exposures, estimation of both individual and cumulative effects at the population level, and screening of potential interactions. Moreover, these ensemble learning methods allow for more flexibility since they relax a priori assumptions regarding the functional forms and presence of interactions, which means that one can include all the exposures without specifying interactions or functional forms (e.g., non-linearities) and the approach will identify the models that yield the best predictions.

4.2 Incorporating Causal Inference Methods When Studying Neurotoxicants

For obvious ethical reasons, it is often impossible to conduct randomized controlled trials in the field of environmental epidemiology. Therefore, most studies investigating potential neurodevelopmental effects of chemicals rely on observational studies. Additionally, other issues arise when studying environmental chemicals that make causal inferences difficult. First, exposure to the chemical of interest is typically ubiquitous, and all individuals in the study are exposed at a detectable level, leading to continuous exposures, making emulation of RCTs nearly impossible. Second, environmental epidemiology relies on measurements of biomarkers to investigate the exposure-outcome relationship; such reliance on biomarkers makes it difficult to insure “consistency” and “positivity,” two major assumptions necessary for inferring causal effects. “The positivity assumption states that there is a nonzero probability of receiving every level of exposure for every combination of values of exposure and confounders that occur among individuals in the population” [134]. In other words, within each level of the exposure, one will need individuals with all levels of all confounders. This is obviously extremely difficult in studying continuous exposures unless a cohort includes millions of participants or exposures are categorized. The consistency assumption is the ability to hypothetically assign a certain level of exposure to a person exposed to a different level. This opens to the issue of defining what intervention could hypothetically lower the measured biomarker level to a specific desired level in an individual, or what other methodologists call “well-defined interventions” [135]. This calls for the identification of what kind of interventions can directly change the concentrations of a neurotoxicant in the tissue of interest (or circulating blood levels that are also just proxies of the level of neurotoxicant in the target tissue) [136]. This is obviously impossible outside of a few cases (e.g., chelation treatment for lead). Many physiological parameters will influence the measures of circulating levels of a neurotoxicant, and therefore impact the relationship between the external intervention (on which we can intervene) and the biomarker concentrations. This will likely violate the consistency assumption thereby impeding valid causal estimates. However, it is worth mentioning that this issue of consistency can be challenged in terms of scientific discovery. Not all studies aim at emulating an intervention when studying health effects of chemicals, therefore, the knowledge acquired from establishing causal links between a biomarker of exposure and neurodevelopment is still valuable and will likely lead to measures that can reduce external exposures at the population level. For more details on this debate, reader can refer to papers by Judea Pearl, Miguel Hernan, and others [135, 137, 138].

There is also a new body of methods that have been proposed recently to tackle some of the issues arising when inferring causality from observational data. One of the promising methods that can be used is Mendelian randomization analysis [139], which uses genetic variants as instrumental variables for exposures of interest. It is a useful method for assessing causal relationships because the allocation of alleles is random; therefore, genotypes are not expected to be associated with measured or unmeasured confounders. The use of single nucleotide polymorphisms (SNPs) that are associated with exposure levels but not with the outcome of interest (except through the exposure) represents an appropriate approach to examine the causal relationship between pollutant exposures and children’s developmental outcomes. The metabolism and excretion of many xenobiotic substances has been linked to many SNPs (e.g., pesticides and paraoxonase or cytochrome P450). Using these SNPs to construct allelic scores robustly related to the exposures as instrumental variables may help to identify causal effects of environmental exposures on neurodevelopmental test scores. One limitation of such approaches is the lack of instrumental variables (SNPs) that are specific to a class of chemicals, but research on this field should yield valid instruments in the coming years. One study used two SNPs in the 10q24.32 region (near AS3MT) that show independent associations with arsenic metabolism efficiency to investigate the causal relationship between arsenic metabolism efficiency and skin lesion risk [140]. Another issue that arises when using such methods is that they rely on very large sample sizes, and this will depend on the strength of the association between the SNP (or the allelic score combining many SNPs) and the biomarker of exposure. The stronger the association, the smaller the sample size that is needed.

Additional methods that can be leveraged are econometric methods that rely on quasi-experimental designs. Some of these methods leverage natural experiments to emulate randomized controlled trials. We recently leveraged the fact that undergrad students at Harvard University are randomly assigned to dormitories to infer causal estimates of high temperatures on cognitive functions among students during a heat wave [141]. A recent study from the National Bureau of Economic Research showed causal effects of air pollution on dementia by capitalizing on quasi-random variation in pollution exposure due to the EPA’s 2005 designation of nonattainment counties for PM2.5 [142]. Closer to our field, some mesalamine medications used to treat inflammatory bowel disease have dibutyl-phthalate (DBP) in their coating, whereas other mesalamine formulations do not. This difference and the fact that the type of medication depends mainly on the prescribing physician and not on other potential confounders have been leveraged to study the potential effect of DBP on semen quality [143]. Creativity in exploiting such random events will be a very useful approach in solving some of the most challenging issues in environmental neuroepidemiology studies.

4.3 Measurement Error in Exposures

Perhaps one of the most critical issues in the field of environmental epidemiology is the error in measuring both exposures and outcomes. Exposure and outcome misclassification can seriously affect estimates from studies investigating neurodevelopmental effects of chemicals. Non-differential exposure misclassification is a common problem, especially for endocrine-disrupting chemicals with short half-lives that tend to vary widely within individual. Under such conditions, an individual’s true exposure will be difficult to establish because of the low precision of the measurement. When error is not related to the outcome of interest, this will often, but not always, bias the association towards the null. Several studies have used multiple samples to characterize exposure to short half-life compounds, yielding a better characterization. However, this will require a large number of samples, increasing costs in terms of follow-up and analyses [144]. Other researchers proposed using calibration and Bayesian methods to account for non-differential misclassification errors, but these are very rarely used [145]. Construction of latent functions based on measurements of a chemical in multiple matrices represents a hypothetical true measure of exposure and could tackle this issue by modeling the shared variance from these multiple measures instead of a single concentration. This approach was used in studies to examine neurodevelopmental effects of mercury and PFASs [23, 33, 146]. Other researchers propose sensitivity analyses to explore the impact of such biases, allowing for inference of interval estimates.

4.4 Holistic Approaches to Study Child Development

Child neurodevelopment is a dynamic process with a complex structure of determinants and feedback loops. Present studies of developmental neurotoxicants are still in their infancy in terms of addressing such complexity and future studies may benefit from incorporating it both in the study design and analyses stages. Currently, there is a strong need for a comprehensive approach that puts epidemiological studies of neurotoxicants within a developmental framework, particularly across the lifespan.

Context information matters and is a key to understanding effects of neurotoxicants on child development. For instance, many non-environmental cofactors play a key role beyond their main effects and can be effect modifiers of the associations between neurotoxicants and neurodevelopment. Many confounders can potentially play an effect modification role. Additionally, every factor that is impacting child neurodevelopment and not associated with the exposure may modify neurotoxicant’s effect. An important example is the stimulating home environment. While it is very important to take this variable into account, it is not acting as a confounder in most of the studies, but mainly as a strong predictor of the neurodevelopmental outcome. It is therefore very useful to test for its modification effect. Another important example is some nutritional factors that co-occur with environmental exposures. Recently, a few studies showed that folic acid supplementation may blunt the deleterious effects of pesticides and air pollutants on autism [147, 148]. Selenium has also been shown to play an important modifying role in the association between mercury and neurodevelopment [149]. Both these examples of effect modifiers including stimulation and nutritional factors show up when investigating effect modification by socioeconomic status.

These contextual factors may be of an utmost interest in areas of the globe in which many environmental, nutritional, and social factors co-act to yield a high burden of developmental disabilities. For instance, around a third of children in low- and middle-income countries are not developmentally on track [150]. The interplay between nutritional deficiencies, lack of child stimulation due to education policies and early childcare, and a high burden of environmental exposures may explain a significant part of the observed delay in children’s neurodevelopment in these countries. These issues cannot be tackled separately, hence the need for holistic approaches to study neurotoxicants, in addition to an urgent need to develop specific and tailored frameworks that can incorporate these complexities. Transposition of western models to the LMIC realities may likely fail if local dynamics and holistic approaches to environmental health are not considered.