Introduction

Pesticides have undoubtedly changed our world, allowing us to meet increasing agricultural demands and mitigate the spread of endemic insect-borne diseases. Pesticide production has increased linearly for more than half a century, with an estimated yearly consumption of 3.5 x 10⁶ tons globally in 2020 [1,2]. In their management of pest populations, pesticides have augmented the global food supply while helping extinguish endemic insect-borne disease. Studies estimate that 50% of malaria cases and 18% of child deaths related to malaria can be prevented with insecticide-treated nets, a more powerful prevention method than untreated nets [3,4]. The use of pesticides not only benefits our ability to feed and protect global populations, but also benefits the economy. Excluding external application costs, it is estimated that farmers who use pesticides gain a $6.50 return on investment for each dollar spent [5].

Despite these advantages, pesticide use is not without harm for population health. Pesticides are used extensively in the agricultural industry – where they enter the global food supply – and in residential areas, which places the applicator and the general public at risk for pesticide-related health complications [6]. The widespread use of pesticides introduces a complex dynamic of global protection compared to personal harm, where their many benefits in food production and vector-borne illness may be tempered by potential risks among chronically exposed communities. Pesticide exposures occur through inhalation, skin absorption, and/or ingestion. High exposure to acutely toxic types of pesticides can be fatal, whereas chronic exposure to pesticides with carcinogenic properties may influence cancer risk. The toxic properties of pesticide chemicals include other complications affecting every organ system in the body. Children are particularly vulnerable to the adverse effects of pesticides due to biological factors and behavioral patterns that increase their unintentional exposure to pesticides (hand-to-mouth, object-to-mouth, and playing outside barefoot). Children also consume more food and liquids per kilogram of their body weight, exacerbating their pesticide encounters. Additionally, many chemicals transfer through breast milk placenta and reside in human fat stores [7,8]. Environmental exposure to pesticides is particularly harmful in early childhood, as the human brain undergoes critical stages of neurodevelopment during this period [9].

Through direct and indirect mechanisms, pesticides can lead to complications in birth outcomes, childhood asthma, pediatric cancers, and importantly, neurocognition [10]. Multiple studies have demonstrated the negative effects of pesticide exposures on various areas of child neurodevelopment, including cognitive, motor, and behavioral deficits; however the majority of studies come from high-income countries [11–13]. Evidence from low-and middle-income countries (LMICs) is critical as these countries are disproportionately affected by pesticide exposure. The World Health Organization reported that an estimated 94% of deaths caused by environmental pollution, including the use of toxic pesticides, occur in LMICs [10]. Further, pesticides use in LMICs does not require the same rigorous standards as expected in high-income countries, due to limited funding for enforcing the regulations and educating applicators [14]. For example, the EcoSalud Project, which will appear throughout articles in this review, aimed to address the lack of government intervention on high rates of acute pesticide poisoning among smallholder farmers in Ecuador, who often use cheap, harmful pesticides and may not understand toxicity labels [15]. Pesticides that are not approved or registered for use in the United States can be lawfully manufactured on U.S. soil and exported to other countries [16]. With looser regulation in LMICs, empty pesticide containers are often left behind where children may encounter them while playing [17]. Furthermore, undernutrition and nutrient deficiencies are more common in LMICs and exacerbate the burden of pesticides on children’s health [18].

These factors, in combination with decreased access to healthcare, heighten the deleterious influence of pesticides on the health of children in LMICs [19]. However, there is no succinct resource that describes the impact they may have on child neurodevelopment within LMICs. The purpose of this study is to systematically review studies examining the effects of pesticide exposure on child neurodevelopment in LMICs and to identify gaps in the literature for future policy and practice.

Materials and methods

We used The Preferred Reporting Items for Systematic review and Meta-Analysis (PRISMA) Protocols 2020 Checklist for this systematic review (see Supporting information). On April 28, 2020, the overarching review protocol on environment and neurodevelopmental outcomes was submitted to PROSPERO (CRD42020166245). In August 2020, an amendment was added to exclude high-income countries and split the review to three focused topic areas: pesticides, heavy metals, and air pollution. The associated review on heavy metals review has been published [20], and a review on air pollution is forthcoming.

Results

Overview of search results

Four separate literature searches yielded a total of 20,908 articles. After combining additional records and removing duplicates, 19,876 unique articles remained. The title and abstract screening further reduced this number to 283 full-text articles, and 31 articles met the inclusion criteria and were included in this review (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA flow diagram of included studies.

https://doi.org/10.1371/journal.pone.0324375.g001

The studies were conducted in 11 different middle-income countries, including Ecuador (n = 8), Mexico (n = 6), China (n = 4), Egypt (n = 3), South Africa (n = 3), Costa Rica (n = 2), Brazil (n = 1), Thailand (n = 1), Bangladesh (n = 1), Colombia (n = 1), and Tanzania (n = 1) [21,22] (see Supporting information). The majority of studies were conducted in upper-middle income countries [24], with the exception of five studies from Egypt [25–27], Tanzania [28], and Bangladesh [29], which are classified as lower-middle income countries [30] (Fig 2). The Ecuadorian studies selected three different subset populations from the Pichincha province [31–38], however it was unclear which proportion of participants were shared among the Ecuadorian studies. For example, four studies from the EcoSalud Project used the same methods to measure cognitive development in a single subset population, but each article presented different elements of the data. Furthermore, two of the EcoSalud studies measured exposure using questionnaires, while the other two relied on the exposure levels of residence areas (i.e., high versus low) [31–34]. A different subset population resided in the Tabacundo-Cayambe region of Ecuador and was part of two different studies [35,36]. The other subset population was derived from the ESPINA study and appears in two articles in this review [37,38]. Of the six studies conducted in Mexico [39–44], four were conducted in Morelos with the same population [39–42].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Geographic distribution of included studies.

Legend: Mexico (n = 6): Direct Observations – Torres-Sánchez (2007), (2009), (2013), Bahena-Medina (2011), Watkins (2016); Indirect Observations – Guillette (1998). Ecuador (n = 8): Direct Observations – Grandjean (2006), Harari (2010), Suarez-Lopez (2017), Espinosa da Silva (2022); Indirect Observations - Handal (2008), (2007) (a), (b), (c). Costa Rica (n = 2): Direct Observations – Lu (2009), Wendel de Joode (2016). Brazil (n = 1): Indirect Observations – Eckerman (2007). Colombia (n = 1): Indirect Observations – Benavides-Piracon (2022). Egypt (n = 3): Direct Observations – Abdel Rasoul (2008), Eadeh (2021), (2023). South Africa (n = 3): Direct Observations – Eskenazi (2018), An (2022); Indirect Observations – Chetty-Mhlanga (2021). Tanzania (n = 1): Indirect Observations – Chilipweli (2021). Bangladesh (n = 1): Direct Observations – Bliznashka (2023). Thailand (n = 1): Direct Observations – Fiedler (2015). China (n = 4): Direct Observations – Guodong (2012), Xue (2016), Zhou (2022), Chen (2022). Reprinted from GeoNames and Open Street Map under a CC BY license.

https://doi.org/10.1371/journal.pone.0324375.g002

The study characteristics and outcomes were subdivided based on methods of measuring the exposure: biomarkers (child and maternal measurements) and region of residence or questionnaire/survey (Tables 1–3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Study Characteristics of Included Articles using Biomarker Measurements.

https://doi.org/10.1371/journal.pone.0324375.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Study Characteristics of Included Articles using Questionnaire Assessments.

https://doi.org/10.1371/journal.pone.0324375.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Exposure and Cognition Assessments of Included Articles.

https://doi.org/10.1371/journal.pone.0324375.t003

Discussion

The purpose of this review was to explore the potential effects of pesticide exposure on neurodevelopment in LMICs to inform protective actions for children’s neurodevelopment. Of the 31 studies in this review, 23 reported significant associations between pesticide exposure and impaired neurocognitive development in at least one domain; three studies reported significant association with impaired behavioral problems. While the data do not provide adequate evidence to implicate a specific pesticide or susceptibility of a specific neurodevelopmental domain, our results do suggest that this is an area that should be further investigated. These results are consistent with existing literature focused primarily on high-income countries [11,12,57,58].

The associations observed between pesticide exposure and neurodevelopment appears to be highly dependent on the method of exposure measurement. Both child- and maternal-level measurements indicate a consistent significant association between pesticide exposure and neurodevelopment with tests of serum measurement of acetylcholinesterase, region of residence, and questionnaires/interviews, but not with child-level urine, maternal serum, and prenatal urine sampling. A possible explanation for the variability within results could be the sensitivity of the diversified measurement methods. One study found that organophosphate levels were more readily detected in sweat than in either blood or urine, and another found moderate reliability within prenatal, intra-individual dialkyl phosphate levels in the urine [59]. A separate study reported that the reliability of urinary analysis varied by pesticide class. Specifically, TCPy did not accurately reflect the application of chlorpyrifos or contact with its residue, while analysis of IMPy and 3-PBA was more sensitive to the application of diazon and pyrethroids [60].

Urine sampling offers unique strengths when measuring recent, acute exposures to rapidly metabolized pesticides exposures, but its varying results represent limitations for capturing the relevant timing of pesticide exposure. For example, in an included study of Thai children living in farming regions, some positive associations were surprisingly shown between exposure to pesticides and motor speed, learning, and memory [45]. Though this relationship was notable, the effect was modest and not consistently observed across domains. This study also reported that TCPy exposure in children was higher during low pesticides use season, rather than high pesticides use season [45], supporting the potential impact of timing on pesticide exposure measurements. As such, without being able to measure chronic pesticide exposure, as well as its intensity and cumulative effect, the neurotoxicity profile of pesticides through urine sampling remains incomplete. This may offer implications for the lack of relationship found between neurodevelopment and urine organophosphates in the included articles.

Although not utilized by any of the studies in this review, hair sampling has been found to be an effective method for measuring pesticide levels and the time course of exposure. Hair sampling may detect the parent compounds an individual is exposed to, rather than only the metabolites detected through urine sampling [59,60], coupled with the ability to identify multiple classes of pesticides from each sample [61]. Hair analysis is not without limitations-- exogenous (from dust and other sources) versus endogenous pesticide contamination is not always clear in pesticide measures from hair analysis. Additionally, expensive equipment to conduct hair sampling remains a barrier for LMICs to conduct, potentially explaining its absence in this review. Specific preparation procedures have been tested, but not yet universally established for hair sampling [62]. However, future studies could consider this as a potential methodology.

It is also important to note that sensitivity can be impacted by specificity. Cholinesterase activity, for example, is only reflective of organophosphate and carbamate exposures [63]. While some chemicals can be directly measured with spot urine sampling, others are metabolized too rapidly in the body to be measured directly. In these cases, metabolites are measured instead, as they break down at a stable rate. In some instances, different chemicals break down to the same metabolite, as with 3-PBA, and a secondary measurement method is needed to uncover which substance the subject was exposed to [64]. Thus, if the biological sampling method is too specific, exposures to other pesticides may go unnoticed. Overall, the findings of this review and numerous studies underscore the importance of further considerations on the best methods for appropriate pesticide detection in humans given a study’s specific research objective or design. More clarification about the benefits and challenges to various exposure measurement techniques should be clearly defined so that the interpretation of resulting data can be more assessable.

The finding that urine and serum organophosphates were rarely associated with adverse neurodevelopmental outcomes suggests that some methods of pesticide measurement, being time-specific, may affect study results. Even after pesticide exposures, cholinesterase levels increasingly normalize in the days-to-weeks following the exposure, as the body creates more acetylcholinesterase enzymes [63]. This consideration is not unique to acetylcholinesterase levels; blood and urine sampling methods are also only indicative of recent exposures. The ability of biomonitoring is dually limited by the sensitivity of the test and ability of the body to rid itself of the toxicant. Additionally, half-life must be considered in carbon-containing compounds [65]. One study reported that pesticide exposure levels measured by spot urine sampling varied more within a single child than between groups of children, demonstrating that single measurements are not reliable indicators of one’s long-term exposure to pesticides [66]. Similarly, a review including various analyses of urinary pesticide metabolites concluded that metabolite levels fluctuate greatly over time, which may weaken their indication of true pesticide exposures with smaller sample sizes [60]. Other research regarding the validity of urine sampling indicates a high degree of intra-individual variability of pesticide measures. This highlights the need for measurements at multiple points in time to accurately portray exposure levels [67,68]. Therefore, subjects who are tested beyond a certain window of exposure may appear as though they were never exposed or exposed at a much lower level than is accurate. Any study utilizing biomonitoring is limited by the understanding of temporal exposure considerations for each pesticide class. This is an important area of concern in pesticide exposure research as approximately half of the studies with biological measurements in this review measured intraindividual pesticide exposure more than once.

All nine studies using survey or questionnaire measurements of exposure levels had uniform conclusions of a negative impact on neurodevelopment. Surveys show strength for determining pesticide exposure across the life course; these studies, however, were limited, as there was no direct measure of exposure. Furthermore, surveys and location-based studies are at risk for recall and ecological bias, respectively. The Improving Exposure Assessment Methodologies for Epidemiological Studies on Pesticides initiative-- a collaborative effort across the Institute of Occupational Medicine, the Institute of Risk Assessment Sciences at Utrecht University, and the Centre for Occupational and Environmental Health of the University of Manchester-- is attempting to document the impact of recall bias on pesticide exposure classification, while also noting the reliability of various surrogate pesticide exposure measurement methods [69]. Two included studies suggested that recall bias is affected by time and the item being recalled [70,71]. Further results may aid in informing the direction of future studies analyzing the impact of pesticides on human health.

The variability of results between different techniques in this review illustrate that multiple methods of measuring pesticide exposure (i.e., biological markers, region of residence, survey questionnaire) may be applied to understand and quantify the exposure. Researchers must consider the benefits and limitations of each type of pesticide analysis according to the exposure and question of interest. Biomonitoring allows for direct and objective measures of exposure that the subject may not be aware of, but also introduces various challenging factors, including specificity, sensitivity, timing, pathway of exposure, and availability. Non-biomarker methods of pesticide measurement, such as surveys or proximity, provide different opportunities to characterize a participant’s exposures, including the timeframe and sources of exposure. Using region of residence as a method of pesticide quantification can provide information on community-based exposures. These methods, however, do not allow for confirmation of exposure. One review, which investigated various methods of quantifying pesticide exposure, underscored the importance of considering multiple exposure analysis methods [72]. The results recommend utilizing multiple biomarker modalities in future research so that less weight is allotted to indirect measures when policymaking for pesticide exposure prevention in children [72].

Many well-identified risk factors for adverse neurodevelopmental outcomes in LMICs are recognized and should be considered when designing studies in this research area. These include maternal education, parental habits such as smoking, nutrition, socioeconomic status, home environment, and other environmental exposures, such as cadmium and mercury, and are known to influence neurodevelopmental outcomes, which can have significant effects on study results [73–78]. The included articles in this review tended to evaluate nutrition (largely via anthropometric measures) and socio-economic status [28,33,47], which appeared to show positive interaction effects when mitigating the relationship between pesticide exposures and neurodevelopmental outcomes. Notably, in three studies, the nature of prenatal or current maternal employment was closely associated pesticide exposures [28,31,34]. From the data available in these studies, the difference in neurodevelopmental outcomes among children exposed to pesticides through pre- versus postnatal maternal employment could be attributable to the positive benefits associated with maternal employment, such as greater income, rather than solely the exposure itself. This dichotomy between pre- and postnatal maternal employment highlights the complex relationship between sociodemographic factors and the timing of pesticide exposures, as it is impossible to discern whether the difference in neurodevelopmental outcomes resulted from the timing of the pesticide exposure (e.g., children exposed prenatally rather than postnatally are more likely to suffer neurodevelopmental deficits). It is likely that both the timing of exposure and sociodemographic factors impact neurodevelopmental outcomes to some degree, but the question remains: do the positive benefits associated with maternal employment – no matter the industry – outweigh the negative neurodevelopmental impacts associated with maternal pesticide exposure? Answering this question is important in establishing guidelines that will protect children’s neurodevelopment.

This review highlights the limited evidence base derived from LMIC settings that addresses pesticide exposure in the early life course. One major gap is the adequate inclusion of controls in some studies and lack of standardized norming data for many of the study populations. It was specifically noted that this data was not available in Ecuador [33]. Without this data, it is difficult to discern whether the children tested in these studies had neurodevelopmental delays or would differ from unexposed peers that were otherwise similar. This is an inherent problem when conducting studies in LMICs, where less funding is available for creating standardized developmental scales for specific populations.

Another gap appears in infant pesticide exposure analyses. Currently, exposure data for infants is based on either maternal pesticide exposure measurement during gestation or subjective maternal recall. For example, the results of the maternal serum studies in Mexico across varying ages showed that the impact of prenatal pesticide exposure on neurodevelopment could be dependent on the age of the subject and the specific tests used [39–42]. Therefore, these exposure quantifications may not accurately reflect infant pesticide exposure levels, depending on extent of transfer to placenta and/or occurrence in breastmilk, thereby compromising true associations with neurodevelopment. Furthermore, some of the studies did not specify when the maternal pesticide measurements were taken (first trimester, second trimester, etc.). As previously discussed, the timing of pesticide measurements may significantly impact the results. Additionally, changes in DNA transcription and brain development occur more rapidly during gestation than any other time of life [79,80], and neurodevelopmental outcomes appear to be highly dependent on the timing of prenatal stressors [81]. Ideally, both maternal and infant measurements would be taken within documented timeframes to determine infants’ exposure to pesticides with more precision.

There is a notable gap in geographical areas where children were examined for the impacts of pesticide exposure on neurodevelopment. All but five of the studies (conducted in Tanzania, Egypt, and Bangladesh) included in this review were conducted in upper-middle income countries [25–29], and no research was gathered on children in low-income countries. As children in low-income countries are at the greatest risk for developmental impacts secondary to pesticide exposures, it is imperative that more research be conducted in these areas. Additionally, most of the research was concentrated in identical or very similar regions of two countries— Ecuador and Mexico. To effectively understand the impacts of pesticide exposure on neurodevelopment, a larger, more diversified body of research is needed.

The current study is limited in that variability within research designs prevented comparative analysis between studies. Ideally, this literature review would have allowed for insights on exposure impacts by pesticide type, age of exposure, age developmental analysis, and the type of developmental domain in question. Unfortunately, no consistent patterns were noted in any of these subgroupings, secondary to large differences in the design protocols such as: (a) study design (cross-sectional vs. cohort), (b) age of subjects, (c) prenatal vs. postnatal measurements, (d) method of measuring exposure, (e) developmental test utilized, (f) type of statistical test utilized, and (g) whether adjustments were made for confounding variables. These distinctions also prevented a meta-analysis from being performed.

To develop policy and practice that protects children from neurodevelopmental deficits, we need well-designed studies that represent the populations of children exposed throughout LMICs. Careful attention to exposure assessment methods and timing of exposure assessment to reflect and inform key research questions, use of well-validated outcome measures conducted by trained data collectors, and inclusion of important covariates are further needed.

Conclusion

In this review, exposure to organochlorines, carbamates, chlorpyrifos, and fungicides were typically associated with worse outcomes in executive functioning, cognition, motor development, and behavior for children, most commonly in Latin America, particularly Ecuador, and sub-Saharan Africa. However, high variability in our findings reflect inadequate data to discern the impact of specific pesticide types and classes, due to limited study areas and variable validity of exposure assessment. Thus, this review exposes limitations in the current evidence base of LMIC-based studies regarding the neurodevelopmental impact of pesticide exposures. LMICs will benefit from access to robust, reliable exposure assessment methods, such as biomonitoring to understand population-level exposures and trends. In terms of policy and practice, this review provides evidence that national and/or local efforts to mitigate harm from well-recognized pesticide exposures in LMICs are likely to have future benefits for the development, health, and well-being of countless children currently impacted by acute or chronic poisoning linked with pesticides. Future coordination in testing and measurement efforts would advance our knowledge in this field.

Supporting information