Pesticide exposure and child growth in low- and middle-income countries: A systematic review

Background In low- and middle-income countries (LMICs), pesticides are widely used in agricultural and residential settings. Little is known about how pesticides affect child growth. Objectives To systematically review and synthesise the evidence on the associations between pesticide exposure and adverse birth outcomes and/or impaired postnatal growth in children up to 5 years of age in LMICs. Methods We searched 10 databases from inception through November 2021. We included cohort and cross-sectional studies investigating associations between self-reported or measured prenatal or postnatal pesticide exposure and child growth (postnatal child linear/ponderal growth, and/or birth outcomes). Two researchers screened studies, extracted data, and assessed certainty using GRADE. The protocol was preregistered with PROSPERO (CRD42021292919). Results Of 939 records retrieved, 31 studies met inclusion criteria (11 cohort, 20 cross-sectional). All studies assessed prenatal exposure. Twenty-four studies reported on birth weight. Four found positive associations with organochlorines (0.01–0.25 standardised mean difference (SMD)) and two found negative associations (–0.009 SMD to –55 g). Negative associations with organophosphates (–170 g, n = 1) and pyrethroids (–97 to –233 g, n = 2) were also documented. Two (out of 15) studies reporting on birth length found positive associations with organochlorines (0.21–0.25 SMD) and one found negative associations (–0.25 to –0.32 SMD). Organophosphate exposure was negatively associated with birth length (–0.37 cm, n = 1). Organophosphate exposure was also associated with higher risk/prevalence of low birth weight (2 out of nine studies) and preterm birth (2 out of six studies). Certainty of the evidence was “very low” for all outcomes. Conclusion The limited literature from LMICs shows inconclusive associations between prenatal pesticide exposure, child growth, and birth outcomes. Studies with accurate quantitative data on exposure to commonly used pesticides in LMICs using consistent methodologies in comparable populations are needed to better understand how pesticides influence child growth.


Introduction
Pesticides are widely used in agricultural and residential settings globally (FAO, 2021; World Health Organization and Food and Agriculture Organization of the United Nations, 2019). In some low-and middle-income countries (LMICs), like those in sub-Saharan Africa, South and Southeast Asia, pesticide use per cropland area remains lower than in high-income countries (FAO, 2021). However, pesticide exposure can be substantially higher in LMICs than high-income countries and populations may be exposed to more toxic chemicals in these settings (World Health Organization and Food and Agriculture Organization of the United Nations, 2019; Pan International, 2021). This stems from: (1) the fact that high-income countries continue to export pesticides that have been banned for use in their own countries; (2) fewer regulations in some LMICs (in place and/or properly enforced); and (3) unsafe pesticide handling, storage, application, and disposal practices which are more commonly used in many LMICs than in high-income countries (World Health Organization and Food and Agriculture Organization of the United Nations, 2019; Pan International, 2021). Dietary intake of pesticide residues from contaminated foods and pesticide use for water chlorination are also among the primary routes of exposure (Mostafalou and Abdollahi, 2017).
Children are especially vulnerable to the harmful effects of pesticides due to their increased exposure relative to their body weight, behavioural factors (they spend more time on the ground, crawling, touching objects, and putting their fingers, toys, and other objects in their mouths), and the fact that their brains and organs are still developing (Eskenazi et al., 1999). Human and animal studies show that several mechanisms explain the effects of pesticide exposure on children, and specifically on child growth outcomes in early life. Pesticides can lead to endocrine, hormonal, thyroid, and placental disruption in pregnancy (Mustieles et al., 2017;Silvia et al., 2020;Chevrier et al., 2008;Milczarek et al., 2016;Cecchi et al., 2021) and to thyroid disruption in neonates (Sun et al., 2022). Pesticides can also influence glucose metabolism (Debost-Legrand et al., 2016), immune regulation (Neta et al., 2011;Tyagi et al., 2016), and the bacterial composition of the gut microbiome (Yuan et al., 2019) as well as contribute to environmental enteric dysfunction (EED), a subclinical condition resulting in increased gut permeability and impaired nutrient absorption (Mapesa et al., 2016).
Studies conducted in high-income countries (primarily the United States) have consistently shown that prenatal exposure to organochlorines is associated with adverse birth outcomes, including preterm birth, small-for-gestational age, smaller head circumference, and shorter femur length (Longnecker et al., 2001;Ouidir et al., 2020). In contrast, evidence suggests that prenatal exposure to organophosphate insecticides and herbicides is not associated with adverse birth outcomes (Reiss et al., 2015;Shirangi et al., 2011;de Araujo et al., 2016). With respect to associations between prenatal pesticide exposure and birth weight and length, evidence is less conclusive. Some studies have linked prenatal exposure to organochlorines with increased birth weight, overweight and elevated body mass index (BMI) in infancy and childhood (Mendez et al., 2011;Pinos et al., 2021;Stratakis et al., 2022), while other studies have shown that prenatal organochlorine exposure is associated with lower birth weight (Wolff et al., 2007;Govarts et al., 2012). Similarly, the literature reports no consistent associations of organochlorines and birth length (Wolff et al., 2007;Kalloo et al., 2020). With respect to organophosphates, a recent meta-analysis which pooled 10 studies (eight from high-income countries and two from LMICs) found that prenatal organophosphate exposure was not associated with birth weight (Khoshhali et al., 2020). Although some additional studies have shown null associations between prenatal organophosphate exposure and birth weight, corroborating the conclusion from this meta-analysis (Balalian et al., 2021), others have shown that prenatal exposure to organophosphates is associated with lower birth weight (Rauch et al., 2012;Gemmill et al., 2013). Whether prenatal organophosphate exposure is associated with birth length remains unclear as well, with studies showing positive (Eskenazi et al., 2004), negative (Rauch et al., 2012;Gemmill et al., 2013), and the null (Khoshhali et al., 2020) associations with birth length.
Though more limited, evidence on postnatal exposure suggests that postnatal pesticide exposure can also be harmful for child growth. Another recent systematic review assessing prenatal and postnatal pesticide exposure and obesity in children included eight studies of postnatal pesticide exposure (five in high-income countries and three in LMICs) (Pinos et al., 2021). This systematic review found that postnatal exposure to organophosphates was associated with higher BMI Z-scores in boys at four years of age and higher waist circumference and obesity in children 6-18 years of age. Postnatal exposure to organochlorines was not associated with weight gain during childhood and adolescence, but was associated with lower BMI Z-score in boys (Pinos et al., 2021). Further, evidence suggests that the effects of prenatal pesticide exposure may manifest later in infancy and childhood. One pooled analysis of seven European cohorts showed that prenatal exposure to organochlorines was associated with increased child growth between birth and 24 months of age (Iszatt et al., 2015), whereas one systematic review showed that prenatal exposure to endocrine-disrupting chemicals (including organochlorines) was associated with obesity later in life (Tang-Péronard et al., 2011).
In contrast to this large body of evidence in high-income countries, less is known about the relationship between prenatal and postnatal pesticide exposure and child growth in LMICs. Therefore, the research question we aimed to answer was: what is the effect of pesticide exposure on child growth in children <5 years of age living in LMICs? With this objective in mind, we systematically reviewed and synthesised the evidence on the associations between pesticide exposure and adverse birth outcomes and/or impaired postnatal growth in children up to 5 years of age in LMICs. Given the limited evidence from LMICs, we kept the review broad: considering any active ingredients or mixtures of active ingredients and a wide range of child growth outcomes at birth and postnatally.

Search strategy, eligibility criteria, and selection process
This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Guidelines (Moher et al., 2009). It was preregistered with PROSPERO prior to commencing (CRD42021292919). The PECO (populations, exposures, comparators, and outcomes) statement for this review was (Morgan et al., 2018): • Population: children <5 years of age in LMICs.
• Exposure: self-reported or measured prenatal (at any time during pregnancy) or postnatal exposure to pesticides. • Comparator: a less exposed reference population (continuously, high/low exposure groups, or exposed/unexposed groups). • Outcomes: any anthropometric measures of child linear or ponderal growth, and/or any of six birth outcomes (birth weight, birth length, low birth weight, preterm birth, small-for-gestational age, large-forgestational age).
We searched 10 electronic databases from inception to November 2021, with no language restriction: PubMed, Cochrane Library, Embase, Scopus, LILACS, Web of Science, CAB abstracts, Global Health (CABI), Global Index Medicus, and SciELO. The selection of databases was informed by databases searched in prior systematic reviews on pesticide exposure and child neurodevelopment (Sapbamrer and Hongsibsong, 2019;González-Alzaga et al., 2014;Muñoz-Quezada et al., 2013) and through consultations with a research librarian. The inclusion of 10 databases was to ensure our review was extensive and comprehensive. Search terms included medical subject headings (MeSH), keywords, and free text words along the following themes: child, pesticides, child growth, and birth outcomes. Search terms were combined using the Boolean operators "AND" and "OR". The search strategy was developed by LB and LMJ. The search strategy was informed by search terms and keywords used in prior systematic reviews on pesticide exposure and child neurodevelopment (Sapbamrer and Hongsibsong, 2019;González-Alzaga et al., 2014;Muñoz-Quezada et al., 2013) and through consultations with a research librarian. The complete search strategy as executed in each database is available in Supplemental Table 1. All databases were searched on December 1, 2021. Searched fields were title, abstract or author keywords. Reference lists of extracted articles were reviewed for any additional studies that may have been missed by the search.
Peer-reviewed articles were included if they met the following criteria: (1) conducted in a LMIC (based on The World Bank categorisation at the time the study was conducted); (2) assessed children <5 years of age; (3) evaluated self-reported or measured prenatal (at any time during pregnancy) or postnatal exposure to pesticides; (4) measured at least one anthropometric measure of child linear or ponderal growth, and/or one of six birth outcomes (see section 2.2). We included prospective cohort studies and cross-sectional studies, and excluded animal studies, case-control studies, simulation studies, case reports, case studies, opinions, editorials, commentaries, letters, conference abstracts, ecological studies, reviews, and systematic reviews. We also excluded studies conducted in high-income countries, assessing children >5 years of age, and focusing solely on insecticide-treated bednets for malaria prevention.
Two reviewers (LB and LMJ) independently screened all titles and abstracts for inclusion using Covidence (n.d). Disagreements were resolved through discussion, and input from a third reviewer (AR) as needed. Full texts of eligible studies were reviewed for inclusion and data extraction. One reviewer (LB) extracted information on: (1) publication details (author names, journal, year of publication), (2) study meta-data (location, setting, design, population, year(s) of data collection, duration of study, inclusion and exclusion criteria, funding source); (3) participant characteristics (number, age, sex, education, occupation); (4) pesticide exposure (types of pesticides, method of exposure assessment, time point of exposure assessment); (5) outcomes; and (6) methodological quality (study type, size, confounding and attempts to correct for confounding, list of confounding factors; and funding sources). Data extraction was reviewed by two additional reviewers (AR and LMJ). Any disagreements regarding the extracted data were resolved through discussion between the three reviewers (LB, AR, and LMJ). When information was unclear or unavailable, we did not attempt to contact the authors of the original study.

Outcome measures
The primary outcomes were any anthropometric measures of child linear or ponderal growth measured at any time point up to 5 years of age: length/height-for-age Z-score, weight-for-age Z-score, weight-forlength/height Z-score, BMI Z-score, and stunting, underweight, wasting, and overweight as defined by individual articles. The secondary outcomes were six birth outcomes: birth weight, birth length, low birth weight, preterm birth, small-for-gestational age, and large-forgestational age as defined by individual articles.

Data analysis
Characteristics of the studies included in the review were summarised narratively by exposure characteristics (types of pesticides, method of exposure assessment, time point of exposure assessment), types of outcomes assessed, and study characteristics (study design, study year, region). Data were summarised narratively using two groupings: outcomes assessed and pesticides assessed. First, we grouped studies by the frequency of outcomes they assessed. Then, by outcome, we grouped studies by the frequency of pesticide types they assessed: organochlorines, organophosphates, pyrethroids, carbamates, and chlorophenols (non-specific biomarkers of organochlorines). Where both unadjusted and adjusted estimates were reported, we summarised evidence on adjusted estimates. For continuous outcomes, we reported mean differences (MD) and standardised mean differences (SMD). For binary outcomes, we reported odds ratios (ORs), relative risks (RR), and proportions in exposed and unexposed groups. We were unable to extract all necessary information to convert and standardise effect estimates across articles. We therefore reported effect estimates as reported by individual articles. Meta-analysis was not conducted due to the heterogeneity of studies: <2 eligible studies were conducted on the same population, assessed the same pesticides, assessed exposure over the same period, and included the same outcomes.

Risk of bias and quality of evidence assessment
Two reviewers (LB and LMJ) independently assessed the study risk of bias and quality of the evidence for outcomes reported in ≥4 studies using the GRADE approach (GRADE Working Group, 2013). The GRADE approach rates the certainty of the evidence for a given outcome based on five criteria: risk of bias, inconsistency, indirectness, imprecision, and other considerations (publication bias, large effect, plausible confounding, and dose-response gradient). The risk of bias assessment was first conducted at the study-level and then at the outcome-level. Under each criterion, the quality of the evidence is considered "not serious" (no study limitations) as a baseline, and can then be downgraded by one level to "serious" or two levels to "very serious" depending on the study limitations (GRADE Working Group, 2013). One limitation of the GRADE approach is that observational studies start as "low-quality evidence" (Guyatt et al., 2011). However, not all observational studies are of low-quality (Viswanathan et al., 2018). Given these considerations and because all the included studies in this review were observational, we assigned an initial risk of bias rating of "not serious" to all studies. We then downgraded studies for limitations specific to observational studies: failure to develop and apply appropriate eligibility criteria, flawed measurement of exposure and outcome, failure to adequately control for confounding, and incomplete follow-up (Guyatt et al., 2011). No other modifications were made to the GRADE approach. Disagreements were resolved through discussion and input from a third reviewer (AR) as needed. We used the GRADEpro Software to conduct the GRADE assessment and to prepare the "Summary of findings" table (GRADEpro GDT, 2021).
One study assessed chlorophenols, non-specific biomarkers of organochlorines .
Four of the eight studies assessing self-reported exposure did not specify the types of pesticides used Wang et al., 2012a;Moreno-Banda et al., 2009;Rahimi et al., 2020). One study assessed both self-reported pesticide exposure and urinary metabolites, and did not specify the types of pesticides used based on self-report . One study collected blood samples but used them to assess acetylcholinesterase and butyrylcholinesterase activity and hormone levels. The types of pesticides used were specified as insecticides of the organophosphate and carbamate families, based on commonly used pesticides . The remaining two studies were conducted in Poland, which was an LMIC at the time the studies were conducted. In one of these studies, which assessed residential exposure, the most frequently reported pesticides were organophosphorus compounds, pyrethroids, phenoxyacetic acid derivatives, and benzene thiosulphonate derivatives (Hanke et al., 2003). In the second study, which assessed occupational exposure at greenhouses, those responsible for chemical protection at the greenhouses reported the use of 17 pesticides primarily used as insecticides, fungicides, and acaricides .
Of the 12 studies that assessed associations between organochlorines and birth weight, nine found significant associations. Four studies found significant positive associations with DDT, DDE, and DDD ranging from 0.008 to 0.25 SMD Arrebola et al., 2016;Steinholt et al., 2020;Bravo et al., 2019). One study found a significant negative association with total DDT of − 0.009 g , and two studies found significant negative associations with HCH and its isomers ranging from − 5.81 to − 55.14 g   (Table 2). One study showed that higher concentrations of α-HCH, DDE isomers, and dieldrin were significantly correlated with lower birth weight (Dwivedi et al., 2022). Another study observed significantly different birth weight by tertile of prenatal concentrations of β-HCH, but no indication of dose response (Gladen et al., 2003). Three studies found no significant associations between prenatal organochlorine exposure and birth weight (Fang et al., 2019;Guo et al., 2014;Abdel Hamid et al., 2020;Van Tung et al., 2016). Four of these 12 studies which assessed the associations between organochlorines and birth weight controlled for maternal BMI at the time of assessment (Guo et al., 2014;Xu et al., 2017;Arrebola et al., 2016;Steinholt et al., 2020), three controlled for pre-pregnancy BMI (Fang et al., 2019;Yang et al., 2020;Gladen et al., 2003), and none controlled for dietary intake.
Two of the studies that assessed associations between pyrethroids and birth weight found significant negative associations ranging from − 96.76 to − 233.3 g Hanke et al., 2003) and two found no significant associations Jaacks et al., 2019) (Table 2). None of the five studies that assessed associations between carbamates and birth weight found significant associations Cecchi et al., 2021;Zhang et al., 2018;Yang et al., 2020;Hanke et al., 2003). The study that assessed chlorophenols non-specific to organochlorines found that higher concentrations of 2,4,6-trichlorophenol (TCP) and pentachlorophenol (PCP) were associated with − 30 g and − 37 g lower birth weight, respectively   (Table 2).
Of the seven studies that assessed the associations between organochlorines and birth length, one study found a significant positive association of 0.21 SMD with higher levels of p,p ′ -DDT (Bravo et al., 2019). Another study found significant negative associations with o,p ′ -DDE and p,p ′ -DDD ranging from − 0.25 to − 0.32 SMD and a 0.25 SMD significant positive association with aldrin . The remaining five studies found no significant associations between prenatal exposure to organochlorines and birth length (Fang et al., 2019;Xu et al., 2017;Arrebola et al., 2016;Abdel Hamid et al., 2020).
Four of the five studies that assessed prenatal organophosphate exposure found no significant associations with birth length Wang et al., 2012b;Liu et al., 2016;Naksen et al., 2015).

Table 2
Summary of findings on the associations between prenatal pesticide exposure and the four most frequently reported outcomes (birth weight, birth length, low birth weight, and preterm birth). Abbreviations: OR; odds ratio; pp, percentage points; Refs, references; RR, relative risk; SMD, standardised mean difference.

Most frequently reported outcomes in included studies
One study found − 0.372 cm lower birth length among those in residential proximity to fruit croplands (where organophosphates and carbamates were frequently used) relative to those not in residential proximity to fruit croplands . Another study found no association between prenatal carbofuranphenol levels and birth length . The one study that assessed associations between pyrethroid metabolites and birth length found no significant associations . The study that assessed chlorophenols non-specific to organochlorines also found no significant associations with birth length .

Associations with low birth weight
Of the nine articles reporting on low birth weight, two assessed associations with organochlorines Van Tung et al., 2016), five with organophosphates Cecchi et al., 2021;Moreno-Banda et al., 2009;Jurewicz et al., 2005;Jaacks et al., 2019), two with pyrethroids Jaacks et al., 2019), two with carbamates Cecchi et al., 2021), and two reporting on occupational exposure did not specify the types of pesticides used Rahimi et al., 2020).
None of the studies that assessed the associations between organochlorines or pyrethroids and low birth weight found significant associations Toichuev et al., 2018;Jaacks et al., 2019;Van Tung et al., 2016). Of the five studies that assessed associations with organophosphates, one found 2.13 (95% CI 1.12, 4.08) higher risk of low birth weight among women with detectable prenatal levels of IMPY relative to women with undetectable levels . Another study reporting on residential exposure to organophosphates and carbamates found that 6% of women assessed during the spraying season had low birth weight infants whereas 0% of women assessed during the non-spraying season had low birth weight infants . The remaining studies assessing organophosphates or carbamates found no significant associations between prenatal exposure and low birth weight Moreno-Banda et al., 2009;Jurewicz et al., 2005). One of the studies reporting on occupational exposure found significantly higher prevalence of low birth weight (11.6%) among greenhouse workers compared to housewives (3.2%) . However, the second study found no significant associations between greenhouse work during pregnancy and birth length .

Associations with pre-term birth
Of the six studies reporting on preterm birth, one assessed associations with organochlorines (Cioroiu et al., 2010), four with organophosphates Cecchi et al., 2021;Jurewicz et al., 2005;Jaacks et al., 2019), two with pyrethroids Jaacks et al., 2019), two with carbamates Cecchi et al., 2021), and one reporting on occupational exposure did not specify the types of pesticides used .
The study assessing associations between organochlorines and preterm birth found significantly higher prenatal levels of total DDT in women with preterm births (715 ± 952 ng/g) than in women with normal births (687 ± 1393 ng/g), but lower prenatal levels of total HCH in women with preterm births (174 ± 210 ng/g) than in women with normal births (207 ± 337 ng/g) (Cioroiu et al., 2010). One of the four studies assessing associations with organophosphates found significantly higher risk of preterm birth among women with higher prenatal concentrations of 4-nitrophenol: RR 3.57 (95% CI 1.65, 7.73) comparing quartile 3 to quartile 1, RR 3.57 (95% CI 1.65, 7.73) comparing quartile 4 to quartile 1; and RR 1.44 (95% CI 1.17, 1.78) per μg/g creatinine adjusted higher 4-nitrophenol . Another study reporting on residential exposure to organophosphates and carbamates found that 6% of women assessed during the spraying season had preterm births whereas 0% of women assessed during the non-spraying season had preterm births . The remaining studies assessing organophosphates, pyrethroids, or carbamates found no significant associations between prenatal exposure and preterm birth Jurewicz et al., 2005;Jaacks et al., 2019;Rahimi et al., 2020).

Associations with other outcomes
Of the three studies that assessed associations with small-forgestational age, one found that higher prenatal concentrations of 4nitrophenol were associated with 4-times higher risk of a small-forgestational age birth, whereas higher prenatal concentrations of TCPY and 3-PBA were associated with 0.12-0.75 times significantly lower risk of a small-for-gestational age birth, respectively . The second found that residential proximity to fruit croplands during pregnancy was not associated with either small-or large-for-gestational age births . The third study found no small-for-gestational births either among women assessed during the spraying season or those assessed during the non-spraying season . With respect to anthropometry at birth, one study showed that higher concentrations of p,p'-DDE in each trimester of pregnancy were not associated with BMI Z-score, length-for-age Z-score, weight-for-age Z-score, or weight-for-length Z-score . Lastly, one study found that working in a greenhouse was not associated with low birth weight in term pregnancies .

Associations with outcomes later in infancy
Of the four studies that reported outcomes later in infancy, two studies found that prenatal exposure to organochlorines was not associated with child height or weight at 8-9 or 12-14 weeks of age , or with child BMI or height at 16 months of age . One study found that higher prenatal concentrations of the organophosphate 4-nitrophenol were significantly associated with lower length-for-age Z-score (− 0.50 ± 0.25 SD comparing quantile 3 to quantile 1) and lower weight-for-age Z-score (with an increasing dose response relationship with higher concentrations ranging from − 0.35 ± 0.17 to − 0.66 ± 0.17 SD) at 1 year of age . Prenatal levels of 4-nitrophenol showed no consistent associations with weight-for-length Z-score at 1 year of age with a significant positive association comparing quantile 2 to quantile 1, but a negative association comparing quantile 4 to quantile 1. At 2 years of age, higher prenatal concentrations of the organophosphate metabolite TCPY were associated with significantly lower length-for-age Z-score and higher prenatal concentrations of 4-nitophenol were significantly associated with lower weight-for-length Z-score. Prenatal pesticide exposure was not associated with stunting at 1 or 2 years of age in that study . Finally, one study found that higher prenatal exposure to a p,p ′ -DDE and total DDT were associated with significantly lower BMI Z-score at 6 months of age, while higher prenatal exposure to p,p ′ -DDT was associated with higher BMI Z-score at 12 months of age . In contrast, higher prenatal concentrations of HCH isomers were associated with significantly higher BMI Z-score at 6, 12, and 24 months of age .

Subgroup analysis
Of the 31 studies included in this review, six conducted subgroup analyses: five by child gender (Fang et al., 2019;Zhang et al., 2018;Guo et al., 2016;Wang et al., 2012b;Liu et al., 2016) and one by paraoxonase 1 phenotype . Fang et al., 2019 found no associations between prenatal exposure to organochlorines and birth weight or length in the full sample; however, in sub-group analysis they found that higher prenatal concentrations of HCH isomers were associated with − 34.48 to − 28.61 g significantly lower birth weight among boys (Fang et al., 2019). Guo et al., 2016 found no association between prenatal exposure to non-specific organochlorines and birth length, but a significant negative association between higher prenatal PCP concentrations and birth length among boys − 0.23 (95% CI -0.41, − 0.01). Sub-group analysis also revealed that the observed associations between higher prenatal concentrations of 2,4,6-TCP and PCP and lower birth weight were only significant among boys and not girls . The remaining three studies found no differences in the associations between prenatal exposure to organophosphates or carbamates and birth weight and length in girls versus boys Wang et al., 2012b;Liu et al., 2016). Naksen et al., 2015 examined the associations between prenatal exposure to organophosphates and birth weight and length by paraoxonase 1 phenotype and found no differences .

Timing of exposure assessment
Of the 31 articles included in this review, 23 assessed exposure at a single time point: 16 at delivery/birth Fang et al., 2019;Zhang et al., 2018;Guo et al., 2014Guo et al., , 2016Wang et al., 2012b;Xu et al., 2017;Yang et al., 2020Yang et al., , 2021Liu et al., 2016;Cupul-Uicab et al., 2010;Dwivedi et al., 2022;Toichuev et al., 2018;Abdel Hamid et al., 2020;Gladen et al., 2003), five during the third trimester of pregnancy Arrebola et al., 2016;Steinholt et al., 2020;Cioroiu et al., 2010;Bravo et al., 2019), and two at the first antenatal care visit when the pregnancy was confirmed (Hanke et al., 2003;Jaacks et al., 2019). Of the remaining eight studies, three assessed exposure once in each trimester of pregnancy Garced et al., 2012;Naksen et al., 2015) and five assessed exposure over the entire pregnancy and preconception period by using occupation or residential proximity as a proxy for exposure Wang et al., 2012a;Jurewicz et al., 2005;Rahimi et al., 2020;Van Tung et al., 2016). None of the studies assessing exposure in each trimester of pregnancy found differences by trimester Garced et al., 2012;Naksen et al., 2015). Of the five studies using proxy measures covering pregnancy and preconception, three found no significant associations with child growth outcomes. One study found significantly lower birth length among exposed than unexposed children (− 0.61 ± 1.16 vs − 0.24 ± 1.04 respectively) , while another found significantly higher prevalence of preterm birth among greenhouse workers than housewives (11.9% vs 6.1%, respectively) .

Certainty of the evidence assessment
The certainty of the evidence was "very low" for all outcomes reported in ≥4 studies (e.g., birth weight, birth length, low birth weight, and preterm birth) due to downgrades for inconsistency, indirectness, and imprecision (Supplemental Table 2).

Discussion
In this systematic review, we sought to assess the effect of pesticide exposure on birth outcomes and/or postnatal growth in children <5 years of age living in LMICs. Although we searched 10 databases and allowed for broad definitions of the exposures and outcomes, our search strategy identified only 653 unique records, highlighting the limited literature on the topic in both high-income countries and LMICs. We included 31 studies from LMICs assessing the associations between prenatal pesticide exposure and child growth outcomes at birth or up to age 5 years. We found no consistent associations between prenatal pesticide exposure and birth weight and birth length for all pesticide classes, including associations for subgroups by child sex. Prenatal exposure to organochlorines appeared to be associated with birth weight, but the direction of this association remained unclear with studies demonstrating both positive and negative associations. Prenatal exposure to pyrethroids may be associated with lower birth weight. Similarly, findings with respect to birth length were not consistent, with positive, negative, and null associations observed in the literature. In addition, we found no consistent evidence of an association between prenatal pesticide exposure and low birth weight and preterm birth. Associations with growth outcomes later in infancy (including the primary outcomes of anthropometric measures of linear and ponderal growth) were also inconclusive given the limited number of studies. The certainty of the evidence was "very low" for all outcomes.
Our findings extend existing evidence on the associations between pesticide exposure and child growth in two key ways. First, to the best of our knowledge, this is the first systematic review to focus only on LMICs. Previous systematic reviews have included both high-income countries and LMICs without comparing and contrasting findings across these two groups (Khoshhali et al., 2020;Tang-Péronard et al., 2011). This distinction is important given that approved and commonly used pesticides vary between high-income countries and LMICs (Donley, 2019a). Second, whereas prior systematic reviews have focused on a single pesticides, e.g., glyphosate (de Araujo et al., 2016), a specific outcome, e.g., obesity (Pinos et al., 2021), or a single exposure period, e.g., prenatally (Khoshhali et al., 2020), we considered all types of pesticides, a wide range of child growth and birth outcomes, and both prenatal and postnatal exposure. Thus, we conducted a broader literature review than prior studies.
Considerable evidence shows that several mechanisms underlie the associations between pesticide exposure and child growth, including endocrine, hormonal, thyroid, and placental disruption in pregnancy (Mustieles et al., 2017;Silvia et al., 2020;Chevrier et al., 2008;Milczarek et al., 2016;Cecchi et al., 2021), thyroid disruption in the neonatal period (Sun et al., 2022), glucose metabolism disruption (Debost- Legrand et al., 2016), immune dysregulation (Neta et al., 2011;Tyagi et al., 2016), disruption of the gut microbiome (Yuan et al., 2019), and increased gut permeability and impaired nutrient absorption (Mapesa et al., 2016). Despite this considerable literature, empirical evidence on the precise mechanisms through which pesticides influence child growth in LMICs is lacking. Of the 31 studies included in this review, only one conducted mediation analysis to formally assess mechanisms and found that prenatal exposure to organochlorines reduced birth weight by disrupting thyroid hormone metabolism and glyceraldehyde metabolism . Two more studies assessed mechanisms without conducting mediation analysis and found that prenatal pesticide exposure was associated with more pregnancy complications, including premature rupture of membranes, intrauterine growth retardation, and threatened miscarriage Cecchi et al., 2021). More work is needed to understand how different mechanisms influence specific child growth outcomes at birth and later in infancy. Mediation models can help quantify these mechanisms, build the empirical base on mechanisms, and inform the design and targeting of interventions to reduce any potential adverse effects of prenatal pesticide exposure on child growth.
Given these plausible biological and physiological mechanisms through which pesticide exposure during pregnancy can influence child growth, the lack of consistent associations in LMICs is likely due to two important limitations of the current literature. One major limitation is the incomplete toxicology picture. Most studies included in this review assessed a few active ingredients and some did not assess specific pesticides but used residential location or occupation as a proxy for pesticide exposure. A recent study identified 239 active pesticide ingredients approved for use in China (Donley, 2019a), the country with the most studies included in this review, which evaluated a small fraction of these approved active ingredients. Moreover, people are not exposed to a single active ingredient but rather to a complex mixture of active ingredients and their metabolites. None of the studies included in this review explored combinations or complex mixtures of ingredients and thus it is not possible to disentangle the effects of individual pesticides from the effects of multiple pesticides including potential additive or synergistic effects (Pinos et al., 2021). Another limitation of the biomarkers assessed in included studies is the fact that some biomarkers are non-specific to pesticides and capture by-products of industrial processes . Further, most of the biomarkers assessed were developed for high-income settings (largely by the US Centers for Disease Control and Prevention) and do not match to frequently used active ingredients in LMICs. For example, one of the included studies from Bangladesh used a standard set of biomarkers used by the Centers for Disease Control and Prevention, several of which were not detected in most participants and one of which was not detected in any participants . Lastly, many included studies evaluated active ingredients such as DDT and other organochlorines which have been banned for use in agriculture in many settings including LMICs for over a decade (Donley, 2019b). Thus, levels of these pesticides are likely to reflect exposure through consumption of animal-source foods such as fatty fish, which tend to have the highest concentrations of organochlorines (Schecter et al., 2010). Despite this evidence, only one of the studies included in this review controlled for dietary intake during pregnancy in assessing the association between prenatal pesticide exposure and weight-related anthropometric Z-scores in infancy . Given that intake of animal-source foods is positively associated with both organochlorine pesticide exposure (Marks et al., 2021) and child growth (Asare et al., 2022), this residual confounding likely biased results towards the null. Future studies should explore additional biomarkers that reflect commonly used pesticides and active ingredients in a specific setting. This could be done by reviewing country-specific lists of approved active ingredients (Donley, 2019a) or by collecting information on pesticides sold in local markets. More accurate quantitative data on commonly used pesticides in LMICs is needed to better understand exposure levels and their associations not only with the child growth outcomes we reviewed here, but also with maternal and child perinatal health outcomes more broadly. Local capacity should also be built and strengthened to support research and address issues around infrastructure and resources to measure pesticide biomarkers in LMICs ( Barnett-Itzhaki et al., 2018). Advances in exposure assessment from high-income settings (e.g., personal passive samplers) can help reduce the cost and participant burden of pesticide assessments in LMICs O'Connell et al., 2014).
The second major limitation of the existing literature is that most studies assessed pesticides at a single time point during the third trimester of pregnancy or at delivery/birth. Moreover, studies of organophosphates, pyrethroids, and carbamates mostly assessed biomarkers with short half-lives. Thus, most of the included studies did not capture chronic exposure levels. Chronic low-level exposure to pesticides has been linked to numerous neurologic functioning, neurodevelopment, and growth problems in animals and adults (Eskenazi et al., 1999). However, chronic low-level exposure in pregnant women and young children has been largely overlooked and dismissed due to lack of visible clinical signs or symptoms (Hertz-Picciotto et al., 2018). In this review, only two studies conducted repeat measures of organophosphates across trimesters to measure chronic exposure during pregnancy. Therefore, evidence is still limited on the short-and long-term effects of chronic low-level exposure during pregnancy and childhood to doses that do not produce overt clinical symptoms (Costa, 2006).
Relatedly, one limitation of our methodology was the inclusion of cross-sectional studies, which do not capture chronic exposure to pesticides with short half-lives and are subject to reverse causation (i.e., the temporal relationship between the exposure and outcome cannot be established conclusively). However, of the 20 cross-sectional studies we included, only one assessed exposure to organophosphates using biomarkers with short half-lives . The remaining 19 studies assessed either exposure to organochlorines (which have long half-lives) or self-reported exposure to pesticides over reference periods covering preconception and pregnancy. However, issues relating to the temporality of exposure and outcome remain. Given that changes in weight, including during pregnancy, are correlated with concentrations of fat-soluble pesticides like organochlorines (Vizcaino et al., 2014;Grimalt et al., 2022;Jansen et al., 2017), it is possible that the outcomes could have influenced exposure levels. Nevertheless, as the evidence on the associations between pesticide exposure and child growth in LMICs is still at an early stage, cross-sectional studies play an important role in generating hypotheses for further investigation.
Although included studies were limited to a single prenatal assessment, their findings do suggest that the effects of prenatal exposure to pesticides may manifest later in life. Specifically, one study found that prenatal exposure to organophosphates was associated with poor child linear and ponderal growth at two years of age , and another study found that prenatal exposure to organochlorines was associated with higher BMI from 6 to 24 months of age . Relatedly, none of the included studies assessed pesticides in children providing no information on concurrent exposure or on chronic or acute exposure in children. Some of the mechanisms through which pesticides may influence child growth (e.g., changes in glucose metabolism, immune dysregulation, EED, and changes in gut microbiome) may be more active and relevant for concurrent exposure within the same individual rather than for prior exposure in the mother. Future studies in LMICs should consider assessing both acute and chronic exposure before and during pregnancy as well as in infancy and early childhood (taking a life-course approach) to better understand the concurrent and chronic short-, medium-, and long-term effects of pesticide exposure on child growth.
Despite the strengths of our methodology, including a pre-registered protocol, a search strategy developed in consultation with a research librarian, the lack of language restrictions, and the inclusion of 10 databases, some limitations of the current literature are worth noting. First, we identified only three studies (out of 653 unique records) conducted in sub-Saharan Africa, none of which met our inclusion criteria. Two of these studies were conducted in South Africa; one assessed prenatal pesticide exposure in the context of insecticide-treated bednets for malaria prevention (Coker et al., 2018), whereas the other was a case report of acute chlorpyrifos poisoning during pregnancy (Solomon and Moodley, 2007). The third study reviewed the contribution of pesticides to the burden of diabetes in sub-Saharan Africa (Azandjeme et al., 2013). Further, we identified no studies in Central America, the Caribbean, or Asia Pacific. This is particularly concerning given that Central America and the Caribbean have the highest pesticide application rates in the world (FAO, 2021). Second, many of the included studies had small sample sizes or participants were recruited at a single location, limiting the generalisability of the findings. More studies in similar populations using consistent methodologies within and across LMICs are needed to better understand if, and how, pesticides influence child growth in LMICs. Third, few studies conducted subgroup analysis and the ones that did only examined child sex and paraoxonase 1 phenotype. Other subgroups may be more important to understand if, and how, the effects of pesticides on child growth vary, to identify the most vulnerable groups, and to inform the design and targeting of intervention to reduce pesticide exposure. One important area for future research may be subgroup analysis by household wealth as wealthier households may be less exposed to pesticides because they have improved water sources, can afford less contaminated foods, and are less likely to work in agriculture and be exposed occupationally, relative to poorer households (Fairburn et al., 2019).
Lastly, none of the studies included in this review examined the causal impact of pesticide exposure on child growth. More conclusive studies identifying causal impacts are needed. One approach could be to conduct randomised controlled feeding trials of organic foods. Three small studies in the United States (sample sizes ranging from 9 to 40 children) have shown that substituting conventional foods with organic foods significantly reduced pesticide biomarkers in children (Lu et al., 2008;Bradman et al., 2015;Fagan et al., 2020;Hyland et al., 2019). Another study of 20 pregnant women, also in the United States, showed that substituting conventional fruits and vegetables for organic significantly reduced pesticide exposure during pregnancy (Curl et al., 2019). However, none of these intervention studies measured birth outcomes or child growth. Moreover, to the best of our knowledge, no organic feeding trials have been conducted in LMICs either in pregnant women or in children. These types of trials have important feasibility, economic, ethical, and sustainability implications related to the availability, accessibility, and affordability of organic foods that should be addressed by any future trials in LMICs. For instance, in many LMICs obtaining "organic" certification is prohibitively expensive. Nonetheless, consumer demand for organic food is increasing around the world including in LMICs such as India (Chandra et al., 2021). Another approach could be to conduct randomised controlled trials of agroecological interventions promoting the adoption of organic, climate-smart agricultural practices. These may be more suitable for LMICs, where one-third of adults are employed in agriculture (The World Bank, 2021). To the best of our knowledge, only one such intervention in Tanzania has been evaluated to date. This nutrition-sensitive agroecological intervention improved child dietary diversity but had no effect on child growth (Santoso et al., 2021). However, this study did not assess pesticide exposure in women or children. Another upcoming trial in Andhra Pradesh, India will assess the effect of an agroecological intervention on pesticide exposure and child growth (ISRCTN Registry, 2021), and thus help build the evidence on the causal effect of reducing pesticide exposure on child growth. Nevertheless, given these challenges in assessing the causal impact of pesticides exposure on child growth, the precautionary principle can be adopted, especially with regards to highly hazardous (Food and Agriculture Organization of the United Nations and World Health Organization, 2016).

Conclusion
This systematic review of pesticide exposure and child growth among children <5 years of age living in LMICs identified 31 articles examining the associations between prenatal pesticide exposure and child growth outcomes at birth and later in infancy. The existing literature suggests no consistent associations between prenatal pesticide exposure and birth weight, birth length, low birth weight, and preterm birth. Many included studies were at serious risk of bias, and the quality of the evidence was "very low" for all outcomes. More studies with accurate quantitative data on commonly used pesticides and their metabolites are needed to better understand if, and how, pesticides influence child growth in LMICs. Ideally, these studies should capture chronic exposure to complex mixtures using consistent methodologies in comparable populations of pregnant women and children. Given the limited resources in LMICs to accurately measure pesticides in various biological matrices following gold-standard methods, capacity and infrastructure strengthening are needed to ensure the feasibility of toxicological cohort studies in LMICs. Precautionary guidelines may be needed given the potentially harmful effects of pesticide exposure on child growth.