Eligible studies published in English through April 2023 were retrieved from the PubMed and EBSCO databases. The reference lists of the retrieved articles were manually searched for additional applicable sources. We used predefined search terms to identify relevant articles (Table 1). After the removal of duplicates, the titles and abstracts were screened by the authors (MF, MK, VW, KW, NAD, MR, and ESH). Articles that met the inclusion criteria were then subjected to full-text screening. Cases of discrepancy were settled by consulting other authors (SRS and YAJ). Consensus was reached upon mutual agreement of all authors. A paper's corresponding author was contacted for clarification if needed. The data collection and extraction processes were conducted using Microsoft Excel 2017 and Mendeley software, and the data were stored in the cloud using a service provided by Universitas Islam Indonesia.

The inclusion criteria were the analysis of the influence of pesticide exposure, regardless of class, on stunting incidence. Pesticide exposure was defined for every child who lived on or near an agricultural facility. We broadened the exposure criteria based on laboratory examinations and observational criteria to capture exposed and nonexposed children. Additionally, stunting in an article must be quantified by a height-age z score (HAZ) measurement, while self-reported information or other measurements were omitted from the analysis. Observational studies (cohort, case-control, and cross-sectional), randomized controlled trials (RCTs), interventional studies, reviews, and proceedings were eligible for inclusion and analysis. However, opinion, perspective, and descriptive data, including case reports, case series, and case studies, that did not examine the association between pesticide exposure and stunting were excluded. A certainty analysis of every included study was performed using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) summary [21].

A quality assessment was performed using the Newcastle-Ottawa Scale (NOS) for non-RCTs, observational studies, cohort, and case-control studies [22]. This method offers a risk of bias (RoB) assessment in 3 domains: selection, comparability, and outcome or exposure. Given the limitation of poor agreement and lack of a comprehensive manual [23], the NOS method is the most applicable RoB tool for nonrandomized studies and meta-analyses. Cross-sectional studies were examined using the Agency for Healthcare Research and Quality guidelines [24]. RCT studies were examined using the Cochrane risk-of-bias tool [25]. Two authors (YAJ and MK) performed the RoB analysis.

Seven authors (MF, MK, VW, KW, NAD, MR, and ESH) extracted each study's author name, research design, population, pesticide type, stunting outcome, and finding. In cases of uncertainty, other authors were consulted (SRS and YAJ) until consensus was reached. The data were analyzed similarly. The data analysis focused on the endpoint estimation effects, including relative risk (RR), odds ratio (OR), mean difference, correlation coefficient, P value, and confidence interval (CI). Studies for which statistical data were missing were removed from the quantitative analysis. A study's corresponding author was contacted as necessary for further information. If no OR or RR values were reported by a study, the calculation of a 2×2 table was warranted. Since this study analysis was also quantitative, heterogeneity was used to assess the diversity of the included studies. A random effects model was used, and a forest plot was created to display effect estimates and CIs for the individual studies. The source of any heterogeneity was identified by meta-regression using OpenMEE software. A sensitivity and validity analysis was conducted using a funnel plot in Revman version 5.4 software (Cochrane, Fredericksburg, VA, USA).

A total of 13 studies were included in the analysis: 2 reviews [26,27], 2 RCT [17,28], 2 cohort studies [18,29], 2 case-control studies [12,30], and 5 cross-sectional studies [11,31-34]. Four studies elucidated the effects of pesticides on stunting [11,12,27,30], 3 focused on organophosphate and carbamate [29,31,32], 5 focused on permethrin, a pyrethroid class [17,18,33,34], and one simultaneously analyzed metabolite herbicide, organophosphate, and pyrethroid [29]. Most studies analyzed stunting using HAZ; only one used the ponderal index [35] and, thus, was omitted from the further analysis. Among the 13 retrieved studies, 6 included a meta-analysis. The detailed baseline characteristics of the included studies are presented in Table 2 and Supplementary Table 1.

Kartini et al. [12] examined high-exposure pesticide-induced stunting among school-aged children who lived near onion farming (adjusted OR [aOR], 3.9; 95% CI, 1.15–13.26). Evidence of stunting was proven by HAZ and low IGF-1 levels (aOR, 8.35; 95% CI, 3.65–19.14). Kartini et al. [12] thoroughly evaluated risk factors for stunting, including nutritional status, low birth weight, infection history, and parents’ education and occupation. Protein and energy intake as well as high exposure to pesticides were associated with stunting. However, the sample size was inadequate, which limited the study's power. The study of Paudel et al. [30] corroborated the findings of Kartini et al. [12] in their adjusted multivariate analysis (aOR, 3.51; 95% CI, 1.33–9.23). The strength association of the results was proven by its adequate sample size, but the study was limited by its use of poor methods of detecting pesticide exposure. A review by Vilcins et al. [30] summarized a similar result to that of Paudel et al. [27] of pesticide-induced stunting (aOR, 3.51; 95% CI, 1.33–9.23). Apart from the previous findings, Grandjean et al. [11] found that a low score on the Stanford-Binet Copying Test, a neurodevelopment test, was associated with pesticide exposure (β=-1.01; standard error [SE]=0.37; P<0.00) and stunting in children (β=-0.74; SE=0.35; P<0.05). Of 37 exposed groups (-1.84±0.83; range, -3.65 to 0.35), 15 stunted children had median dimethyl metabolite and diethyl metabolite levels of 0.2 nmol/kg/day and 0.27 nmol/kg/day, respectively [11].

In general, pesticide exposure was associated with stunting. However, the limited sample sizes and inadequate pesticide exposure assessments prevented acceptance of the stated hypothesis.

Low use of organophosphate and carbamate was correlated with HAZ in children compared to their high and intermediate use (-1.95±1.86, -1.05±1.5, -1.25±1.27, F=4.06, P=0.0194, respectively) [31]. This finding seems different because the low carbamate and organophosphate use levels were 0.65±0.93 kg and 0.64±0.74 kg, respectively, compared with the high (0.28±0.27 and 0.60±0.47) and intermediate (0.45±0.33 and 0.91±0.79) levels [31]. Additionally, significantly more carbamate was used than organophosphate (P=0.04 vs. P=0.12). Overall, stunting was noted in 11 of 41 children in the low use group versus 2 of 40 in the intermediate use group and 4 of 54 children in the high use group. Unfortunately, this study did not examine the correlation between urinary pesticide exposure and HAZ in children [31]. Handal et al. [32] also reported that organophosphate and carbamate pesticide exposure was not correlated with stunting incidence in children (OR, 0.04; P=0.84). Stunting, however, affects a child's gross motor function, making it a moderate variable or independent factor in an individual's neurodevelopment. The prevalence of stunting was increased after high or low pesticide exposure, and children in both regions replace with in both exposure level resided in rural areas with lower education and income levels [32]. Another study reported on the effects of exposure to 4 organophosphate metabolites on stunting [29], including 3,5,6-trichloro-2-pyridinol (TCPy; chlorpyrifos metabolite), 4-nitrophenol (parathion and methyl parathion metabolite), malathion dicarboxylic acid (malathion metabolite), and 2-isopropyl-4-methyl-6-hydroxypirimidine (IMPY; diazinon metabolite). This study found that organophosphate exposure was not correlated with stunting, even after 2 years of follow-up (RR at Q4: TCPy, 1.16 [0.8–1.68]; 4-nitrophenol, 0.85 [0.61–1.19]; IMPY, 1.04 [0.73–1.49]) [29]. Exposure to organophosphate, regardless of class, was not associated with stunting. However, this study was limited by its small sample size, exposure assessments used, and poor design. Jaacks et al. [29] used the most applicable research method that seemingly required a larger sample size and long-term cohort evaluation to prove the relationship between exposure and outcomes.

Permethrin, a class of pyrethroid, was the most commonly used pesticide for treating malaria, namely insecticide-treated nets (ITN), according to some retrieved studies [17,18,29,33,34]. Stunting cases increased parallelly with high exposure to malaria among children with HIV (RR, 1.24 [1.06–1.48]) [18], malaria treatment with ITN caused malnutrition, including stunting (-0.20±0.09) [33], and ITN coverage was less effective at preventing stunting (OR, 0.02 or mean difference, 0.19; range, 0.18–0.19) [34]. An RCT by Friedman et al. [28] found that use of the ITN intervention did not correlate with stunting (mean pooled data, -0.95; 95% CI, -1.07 to -0.86). A similar RCT by Ter Kuile et al. [17] indicated that use of the ITN intervention versus control was correlated with HAZ (mean ITN, -1.40 vs. control, -1.76; mean difference, 0.36; range, 0.10–0.62; P=0.001). Another study of the more specific metabolite pyrethroid 3-PBA found no correlation with stunting incidence at the first or second follow-up (mean difference, 1.01; range, 0.83–1.24; and mean difference, 0.95; range, 0.69–1.32, respectively) [29].

Although an RCT design and large sample size were proposed in the ITN study on stunting, the result regarding the relationship between exposure and outcome was debatable. No suitable pyrethroid assessment was performed in either of the studies, so a questionnaire was used to collect the information. The robust method used in the Jaacks et al. cohort study inadequately proved an association between pesticide contamination and stunting. Thus, pyrethroid exposure was not associated with stunting incidence.

Six studies were included in the quantitative analysis, as presented in Figs. 2 and 3 [11,12,28,30-32]. Of 445 subjects who used ITN treatment to eradicate malaria in the study of Friedman et al. [28], 79 (16.7 %) denoted stunting, and 76 of 393 controls were positive for stunting. The study of Grandjean et al. [11] reported that 15 of 37 children with prenatal pesticide exposure versus 16 of 35 unexposed children were stunted. A history of pesticide applications was stratified as high, moderate, low, and unexposed; of them, 13, 14, 13, and 8 were stunted, respectively. On the other hand, similar results were seen among control subjects (14, 31, 39, and 28, respectively). Thus, 40 of 124 exposed versus eight of 36 controls were stunted [12]. Orozco et al. [31] stratified pesticide use intensity into more, intermediate, and less levels. We defined high pesticide exposure as high and moderate intensity and low exposure as less intensity. The authors reported that 28 of 94 children with high exposure versus 21 of 41 children with low exposure were stunted [31]. In a flower plantation area in the study of Handal et al. [32], greater pesticide exposure led to stunting in 83 of 154 children, while lesser exposure led to stunting in 68 of 129 children. Paudel et al. [30] also clarified the presence of 58 of 85 stunted children in pesticide-exposed groups versus 60 of 151 stunted children in nonexposed groups.

Total events were 278 of 845 population in the exposed group versus 228 of 774 in the unexposed group. Although the funnel plot to verify sensitivity confirmed symmetrical data, heterogeneity existed (I²=75%, P=0.003). Additionally, a meta-regression determined that age was a significant (P≤0.001) source of heterogeneity. Therefore, pesticide exposure was not associated with stunting in children (P= 0.28).