Comparing Impact of Pesticide Exposure on Cognitive Abilities of Latinx Children from Rural Farmworker and Urban Non-Farmworker Families in North Carolina

Abstract

Pesticide exposure remains a health hazard despite extensive study into adverse effects. Children in vulnerable populations, such as Latinx children in farmworker families, are particularly at risk for exposure. Several studies have demonstrated the detrimental cognitive effects of prenatal exposure to pesticides, particularly organophosphates (OPs) within this high-risk group. However, results from studies investigating the cognitive effects of early childhood pesticide exposure are equivocal. Most studies examining the effects of pesticide exposure have used correlative analyses rather than examining populations with expected high and low exposure. The current study compares 8-year-old children from rural families of farmworkers and urban, non-farmworker families. We used the Weschler Intelligence Scale for Children – Fifth Edition (WISC-V) to assess cognitive performance in these children. We designed this study with the expectation that children from farmworker families would have greater exposure to agricultural pesticides than urban, non-farmworker children. This assumption of exposure to agricultural pesticides was confirmed in a recent report that assessed exposure probabilities using life history calendars. However, data from passive wristband sampling of acute (1-week) pesticide exposure from these same children indicate that both study populations have considerable pesticide exposure but to different chemicals. As expected the children of farmworkers had greater OP exposure than non-farmworker children, but the non-farmworker children had greater exposure to two other classes of insecticides (organochlorines [OCs] and pyrethroids). Our analyses considered these findings. A comparison of the cognitive scores between groups revealed that children from farmworker families had slightly higher performance on the Visual-Spatial Index (VSI) and Verbal Comprehension Index (VCI) when compared to children from non-farmworker families. Regression analyses where pesticide exposure was included as covariates revealed that OC exposure accounted for the largest portion of the group differences for both VSI and VCI. However, a post-hoc moderation analysis did not find significant interactions. The main study outcome was that the non-farmworker children exhibited lower WISC-V scores than the children from farmworker families, and the analyses incorporating pesticide exposure measures raise the hypothesis the that pervasive and persistent nature of a variety of pesticides may have adverse effects on the neurodevelopment of young Latinx children whether living in rural or non-farmworker environments.

1. Introduction

The neurotoxic effects of many pesticides have resulted in some restricted use and application of such chemicals (Costa, 2006, 2015; Richardson et al., 2019). The continued use and existing accumulations of various pesticides in the environment still pose a significant health risk, particularly for certain populations. Families that live and work in agricultural communities come into contact with pesticides, particularly agricultural pesticides such as organophosphates (OPs), more regularly than those who do not (Curwin et al., 2005; Hyland and Laribi, 2017; Quandt and Arnold, 2020). Increased exposure in these groups extends beyond farmworkers (FW), specifically to their children (Curl et al., 2002), as pesticides can enter the children’s environment through drift from nearby fields and direct contamination from exposed adults living in the home (Arcury et al., 2020; Fenske et al., 2013). The hazardous neurotoxic effects of OPs are generated mainly through the inhibition of acetylcholinesterase (Richardson et al., 2019), which can result in difficulties with motor coordination, attention, and respiratory issues (Costa, 2006; Naughton and Terry, 2018; Richardson et al., 2019; Roldan-Tapia et al., 2006). Children are especially vulnerable to the toxic effects of pesticides (Hyland and Laribi, 2017) because of their developing nervous system (Weiss, 2000), immature metabolism, and increased exposure through frequent hand-to-mouth behavior and playing on contaminated surfaces, such as the floor (Landrigan et al., 1999; Roberts et al., 2012).

A growing number of studies have examined the neurodevelopmental effects of pesticides, particularly OPs, in children with a large portion of the investigations examining the impacts of prenatal pesticide exposure on later neurocognitive development (Bouchard et al., 2010; Bouchard et al., 2011; Furlong et al., 2014; González-Alzaga et al., 2014; Marks et al., 2010). Probably the most unequivocal adverse effect of prenatal OP pesticide exposure identified in children is a reduced intelligence quotient (IQ) scores. A series of studies published in 2011 showed IQ reductions in children aged 6–9 years with increases in maternal metabolites for OP pesticides measured at birth or during pregnancy (Bouchard et al., 2011; Engel et al., 2011; Rauh et al., 2011). In addition, children with high prenatal OP exposure (specifically chlorpyrifos) had altered cerebral cortex volume and thickness compared to children with low exposure. Notably, the differences in cortical anatomy were associated with differences in IQ scores (Rauh et al., 2012). Prenatal OP pesticide exposure has been related to many other neurodevelopmental deficits, including motor impairments (Bouchard et al., 2011), behavioral issues (Bouchard et al., 2010; Marks et al., 2010), attention deficit hyperactive disorder (Bouchard et al., 2010; Bouchard et al., 2011; González-Alzaga et al., 2014; Marks et al., 2010), and disrupted social development (Furlong et al., 2014).

While abundant information links neurocognitive developmental effects to prenatal OP pesticide exposure, an understanding of the neurobehavioral impact of early childhood OP pesticide exposure is inconsistent (Bouchard et al., 2011). Studies comparing prenatal and early childhood pesticide exposure have had mixed results (Bouchard et al., 2011; Cartier et al., 2015; Marks et al., 2010; Viel et al., 2015). A few studies have found similar associations between neurocognitive function and prenatal and early childhood exposure. Marks et al., (2010) found prenatal and early childhood exposure was associated with the increased occurrence of ADHD, and Cartier et al. (2015) observed changes in cognitive performance, although the details of cognitive aspects differed when comparing associations with maternal and child pesticide metabolites. However, the findings for most studies comparing exposure time points vary between prenatal and early childhood effects, with negative changes being primarily associated with exposure during the prenatal but not early childhood period (Bouchard et al., 2011; González-Alzaga et al., 2014). Specific studies focusing on early childhood OP exposure are not as consistent as those for prenatal exposure. Some studies have reported no association between child pesticide exposure and cognition or behavior (Bouchard et al., 2011; Butler-Dawson et al., 2016; Rohlman et al., 2005). Others report early childhood OP exposure to be related to deficits in working memory (Cartier et al., 2015; Ruckart et al., 2004; Viel et al., 2015) and attention problems such as ADHD (Bouchard et al., 2011; Lizardi et al., 2008; Marks et al., 2010; Ruckart et al., 2004).

The current study examines the cognitive impacts of environmental pesticide exposure by comparing children from rural FW communities and those from urban non-farmworking (NFW) areas. Our study design was based on the expectation that children from FW families would have greater pesticide exposure than children from NFW families. We have recently confirmed that the children from FW families have life histories indicating a higher probability of agricultural pesticide exposure in both prenatal and early childhood periods compared to the urban children from NFW families (Quandt et al., 2020). We also recently confirmed that acute (1 week) exposure to OPs was higher in children from FW families. However, children from NFW families had an increase in exposure to other classes of pesticides (organochlorines (OC) and pyrethroids) when compared with children from FW families (Arcury et al., 2020).

The three studies that provide strong evidence of detrimental neurological effects from prenatal pesticide exposure all used the Wechsler Intelligence Scale for Children – Fourth Edition (WISC-IV) to assess the intelligence quotient, or IQ (Bouchard et al., 2011; Engel et al., 2011; Rauh et al., 2011). The present study also utilized the WISC-V to assess cognition with the newly standardized WISC-V (Weiss, 2019). The previously reported influence of exposure on cognitive development led to the hypothesis that children from FW families would score lower on the WISC-V than children from NFW families. Analyzing acute pesticide exposure data allowed the investigation of possible mechanisms contributing to the study outcomes; specifically, the association between pesticide exposure and cognitive performance.

2. Methods

We designed the Preventing Agricultural Chemical Exposure (PACE5) study to identify and compare the effects of overall pesticide exposure on cognitive and brain development in Latinx children from FW and NFW families. PACE5 is a longitudinal community-based participatory research (CBPR) project performed as a partnership between Wake Forest University School of Medicine and North Carolina Farmworkers Project (Benson, NC; http://ncfwp.org). The current paper reports cognitive test results from baseline testing performed when the children were 8 years old with the aim to compare the cognitive outcomes associated with OP exposure between children from FW and NFW families. The Wake Forest University School of Medicine Institutional Review Board, approved all PACE5 protocols and procedures and the study has received a Certificate of Confidentiality from the National Institutes of Health.

2.1. Participant Recruitment

To be included in this study, children must have been 8 years of age and have completed the first grade in the US at the time of enrollment. All participants were from self-identified Latinx families whose household income was below 200% of the US federal poverty line. To be eligible for the FW group, at least one adult from the participant’s household must have worked on non-organic farms for the past three years. The majority of adults from FW families had a stable history of farm work from the child’s birth to the time for study enrollment. For NFW participants, any adults living in the home could not have had routine exposure to pesticides through employment or lived adjacent to agricultural fields within the previous three years. Children were excluded from the study if they had a life-threatening illness, history of neurological, physical, or developmental disorder that would interfere with completing the cognitive testing or a brain scan, which was part of the larger study. At the final study visit, the date of birth of one NFW child was found to be different than originally reported at the initial baseline visit. It is possible that this child may have been 7 years old, rather than 8 years old, at baseline. However even following further communication with the family by the native Spanish speakers on our team, we were unable to clarify their true date of birth.

The North Carolina Farmworkers Project and local recruiters in Winston-Salem, as well as other community organizations identified potential participants. Family recruitment took place between March 2018 and December 2019. A bilingual staff member explained the overall study procedures to families and answer any questions. The staff member obtained signed informed consent from a parent/guardian, and the children provided assent. We recruited a total of 76 children from families of FWs from eastern North Carolina and 65 from NFW families located in central North Carolina. Because interviewers worked through community partners and community events, the number of potential participants or their parents who refused to participate is unknown.

2.2. Demographic and life history data collection and measures

Mothers of the participants completed baseline questionnaires administered by a study interviewer. All interviewers were native Spanish speakers, bilingual in English and Spanish, and received appropriate training prior to data collection. These baseline questionnaires gathered detailed information about the child participant as well as the family and home environment. These questionnaires and their data were recorded through Research Electronic Data Capture (REDCap) (Harris et al., 2009). Interviewers used life history calendars to detail the life of the child participants, extending from prenatal events to the time of the interview (Quandt et al., 2020). These calendars captured information concerning living conditions and proximity to agricultural fields, dwelling type, family work history, and water source.

Based on the review of prior literature, we identified a list of potential covariates that were available to include in our analyses based on the questionnaires and life history calendars collected for this study. To assure the results of cognitive testing were representative of differences between FW and NFW groups, this list of relevant measures was developed and included the following: child sex (Bouchard et al., 2011; Donauer et al., 2016; Marks et al., 2010), if the child had pre-k education experiences (Cartier et al., 2015; Khan et al., 2014; Marks et al., 2010), season of WISC-V assessment (Bouchard et al., 2011), highest maternal education completed in years (Butler-Dawson et al., 2016; Cartier et al., 2015; Donauer et al., 2016; Engel et al., 2011; Viel et al., 2015) and depression (Marks et al., 2010), prenatal alcohol and tobacco use, presence of father in the home, number of siblings, housing conditions (apartment, trailer, or detached home), water sources (city or well) (Butler-Dawson et al., 2016; Engel et al., 2011; Stein et al., 2016), and exposure to adverse life experiences as reported through Adverse Childhood Experiences questionnaire (ACEs) (Engel et al., 2011; Stein et al., 2016). We used maternal education as a surrogate for maternal intelligence as we did not have measures of their IQ (Stein et al., 2016). Information regarding learning disabilities or ADHD was also recorded based on maternal response to the questionnaire.

2.3. Cognitive Assessment

Children were administered the Wechsler Intelligence Scale for Children, Fifth Edition (WISC-V, www.pearsonassessments.com/wisc-v) in their more proficient language (English or Spanish) to assess cognitive abilities. Language proficiency was determined using a language/school report. Supplemental Table 1 outlines all language-based questions. Two clinical psychologists (one bilingual and one English-speaking) and one bilingual specialist in school psychology completed all assessments. The English-speaking psychologist tested six (6) children who were most proficient in speaking English. The remaining children were tested in their proficient language by the bilingual testers. The specialist in school psychology was one of the team members who worked on the standardization of the Spanish version of the WISC-V with Pearson’s Clinical Assessment group. The test administration took place using the Q-interactive format. Q-interactive is a 1:1 iPad-based testing system that helps administer, score, and report the WISC–V. Test administration on Q-interactive takes place via two iPads connected via Bluetooth. This allows the tester to explain instructions, record and score responses, take notes, and control visual stimuli while the child views and responds to the prompts provided. The child and tester seated on opposite sides of a table in a distraction-free, quite environment during testing. Parents were not in the room during the testing procedures to ensure that testing was not interrupted and in keeping with clinical standards.

The WISC-V is the standard instrument for the neurobehavioral assessment of a child’s cognitive abilities. It is the latest edition and replaces the WISC-IV. It has more interpretive power, is more efficient and user-friendly than previous versions, and has updated psychometric properties (Weiss, 2019). The WISC-V offers updated theoretical foundations that increase the breadth of coverage by investigating and developing visual-spatial, fluid reasoning, and visual working memory subtests. These subtests measure additional processes related to learning and additional cognitive processes related to learning disabilities and represent an intellectual function in specific cognitive domains, which are quantitatively assessed with the following five indices: Verbal Comprehension Index (VCI), Visual-Spatial Index (VSI), Fluid Reasoning Index (FRI), Working Memory Index (WMI), and the Processing Speed Index (PSI). The WISC-V also provides a combined index score representing general intellectual ability, termed Full-Scale Intellectual Quotient (FSIQ). These five indices and the FSIQ were used in the current study. Supplemental Table 2 lists the specific subtests utilized in this study with a description of their measurements. Scoring is performed by the electronic version of the test with additional verification performed by the test administrators. After the clinician verified the scores, they were downloaded from the Pearson website where raw scores, normed composite scores, and percentiles are generated. The scores used in this study are composite scores. It should be noted that the baseline testing used the fully-automated WISC-V version on the iPad. In 2020, after these baseline data was collected, two subtests (coding and symbol search) were taken off the iPad and are now administered and scored manually (consult the Pearson website for further details). Thus, caution should be used when comparing results from this study with studies that used the WISC-V version with the manual subtest administration. In follow-up visits it became clear that two NFW children had errors in their date of birth entered into the WISC-V software. We were unable to go back and rescore these tests with the correct birth date due the change in the testing/scoring procedure as described above. However, the age differential due to the error was within a year for both participants and should not have affected the scoring in a meaningful way. At a later date, following testing, the clinician performing the test generated a summary report that was provided to the parent/guardian. If necessary, the parent was allowed to discuss the report with the clinician.

2.4. Pesticide exposure monitoring

The analysis examined the role of pesticide exposure on cognitive test scores used measures of passive pesticide exposure monitored through silicone wristbands that have proven effective for detecting exposure to environmental chemicals (Arcury et al., 2020; Dixon et al., 2019). Exposure monitoring with wristbands can only detect the exposure to pesticides in the environment and not the direct metabolites resulting from absorption of environmental exposure. However, pesticides assessed in this study are all absorbed through the skin and subsequently through ingestion (hand to mouth). Several studies have reported significant correlations between chemical concentrations in wristbands and the respective biological metabolite concentration (urine or serum), providing evidence that wristbands can capture the bioavailable fraction of organic chemicals in the personal environment (Dixon et al., 2018; Hammel et al., 2020; Hammel et al., 2016; Hammel et al., 2018; Levasseur et al., 2021; Quintana et al., 2019). Further, prior studies examining pesticide exposure for children in North Carolina farmworker families found that most children had detections for organophosphate pesticides in urinary metabolites (Arcury et al., 2007; Arcury et al., 2006; Arcury et al., 2005).

A recent publication documenting the findings from this study population (Arcury et al., 2020) describes detailed results from the baseline pesticide exposure monitoring. We briefly describe the procedure for collecting and analyzing the samples here for completeness, but further details can be found in prior publications (Anderson et al., 2017; Arcury et al., 2020; Bergmann et al., 2017; Donald et al., 2016). At the end of the interview, interviewers gave participants wristbands and provided instructions on how they should be worn (Anderson et al., 2017). The participants wore the wristbands for a seven-day period. After the seven-day sampling period, the wristbands were removed from the child and placed in a Teflon bag, where they remained until retrieved by interviewers and shipped to the laboratory for analysis. The laboratory utilized validated cleaning and extraction procedures before analyzing the pesticides with dual micro-electron detector gas chromatography, similar to those previously performed (Anderson et al., 2017; Arcury et al., 2020; Bergmann et al., 2017; Donald et al., 2016). The exposure data used for the analyses in this manuscript included the three neurotoxic classes (OPs, OCs, and pyrethroids) that had the highest frequency of detection within our study populations (Arcury et al., 2020). We quantified each pesticide class differently due to different distributions of exposure across participants. We quantified the OCs as a continuous measure as there was a range of exposure from 0–7 different OCs. We quantified the pyrethroids in the categorizes of 0, 1, 2, or ≥3. As there were three NFW and zero FW children with four detected pyrethroids and zero NFW and one FW child with six detected pyrethroids, the individuals with three, four, or six detections were combined to create a ≥3 category. Because most children either did not have OP detections or had only one detection, we classified the OPs as a dichotomous variable (exposed or not exposed). Only one NFW and 5 FW children had two detections.

2.5. Data Analysis

We conducted bivariate analyses to assess the difference by FW status for candidate covariates and other demographic characteristics described in Section 2.2. Chi-square tests or Fisher’s exact tests were used when appropriate. Any variables identified with a significance level of <0.1 were included as covariates in statistical models. We did not use birth weight and learning disabilities as covariates because they may be causal intermediaries (Ananth and Schisterman, 2017) for the effects of pesticides on cognitive function. In other words, it may be that pesticides cause learning disabilities and these learning disabilities are a major contributor to cognitive test scores. We recognize that ADHD could be associated with WISC-V scores, and three NFW children were reported to have ADHD. However, this variable was not included in the statistical models since the low frequency of ADHD would likely lead to spurious results. Statistical comparisons were not performed on housing quality or water source due to uniform group distribution for each variable. To test the main study hypothesis for each cognitive measure, we fit a sequence of general linear models to examine the group differences by FW status. One set of analyses did not include covariates (unadjusted model), and the other set of analyses included covariates identified with p < 0.1 in the bivariate analysis (adjusted model). We also examined associations between acute pesticide exposure measured through the silicone wristbands and cognitive performance. Specifically, we used general linear models to investigate group differences in VCI and VSI while accounting for detections of the three most prominent classes of pesticides identified within these populations (Organochlorines (OC), Organophosphates (OP) and Pyrethroids), each in separate analyses. All analyses reported the least square means and standard errors (SE). We calculated the magnitude of change in the parameter estimates for VCI and VSI for models with pesticide exposure adjustment as the percentage change in the difference between FW and NFW between adjusted and exposed models. We also performed moderation analyses for OC exposure on the group differences in VCI and VSI by modeling an interaction between pesticide exposure and FW status.

3. Results:

A total of 126 participants (71 FW and 55 NFW) successfully completed the baseline questionnaires, life history calendars, WISC-V assessments, and acute pesticide exposure monitoring. Group comparisons of the participants reveal FW and NFW children to be comparable across the majority of the selected demographics (Table 1). The two groups differed in regards to the season of the year that testing occurred (p<0.001), education experience (such as pre-k or Head Start) (p = 0.0035), and maternal education (p = 0.0097). Our inclusion criteria of variables with group difference of p<0.1 expanded the group differences to additionally include the child’s testing language (p= 0.0708) and the presence of a father in the home (p = 0.0721). We did not perform group comparison for housing type or water source as nearly all the FW children resided in trailers, while NFW varied between houses and apartments. Similarly, almost all of the NFW children consumed city water while FW children had access to either city or well water sources. The analysis did not use sex as a covariate as groups had an equal number of females and males (50.9% female and 49.1% male for NFW, 49.3% and 50.7% for FW).

Table 1.

Personal and Family Characteristics of Latinx Non-Farmworker and Farmworker Families

Variable	Non-Farmworker	Farmworker
	N = 55 (%)	N = 71 (%)	p-value
Sex
Male	28 (50.9)	35 (49.3)	0.8574
Female	27 (49.1)	36 (50.7)
* Testing Language
English	14 (25.5)	29 (40.8)	0.0708
Spanish	41 (74.5)	42 (59.2)
* Testing Season
DEC-FEB	10 (18.2)	15 (21.1)	<0.0001
MAR-MAY	5 (9.1)	34 (47.9)
JUN-AUG	31 (56.4)	6 (8.5)
SEP-NOV	9 (16.4)	16 (22.5)
* Pre-Kindergarten	26 (47.3)	16 (22.5)	0.0035
ADHD	3 (5.5)	0 (0.0)	0.0806
Learning Disability	9 (16.4)	2 (2.8)	0.0101
Maternal Birth Country			0.5077
US	41 (74.5)	58 (81.7)
Mexico	5 (9.1)	3 (4.2)
Other	9 (16.4)	10 (14.1)
* Maternal Education (years)			0.0097
0–6	13 (23.6)	32 (45.1)
6–13	33 (60.0)	36 (50.7)
13 +	9 (16.4)	3 (4.2)
Maternal Depression	6 (11.8)	4 (6.0)	0.2629
Prenatal Substance Use
Tobacco	1 (1.8)	0 (0.0)	1.00
Alcohol	1 (1.8)	1 (1.4)	0.4365
* Father in Home	51 (92.7)	58 (81.7)	0.0721
Number of siblings
0	3 (5.7)	5 (7.0)	0.9823
1	14 (26.4)	17 (23.9)
2	17 (32.1)	23 (32.4)
3+	19 (35.8)	26 (36.6)
Adverse Experience ǂ			0.3771
0	22 (40.0)	37 (52.1)
1	19 (34.5)	21 (29.6)
2+	14 (25.5)	13 (18.3)
Pesticide Exposure Detections
Organochloridea	3.8 (1.5)	2.4 (1.8)	<.0001
Pyrethroid
0	15 (28.3)	30 (43.5)	0.0868
1	11 (20.8)	17 (24.6)
2	8 (15.1)	3 (4.3)
3+	19 (35.8)	19 (27.5)
Organophosphate
0	27 (50.9)	20 (29.0)	0.0135
1	26 (49.1)	49 (71.0)

^*Variables with p<0.1, included as covariates for WISC-V Analyses

^ǂAdverse Experience based on scores from Adverse Childhood Experience questionnaire (ACEs).

^aOrganochloride is reported as a continuous variable, mean (sd).

A direct comparison between groups found no difference in WISC-V performance with any index measured or FSIQ. After adjusting for differing covariates between our groups (Language of WISC-V administration, Season of WISC-V administration, Pre-K experience, Maternal Education, and Father in home), we found that NFW children had scores significantly lower than FW children in VCI (FW: 92.85 ± 2.74 vs. NFW: 86.13 ± 2.78, p = 0.0370) (Table 2). NFW children also marginally underperformed on VSI (FW: 95.51 ± 2.06 vs NFW: 91.29 ± 2.09 p = 0.0814). Average scores of the indices of the WISC-V range from 90–109 within the general population, with Latinx children performing on average 2–6 points lower across the indices (ranging from 94.2 to 98.3). Children in the FW group generally performed within this range on all indices when adjusting for covariates. On average NFW children marginally underperformed with the VCI (86.13 ± 2.78) and WMI (89.50 ± 2.52) indices, as well as FSIQ (89.62 ± 2.33).

Table 2.

Analysis of group differences between FW and NFW Children for WISC-V indices.

	Unadjusted			Adjusted*
	Mean (SE)		p-Value	Mean (SE)		p-Value
Indices	Non-Farmworker	Farmworker		Non-Farmworker	Farmworker
FSIQ	88.62 (1.66)	90.86 (1.46)	0.313	89.62 (2.33)	93.50 (2.30)	0.1495
FRI	92.76 (1.64)	91.13 (1.44)	0.455	93.54 (2.42)	90.49 (2.39)	0.2755
PSI	100.38 (1.92)	103.74 (1.71)	0.194	102.93 (2.81)	106.70 (2.85)	0.2501
VCI	86.44 (1.97)	88.66 (1.74)	0.399	86.13 (2.78)	92.85 (2.74)	0.0370
VSI	90.76 (1.51)	91.48 (1.33)	0.722	91.29 (2.09)	95.51 (2.06)	0.0814
WMI	89.89 (1.76)	90.63 (1.55)	0.112	89.50 (2.52)	93.02 (2.49)	0.2257

^*Covariates included Maternal Education, Father in home, Pre-Kindergarten experience, Season of WISC-V administration and language of WISC-V administration.

Given that children from FW families had higher probable exposure to agricultural pesticides, the primary outcome indicating children from FW families had higher cognitive scores than children from NFW families was unexpected. To gain deeper insight into this finding, we performed an additional analysis comparing VSI and VCI by FW status controlled for pesticide detections. Exposure detections for each pesticide class are from a 1-week period and described in Table 1, and complete exposure data has been recently published elsewhere (Arcury at al., 2021). We separately assessed the impact of including measurements of acute detections for three different pesticide classes (pyrethroids, OPs, and OCs) measured contemporaneously with the WISC-V testing. Table 3 outlines the results from these analyses along with the original adjusted group differences (based on covariates identified in Table 1) in the WISC-V indices. There was a modest reduction of group differences when controlling for pyrethroid or OP exposure for both VCI and VSI, reducing the significance of group differences for these WISC-V indices when accounting for each of these pesticide classes. Controlling for OC exposure had the largest effect on the group differences for VCI and VSI. The difference between the two groups was substantially lower when accounting for OC, with the mean group differences reduced by 32.59% and 25.11%for VCI and VSI, respectively. Given that OC exposure had the largest effect on group differences, post-hoc moderation analysis assessed interactions between OC exposure and group for VCI and VSI. Neither interaction proved statistically significant. However, the data clearly demonstrated a consistent negative relationship between OC exposure and VCI/VSI scores in both study groups. We generated estimated VCI and VSI means for 0, 3, and 6 OC detections and both scores consistently decreased as number of OC exposures increased. In all cases, the scores for FW children were higher than for NFW children. Although not significant, we present these results in Supplemental Table 3 for completeness.

Table 3.

Differential Adjustment of Group Differences in WISC-V Performance by Pesticide Class.

	Non-Farmworker	Farmworker	p-Value	Difference (%)
VCI	Mean (SE)	Mean (SE)
Adjusted*	86.13 (2.78)	92.85 (2.74)	0.0370 *	-
Organochloride	88.22 (2.83)	92.77 (2.76)	0.1713	32.59
Pyrethroid	89.09 (2.82)	95.23 (2.87)	0.0581	9.04
Organophosphate	87.45 (2.79)	93.36 (2.86)	0.0743	12.44
VSI
Adjusted	91.29 (2.09)	95.51 (2.06)	0.0814	-
Organochloride	91.90 (2.18)	95.06 (2.12)	0.2151	25.11
Pyrethroid	91.57 (2.22)	95.41 (2.26)	0.1305	9.00
Organophosphate	91.01 (2.17)	95.77 (2.22)	0.0643	12.80

^*Adjusted indicates scores from Table 2 that account for the identified covariates: Maternal Education, Father in home, Pre-K experience, Season of WISC-V administration and Language of WISC-V administration. Percent difference indicates the reduction in group differences that resulted after further adjustment of WISC-V scores to account for number of detections identified for each class of pesticide, as quantified in Table 1.

4. Discussion

It is vitally important that researchers continue to uncover the neurobehavioral effects associated with children encountering pesticides. Here, we present preliminary evidence that pesticide exposure during early childhood is associated with lower performance with cognitive tasks and may vary by pesticide type. The seminal studies examining pesticide exposure in children centered on correlating prenatal OP metabolite levels in serum (Bouchard et al., 2011; Engel et al., 2011), urine (Bouchard et al., 2011; Cartier et al., 2015; Donauer et al., 2016; Marks et al., 2010; Stein et al., 2016; Viel et al., 2015) and cord blood (Engel et al., 2011) with childhood IQ. The findings from these studies strongly implicate prenatal pesticide exposure (primarily OPs) with reductions in cognitive function. Fewer studies have assessed the effects of early childhood pesticide exposure on neurobehavioral function, and the results are less consistent (Bouchard et al., 2010; Bouchard et al., 2011; Butler-Dawson et al., 2016; Cartier et al., 2015; Lizardi et al., 2008; Marks et al., 2010; Rohlman et al., 2005; Ruckart et al., 2004; Viel et al., 2015). The current study directly compared cognition in populations of children with expected high and low general pesticide exposure, FW vs. NFW respectively, to further investigate this issue. Our recent work using life history calendars confirmed that children from FW families had greater risk for agricultural pesticide exposure during prenatal and early childhood periods (Quandt et al., 2020).

Given the high prevalence of pesticides in FW communities, as compared to NFW, and their known neurotoxic effects (Bouchard et al., 2011; González-Alzaga et al., 2014; Marks et al., 2010; Stein et al., 2016), we hypothesized that children from FW families would have lower cognitive function than children from NFW families. Studies utilizing the WISC-IV have found prenatal OP exposure to adversely affect FSIQ, (Bouchard et al., 2011; Engel et al., 2011; Rauh et al., 2011), WMI (Cartier et al., 2015; Rauh et al., 2011), and Perceptual Reasoning (Bouchard et al., 2011; Engel et al., 2011). In the current study we found that children from NFW families exhibited poorer performance on two of the WISC-V indices (VCI and VSI) compared to children from FW families. The NFW children marginally underperformed for VCI, WMI, and FSIQ as compared to the general population. However, WISC-V performance can also vary depending on ethnicity and parental education status. When parents have a high school or lower level of education, as was the case for the majority of the mothers of the NFW children, the FSIQ score for Latinx children ranges between 88 and 92. Thus, these children were within the lower end of this normal range based on maternal education (Muñoz, 2019). We did not expect these findings, given the prior life history exposure findings and more recent data showing that FW children from this study had higher acute exposure to OP pesticides than NFW children (Arcury et al., 2020). However, it is important to note that, while the NFW children exhibited some OP exposure, the levels of exposure to pyrethroids and OCs among NFW children were higher than among FW children.

Although the most conclusive evidence for adverse cognitive effects from pesticides is based on exposure to OPs, Veil et al. (2015) described a decrease in VCI and WMI with increased pyrethroid exposure. There is also some evidence for deficits associated with OC exposure (Dorea, 2021; Gaspar et al., 2015; Roberts et al., 2012; Sagiv et al., 2012). Sagiv et al. (2019) have reported increased exposure with OCs to coincide with lower scores for PSI on the WISC-III. Studies from the CHAMACOS cohort investigating prenatal exposure to DDT and DDE (an OC pesticide and a common OC metabolite) saw decreased PSI in 7 year old children and an additional reduction in FSIQ in 7 year old girls (Gaspar et al., 2015). Results from our study did not reveal any change to PSI. However, our analysis revealed NFW children underperformed on the VCI and VSI indices compared to children from FW families. Analyses adjusted for acute exposure to OPs, OCs and pyrethroids revealed that OC exposure accounted for the largest difference between the two groups. Analyses examining how OC exposure may moderate the group differences in VCI and VSI showed consistent decreasing scores with increasing exposure, but the results did not achieve significance. Thus, our findings suggest that OC exposure may account for the difference in performance between the two groups. However, it is important to note that we did not power this study for moderation analyses and these findings need to be explored and confirmed by future studies to determine the differential effects of these three classes of pesticides on cognition in children.

Our study design identified high and low risk groups originally based on FW status which correlates with general OP pesticide exposure. It is reasonable to expect that the children from FW families will have higher risk of exposure to OP pesticides due to their ban from use other than for agricultural applications. A recent publication (Arcury et al., 2020) confirmed a single acute exposure measurement in this study population. That same manuscript also examined OC and pyrethroid exposure and noted detections were greater in the NFW than those from FW families. Each of these pesticide classes has known neurotoxic effects. The OPs exhibit hazardous effects by inhibiting acetylcholinesterase (Richardson et al., 2019) and altering gene expression (Naughton and Terry, 2018). They have been known to cause changes in motor coordination, vision, attention, loss of memory, and respiratory issues (Costa, 2006; Naughton and Terry, 2018; Richardson et al., 2019; Roldan-Tapia et al., 2006). The OCs and pyrethroids both affect neurotransmission through Na channels. OCs additionally interfere with the inhibitory effects of GABAα receptors, causing hyperexcitability and in extreme cases, seizures (Costa, 2015; Jayaraj et al., 2016). There is evidence that pyrethroids and OCs may initiate cascades that may affect levels of neurotransmitters (Costa, 2015; Richardson et al., 2019). Nevertheless, pyrethroids remain a standard pest control product used by homeowners to control various pests (lice, mosquitos, bed bugs and more) (Jayaraj et al., 2016; Richardson et al., 2019). It is not unusual that they would be found in high concentrations in urban settings, particularly in low-income areas. The hazardous effects of OCs have been abundantly documented, and these substances have even been implicated in neurodegenerative disorders (Costa, 2015; Jayaraj et al., 2016; Richardson et al., 2019). The EPA officially banned their use in the 1970s, over 4 decades ago (Costa, 2015), making their continued presence quite alarming if they have deleterious effects on cognitive development. Detection of these compounds in the current study population is most likely due to the extreme persistence and bioaccumulation of OCs throughout the environment. While not examined as part of the current study, it is important to note the compounded effects that arise from mixtures of various pesticides (Dorea, 2021; Grandjean and Landrigan, 2014; Mostafalou and Abdollahi, 2017). Table 1 and Arcury et al. (2021) demonstrate that while the study design successfully identified high (FW) and low (NFW) OP exposure groups, there was evidence of additional pesticides in the two groups as well. We did not design the present study to investigate mixture effects resulting from concurrent exposure to multiple pesticides. However, the results presented here stress the continued need for future studies to include assessments of pesticide mixtures in the study design.

Despite the established mechanisms of actions for these pesticides, little data exists about their direct association with neuroanatomical anomalies or other adverse behavioral effects. We know that the two indices found in the current study, VSI and VCI, are associated with reasoning and concept formation. Specifically, the VSI measures a child’s nonverbal reasoning and concept formation, visual perception and organization, visual-motor coordination, and ability to analyze and synthesize abstract information. The VCI specifically measures verbal reasoning, understanding, concept formation, crystallized intelligence, or the knowledge acquired through life experiences and learning. Given this understanding, it is not surprising that both scores changed, as they represent different aspects of similar cognitive traits. However, the current study cannot infer the direct mechanism of how exposure to pesticides might affect performance on neurobehavioral assessments.

Additionally, little knowledge exists about the neural correlates related to cognitive functions that underlie WISC-V performance. A few studies have attempted to link the effects of pesticide exposure on cognitive function back to specific brain regions. Sagiv et al. (2019) utilized functional near-infrared spectroscopy (fNIRS), an imaging method that allows the assessment of neural activity, to examine the effects of prenatal pesticide exposure. Participants with high levels of exposure showed a decrease in the prefrontal cortex during a cognitive flexibility task. Rauh et al. (2012), found that children with high chlorpyrifos (an OP pesticide) exposure had differences in anatomy in the superior temporal and inferior frontal cortices and areas associated with the default mode network compared to a low-exposure group. The brain differences in children with high exposure were correlated with lower IQ scores (Rauh et al., 2012). Further investigation of neuroanatomy and neuronal networks combined with neurobehavioral assessments and pesticide exposure will be necessary for further understanding the functional impact of prenatal and early childhood pesticide exposure and the mechanisms by which different pesticide classes alter brain function.

5. Potential Limitations

The current study is distinctive in its direct assessment of cognitive performance among children with high (rural, FW) and low (urban, NFW) agricultural pesticide exposure. While providing additional evidence for the adverse impact of pesticide exposure on child cognition, our study design encompasses a narrow and specific population of children and may not directly reflect exposure effects within the general public. The inclusion criteria of both ethnic and economic status can impact our results, as demonstrated by the marginal underperformance of these children compared to the general population averages for the WISC-V. We adjusted for various covariates in our study to account for these effects but acknowledge this limitation of our study design. A further limitation of the study is that we cannot infer the cause of the differences in cognitive function between the two study populations. Generally, we strived to successfully recruit populations on many compatible variables except for belonging to FW families. However, a few potential causal intermediaries (birth weight and learning disabilities) and univariate demographic features (housing, water quality, and ADHD diagnosis) have the potential to influence cognitive function despite omission in our analysis. Additionally, as we have recently shown (Arcury et al., 2020), both groups of children have acute pesticide exposure, and agricultural pesticides were not the only compounds detected. All pesticide measurements were made closely in time with the cognitive testing, but they do only represent a single (1-week) exposure period. While data from our early life questionnaires indicate long-term exposure was likely for the children tested, based on the wristband exposure data, the observed effects may not reflect continuous exposure effects. Further, FW and NFW children had documented differences in the seasons of measurement.

With these limitation in mind and noting that our study design did not include a moderation analysis, our analyses did find OCs to account for a reasonable percentage of the differences in cognitive performance observed between FW and NFW children. Our post-hoc moderation analysis showed trends consistent with increases in exposure being related to decreases in cognition, but the findings were not statistically significant. Furthermore, the ecological design of our study only allows for an exploration of associations between exposure and cognitive changes. While this study did not have a sufficient sample size or pesticide dosage data for establishing a concentration-dependent relationship between pesticide exposure and cognitive performance, future studies would be strengthened by including biological markers of exposure as has been performed elsewhere (Bouchard et al., 2011; Cartier et al., 2015; Engel et al., 2011; Rauh et al., 2011). Future studies should consider such limitations to better address the significance of pesticide exposure on group differences. Finally, pesticide exposure is likely not the only factor associated with cognitive performance in these groups, and observational group comparisons cannot establish causality. Currently, longitudinal investigations are ongoing within these populations to assess changes in cognition and multi-year pesticide exposure to potentially address some of these limitations.

6. Conclusion

Pesticides remain present in the environment and pose significant health concerns, specifically to vulnerable communities. This study compared cognitive performance between populations expected to have high (FW) and low (NFW) pesticide exposure and found, surprisingly, that the NFW population demonstrated lower performance than FW. NFW children exhibited lower performance than FW children FSIQ and all indices on the WISC-V, with differences statistically significant for VCI and VSI. Follow up analyses to further investigate this finding raise the possibility that OC exposure may be contributory given the greater proportion of children with detectable exposure in the NFW children. The prohibition of this class of pesticides has been in place for several decades, making these results intriguing. It is advisable to relate cognitive differences with pesticide exposure as a preliminary conclusion. However, they raise the crucial hypothesis that the potent bioaccumulation of these chemicals in the environment results in harmful exposure and alters cognitive function in children in these communities. Additionally, these results highlight the importance of performing future assessments of multiple pesticide classes within subject populations, as several pesticide types are of great concern to the neurodevelopment of children.