Because of these findings and our previously reported positive associations between AChE activity and neurobehavioral performance in boys and transient alterations in neurobehavioral performance in relation to a peak pesticide spray season , we hypothesized that children living near greenhouse floricultural production sites, assessed during the same time period with respect to harvest season, had worse neurobehavioral performance compared to children living farther away. Seventy three percent of ESPINA participants were recruited using contact information from their participation in the 2004, county-wide, Survey of Access and Demand of Health Services in Pedro Moncayo County, a large representative survey of Pedro Moncayo County conducted by Fundacion Cimas del Ecuador in collaboration with the communities of Pedro Moncayo County. Using home addresses from this survey, 419 participants were contacted and invited to participate in the ESPINA study. Participant losses in this case were mostly due to missing or inaccurate information or changes in residential address. The remaining 27% of participants were recruited by word of mouth or community announcements. Children were selected after parents completed a pre-survey, which provided enough information to classify children who lived with a floricultural worker or without any agricultural workers. To be included, children who lived with a floricultural worker must have done so for at least one year. Children living without floricultural workers included children who had never cohabitated with any agricultural worker,10 plastic plant pots had never inhabited a home where agricultural pesticides were stored and had no previous direct contact with agricultural pesticides. In 2008 a total of 313 children between the ages of 4 and 9 were examined in the months of July and August.

Further assessment included parental surveys at the subjects’ homes and examinations of child participants in seven schools in Pedro Moncayo County. The present analyses include 304 participants who had information for all covariates of interest. Among the participants, informed consent was obtained from at least one parent in each case and verbal assent was obtained from all children older than seven years of age. Cases where parental consent or child assent were missing are not included in this study. Participant recruitment is described in more detail in Suarez-Lopez et al., 2012. The ESPINA study was approved by the Institutional Review Boards of the University of Minnesota, The University of California San Diego, Universidad San Francisco de Quito and the Ministry of Public Health of Ecuador, and is endorsed by the Commonwealth of Pedro Moncayo County. Geographical coordinates of Pedro Moncayo County homes were obtained through portable global positioning system receivers in 2004, 2006 and 2010 by Fundacion Cimas del Ecuador, as part of the System of Local and Community Information . Flower plantation edges were created using satellite imagery from 2006 . Distance between the child’s home and the nearest 1m segment of the nearest flower plantation perimeter was calculated using ArcGIS 9.3. We also calculated the total area of flower crops within 100m of participants’ homes.Neurodevelopmental assessments were conducted using the NEPSY-2, a standardized and validated neurodevelopment test battery for children of ages 3 to 16 years . Children were tested in 11 age-appropriate sub-tests in five domains: Attention & Inhibitory Control , Language , Memory & Learning , Sensorimotor and Visuospatial Processing .

Descriptions of each sub-test have been described elsewhere . Three sub-tests required translation into Spanish using terminology appropriate for the local population . Translation of the NEPSY-2 test has been found to be relatively unaffected by language and culture . The neurodevelopmental assessments were conducted by trained examiners in a quiet and controlled setting, and the evaluations lasted for no more than two hours. Only one child and one examiner where allowed in a room. A parent was allowed in the examination room if the parents or children requested so. In such cases, the parent was required to be in complete silence and outside of the participant’s line of sight. Additional details of these assessments are described in previous publications . Neurobehavioral sub-test scaled scores were calculated using the NEPSY-2 scoring assistant software . Most sub-test scores consisted of primary scaled scores, which are age-adjusted values based on a national normative sample of US children . Higher scores represent better performance. Neurobehavioral scaled scores were designed to range from 1 to 19 , with scores of 6 or 7 considered borderline and scores less than 6 considered below expected. For sub-tests that were composed of more than one primary scaled score , we used the average of all available primary scaled scores per sub-test for the domain score calculation. For sub-tests that included both correct and error components or time and error components , we used the combined scaled scores as primary scaled scores. The only sub-test within the Sensorimotor domain was Visuomotor Precision; hence, the Sensorimotor domain score is equal to the Visuomotor Precision scaled score. We calculated a total neurobehavioral summary score as the average of primary scores of all eleven sub-tests.

Domain scores and the neurobehavioral summary score were used as the measures of neurobehavioral performance, to reduce the number of associations tested thereby reducing the likelihood of type 1 errors. Children’s height was measured to the nearest 1 millimeter, using a height board and following recommended procedures . Because height for age is a better indicator of under-nutrition than BMI , we calculated z-scores for height -for-age using the World Health Organization growth standards and used this variable as an indicator of malnutrition . Hemoglobin concentration was measured from a drop of blood from a finger-stick sample using the EQM Test-mate ChE Cholinesterase Test System 400 . Other covariates, including self-reported maternal education, race, and cohabitation with a flower plantation worker, were obtained through interviews conducted at participants’ homes. Because previous ESPINA analyses have shown that the season during which a child was assessed may affect both exposure indicated by acetylcholine esterase activity as well as neurodevelopmental scores , time between Mother’s Day harvest and date of outcome assessment was also considered. To include children with missing information in multi-variable analyses, we created a “missing” race category, and imputed information of maternal education for 15 children with missing information by substituting the “household head’s” education reported in the SILC 2004. In primary analyses used linear regression analyses to test the associations between neurobehavior scores and residential distance to the nearest flower plantation. We also estimated odds ratios of low neurodevelopmental scores associated with distance to the nearest flower plantation using multivariate logistic regression models. In both linear and logistic regression analyses, our main adjustment model included adjustment for child’s age, sex, race, maternal education,plastic pots large cohabitation with a flower plantation worker and 2 constructs of nutrition: height-for-age z-score and hemoglobin concentration. These variables are known to affect or correlate with neurobehavioral development in children. In sensitivity analysis, we assessed robustness of results to choice of covariates by additionally including income and household pesticide use. We also assessed effect modification by child sex using a multiplicative interaction term since we previously observed that the associations between AChE and neurobehavior were stronger among boys than girls in this study population . In additional analyses, we assessed possible nonlinearities in the exposure-response relationship in a number of ways. We modeled categorical exposure as an alternative to analyzing exposure as a continuous measure . In all cases we used the same main adjustment models described above.

We first analyzed exposure as quartiles of distance to the nearest plantation. Furthermore, considering that flower crops are enclosed within greenhouses, which reduce the amount of pesticide drift, we modeled residential proximity using finer categories within close proximity: 0-50m, 51-100m, 101-500m, and >500m. To estimate associations with binary outcomes, we combined the categories of 0-50m and 51-100 to improve the statistical power of analyses considering the small number of participants within each of the categories . We additionally applied general additive models , using smoothing spline functions to fit the relationship between continuous distances and sub-scale scores, adjusted for covariates. In order to assess whether the amount of crop area near a residence is associated with the outcome we examined total flower crop areas within a 100m buffer around participants’ homes as a continuous variable and as a 3-category variable: no growing areas within 100m of the residence , crop area within 100m below the median and crop area within 100m above the median. In sensitivity analyses, we used the same approach with 150m and 200m buffers. Analyses were conducted using proc glm and proc logistic in SAS software, Version 9.4 and GAM analyses were done in R 3.6.0 using the mgcv package. The mean age of children at the time of assessment was 6.6 years ; 51% were male, 72% reported mestizo as their race, 21% indigenous, 2% other and 5% had missing information . of the 307 participants lived with a floricultural worker. The means of height-for-age z-score and hemoglobin concentration were −1.25 and 12.7 mg/dL , respectively. On average, children in this sample lived 446 m from the edge of the nearest flower crop . In this sample of children, 6% lived within 50m, 11% lived within 100m, 27% lived within 200 m, and 61% lived within 500 m of a plantation’s edge. The average quantity of growing area within 100m of the residence was 292 m2 . The neurobehavioral sub-test scores for subjects in our study sample were lower but with similar variability to those of the NEPSY-2 normative sample of U.S. children . The neurobehavioral domain scores of our participants ranged from 4 to 13, with mean scores of 6.6 in Language, 8.8 in Memory and Learning, 9.9 in Sensorimotor, 9.6 in Visuospatial and 8.5 in the Attention and Inhibitory domain. The lowest domain scores were observed in the Language and Attention & Inhibitory Control domains. The percentages of children with scores of clinical concern for their age group were: 10% in the Memory domain, 12% in Visuospatial Processing, 15% in Attention & Inhibitory Control, 17% in Sensorimotor, and 33% in the Language domain . A decrement in the Language domain score was observed per 100m decrease in distance from the child’s residence to the edge of the nearest flower crop of −0.08 units . An association with the Attention & Inhibitory Control domain was also observed, although it was weaker . No associations were observed with Memory & Learning, Sensorimotor or Visuospatial Processing in analyses of neurobehavior scores as continuous outcomes and proximity to floricultural crops as a continuous exposure, adjusting for age, sex, race, height-for age z-score, hemoglobin, maternal education, and cohabitation with a flower plantation worker. Additional adjustment for income or residential pesticide use did not meaningfully change the results . We assessed evidence for effect modification by child sex and saw no statistically significant interaction . When using clinical thresholds to define a binary outcome we observed that for every 100m of closer residential distance to floricultural crops, the odds of low overall neurobehavioral score increased by 10% . For every 100m closer in proximity to treated floricultural crops, the odds of low Language score increased by 9% , and the odds of low Memory & Learning score increased by 24% . In the Sensorimotor domain, we observed results in the direction opposite to that hypothesized: children living closer to crops had significantly lower odds of a low Sensorimotor score . No clear threshold effect was observed using quartiles of exposure . However, using categories to distinguish those living closest to floricultural crops, we observed that participants who lived within 50 m of a flower crop had the lowest neurobehavioral scores in three out of the five domains assessed . Compared to children living >500m from a flower crop, children living within 50m of a crop had statistically significant lower scores in the domains of Attention & Inhibitory Control, Language, and Memory & Learning. A similar trend was observed in logistic regression analyses of the dichotomous outcomes . In these analyses, categories of 0-50m and 51-100m were combined to improve statistical power and model stability. Nonetheless, estimates still had wide confidence intervals for the overall score and the Memory & Learning domain.