To our knowledge, there have not been any previous studies evaluating residential proximity to reported agricultural pesticide use and more subtle neurodevelopmental outcomes such as cognitive function in children. In the present study, we evaluated the relationship between prenatal residential proximity to agricultural use of a variety of neurotoxic pesticides and neurodevelopment using WISC assessed at 7 y of age because this test provides a more specific and reliable measure of cognition than earlier neurodevelopmental assessments conducted in our cohort and may have greater implications for school performance.We estimated agricultural pesticide use near each woman’s residence during pregnancy using California PUR data from 1999– 2001 . We selected potentially neurotoxic pesticides with agricultural use in our study area during the prenatal period, including fifteen OPs and six carbamates , two manganese -based fungicides , eight pyrethroids , and one neonicotinoid . The PUR data include the amount of active ingredient applied, the application date, and the location, defined as a 1-mi2 section defined by the Public Land Survey System . We edited the PUR data to correct for likely outliers that had unusually high application rates by replacing the amount of pesticide applied based on the median application rate for that pesticide and crop combination . For each woman, we estimated the amount of all pesticides in each pesticide class used within a 1-km radius of the pregnant woman’s residence using the latitude and longitude coordinates and a geographic information system.
In all cases, the 1-km buffer around the home included more than one PLSS section; thus,blueberries in containers growing we weighted the amount of pesticide applied in each section by the proportion of land area that was included in the buffer. We selected a 1-km buffer distance for this analysis because it best captures the spatial scale most strongly correlated with measured agricultural pesticide concentrations in house-dust samples . Detailed descriptions of the equations and methods that we used to calculate nearby pesticide use have been published previously . We estimated pesticide use within 1 km of the maternal residence during each trimester of pregnancy for participants with residential location information available for two or more trimesters and computed the average pesticide use during pregnancy by summing the trimester-specific values and dividing by the number of trimesters included. We also created individual variables for nearby use of each of the five individual OP pesticides with the highest use in our study area during the prenatal period .In addition to examining the simple sum of pesticide use in each class, we also used relative potency factors to generate class-specific toxicity-weighted sums to account for differences in the neurotoxicity of individual pesticides in OP and carbamate classes. The RPF of a chemical is the ratio of the relevant toxicological dose of an index chemical to the relevant toxicological dose of the chemical of interest. Currently, RPFs are available for OP and carbamate pesticides, but not for neonicotinoids, pyrethroids, or Mn-fungicides. Thus, we were only able to create toxicity-weighted sum variables for OPs and carbamates. We calculated the toxicity-weighted use for each OP or carbamate pesticide, expressed as kg-equivalents of chlorpyrifos, by multiplying the kilograms of pesticide used within 1 km of the maternal residence during each trimester for each pesticide by the RPF of that pesticide, and summing to create the toxicity-weighted use for the fifteen OP and six carbamate pesticides.
Additionally, because these pesticides share a common mechanism of toxicity, we also generated a toxicity-weighted sum for OPs and carbamates combined . The RPFs, total kilograms and toxicity-weighted kilograms of use in the Salinas Valley in 2000 for each OP and carbamate pesticide are provided in Table S1. Finally, we created a rank index of neurotoxic pesticide use that included the five pesticide classes of interest by generating a percentile rank of the participants from lowest to highest use for each neurotoxic pesticide class and then calculating the average percentile rank across the five classes. We also explored principal components analysis as a method for combining pesticide use across the five different classes of neurotoxic pesticides.We log10-transformed continuous pregnancy average and trimester specific sums of pesticide use to reduce heteroscedasticity and the influence of outliers and to improve the linear fit of the model. The scores for Full-Scale IQ and for the four subdomains were normally distributed and were modeled as continuous outcomes. We selected model covariates a priori based on factors associated with infant neurodevelopment in previous analyses [i.e., child’s exact age at assessment, sex, maternal PPVT score and maternal education ]. We considered the following variables as additional covariates in our models : maternal country of birth, maternal age at delivery, marital status at enrollment, and maternal depression [using the Center for Epidemiologic Studies Depression Scale ] at the child’s 7-y visit. In addition, we considered covariates collected at each visit including housing density , HOME score , household poverty level , presence of father in the home , maternal work status, location of assessment , and season of assessment. We imputed missing values at a visit point using data from the nearest available visit. We retained covariates that were significant at any time point in the multivariate regression models and used the same covariates in all models.
We fit separate regression models for each pesticide class or individual pesticide; we also fit models including multiple classes or pesticides. We used generalized additive models with a three degrees-of-freedom cubic spline function to test for nonlinearity. None of the digression from linearity tests was significant ; therefore, we expressed neurotoxic pesticide use linearly in regression models.We observed an inverse relationship between the agricultural use of OP pesticides within 1 km of maternal residences during pregnancy and cognitive development in children at 7 y of age. For each standard deviation increase in agricultural use of total OPs or toxicity-weighted OPs, there was a 2-point decrease in Full-Scale IQ. To put these findings in perspective, other authors have estimated that each 1- point decrease in IQ decreases worker productivity by ∼2% , reducing lifetime earnings by US$18,000 in 2005 dollars . The results were independent of prenatal urinary DAP concentrations in the model, and the effect estimates of nearby OP use and urinary DAPs were of similar magnitude. These independent associations suggest that our previous observation of a relationship between prenatal urinary DAPs and IQ did not completely account for exposure to OP pesticides during pregnancy and that using both urinary DAPs and PUR data seems to provide a morecomplete characterization of OP pesticide exposure. Urinary DAP concentrations provide an estimate of exposure to some, but not all, of the OP pesticides we evaluated using PUR data and primarily reflect dietary exposures . The two individual OP pesticides that had the strongest inverse relationship with Full-Scale IQ were acephate and oxydemeton-methyl,planting blueberries in containers but agricultural use of these two pesticides was highly correlated . It is important to note that although oxydemeton-methyl devolves to urinary DAPs, acephate does not; it is also important to note that oxydemeton-methyl is the most toxic of all the OPs used in the Salinas Valley . Agricultural use of other potentially neurotoxic pesticide classes was correlated with the use of OPs, and there were also significant inverse associations between Full-Scale IQ and nearby agricultural use of pyrethroid insecticides, Mn-based fungicides , and a neonicotinoid insecticide . The combined agricultural use of pesticides from five neurotoxic pesticide classes based on an average rank index and PCA produced similar results to those observed for toxicity-weighted OP pesticide use alone, making it difficult to determine whether OP pesticide use alone is driving the relationship or if the results are caused by the combined use of neurotoxic pesticides that are highly correlated. This is the first study to evaluate the relationship between cognitive abilities in children and reported agricultural use of neurotoxic pesticides near maternal residences during pregnancy. A recent study conducted in Spain that used residential proximity to agricultural fields as a proxy for pesticide exposure observed an inverse relationship between postnatal, but not prenatal, hectares of crops near the residence and Full-Scale IQ, Verbal Comprehension, and Processing Speed in children 6–11 y of age .
A study in California utilizing PUR data found that any agricultural use of OPs or pyrethroids within 1.5 km of maternal residences during the third trimester of pregnancy compared with no agricultural use of these pesticides was associated with an approximately doubled risk of autism spectrum disorder . Higher concentrations of pyrethroid metabolites in children’s urine have been associated with an increased risk of behavioral problems in school-age children and with attention deficit/ hyperactivity disorder in one study but not in another . Previous studies have observed inverse associations between children’s cognition and levels of manganese in blood and hair . Proximity to agricultural use of the neonicotinoid imidacloprid during pregnancy has been associated with an increased risk of neural tube deffects , but there are no previously published studies evaluating cognition in children. The main strength of this study is the use of PUR data, which provide the amount of active ingredients in and the locations of all agricultural pesticide applications, and which represent a major improvement in exposure classification compared with using only crop locations, as in the recent Spanish study . We also had extensive information on potential confounders and other chemical exposures available for the CHAMACOS cohort. However, there are some limitations of our study. We were able to determine proximity to agricultural pesticide use only at the maternal residence, but not at other locations where the mother may have spent time. Although we used residential proximity to agricultural pesticide use as a proxy for pesticide exposure, previous studies have shown that PUR data are correlated with environmental pesticide concentrations , suggesting that these data provide a meaningful indicator of pesticide exposure. We did not account for prenatal exposure information from other potential sources of pesticide exposure including home use, occupational take-home, and dietary intake, but our models included prenatal urinary DAPs, which we believe primarily reflect both dietary and residential exposures . In general, these limitations would likely lead to exposure misclassification and would bias our results toward the null. People living in agricultural communities are exposed to a complex mixture of many individual pesticide active ingredients as well as to potentially neurotoxic adjuvants included in the formulation. Better methods are needed for toxicity weighting across neurotoxic pesticide classes. To improve pesticide exposure assessment based on PUR data, exposure models should be optimized using measured pesticide concentrations in air or house-dust samples. In future analyses, we intend to use more complex statistical methods to address collinearity of exposures, such as weighted quantile sum regression and hierarchical Bayesian models . The results from the present study need to be replicated in studies that include children living in both agricultural and non-agricultural communities to obtain more variability in exposure to pesticide mixtures.In 2011, around 4200000 hectares of land were devoted to tobacco growing, representing less than1% of total arable land globally; however, in several low- and middle-income countries, the percentage of arable land devoted to tobacco growing has recently increased.1 For example, it has almost doubled in China, Malawi and the United Republic of Tanzania since the 1960s. Deforestation for tobacco growing has many serious environmental consequences – including loss of biodiversity, soil erosion and degradation, water pollution and increases in atmospheric carbon dioxide. Tobacco growing usually involves substantial use of chemicals – including pesticides, fertilizers and growth regulators.These chemicals may affect drinking water sources as a result of run-off from tobacco growing areas. Research has also shown that tobacco crops deplete soil nutrients by taking up more nitrogen, phosphorus and potassium than other major crops. This depletion is compounded by topping and de-suckering plants, which increase the nicotine content and leaf yields of tobacco plants.3 Land used for subsistence farming in low- and middle-income countries may be diverted to tobacco as a cash crop. Intensive lobbying and investments by multinational tobacco companies and leaf buyers along with market liberalization measures have encouraged the expansion of tobacco agriculture in low- and middle-income countries.