Integrating NHANES and Toxicity Forecaster Data to Compare Pesticide Exposure and Bioactivity by Farmwork History and US Citizenship

toxicology, bioactivity, environmental health, human health, pesticides, occupational 4 health, farmworkers 5 6 Authors: Chanese A. Forté 1,2, Jess A. Millar 3,4, and Justin Colacino 1,5,6 7 8 Author Affiliations: 1. The University of Michigan School of Public Health, Department of 9 Environmental Health Sciences, Ann Arbor, MI, USA 10 2. The University of Michigan, College of Engineering, Michigan Institute of Computational 11 Discovery and Engineering, Ann Arbor, MI, USA 12 3. The University of Michigan School of Public Health, Department of Epidemiology, Ann Arbor, 13 MI, USA 14 4. The University of Michigan Medical School, Department of Computational Medicine and 15 Bioinformatics, Ann Arbor, MI, USA 16 5. The University of Michigan School of Public Health, Department of Nutritional Sciences, Ann 17 Arbor, MI, USA 18 6. University of Michigan College of Literature, Sciences, and the Arts, Program in the 19 Environment, Ann Arbor, MI, USA 20 21 ORCIDs: 22 Chanese A. Forté, 0000-0002-6540-3180 23 Jess A. Millar, 0000-0001-8945-3396 24 Justin Colacino, 0000-0002-5882-4569 25 26 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. 2 Corresponding Author: 27 Justin Colacino, 28 +1 734 – 647 – 4347 29 [email protected] 30 6611D SPH1, 1415 Washington Heights 31 Ann Arbor, MI, 48104 32 33 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 3

34 35 Introduction— Farmworkers in the United States, especially migrant workers, face unique 36 barriers to healthcare and have documented disparities in health outcomes. Exposure to 37 pesticides, especially those persistent in the environment, may contribute to these health 38 disparities. 39 40 Methods—We queried the National Health and Nutrition Examination Study (NHANES) from 41 1999-2014 for pesticide exposure biomarker concentrations among farmworkers and non-42 farmworkers by citizenship status. We combined this with toxicity assay data from the US 43 Environmental Protection Agency’s (EPA’s) Toxicity Forecast Dashboard (ToxCast). We 44 estimated adverse biological effects that occur across a range of human population-relevant 45 pesticide doses. 46 47 Results—In total, there were 1,137 people with any farmwork history and 20,205 non-48 farmworkers. Of the 14 commonly detectable pesticide biomarkers in NHANES, 2,4-49 dichlorophenol (OR= 4.32, p= 2.01x10-7) was significantly higher in farmworkers than non-50 farmworkers. Farmworkers were 1.37 times more likely to have a bioactive pesticide biomarker 51 measurement in comparison to non-farmworkers (adjusted OR=1.37, 95% CI: 1.10, 1.71). 52 Within farmworkers only, those without U.S. citizenships were 1.31 times more likely to have 53 bioactive pesticide biomarker concentrations compared those with U.S. citizenship (adjusted OR 54 1.31, 95% CI: 0.75, 2.30). Additionally, non-citizen farmworkers were significantly more 55 exposed to bioactive levels of /g2010-hexachlorocyclohexane (BHC) (OR= 8.50, p= 1.23x10-9), p,p-56 DDE (OR= 2.98, p= 3.11x10-3), and p,p’-DDT (OR= 10.78, p= 8.70x10-4). 57 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 4 58 Discussion— These results highlight pesticide exposure disparities in farmworkers, particularly 59 those without U.S. citizenship. Many of these exposures are occurring at doses which are 60 bioactive in toxicological assays. 61 62 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 5 1.1 Introduction 63 Pesticide exposure has been linked to a myriad of human health outcomes such as obesity, 64 immune alteration, cancer, neurological conditions, type II diabetes mellitus, and death (Wei et 65 al. 2014; Zong et al. 2018; Medehouenou et al. 2019). More specifically, many pesticides are 66 strong endocrine disruptors because they mimic hormones like estrogens and androgens (Briz 67 et al. 2011; Wong et al. 2019). Persistent pesticides last in the environment and human body for 68 years or even decades and can bioaccumulate and bioconcentrate. Persistent pesticides 69 include organochlorines like dichlorodiphenyltrichlorethane (DDT), Lindane, Chlordane, Dieldrin, 70 Heptachlor and their metabolites. Non-persistent pesticides include organophosphates, 71 carbamates, pyrethroids, chlorinated phenols, acyl alanine fungicides and more chemical 72 groups, and were thought to be the less harmful answer to previously used persistent chemicals 73 (e.g. organochlorines) (Abubakar et al. 2020). However, non-persistent chemicals still affect 74 human health. While pesticides are associated with endocrine disruption, cancers, and motor 75 neuron disorders, there is still a lack of human health data on the dose-response, toxicological 76 mechanisms, or how population exposure concentrations relate to social determinants of health 77 (Mostafalou and Abdollahi 2013; Dhananjayan and Ravichandran 2018). 78 Social determinants of health like occupation or citizenship can alter both exposure and 79 health outcomes related to chemicals like pesticides. Healthcare policy and services are limited 80 to non-existent for immigrants and especially migrant workers residing in the United States 81 (US). For example, many policies that on the surface appear highly beneficial for the American 82 people like the Affordable Care Act of 2010, actually exclude immigrants completely from 83 accessing care (Quesada et al. 2011). In addition, agreements like the North American Free 84 Trade Agreement between the US, Canada, and Mexico limit migrant worker rights (Barnes 85 2013). Moreover, migrant worker health is often unprotected by the law and workplace 86 discrimination leaves migrant workers very vulnerable (Quesada et al. 2011; Ramos et al. 2016; 87 Ramos 2018; Saxton and Stuesse 2018). Prior research on migrant workers in the US Midwest 88 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 6 found factors like economics, logistics, and health significantly affected the mental health of 89 migrant workers (Ramos et al. 2015). Overall, a gap exists in the quantification of pesticide 90 exposure among farmworkers and migrant workers, and specifically how these exposures may 91 differ by worker category or US citizenship status. 92 A major challenge in the field of occupational and environmental health is understanding 93 and predicting the health effects of exposure to chemicals like pesticides. There are currently 94 85,000 chemicals on the global market that Toxic Substances Control Act (TSCA) has listed in 95 its inventory of substances, and there is little to no experimental toxicology or epidemiology data 96 on many of them (Attene-Ramos et al. 2013; Adeola 2021). In 2008, the US Environmental 97 Protection Agency (EPA) collaborated with multiple other federal agencies including the Food 98 and Drug Administration and the National Institute of Environmental Health Sciences to create 99 the Toxicology in the 21st Century (Tox21) program (Thomas et al. 2018). The goal of Tox21 is 100 to develop high throughput testing methods to determine the safety of chemicals such as food 101 additives and pesticides. Additionally, Tox21 quantifies the biological mechanisms that 102 chemicals alter to prioritize the chemicals being tested and generate a wealth of data to predict 103 toxicological responses in the human body (Attene-Ramos et al. 2013; Thomas et al. 2018). 104 These data are a rich, but untapped, resource to characterize the dose-dependent effects of 105 exposure to pesticides in the context of social determinants of health like occupation and 106 citizenship. This data is then presented in the Toxicity Forecast Dashboard (ToxCast). 107 To address these gaps and understand how pesticide exposure and effects vary by 108 occupation and citizenship, this study’s goal is to determine if people residing in the US are 109 exposed to bioactive concentrations of pesticides. This project has the following aims: 1) 110 quantify and compare pesticide biomarkers among farmworkers and non-farmworkers, 2) 111 quantify and compare pesticide biomarkers between citizen and non-citizen farmworkers, 3) 112 compare exposure concentrations to known bioactive benchmark concentrations in the Tox21 113 high throughput toxicity data (ToxCast). We hypothesized that on average farmworkers will have 114 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 7 higher concentrations of pesticides biomarkers than non-farmworkers. Furthermore, among 115 farmworkers, we hypothesize that non-citizens will have higher pesticide biomarker 116 concentrations than US citizens. Additionally, we hypothesize people residing in the US will be 117 exposed to bioactive concentrations of pesticides. Moreover, we hypothesize farmworkers will 118 be exposed to bioactive concentrations of pesticides more frequently than non-farmworkers. 119 120 1.2 Methods 121 Our overall study design involves comparing the distributions of chemical biomarker 122 concentrations in The National Health and Nutrition Examination Survey (NHANES) with the 123 distributions of doses for those chemicals which exhibit bioactivity in ToxCast. In addition, we 124 quantify which cellular target families are most often affected by these pesticides and look to 125 see how these target families differ by history of farmwork and U.S. citizenship status. 126 127 1.2.1 The National Health and Nutrition Examination Survey (NHANES) 128 NHANES is a cross-sectional study representative of the US population with 129 oversampling weights for minoritized populations. NHANES is a cross sectional assessment of 130 the health and nutrition of adults and children residing within the US. The current iteration of the 131 continuous study began in 1999. Study participants are enrolled on a continuous basis, with 132 data analyzed and deposited in two-year windows. NHANES collects extensive information on 133 the study participants such as self-reported occupation, urinary and serum biomarkers, and self-134 reported demographics such as age, gender, citizenship, poverty index ratio, and education. 135 136 1.2.2 Study Population 137 This study included NHANES study participants aged 18 years and older who also had 138 occupation and pesticide exposure data present between 1999 and 2014. This study integrated 139 29 datasets from NHANES laboratory data to understand pesticide exposure, occupation, and 140 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 8 demographics of the study population. From the Industry and Occupation Survey, individuals 141 were coded as “farmworker” or “non-farmworker” using the Current Industry (OCD230=1, 142 OCD231=1), Current Occupation (OCD240=18, OCD241=18), Longest Industry (OCD390=1, 143 OCD391=1), and Longest Occupation (OCD392=18), where all participants who put 144 “Agriculture, Forestry and Fishing” were coded as a farmworker. 145 From the demographics data, DMDEDUC2 (older than 18 years of age) and DMDEDUC3 (l8 146 years of age and younger) were combined to create one education level based on the 147 DMDEDUC2 categories. The US citizenship variable (DMDCITZN) is defined as 1= “Citizen by 148 Birth or naturalization” and 2= “Not a citizen of the US”, and we removed anyone who 149 responded with “Refused”, “Don’t Know”, or skipped the question. 150 151 1.2.3 Biomonitoring Samples and Detectability 152 NHANES performs chemical biomonitoring in study participants urine and blood. 153 Participants provided partial urine void in a sterile sampling cup at the mobile examination 154 center. Blood samples are collected by certified laboratory professionals. Urine and blood 155 samples are then analyzed for chemical metabolites using isotope dilution gas chromatography 156 high-resolution mass spectrometry (GC/IDHRMS). Pesticide biomarkers measured in blood 157 samples and reported as either 1) fresh weight basis (i.e., pg/g serum) and 2) lipid weight basis 158 (i.e., ng/g lipid). The lipid adjusted values account for blood lipid concentrations and are of 159 particular importance for the accurate quantification of lipophilic pesticides (Barr et al. 2005). 160 All urinary biomarker measurements were adjusted for urinary creatinine, and all blood 161 pesticide biomarker measurements were blood lipid adjusted. Detectability percentages were 162 calculated by dividing the total number of measurements above LOD by the total number of the 163 chemical’s measurements in NHANES. To ensure that we included chemicals with values 164 above the limit of detection in most of the study participants, detection frequency percentages of 165 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 9 50% and higher across the population were maintained which resulted in 14 chemicals of 166 interest (Silver et al. 2018). 167 These chemicals included the following: 2,4-Dichlorophenol (24DCP), 2,4-168 Dichlorophenoxyacetic acid (24D acid), 2,5-Dichlorophenol (25DCP), 3,5,6-Trichloropyridinol 169 (TCP), 4-Nitrophenol, β -hexachlorocyclohexane (β -HCH), diethyltoluamide acid (DEET acid), 170 Dieldrin, Heptachlor Epoxide, 3-phenoxybenzoic acid (3-PBA), p,p’-DDE, and p,p’-DDT. 171 Additionally, the measurements of TCP, a chlorpyrifos metabolite, were compared to the 172 ToxCast toxicity data for both CPF and chlorpyrifos-oxon (CPO). 173 174 1.2.4 Toxicity Forecast Dashboard Data 175 The US EPA’s Toxicity Forecast Dashboard (ToxCast) is a collection of publicly 176 available high throughput toxicity data intended to make chemical assessment more accessible 177 by allowing researchers to search which chemicals show toxicological effects more easily within 178 human tissue. High throughput toxicity screening initiatives have been developed to quantify 179 biological effects of chemicals, including pesticides, in vitro. Dose response curves are created 180 for each chemical and assay, and from these curves the activation concentrations and positive 181 hitcalls are defined. ACC is the concentration at which the model reaches the cut-off values for 182 the chemical to be considered active and is based on the levels of significance for the dose 183 curve response. The ACC can be used as a proxy of potency to determine the genes, proteins, 184 enzymes, effects on biological pathway and viabilities at which chemicals are active. 185 186 1.2.5 Comparing NHANES and ToxCast 187 Using the corresponding Chemical Abstracts Service Registry Numbers (CASRNs) 188 obtained from PubChem, data from ToxCast were matched to NHANES. From this new dataset, 189 we created pesticide concentration distribution boxplots by the chemical and farmwork history or 190 U.S. citizenship in the tidyverse using the ggplot2 R package (Wickham 2016). Pesticide 191 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 10 distributions were overlaid unto the same axis to quantify overlap between the pesticide 192 concentration distributions of exposure in NHANES participants and bioactivity in ToxCast. To 193 visualize the distribution of exposure in comparison to pesticide bioactivity concentrations, 194 ToxCast ACCs and NHANES biomarker concentrations were plotted as boxplots using molarity 195 units. 196 197 1.2.6 Statistical Analysis 198 All data management and analysis were completed in R version 4.1.3. All code for our 199 work can be found on our GitHub repository (Millar and Forté 2023). Graphics were created 200 using the ggplot2 package library (Wickham 2016). All NHANES data was downloaded using 201 the RNHANES packaged in R (Susmann 2016). The main outcomes of this project include 1) 202 quantifying the distribution of the pesticide concentrations across NHANES and ToxCast, 2) 203 quantifying the demographics of people with and without bioactive measurements, and 3) 204 investigating how bioactivity differs by chemical, farmwork history, and US citizenship status. 205 These outcomes inform the overarching project question of whether people residing in the US 206 are exposed to bioactive levels of pesticides, how these bioactive pesticides affect the body, 207 and whether the rates of exposure to bioactive pesticide concentrations vary based on 208 sociodemographic factors. 209 We labeled anyone who had at least one chemical measurement equal to or above the 210 minimum ToxCast ACC for that chemical as being “bioactive”. Anyone who did not fit this group 211 was defined as “non-bioactive.” Demographics were quantified by bioactivity status among all 212 study participants and then among farmworkers only. For continuous variables like body mass 213 index (BMI) or age in years, we present the mean and standard error, and for all categorical 214 variables, the stratified frequencies and sub-group percentages are provided. 215 Differences in demographic factors by group or citizenship were tested using a 216 Pearson’s chi-square test, using a Rao and Scott Adjustment where necessary for categorical 217 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 11 variables. Low response was defined as 8 or less respondents within one stratum. And for 218 continuous variables, a Wilcoxon Rank test was used to test group means, with a Kruskall-219 Wallis Correction. All significance testing was completed using the NHANES Full Sample 2 and 220 4 Year MEC Exam Weights. A new weight variable titled “MEC16YR” was created using the 221 weighted MEC 2- and 4-year measurements to represent the weights used from 1999-2002 and 222 each year after, respectively. 223 Non-citizen status was determined by the NHANES variable DMDCITZN. We calculated 224 bioactivity by the chemical and marked measurements as bioactive based on their hitcall 225 equaling 1. For model outcomes this bioactivity status by chemical was used as the outcome 226 variable for logistic regression models used to investigate how the odds of being a farmworker 227 and having at least one bioactive measurement differ from non-farmworkers by the chemical. 228 These models were adjusted for BMI, age, poverty index ratio (PIR), survey year, gender, racial 229 ethnicity, U.S. citizenship status, farmwork history, country of birth and education level. After 230 comparing all study respondents’ odds of having a bioactive measure, we created logistic 231 regression models comparing U.S. citizenship status. These models were also adjusted for BMI, 232 age, PIR, survey year, gender, racial ethnicity, country of birth, and education level. 233 Education status was constructed NHANES variables DMDEDUC2 and DMDEDUC3 to 234 include four categories: Less than 9th grade, 9-11th grade (Includes 12th grade with no 235 diploma), High school grad/GED or equivalent, and More than high school. Farmworker status 236 was constructed using NHANES industry or occupation group codes for current job (OCD230, 237 OCD231) or longest job (OCD390, OCD391, OCD392) that included the terms 238 agriculture/agricultural or farming. 239 For lipid adjusted blood measurements, molarity was calculated by multiplying the 240 measurement by serum density of 1.024 g/mL and dividing by molecular weight (Sniegoski and 241 Moody 1979). Urinary measurements were calculated by diving the measurement by molecular 242 weight. All measurements of molarity have units of μ mol/L. 243 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 12 Data from the 1999-2002, 2003-2004, 2005–2006, 2007–2008, 2009–2010, 2011–2012 244 and 2013-2014 data collection cycles were appended, and the sampling weights modified as 245 directed in NHANES documentation. Removal of observations with missing data was done for 246 all analyses. Statistical analysis was done with the R survey package (v4.1-1) to handle 247 complex survey designs present in NHANES. The function survey::svydesign was used to 248 handle sampling weights, with primary sampling units nested within each stratum. 249 Wilcoxon Mann Whitney U test was conducted on individual chemicals in relation to 250 farmworker or non-citizen status using the survey::svyranktest function. The outcome variable 251 for chemicals was calculated as the log molarity for blood measurements and the log of the ratio 252 of the chemical molarity to creatine molarity for urinary measurements. P-values for all tested 253 chemicals were FDR adjusted and AUCs were calculated using the U statistic (Mason and 254 Graham 2002). 255 Both unadjusted and adjusted logistic regression was conducted on individual chemicals 256 in relation to farmworker or non-citizen status using the survey::svyglm function using a quasi-257 binomial model with a logit link. The outcome variable for chemicals was constructed as an 258 indicator variable, with a 1 indicating the measurement was considered chemically bioactive. 259 Adjusted logistic regression included variables for age at screening, race-ethnicity, BMI, 260 education, and survey year for all chemicals, and the additional inclusion of creatine molarity for 261 urinary measurements. P-values for all tested chemicals were FDR adjusted and AUCs were 262 calculated using the WeightedROC R package (v2020.1.31) (Hocking 2020). 263 Initially, the list of pesticides under investigation included 96 different biomarkers present 264 in NHANES, but after removing chemicals with detectability percentages below 50%, we were 265 left with 16 chemicals for analysis (Supplementary Table 1). Assay data for these chemicals 266 from NHANES were then extracted from the ToxCast database. We retrieved the hitcall 267 (representative of an active assay), the activity concentration at cutoff (or ACC), and the 268 intended target family of each ToxCast assay based on the 16 pesticides from NHANES. Using 269 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 13 the hitcall variable, we labeled assays as positive (hitcall==1) or negative (hitcall==0) to mean 270 that an assay did or did not show bioactivity by the pesticide. We created a bioactivity ratio per 271 chemical by dividing the number of positive assays by total number of assays. All chemicals in 272 NHANES were present in ToxCast. However, trans-nonachlor was not maintained in the study 273 because there were only 8 completed assays in ToxCast and none of those assays were active. 274 275 1.3 Results 276 We first assessed demographic features of the study participants based on whether the 277 participant had a history of farmwork or not (Tables 1 and 2). In total, there were 1,137 people 278 who reported any farmwork history, and 20,205 who were categorized as non-farmworkers. The 279 farmworker group was mostly women (N=697, 61.3%), Non-Hispanic White (N=635, 55.8%), 280 U.S. Citizens (N=934, 82.1%) and 26.6% reported some college education or an associate’s 281 degree (N=302). The non-farmworker group had similar mean BMI, age, and poverty index 282 ratio. The non-farmworker group is predominantly men (N=10,187, 50.4%), Non-Hispanic White 283 (N=9,167, 45.4%), had U.S. Citizenship (N=17,626, 87.2%), and 19.2% reported some college 284 or an associate’s degree (N=3,885). 285 To better understand how each of the chemicals relate to each other, Table 3 outlines 286 the pesticides by persistence and frequencies of activity of ToxCast assays. In total, there are 287 15 pesticides that are detectable in NHANES study participants and also assayed in ToxCast. 288 Overall, there were 5 persistent organic pesticides and 10 non-persistent pesticides included in 289 this study. The top three most bioactive pesticides in ToxCast were heptachlor epoxide had the 290 highest percentage of assays which were “active” (39.85%), followed by p,p’-DDT (35.73%) and 291 p,p’- DDE (26.78%). The bioactivity threshold is the lowest ACC of the active assays for a given 292 chemical. These values ranged from 6.5nM (2,4-Dichlorophenoxyacetic acid) to 1.45μM 293 (chlorpyrifos). 294 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 14 Next, we wanted to compare the concentrations of chemicals required to activate the 295 ToxCast assays to the biomarker concentrations measured in people in NHANES. Figure 1 296 presents the distribution of pesticide concentrations among people residing in the United States 297 in orange (retrieved from NHANES), and in blue, the ACCs of active assays retrieved from 298 ToxCast. In this figure, where the pesticide distributions of exposure and bioactivity overlap 299 represents pesticide exposures among the US population that are “bioactive”. Additionally, 4-300 nitrophenol is the only pesticide biomarker in NHANES that does not have human 301 measurements that overlap with the bioactive distribution in NHANES. 302 We present the Mann-Whitney-U Rank Test outcomes by chemical in Supplementary 303 Table 2 to test for differences in biomarker concentration by farmworker status, or within 304 farmworkers, comparing between farmworkers with and without US citizenship. When 305 quantifying the odds of having a bioactive measurement (unadjusted outcomes in Supplemental 306 Table 3, fully adjusted outcomes presented in Figure 2 and Supplementary Tables 4 and 5), we 307 found farmworkers were 4.3 times more likely to have a bioactive measurement in comparison 308 to non-farmworkers for 2,4-D (p=2.0x10-7) while farmworkers were significantly less likely to 309 have a bioactive measurement of 4-Nitrophenol (p= 2.7x10-4). Next, we narrowed our analyses 310 to farmworkers only and found farmworkers living without U.S. citizenships were significantly 311 more likely to be exposed to a bioactive measurement of BHC (OR=8.4, p-value=1.2x10-9, 312 U=13.95), p,p’-DDE (OR=3.0, p-value=3.1x10-3, U=9.43), p,p-DDT (OR=10.8, p-value. =8.7x10-313 4, U=6.56). 314 When trying to understand what intended target families are most affected by these 315 chemicals, Supplementary Table 6 provides the frequency of intended target families by the 316 pesticide. Based on individual intended assay target count, cell cycle (N=487), nuclear receptor 317 (N=318), cytokine (N=143), DNA binding (N=172), and cell adhesion molecules (N=65) were the 318 most frequent targets of the pesticides. Overall, p,p’-DDE (N=305) had the most intended target 319 family counts based on positive assays, followed by p,p’-DDT (N=278), heptachlor epoxide 320 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 15 (N=259), and chlorpyrifos (N=126). Heptachlor epoxide had the highest number of positive 321 assays targeting the cell cycle (N=123) and p,p’-DDT had the second most (N=120). 322 Additionally, for p,p’-DDE had mostly nuclear receptor targeting positive assays (N=102), 323 followed by the cell cycle (N=74) and DNA binding (N=64). 324 325 1.3.1 Discussion 326 When looking at individuals who have pesticide biomarker concentrations at these 327 bioactive levels, demographics statistically differed based on bioactivity, farmwork history and 328 citizenship status. We found NHANES participants are broadly exposed to bioactive 329 concentrations of pesticides. Heptachlor epoxide, p,p’-DDT, and p,p’-DDE were the most 330 bioactive pesticides in ToxCast based on overall percent of positive assays. Disproportionate 331 exposures to bioactive concentrations of pesticides were particularly evident in farmworkers 332 without U.S. citizenship, particularly for persistent pesticides. 333 Pesticide exposures have been associated with increased mortality due to cancer, 334 diabetes mellitus, poisonings, and tuberculosis and other lung infection (Mills et al. 2006; Fry 335 and Power 2017). Pesticide exposure throughout the life course has been associated with 336 breast cancer and dysregulated mammary gland development. For example, mothers with the 337 highest p,p-DDT concentrations were 3.7 times more likely to have daughters who developed 338 cancer by the age of 52 in comparison to mothers with the lowest p,p-DDT blood concentrations 339 (Cohn et al. 2015). Women who are farmworkers and not US citizens could be at increased risk 340 of exposure-associated diseases like breast cancer – these findings warrant further 341 investigation in this area. 342 Citizenship status is also a known barrier to health insurance and treatment (Guadamuz et 343 al. 2020; Chasens et al. 2020), potentially compounding adverse effects of exposure to toxic 344 chemicals like pesticides. In a study of 2,702 participants living with diabetes, non-citizens had a 345 greater risk for poor glycemic management (OR=5.16, 95% CI: 3.73, 6.04) in comparison to 346 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 16 citizens by birth (Chasens et al. 2020). Additionally, citizens by naturalization were also at an 347 increased risk of poor glycemic management (OR=1.95, 95% CI: 1.49,2.55) (Chasens et al. 348 2020). Additionally, this study found that individuals with diabetes and without health insurance 349 were almost twice as likely to have poor glycemic management compared to insured people 350 (OR=1.99, 95% CI: 1.53-2.59). Similar outcomes have also been noted in cardiovascular 351 disease. Using NHANES, researchers retrieved data from 2011 to 2016 to investigate 352 prevalence, treatment, and control of hypercholesterolemia, included 11,680 US-born citizens, 353 2,752 foreign born citizens, and 2,554 non-citizens (Guadamuz et al. 2020). In that study, over 354 half of non-citizens did not have health insurance (52.2); which was significantly more than US-355 born citizens (13.6%, p<0.001) (Guadamuz et al. 2020). 356 Non-citizens also had significantly higher prevalence of diabetes (15.7% vs. 12.8%, 357 p<0.001) (Guadamuz et al. 2020). Treatment percentages were also significantly lower among 358 non-citizens than US-born citizens with hypercholesteremia (16.4% vs 45.5%), hypertension 359 (60.3% vs. 81.1%), and diabetes (51.2% vs. 69.5%) (p<0.001) (Guadamuz et al. 2020). Among 360 noncitizens, those without a usual source of health care or health insurance had lower treatment 361 percentages for hypercholesteremia (2.7% and 8.1%), hypertension (22.2% and 39.1%), and 362 diabetes (15.5% and 28.6%) (Guadamuz et al. 2020). It is very important to understand that 363 overall, environmental risk factors of the many pesticides on the global market are still poorly 364 characterized across the literature. 365 366 1.3.2 Limitations and Strengths 367 Our research shows that NHANES respondents are exposed to multiple pesticides and 368 pesticide types. Quantifying chemical mixtures across a population is complex and methodology 369 for understanding these mixtures is still an emerging area of research. However, there is still 370 plenty of research to be done in understanding chemical mixtures. Much of the research on 371 chemical health outcomes focuses on one chemical at a time, including our study, but people 372 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 17 are often exposed to more than one chemical, chemicals can interact with each other to create 373 new chemicals and once chemicals are in the environment, they can also react with the ambient 374 air or be degraded by the sun’s rays. All these changes to chemicals in relation to mixtures and 375 being in the environment create nuanced exposures and further research is needed to 376 understand how these mixtures may uniquely affect the human body. 377 Some pesticides which did not meet our inclusion criteria could have different exposure 378 based on farmwork occupational status. Oxypyrimidine (7.88% vs. 13.76%, 0.033), desethyl 379 hydroxy DEET (17.37% vs. 11.30%, p =0.015), and DEET (9.17% vs 6.25%, p = 0.036) were 380 significantly different between farmworkers and non-farmworkers, respectively. However, all of 381 these chemicals had detectability percentages below the cutoff for inclusion in our study. It is 382 possible that by restricting the chemicals included we are missing some important differences in 383 pesticide exposure between farmworkers and non-farmworkers. Studying exposures and effects 384 of these less commonly detected pesticides could be an important area of investigation. 385 One of the major limitations of this project is that while NHANES is thorough, reliable, 386 and valid study, it is still cross-sectional. This means the measurements within it are a single 387 measurement in time and cannot be fully representative of chronic exposures or chronic 388 symptomology due to exposures. Another limitation includes most farmworkers being recruited 389 between 1999 and 2004 (N= 1,775, 69.6%), which is of importance since the recruitment and 390 laboratory methods have been updated since 2003. Newer methods for quantifying chemicals 391 from blood and urine samples are more sensitive and can detect lower quantities of chemicals. 392 Additionally, farmworkers living without citizenship had significantly lower BMI as well, which 393 may impact metabolism and accumulation of chemicals in the body. 394 An additional limitation of this study is that not every chemical is measured in every 395 participant, and that not every assay is completed in each chemical. This limitation makes direct 396 comparisons impossible and therefore our results are somewhat limited to group means. There 397 are some known limitations to the ToxCast dataset such as interference of cytotoxicity. Non-398 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 18 specific cell stress can interfere with the frequency reading since the cell is overworking to re-399 gain homeostasis after chemical exposure. ToxCast assays are often assessing effects in a 400 single tissue cell type, which may not accurately reflect chemical sensitivity across organ 401 systems or within particularly susceptible individuals. Moreover, while ToxCast maintains a 402 robust suite of assays measuring effects across a broad spectrum of potential toxic outcomes, 403 not every chemical is tested for every assay and not all potential biological outcomes following 404 chemical exposure are captured. 405 Other limitations inherent to interpreting bioactivity also exist. For starters, urine and serum 406 concentrations reflect excreted or circulating concentrations, respectively, but may not be 407 representative of concentrations in target organs like fat, liver, kidneys, or brain. This is 408 important because many chemicals target specific organs (e.g., organochlorines targeting the 409 central nervous system) or bioaccumulate in specific tissue types like lipids. There are also 410 challenges to being able to relate metabolites to their parent compounds since some chemicals 411 can have more than one parent compound (e.g. the pyrethroid metabolite 3-PBA). This can 412 make ascertaining what active ingredient is bioactive in the human body difficult, and even if 413 considering a limited number of chemicals, there is no way to calculate a direct contribution of 414 each parent compound to a non-specific metabolite. 415 A strength of our study is that it is the first to provide a comprehensive quantification of all 416 the pesticide exposure concentrations within the US population using NHANES from 1999 to 417 2014 and to then stratify these concentrations by social determinants of health with a focus on 418 farmwork, fishing, and forestry work history and U.S. citizenship. By considering all the 419 pesticides within NHANES and narrowing down to those with at least 50% detectability, we find 420 that even within NHANES a small portion (15%) of these chemicals are detected in a majority of 421 NHANES participants. ToxCast & NHANES are both validated, reliable study datasets created 422 by the US government to assess chemical bioactivity and examine the health of people residing 423 in the US. By integrating these two datasets, the results are more generalizable to the U.S. 424 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 19 population. Additionally, this study is one of few to consider health disparities associated with 425 occupation or citizenship and how they may affect pesticide exposure and potential resultant 426 health effects. This project can inform evidence-based guidelines and policies that are focused 427 on reducing pesticide exposure concentrations among people residing within the United States. 428 429 1.3.3 Future Directions 430 While NHANES quantifies many chemical biomarker concentrations for each study 431 participant, these measures do not fully capture how many chemicals each person may be 432 exposed to since every chemical is not tested for in every person. Moreover, toxicological 433 research should continue to focus on novel methods for assessing toxicity of chemical mixtures 434 and interactions to better understand population pesticide exposure and bioactivity of combined 435 pesticide exposures in at-risk individuals. Currently, research looks at predominantly the active 436 ingredients of pesticides, but inactive ingredients used to create pesticides may also influence 437 human health, this is currently being missed in many toxicological studies. Future research can 438 also include temporal data on pesticide exposure. Both NHANES and ToxCast include singular 439 exposure time points in humans and in vitro, respectively. However, for many farmworkers, 440 pesticide exposure is chronic and happens over multiple exposure incidents. 441 Expanding this research to disease biomarkers, symptoms, and diagnoses will also be an 442 important future direction. This way we can better connect target families of ToxCast assays to 443 health outcomes and then stratify findings by occupation and social determinants of health like 444 income, gender, citizenship, and country of birth. In this same vein of understanding social 445 determinant effects on health, more research on how these biomarker concentration 446 distributions differ based on residing or working in a low versus high income country will be 447 important because laws within a nation can alter the health and exposure for many. 448 449 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 20 Works Cited 450 451 Abubakar Y, Tijjani H, Egbuna C, et al (2020) Pesticides, History, and Classification. In: Natural 452 Remedies for Pest, Disease and Weed Control. Elsevier, pp 29–42 453 Adeola FO (2021) Global impact of chemicals and toxic substances on human health and the 454 environment. In: Kickbusch I, Ganten D, Moeti M (eds) Handbook of Global Health. 455 Springer International Publishing, Cham, pp 2227–2256 456 Attene-Ramos MS, Miller N, Huang R, et al (2013) The Tox21 robotic platform for the 457 assessment of environmental chemicals – from vision to reality. Drug Discov Today 458 18:716–723. https://doi.org/10.1016/j.drudis.2013.05.015 459 Barnes N (2013) Is health a labour, citizenship or human right? Mexican seasonal agricultural 460 workers in Leamington, Canada. Glob Public Health 8:654–669. 461 https://doi.org/10.1080/17441692.2013.791335 462 Barr DB, Wilder LC, Caudill SP, et al (2005) Urinary Creatinine Concentrations in the U.S. 463 Population: Implications for Urinary Biologic Monitoring Measurements. Environ Health 464 Perspect 113:192–200. https://doi.org/10.1289/ehp.7337 465 Briz V, Molina-Molina J-M, Sánchez-Redondo S, et al (2011) Differential Estrogenic Effects of 466 the Persistent Organochlorine Pesticides Dieldrin, Endosulfan, and Lindane in Primary 467 Neuronal Cultures. Toxicol Sci 120:413–427. https://doi.org/10.1093/toxsci/kfr019 468 Chasens ER, Dinardo M, Imes CC, et al (2020) Citizenship and health insurance status predict 469 glycemic management: NHANES data 2007–2016. Prev Med 139:106180. 470 https://doi.org/10.1016/j.ypmed.2020.106180 471 Cohn BA, La Merrill M, Krigbaum NY, et al (2015) DDT Exposure in Utero and Breast Cancer. J 472 Clin Endocrinol Metab 100:2865–2872. https://doi.org/10.1210/jc.2015-1841 473 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 21 Dhananjayan V, Ravichandran B (2018) Occupational health risk of farmers exposed to 474 pesticides in agricultural activities. Curr Opin Environ Sci Health 4:31–37. 475 https://doi.org/10.1016/j.coesh.2018.07.005 476 Fry K, Power MC (2017) Persistent organic pollutants and mortality in the United States, 477 NHANES 1999–2011. Environ Health 16:105. https://doi.org/10.1186/s12940-017-0313-478 6 479 Guadamuz JS, Durazo-Arvizu RA, Daviglus ML, et al (2020) Citizenship Status and the 480 Prevalence, Treatment, and Control of Cardiovascular Disease Risk Factors Among 481 Adults in the United States, 2011–2016. Circ Cardiovasc Qual Outcomes 13:e006215. 482 https://doi.org/10.1161/CIRCOUTCOMES.119.006215 483 Hocking TD (2020) Package “WeightedROC”: Fast, weighted ROC curves. In: Compr. R Arch. 484 Netw. https://cran.r-project.org/web/packages/WeightedROC/WeightedROC.pdf 485 Mason SJ, Graham NE (2002) Areas beneath the relative operating characteristics (ROC) and 486 relative operating levels (ROL) curves: Statistical significance and interpretation. Q J R 487 Meteorol Soc 128:2145–2166. https://doi.org/10.1256/003590002320603584 488 Medehouenou TCM, Ayotte P, Carmichael P-H, et al (2019) Exposure to polychlorinated 489 biphenyls and organochlorine pesticides and risk of dementia, Alzheimer’s disease and 490 cognitive decline in an older population: a prospective analysis from the Canadian Study 491 of Health and Aging. Environ Health 18:57. https://doi.org/10.1186/s12940-019-0494-2 492 Millar JA, Forté CA (2023) ColacinoLab / Farmwork_Citizenship_Pesticides: Code and data 493 used to to compare pesticide exposure and bioactivity by farmwork history and US 494 citizenship. In: GitHub. 495 https://github.com/ColacinoLab/Farmwork_Citizenship_Pesticides 496 Mills PK, Beaumont JJ, Nasseri K (2006) Proportionate Mortality Among Current and Former 497 Members of the United Farm Workers of America, AFL-CIO, in California 1973-2000. J 498 Agromedicine 11:39–48. https://doi.org/10.1300/J096v11n01_05 499 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 22 Mostafalou S, Abdollahi M (2013) Pesticides and human chronic diseases: Evidences, 500 mechanisms, and perspectives. Toxicol Appl Pharmacol 268:157–177. 501 https://doi.org/10.1016/j.taap.2013.01.025 502 Quesada J, Hart LK, Bourgois P (2011) Structural Vulnerability and Health: Latino Migrant 503 Laborers in the United States. Med Anthropol 30:339–362. 504 https://doi.org/10.1080/01459740.2011.576725 505 Ramos A, Carlo G, Grant K, et al (2016) Stress, Depression, and Occupational Injury among 506 Migrant Farmworkers in Nebraska. Safety 2:23. https://doi.org/10.3390/safety2040023 507 Ramos AK (2018) A Human Rights-Based Approach to Farmworker Health: An Overarching 508 Framework to Address the Social Determinants of Health. J Agromedicine 23:25–31. 509 https://doi.org/10.1080/1059924X.2017.1384419 510 Ramos AK, Su D, Lander L, Rivera R (2015) Stress Factors Contributing to Depression Among 511 Latino Migrant Farmworkers in Nebraska. J Immigr Minor Health 17:1627–1634. 512 https://doi.org/10.1007/s10903-015-0201-5 513 Saxton DI, Stuesse A (2018) Workers’ Decompensation: Engaged Research with Injured 514 Im/migrant Workers: Anthropology of Work Review. Anthropol Work Rev 39:65–78. 515 https://doi.org/10.1111/awr.12147 516 Silver MK, Arain AL, Shao J, et al (2018) Distribution and predictors of 20 toxic and essential 517 metals in the umbilical cord blood of Chinese newborns. Chemosphere 210:1167–1175. 518 https://doi.org/10.1016/j.chemosphere.2018.07.124 519 Sniegoski LT, Moody JR (1979) Determination of serum and blood densities. Anal Chem 520 51:1577–1578. https://doi.org/10.1021/ac50045a052 521 Susmann H (2016) Package “RNHANES”: Facilitates analysis of CDC NHANES data. In: 522 Compr. R Arch. Netw. https://cran.r-523 project.org/web/packages/RNHANES/RNHANES.pdf 524 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 23 Thomas R, Paules RS, Simeonov A, et al (2018) The US Federal Tox21 Program: A strategic 525 and operational plan for continued leadership. ALTEX 35:163–168. 526 https://doi.org/10.14573/altex.1803011 527 Wei Y, Zhu J, Nguyen A (2014) Urinary concentrations of dichlorophenol pesticides and obesity 528 among adult participants in the U.S. National Health and Nutrition Examination Survey 529 (NHANES) 2005–2008. Int J Hyg Environ Health 217:294–299. 530 https://doi.org/10.1016/j.ijheh.2013.07.003 531 Wickham H (2016) ggplot2: Elegant graphics for data analysis, 3rd edn. Springer International 532 Publishing, Cham 533 Wong HL, Garthwaite DG, Ramwell CT, Brown CD (2019) Assessment of occupational 534 exposure to pesticide mixtures with endocrine-disrupting activity. Environ Sci Pollut Res 535 26:1642–1653. https://doi.org/10.1007/s11356-018-3676-5 536 Zong G, Valvi D, Coull B, et al (2018) Persistent organic pollutants and risk of type 2 diabetes: A 537 prospective investigation among middle-aged women in Nurses’ Health Study II. Environ 538 Int 114:334–342. https://doi.org/10.1016/j.envint.2017.12.010 539 540 541 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 24 Statements and Declarations 542 543 Funding: The researchers included on this study were supported by the National Institute of 544 Occupational Safety and Health Education Research Center (Grant# T42 OH 008455), the 545 National Institute of Environmental Health Sciences Environmental Toxicology and 546 Epidemiology Program (Grant# T32 ES007062), the National Science Foundation Graduate 547 Research Program (Grant # DGE-1256260), and the National Institutes of Health (R01 548 ES028802, P30 ES017885, R01AG072396). 549 550 Competing Interests: The authors do not declare any financial conflicts of interest. 551 552 Author Contributions: Justin Colacino and Chanese Forté contributed to the study conception 553 and design. Material preparation and data collection were performed by Chanese Forté. 554 Analysis was performed by Jess Millar and Chanese Forté. The first draft of the manuscript was 555 written by Chanese Forté and all authors commented on subsequent versions of the 556 manuscript. All authors read and approved the final manuscript. 557 558 Data Availability: The NHANES and ToxCast datasets analysed during the current study are 559 available from the CDC, https://wwwn.cdc.gov/nchs/nhanes/, and the EPA, 560 https://www.epa.gov/chemical-research/exploring-toxcast-data. 561 562 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 25 Tables 563 564 Table 1. Stratified Demographics of NHANES Participants, by Farmwork Category 565 Non-Farmworker Farmworker Variable Mean Standard Error Mean Standard Error p-value Body Mass Index 28.4 6.7 28.32 6.08) 0.584 Age in years 45.88 19.5 48.63 18.81 2.15x10-4 Poverty Index Ratio 2.5 1.63 2.82 1.72) 5.99x10-5 Survey Year N=20,205 Percent N=1,137 Percent < 2.2x10- 16 1999-2000 1,404 6.9 159 14 2001-2002 1,691 8.4 219 19.3 2003-2004 2,890 14.3 358 31.5 2005-2006 1,654 8.2 32 2.8 2007-2008 3,626 17.9 87 7.7 2009-2010 3,831 19 154 13.5 2011-2012 3,278 16.2 96 8.4 2013-2014 1,831 9.1 32 2.8 Gender 1.76x10-10 Men 10,187 50.4 440 38.7 Women 10,018 49.6 697 61.3 Racial Ethnicity < 2.2x10- 16 Mexican American 3,517 17.4 278 24.5 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 26 Other Hispanic 1,577 7.8 36 3.2 Non-Hispanic White 9,167 45.4 635 55.8 Non-Hispanic Black 4,435 22 135 11.9 Other Race 1,509 7.5 53 4.7 Country of Birth 0.538 Born in 50 US states or DC 606 90.2 0 - Born in Mexico 30 4.5 71 74 Born elsewhere 36 5.4 25 26 U.S. Citizenship 4.04x10-4 Non-Citizen 2,579 12.8 203 17.9 Citizen 17,626 87.2 934 82.1 Education Level < 2.2x10- 16 Less than 9th grade 2,004 9.9 233 20.5 9-11th grade 4,021 19.9 147 13 Highschool 4,707 23.3 210 18.5 Graduate/GED 5,566 27.6 243 21.4 Some College or AA 3,885 19.2 302 26.6 566 P-values are derived from a chi-square test, using a Yate's Correction where necessary, and for 567 continuous variables, a Wilcoxon Rank Test was used with a Kruskall-Wallis Correction (as 568 needed). Percentages are out of the total number of respondents for that specific question. In 569 this table, other race includes multi-racial. In this study, 9-11 grad includes 12th grade 570 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 27 completion without a high school diploma. All values in this dataset are weighted and stratified 571 according to NHANES guidelines. 572 573 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 28 Table 2. Stratified Demographics of NHANES Participants with a History of Farmwork, by Citizenship 574 Variable Citizen Non-Citizen Mean Standard Error Mean Standard Error p-value Body Mass Index 28.52 6.33 27.37 4.66 0.038 Age in years 49.9 18.9 42.74 17.07 7.57x10-5 Poverty Index Ratio 3.13 1.68 1.41 1.07 < 2.2 x10-16 Variable N=1,007 % N=237 % Survey Year 1999-2000 139 14.9 20 9.9 9.41x10-05 2001-2002 188 20.1 31 15.3 2003-2004 325 34.8 33 16.3 2005-2006 20 2.1 12 5.9 2007-2008 66 7.1 21 10.3 2009-2010 105 11.2 49 24.1 2011-2012 68 7.3 28 13.8 2013-2014 23 2.5 9 4.4 Gender Men 378 40.5 62 30.5 0.013 Women 556 59.5 141 69.5 Racial Ethnicity Mexican American 125 13.4 153 75.4 < 2.2 x10-16 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 29 Other Hispanic 23 2.5 13 6.4 Non-Hispanic White 622 66.6 13 6.4 Non-Hispanic Black 129 13.8 6 3 Other Race 35 3.7 18 8.9 Country of Birth Born in 50 US states or DC 606 90.2 0 0 < 2.2 x10-16 Born in Mexico 30 4.5 71 74 Born elsewhere 36 5.4 25 26 Education Level Less than 9th grade 110 11.8 123 60.6 < 2.2 x10-16 9-11th grade 115 12.3 32 15.8 Highschool 183 19.6 27 13.3 Graduate/GED 234 25.1 9 4.4 Some College or AA 290 31.1 12 5.9 575 P-values are derived from a chi-square test, using a Yate's Correction where necessary, and a Wilcoxon Rank Test was completed 576 with a Kruskall-Wallis Correction. Percentages are out of the total number of respondents for that specific question. In this table, 577 other race includes multi-racial. In this study, 9-11 grad includes 12th grade completion without a high school diploma. 578 579 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 30 Table 3. Bioactivity of pesticides cross-listed between NHANES and ToxCast, by pesticide and 580 persistence 581 Common Name CAS-RN Total Assays Positive Assays Bio-active Assay Percentage Bioactivity Threshold (µM) 2,4-Dichlorophenol 120-83- 2 678 27 3.98 0.34 2,4- Dichlorophenoxyacetic acid 94-75-7 807 18 2.23 6.49x10-3 2,5-Dichlorophenol 583-78- 8 599 14 2.34 0.33 3-Phenoxybenzoic acid 3739- 38-6 622 11 1.77 0.23 3,5,6-Trichloropyridinol 6515- 38-4 433 21 4.85 1.35 4-Nitrophenol 100-02- 7 682 43 6.30 8.63x10-3 ß- hexachlorocyclohexane a 319-85- 7 654 24 3.67 0.03 Chlorpyrifos 2921- 88-2 639 126 19.72 1.45 Chlorpyrifos-oxon 5598- 15-2 693 132 19.05 0.04 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 31 DEET Acid 134-62- 3 1025 13 1.27 0.17 Dieldrin a 60-57-1 549 121 22.04 0.32 p,p'-DDE a 72-55-9 1139 305 26.78 0.31 p,p'-DDT a 50-29-3 778 278 35.73 0.43 Heptachlor Epoxide a 76-44-8 650 259 39.85 1.31 582 aPersistent Organic Pollutant. 583 A positive assay is defined as hitcall==1. The bioactivity assay percentage is created by dividing 584 the total number of positive assays by the total number of assays and multiplying by 100%. 585 Bioactivity ratio per chemical was calculated by dividing the count of positive assays by the total 586 number of assays within the US Environmental Protection Agency’s Toxicity Forecast 587 Dashboard database. 588 589 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 32 Figures 590 591 592 Figure 1. Comparing the chemical molarity of of NHANES subjects with bioactivity threshholds 593 taken from chemical assays. ACC is the activity concentration at cut-off for a specific assay 594 where a chemical is considered active. 595 596 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint 33 597 Figure 2. Comparing the odds of having a bioactive pesticide biomarker concentration by 598 farmwork history and for farmworkers only by citizenship. This figure presents the outcomes of 599 the regression model of farmworker and non-farmworker health outcomes. Bioactive was 600 defined as having at least one pesticide biomarker concentration that was the same or higher 601 concentration than the minimal concentration needed to see an effect. The data for this table 602 was retrieved from the U.S. EPA’s Toxicity Forecast Dashboard and the National Health and 603 Nutrition Examination Survey. 604 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 28, 2023. ; https://doi.org/10.1101/2023.01.24.23284967doi: medRxiv preprint