1. Introduction
As the human population continues to grow exponentially, food security is emerging as a significant source of concern [
1]. In response to this, artificial techniques, including agrichemical use, have been introduced into farming to improve agricultural yield. While agrichemicals may have enhanced food security for the growing population, it is feared that the toxic effects from agrichemical residues deposited in water, plants, land, and animals may outweigh its benefit. Hence, the tradeoff between agrichemical use and food security may transcend environmental degradation to humans’ toxicological outcomes. While the scope of agrichemicals is broad, this study will only focus on pesticides. The origin of synthetic pesticides dates to approximately nine decades ago in the United States, with dichloro-diphenyl-trichloroethane (DDT) as the first pesticide to be used [
2]. Although DDT was originally intended to combat insect-borne human diseases such as malaria, its application soon included pest control in agriculture [
3]. Even though DDT is now wholly restricted in the United States and many parts of the world due to its toxicity [
4], other pesticides have continued to emerge over the years [
5]. Despite replacing DDT with the new pesticides, their toxicological effects may not differ significantly from DDT. Some examples of commonly used agrichemicals are acephate, acetochlor, alachlor, aldrin, atrazine, glyphosate, metaldehyde, diazinon, and malathion [
6]. These pesticides have been named in many toxicological effects, which may manifest from acute or chronic exposure. For example, malathion was implicated in acute toxic effects related to gastrointestinal discomfort [
7] and glyphosate in chronic toxic effects such as cancer [
8]. Moreover, some of these pesticides have recently begun to gain attention as endocrine disruptors. This is because pesticides can alter the endocrine system’s normal functioning by acting as an agonist/antagonist for endocrine receptors, activators/inhibitors for endocrine biogenesis, or induce epigenetic reprograming during estrogen-induced development [
9].
The most implicated hormone during pesticide endocrine disruption is estrogen [
9]. Despite strong evidence of estrogenic disruptions of these pesticides in in vitro studies [
10,
11,
12,
13], strong evidence linking pesticide as an independent risk factor for the estrogen-related disease has not been clearly elucidated. Hence, the need for studies highlighting the role of pesticides in carcinogenic processes.
Nebraska is one of the United States’ agricultural States with a robust repository for data on agrichemical contaminated groundwater. Therefore, this study will take advantage of this database to explore whether there is an ecological correlation between selected estrogen disrupting pesticides (EDP) and estrogen-related cancer (ERC) across Nebraska’s 93 counties. Hence, this study examined the ecological distribution of EDP across Nebraska counties and its association with ERC.
2. Materials and Methods
The Quality-assessed Agrichemical contaminant for the Nebraska Groundwater database, a publicly available repository for agrichemicals detected in Nebraska groundwater/wells, was examined to address our objectives. The concentration of agrichemicals in this database was evaluated based on well-defined criteria. A detailed description of this database is published elsewhere [
10]. Forty-seven EDP was obtained from the Quality-assessed Agrichemical Contaminant for Nebraska Groundwater Database from 1 January 1974–31 December 2012. However, only four pesticides-acetochlor, atrazine, deethylatrazine (DEA), and de-isopropyl atrazine (DIA) were substantially detected. While the specific pesticide exposure timeline for carcinogenesis is unknown, exposure timeline which preceded the timeline for the anticipated outcomes according to Bradford Hills criteria [
14] was used. These pesticides were tested multiple times within the set timeline from 33,593 unique wells across the 93 counties of Nebraska. All water supply well types such as commercial, irrigation, livestock, domestic, and monitoring wells were included in the data for analysis (
Figure 1). These wells were identified using the clearinghouse numbers.
The outcome variable is the age-standardized incidence rate for selected ERC (breast cancer, uterine cancer, and prostate cancer) of the 93 counties in Nebraska. The age-standardized incidence rates for these cancers were obtained from the State reports of the National Cancer Institute (NCI) between 1 January 2013 and 31 December 2017. The incidence rates for the selected ERC were defined as cases per 100,000, adjusted by age according to the U.S. standard population in the year 2000. These rates were calculated using the SEER*Stat (NCI, Bethesda, MD, USA). The denominator of the incidence rates was obtained from the U.S. census population count between 1969 and 2017. All cancers selected for this study were invasive.
Given that carcinogenesis is a complex relationship between several factors, other potential confounders were included in the analysis. Hence, county-level potential confounders for the selected ERC were obtained from the County Health Rankings database for 2010. County Health Rankings is a collaborative program of Robert Wood Johnson Foundation and the University of Wisconsin Population Health Institute. It provides county by county health determinants and outcomes. From this database, we obtained five potential confounders of ERC, such as physically unhealthy days per county, % of adult who smoke per county, % of obese adult per county, % binge drinking per county and % of uninsured per county. The aforementioned county-level confounders were included in the analysis because their carcinogenesis risk is well established [
15,
16,
17,
18,
19,
20,
21]. Meanwhile, physically unhealthy days were defined by County Health Rankings as the average days in the last 30 days that an adult in a county reported poor physical health. Physically unhealthy days were obtained from responding to the question: “Thinking about your physical health, which includes physical illness and injury, for how many days during the past 30 days was your physical health not good?”.
Data Analysis
Water supply wells were sampled multiple times for the measurement of EDP. As a result, the pesticides in each water supply well had multiple measurements. While the multiple measurements of the pesticide at different time points provided a detailed history of pesticide contamination in each well. It was redundant information for a time trend analysis of the pesticides because they were not uniform across the wells. However, we were able to identify wells that had tested positive at least once for the pesticides of interest, which met this current study’s goal. Therefore, the percent of wells positive for the selected EDP were calculated per county and included in the analysis as the exposure variable. Moreover, this data’s continuous variables were described using minimum, maximum, 25th-percentile, 75th-percentile, mean, and median. While the categorical variables were described using frequencies and percentages. To demonstrate EDP’s ecological distribution and its association with ERC, a GIS mapping of Nebraska was performed. A county-level choropleth maps were created for age standardized rates of each cancer type (breast, prostate, and uterine) using ArcGIS Pro V2.7. Graduated symbols were used to map the percentage of water supply wells positive for the four types of EDP. Both variables were categorized using the equal distance method.
To test the relationship between EDP and ERC, all three data sources- Nebraska groundwater database, State profile of the NCI, and County Health Rankings—were merged by county to form a single data used for analysis. Given that all variables included in the analysis were continuous variables, the LINE (linearity, independence, normality, and equality of variance) assumptions for linear regression analysis were evaluated. While almost all assumptions were met for all the variables, % pesticide positive wells deviated from normality because several counties have % pesticide positive wells of zero values. Since % pesticide positive wells were the primary exposure for this study, they were subjected to square root transformation to improve its skewed distribution. Square root transformation was applied as the transformation procedure because it works best to transform variables with many zero values [
22]. After completing the check for LINE assumptions and transformations, correlations between the variables were determined using the following steps: First, the relationship between age-standardized incidence rates for the ERC and % EDP positive water supply wells were examined in a scatter plot. Secondly, the correlation between potential confounder data from County Health Rankings and the age-standardized ERC rates was calculated using Pearson correlation coefficient, r. To be conservative and parsimonious in the analysis, potential confounding variables with r of at least ±0.2 which were significant at α ≤ 0.1 with any of the ERC were adjusted in a linear regression model between ERC (outcome) and EDP (exposure). Thirdly, collinearity among % EDP positive water supply wells was determined for variables with a correlation coefficient of at least r = 0.50 [
23,
24,
25].
To determine if EDP was associated with ERC, a multiple linear regression model of age-standardized cancer rates as outcome, % pesticides positive wells as exposure and county-level confounders was performed. A significant association between the age-standardized ERC rates and EDP was considered at α ≤ 0.05. Data restructuring and all analysis were performed on Statistical Package for Social Sciences (SPSS) version 26, while the pie chart was done on Microsoft excel 2016.
4. Discussion
The purpose of this paper was to determine the ecological distribution of EDP and its association with ERC by using Nebraska county-level data. Meanwhile Nebraska is an ideal location for this study, given that agrichemicals were previously reported in Nebraska ground and surface water. In fact, most of Nebraska overlies the high plains aquifer providing approximately 88 percent of Nebraska’s drinking water despite its susceptibility to pesticides applied to the land surface [
26]. Shallow depths to groundwater levels, sandy soils, and intensely irrigated cropland all contribute to the high occurrence of the pesticides of interest in the water supply wells [
27]. While these pesticides are diverse and numerous, in this study, atrazine and its metabolites and acetochlor were detected in high concentration in Nebraska groundwater. Moreover, atrazine or its metabolites are among the most prevalent groundwater contaminant in Nebraska [
28].
Despite our ecological study design, we made some profound observations connecting EDP to ERC. Meanwhile, this is not the first study to observe EDP and ERC’s association [
29,
30,
31]. For example, the correlation between DDT, its metabolites, and ERC has been famously described [
13,
32,
33,
34,
35]. In fact, it was on this account DDT was banned by the United States Environmental Protection Agency (EPA) in 1972 [
2]. Today other pesticides with EDP have emerged. However, strong evidence linking adverse estrogenic effects of pesticides is inadequately observed. Hence, this emphasizes the importance of this current study.
Nebraska is one of the leading states in the US concerning agricultural activities. Due to this, agrichemical production has also become one of the most important manufacturing sectors in the State [
36]. Consequently, agrichemicals are frequently detected in watersheds and even groundwater [
37]. Previous studies have observed the association between co-occurring agrichemicals in Nebraska groundwater and non-Hodgkin lymphoma [
38]. Additionally, exposure to pesticides in Nebraska was shown to increase the risks of lung, skin, and hematological cancer [
37]. Another population-based case-control study conducted in Nebraska found an increased risk of glioma due to male farmers’ pesticide exposure [
39]. Moreover, three pesticides (glyphosate, diazinon, and coumaphos) were found to increase non-Hodgkin lymphoma risk among farmers in four midwestern states, including Nebraska [
40]. Here, we will focus on characterizing EDP, given that most of the pesticides detected in Nebraska groundwater between 1974 and 2012 have estrogen disrupting properties.
In this current study, the prevalence of atrazine and DEA contaminated wells were higher in Nebraska’s South and South-Eastern districts. This also is the area with the highest incidence of breast and prostate cancer in the State of Nebraska. While this area is urbanized, significant farming activities occur which may be the source of pesticide contamination in the water supply wells. Also, EDP’s presence in these districts may be contributed by the Platte river that runs to the South East from the West of Nebraska, where the population is very sparse, with robust agricultural activities. Other features in the East and South of Nebraska that may trigger pesticide runoff through the Platte River from the West are its dissected till plains, deep soils, and frequent precipitation [
41]. In fact, the high rates of pesticide runoff in Eastern Nebraska were previously reported [
42].
Furthermore, ERC in Nebraska were found to be higher than the national rates. A positive linear relationship between breast cancer and % water supply wells positive for acetochlor, atrazine, and DIA was observed. Moreover, wells positive for atrazine and its metabolites (DEA) were observed in counties with elevated breast cancer rates, as indicated on the maps. A similar relationship was previously reported in an ecological study conducted in Kentucky [
43]. Although, findings from a different epidemiological study did not observe a significant association between breast cancer and estrogen disrupting pesticides among Latinos in California [
44], which is congruent with our observation when breast cancer rates were modeled with % pesticides positive wells in a multiple linear regression analysis. Furthermore, Muir et al., observed similar observation using both epidemiological and ecological designs. Muir et al.’s ecological design observed a correlation between breast cancer and atrazine, whereas there was no observed statistical association between breast cancer and atrazine in the epidemiological study [
45]. In contrast, in vitro evidence disclosed the upregulation of GPR30, a G-protein coupled receptor for atrazine, on breast cancer cell lines exposed to atrazine even at doses below the maximum contaminant level for atrazine [
12]. Maybe, the disparity in results for atrazine and breast cancer may emerge from differences in study methodologies as ecological studies may not account for other potential risk factors for breast cancer, which are adequately controlled in an epidemiological study. It is possible that atrazine may not be an independent etiology for breast cancer.
Meanwhile, atrazine in aromatase induction, which mediates estrogen synthesis is apparent [
46,
47]. Although animal studies did not find any causal relationship between atrazine and breast cancer [
48], animal models may not sufficiently mimic atrazine mechanisms in humans. While the relationship between atrazine and breast cancer remains inconclusive, atrazine metabolites are another area of great concern in terms of atrazine’s carcinogenicity. Two atrazine metabolites were significantly detected in Nebraska groundwater, this includes, DEA and DIA. While previous epidemiological study conducted on DEA and DIA did not observe potential carcinogenicity [
49], DEA positive wells in this study were found in counties with elevated breast cancer rates (from the map). However, this was not accompanied by a positive association in the linear regression analysis. Again, the differential effects of DEA on the map and the linear regression confirm the role of study designs to determine the relationship between environmental carcinogens and health outcomes accurately. Meanwhile, DIA did not produce any breast cancer effects in the linear model and on the map. This may suggest that DEA is a more toxic metabolite of atrazine than DIA, which is also a metabolite of simazine [
50]. Hence, additional studies are required to explore the differential carcinogenicity of DEA and DIA. Furthermore, a significant association was previously found between breast cancer and other organochlorines [
44,
51,
52,
53], which was not replicated in this current study for acetochlor and breast cancer.
Uterine cancer is another ERC of interest. We observed a positive linear relationship between uterine cancer and all the EDP (atrazine, acetochlor, DEA, and DIA). Which was supported by a study that observed increased uterine fibroids due to atrazine-induced aromatase over-expression [
54]. Moreover, an ecological study among the Mayan populace with a high prevalence of uterine cancer reported a high serum concentration of organochlorines [
55,
56]. This validates our study’s findings, which observed a significant association between uterine cancer and acetochlor after adjusting for physically unhealthy days and % of binge drinking per county in a multiple linear regression analysis. Based on these findings and previous experience with the use of DDT [
57,
58,
59], another organochlorine, it is feared that acetochlor may have detrimental health outcomes.
Among the ERC selected for this study, prostate cancer is the only predominant male cancer. Evidence from animal models has linked prostate cancer to estrogen [
60]. Moreover, EDP’s carcinogenic effects on prostate cell lines have been observed in vitro [
61,
62]. Meanwhile, it must be noted that estrogen and androgen (male hormones) are connected physiologically. In fact, androgens are estrogen precursors mediated by the enzyme aromatase. Moreover, androgen has been shown as an independent etiology for prostate cancer. It was observed in an epidemiological study that African American men with elevated serum androgen have an increased risk of prostate cancer compared to their counterpart Japanese men with low serum androgen [
63].
Additionally, prostate treatment’s effectiveness using an androgen deprivation regimen is proof of androgen’s role in the incidence and progression of prostate cancer. Hence, it is not clear what the roles of estrogen are in prostate cancer. However, an animal study revealed that 5α-dihydrotestosterone, which is impossible to convert to estrogen by aromatase, increased prostate cancer risk in animals by 5%. However, when estrogen was added to 5α-dihydrotestosterone, the risk of prostate cancer increased by 3 folds [
64]. This suggests that estrogen and androgen conversion may not be related to the incidence of prostate cancer. Another pathway mediated by androgen and estrogen interaction may be responsible for prostate cancer. However, no studies to our knowledge indicated estrogen as an independent etiology for prostate cancer. This may explain why we observed a linear negative association between prostate cancer and wells positive for all the estrogenic pesticides. Although atrazine and DEA positive wells were observed in counties with elevated prostate cancer rates as indicated on the map. Hence, our results may be more consistent in the presence of other androgenic pesticides, which were not observed in this study.
Lumping cancer rates into an age-standardized rate fails to consider the potential differences in breast cancer among younger and older age groups. Moreover, our analysis may have been underpowered by our limited sample size and numerous missing observations in the data. These are the major limitations for the findings of this study. Additionally, county level measurement or assessment of pesticide exposure and outcomes may be flawed by consistent migration of county residents, which were not accounted for in this study’s analysis. Moreover, this study methodology was strictly proxy, which impaired our ability to make conclusions concerning the relationship between pesticides and ERC. In response to this, a direct assessment of consumed pesticides which will consider other potential risk factors for carcinogenesis will significantly improve study outcomes. While this study focused on the possible effects of pesticides predominantly detected in the groundwater during the study period, the potential impact of unaccounted co-occurring pesticides, which may be present in limited quantities, cannot be undermined. Meanwhile, the interactions of co-occurring chemicals is another area of future interest. Despite this study’s limitations, the observed linear relationships between pesticide-positive wells could strongly impact our approach for considering these pesticides in future studies.