What this study adds
For the first time in a large agricultural cohort, we assessed exposures to nitrate and disinfection byproducts in drinking water. Most (71%) participants used unregulated private wells and nitrate levels were estimated using geographic information system-based models. Twelve percent of Iowa well users had nitrate concentrations exceeding the Environmental Protection Agency regulatory limit compared with <1% in North Carolina. Among those (25%) using public water supplies, median trihalomethane concentrations were 2.9-fold higher in North Carolina than Iowa. Future evaluation of cancer and other health effects in relation to these exposures will inform regulatory standards.
There is increasing evidence that elevated exposures to nitrate and disinfection byproducts (DBPs) through drinking water pose health risks, including for specific cancers and adverse reproductive outcomes.1–4 Although nitrate and some DBPs are regulated in public water supplies (PWS) by the US Environmental Protection Agency (EPA), private wells remain unregulated. Approximately 43 million people in the United States rely on self-supplied water at their residence, 98% of whom are private well users.5 The EPA defines PWS as any system having at least 15 service connections or serving 25 or more people for a minimum of 60 days per year.6 Water supplies not meeting these criteria are considered private and are not subject to federal regulations. Consequently, private well owners are responsible for maintenance of their wells and testing for multiple contaminants, which can prove costly. Furthermore, groundwater in agricultural settings has been shown to have elevated nitrate levels compared with nonagricultural settings.7 Therefore, private well users in agricultural settings are more likely to have elevated drinking water nitrate levels compared with users of PWS.
Nitrate levels in our water sources have increased over the past century due to human activities, including increased use of nitrogen fertilizers, animal feeding operations, fossil fuel combustion, and nitrogen-fixing crops.8–11 The resulting increase in exposure through drinking water is especially of concern under conditions favoring endogenous nitrosation (e.g., smoking and low antioxidant intake), which allow the formation of N-nitroso compounds (NOC), many of which are carcinogens and teratogens. The International Agency for Research on Cancer (IARC) classified nitrate and nitrite as probable carcinogens when ingested under conditions that increase endogenous nitrosation.12 At this time, the strongest evidence supports associations between drinking water nitrate intake and infant methemoglobinemia, thyroid disease, neural tube defects, and colorectal cancer, though other associations have been reported.1,13 The EPA drinking water maximum contaminant level (MCL) for nitrate (10 mg/L NO3-N) was set to protect against infant methemoglobinemia, an acute condition,14 but other adverse health effects have been observed at exposure below this level.1
DBPs, including haloacetic acids (HAAs), trihalomethanes (THMs), and oxyhalides (bromate and chlorite), are formed in drinking water supplies due to use of chlorine and other disinfectants (ozone, chlorine dioxide, chloramines, etc.), which react with organic compounds in the water to form DBPs.2,15 Exposure can occur through ingestion, inhalation, or dermal absorption. Although more than 700 DBPs are known, only HAAs, THMs, and oxyhalides are currently regulated by the US EPA.2,15 Epidemiologic studies have linked elevated exposure to DBPs with bladder and colorectal cancers, though evidence is strongest for bladder cancer.2,15,16
The Agricultural Health Study (AHS) is a prospective cohort that enrolled pesticide applicators and their spouses in North Carolina (NC) and Iowa (IA), with the goal of assessing health risks associated with pesticide and other agricultural exposures.17 The AHS is unique in that most of the participants live on farms and rely on private wells for their drinking water. It is also one of few cohort studies that has collected information on drinking water source, allowing us to evaluate health risks in relation to exposure to drinking water contaminants. Characterizing drinking water exposures in the AHS will facilitate analyses that could provide valuable inference about health effects in relation to drinking water exposures in agricultural settings. The primary objectives of this study were to describe residential drinking water sources for AHS participants and to evaluate the quality of the drinking water supplies by assessing available nitrate and DBP measurements for participants’ water sources. This descriptive analysis will provide the groundwork for future etiologic studies related to water contaminants in the AHS.
During the enrollment period (1993–1997), individuals applying for restricted-use pesticide licenses in the states of IA and NC were recruited to participate in the study. In IA, both commercial and private pesticide applicators (i.e., farmers) were recruited; however, in NC, only private applicators were recruited. Spouses of private applicators in each state were also invited to participate by completing a questionnaire that included questions about current and past residential drinking water source.17 The AHS population consists of 89,655 participants: 58,563 in IA (4916 commercial applicators, 31,876 private applicators, and 21,771 spouses) and 31,092 in NC (20,518 private applicators and 10,574 spouses). To date, participants have completed questionnaires at four phases (P): enrollment or P1 (1993–1997), P2 (1999–2005), P3 (2005–2010), and P4 (2013–2015), except for commercial applicators (only P1 and P2). Data releases were 20,1701.00 for phases 1 and 2 and 20,1808.00 for phase 3. Additionally, to assess environmental exposures, we collected and geocoded AHS participants’ addresses at each study phase.18,19 Version 1 geocodes were used for assigning drinking water source and modeling nitrate concentrations in private wells (described below).
Drinking water source assignments using agricultural health study questionnaires
We assigned the enrollment drinking water source to each participant using the self-administered questionnaires at P1 and computer-assisted telephone interviews at P2–P3 (Supplemental Figures 1 and 2; https://links.lww.com/EE/A189 for IA and NC, respectively).20 The order of assignments was based on the detail and accuracy of data reported at each phase, since phrasing of questions about drinking water source varied across phases. We only assigned sources based on later phases if participants had not moved since P1 and we only assigned sources at later phases based on partner (spouse or applicator) reporting if the couple reported still living together.
At P1, only spouses were asked to report whether their primary residential drinking water source was a private well, public (community) water supply, bottled water, or some other source. Those who reported using private wells were also asked to report well depth and whether the well was cased (lined to maintain well structure and protect well water from debris).21 We assumed that spouses and private applicators lived together at enrollment and both were assigned the spouse-reported drinking water source. We assigned 32,297 (55.1%) IA private applicators and spouses, and 13,296 (42.8%) NC private applicators and spouses to their drinking water source based on spouse report at P1. This included 28 IA and 12 NC participants who had information on well depth at P1 but no reported water source at any phase and were assumed to be using private wells.
We used information from P2 and P3 for those missing P1 water source (all commercial applicators, farmers without an enrolled spouse, farmers and spouses with missing information). We assigned their source based on P3 reporting since information at P2 was less specific and the P3 questions were similar to those at P1. We assigned 6592 (11.3%) IA and 5587 (18.0%) NC participants their source based on P3. At P2, participants only reported whether their primary drinking water source was a private well or spring (yes/no). Springs were reported by only 2% of participants based on the responses to P3, so we assumed private well use if “private well or spring” was reported. We assumed that participants not using private wells or springs were using PWS. We assigned 8879 (15.2%) IA and 4839 (15.6%) NC participants’ sources based on P2. The remainder with missing water source whose partners reported water source at P3 or P2, were assigned their partner’s reported source. A total of 943 (1.6%) IA and 791 (2.5%) NC participants’ sources were based on their partner’s report at P3, whereas 1030 (1.8%) IA and 665 (2.1%) NC participants’ water sources were based on P2. In sum, we assigned drinking water sources for 74,919 (83.6%) participants, including 84.9% of those in IA and 81.0% in NC.
Public water supply linkages
To ascertain participants’ exposures to regulated water contaminants (e.g., nitrate and DBPs), we linked participants using PWS, including rural water supplies (utilities providing drinking water to residents in rural communities), to the PWS serving their homes, based on their geocoded address at enrollment.
We obtained water quality monitoring data for IA PWS for years 1987–2018 from the Center for Health Effects of Environmental Contamination at the University of Iowa. We linked most towns to PWS by town name, and for those unable to be matched based on name, we compared participants’ addresses with maps of IA rural water cooperatives and a database of private wells to determine the PWS serving those locations. In total, we linked 13,191 (96.6%) IA participants residing in 733 towns to their PWS systems.
We obtained water quality monitoring data for NC PWS for years 1977–2018 from the North Carolina Department of Environmental Quality (NCDEQ). We identified most PWS by town name and the remaining were matched through internet searches of municipalities’ utility companies. For unincorporated towns, we identified the counties in which the towns were located, nearby towns, or nearby utility companies using Google maps and relevant websites. We used the Drinking Water Resilience Interactive Project mapping tool to identify county water systems and companies’ contact information.22 We then called the utilities and county water supply systems to confirm service to participants’ towns. If multiple companies served a town, we called utilities or the NCDEQ Regional Office to ascertain the utility serving AHS addresses. If cities had switched water systems over time, we selected the water system that was active in the 1990s. In total, we linked 5082 (96.0%) NC participants residing in 429 towns to their PWS systems.
Estimating nitrate levels in private wells
Prior work was conducted to develop geospatial models that predicted private well nitrate concentrations in IA23 and NC.24 For IA, we used 34,084 available measurements (not specific to AHS participants) of nitrate in private wells, collected in 1980–2011, to train and test random forest models.23 Similarly, we used 22,000 measurements (not specific to AHS participants) of nitrate in private wells in NC, collected in 1990–2011, to train and test multiple random forest classification models.24 We used AHS participants’ geocoded enrollment addresses to approximate the locations of their private wells and to derive the model covariates, which were used in the predictive models to estimate AHS participants’ private well nitrate levels. The modeled nitrate levels were based on the well depth and location, surrounding land use, nearby nitrogen inputs (e.g., animal waste, fertilizers, and septic systems), population density, soil characteristics, aquifer characteristics, whether crops used irrigation, and other geologic and meteorologic factors.23,24 Top predictors in IA consisted of well depth, topography, distance to animal feeding operations, and distance to sinkholes, while top predictors in NC included distance to swine lagoons agricultural land use, deciduous forest cover, soil type, and runoff potential. Of those using private wells, 8846 (26.5%) participants in IA and 4342 (23.0%) in NC were missing well depth information, an important predictor in our private well nitrate models. For these participants, we used the median well depth for the state in the nitrate prediction models. We were unable to predict nitrate estimates for 1628 (4.8%) participants in IA and 2589 (13.7%) in NC due to missing information on one or more variables in the final models and these participants were excluded from our analyses.
We conducted all statistical analyses using SAS 9.4 software (Cary, NC). We described the proportion of private well, PWS, bottled water, and other water source users in each state at the time of enrollment by participant type. We described demographic, lifestyle, and home characteristics reported at enrollment by drinking water source. We also described the characteristics of private wells, including well depth, if the well was cased and whether the well was tested for nitrate, distance from pesticide mixing and pesticide application sites. We relied on information collected at later phases when information was missing at P1. Information from P2 and P3 was used only if there was no change in drinking water source since enrollment. We excluded commercial applicators from some analyses that relied on spouse information from P1, since they had no spouses enrolled. We also assessed whether any demographic, lifestyle, or other factors were related to participants testing their wells for nitrate.
We computed averages of annual PWS measurements over the period 1990 to 2009 for the sum of 5 haloacetic acids (HAA5), total trihalomethanes (TTHM), and nitrate-nitrogen (NO3-N), overall and by state. If there were multiple measurements reported in a year, we first computed an annual average. For the PWS long-term averages and NO3-N estimates for private wells, we describe the distributions using the median, interquartile range (IQR), and other percentiles since the data were lognormally distributed. For analyte measurements below the detection limit (DL), we imputed a single concentration value based on a log normal distribution.25 For IA PWS measurements, we used the DLs reported for each sample as the upper bound for the imputation; however, where missing, we used the median of the DLs across all years (NO3-N: 0.5, IQR: 0.1–1.0 mg/L; HAA5: 6.0, IQR: 4.0–6.0 µg/L; TTHM: 2.0, IQR: 0.5–2.0 µg/L). DLs were missing for most NC samples, so we used the lowest reported measurement across PWS for each year as our upper bound (range: NO3-N: 0.01–0.29 mg/L, HAA5: 1–25 µg/L, TTHM: 0.1–10 µg/L). In IA, 43%, 14%, and 5% of participants’ average PWS concentrations for HAA5, TTHM, and NO3-N, respectively, were based entirely on values below the DL (imputed), compared with 3%, 3%, and 0% in NC. We also computed the number of years annual average concentrations were at or above the EPA MCLs and one-half the MCL, since past studies reported increased cancer risks for individuals experiencing long-term exposures above half the MCLs.26
Among IA participants, 96.4% (20,983) of spouses, 81.8% (26,072) of private applicators, and 54.6% (2,686) of commercial applicators had drinking water source data at enrollment (Table 1). Among NC participants, 95.0% (10,046) of spouses and 73.7% (15,132) of private applicators had drinking water source information. In IA, 71.4% of spouses and 70.6% of private applicators used a private well or spring at enrollment, while 23.9% of spouses and 25.5% of private applicators used PWS. Among commercial applicators, 26.6% used private wells as their drinking water source and 73.4% were assumed to be using PWS. In NC, 74.7% of the spouses and 74.9% of private applicators were using private wells or springs at enrollment, while 20.2% of spouses and 21.6% of private applicators were using PWS. Few participants reported using bottled water as their drinking water source (≤2% of participants in both states) and other water sources (e.g., cisterns) were also rare (<5%).
Table 1. -
Drinking water sources at enrollment (1993–1997) by AHS participant type, in IA and NC.
||IA (n = 58,563)
||NC (n = 31,092)
||Spouse (n = 21,771)
||Private applicator (n = 31,876)
||Commercial applicatora (n = 4,916)
||Spouse (n = 10,574)
||Private applicator (n = 20,518)
|Other (e.g., cistern)
aCommercial applicators who reported that their drinking water source was “not a private well/spring” were assumed to be using PWS; Percentages were computed after excluding those with unknown/missing information.
The mean age of IA participants at enrollment was younger than NC participants for both those using private wells (IA: 46; NC: 50 years) and PWS (IA 45; NC: 49) (Table 2). Almost all (99%) of IA participants were non-Hispanic white, compared with about 94% in NC, and in both states, about 1% of the participants reported Hispanic ethnicity, with no major differences by drinking water source. There were more Black participants in NC (3.5% on private wells, 5.0% PWS) compared with IA (<0.1%). About 40% of participants in both states were women. Compared with IA participants, higher percentages of NC participants reported their highest educational level as less than high school. In both states, lower percentages of participants using private wells completed college compared with PWS users. Higher percentages of participants in NC reported being current or past smokers at enrollment (almost 50% in NC compared with 35% in IA), in both water source groups. Fewer IA participants reported no alcohol consumption in the past year than in NC. Almost all applicators using private wells (98% in IA and 93% in NC) owned or worked on a farm compared with lower percentages using PWS (81% in IA and 84% in NC). Slightly higher percentages of private applicators using private wells reported applying natural or chemical fertilizer to crops compared with PWS users.
Table 2. -
Characteristics of AHS participants with a known source, by drinking water source (PWS vs. private well) for IA and NC at enrollment.
||Total population (n = 71,902)
||IA private well (n = 34,106)
||IA PWS (n = 13,651)
||NC private well (n = 18,852)
||NC PWS (n = 5,293)
|Age, mean (SD)
|Race and ethnicity, n (%)
| White, non-Hispanic
| White, Hispanic
| Black/African American, non-Hispanic
| Black/African American, Hispanic
| Other, non-Hispanic
| Other, Hispanic
| Missing (n)
|Sex, n (%)
|Education, n (%)
| <High school
| High school graduate
| Some college or 1–3 years vocational education after high school
| College graduate
| >1 year of graduate or professional school completed
| Missing (n)
|Smokinga, n (%)
| Never smokers
| Current smokers
| Past smokers
| Missing (n)
|Alcoholb, n (%)
| <1 time a month
| 1–3 times a month
| 1 time a week
| 2–4 times a week
| Almost every day to every day
| Missing (n)
|Owned or worked on a farmc, n (%)
| Missing (n)
|Applied natural or chemical fertilizer to farmlandd, n (%)
| Missing (n)
|Age of homed, n (%)
| <10 years
| 10–20 years
| 21–40 years
| 41–99 years
| 100+ years
| Missing (n)
|Housing foundation typed, n(%)
| Crawl space
| Combination crawl space/basement
| Full basement
| Did not know
| Missing (n)
Percentages were computed after excluding those with unknown/missing information.
aCurrent smokers at enrollment.
bHow often participants drank any alcoholic beverages in the past year.
cTotal = 42,370, spouses excluded due to no reporting.
dTotal = 69,216, commercial applicators excluded due to no reporting.
Characteristics of the participants’ private wells are described in Table 3. Approximately 75% of private well users (private applicators and spouses only) in each state reported having cased wells (Table 3). This question was asked only of spouses at P1, resulting in missing information for 37% of private well users. About 20% of the participants in each state had wells less than 50 feet (ft); well depth was missing for 27% of IA and 23% of NC participants. Higher percentages of private well users in NC (60%) reported that their wells were >300 ft away from areas where pesticides were mixed compared with IA (44%). Likewise, more IA well users had wells located within 200 ft of a pesticide application site (36% vs. 27% in NC). Higher percentages of IA participants reported testing their wells for nitrate (79%), compared with NC (44%). Of those who had their well tested, test results indicated nitrate levels were safe for bottle-fed infants (i.e., <10 mg/L NO3-N) for 99% of those in NC compared with 88% in IA.
Table 3. -
Characteristics of private wells among AHS participants, excluding commercial applicators.
||Population (n = 52,244)
||NC (n = 18,852)
|Cased wella (P1)
| Yes, n (%)
| No, n (%)
| Not applicable (used a spring)d
| Do not know if well was casede
|Well depthb (P1–P3), n (%)
| <50 ft
| 50–150 ft
| >150 ft
| >100 ft (P2)
|Well distance from pesticide mixing sitec (P1–P2)
| No pesticides mixed on farm
| <150 ft
| 150–300 ft
| >300 ft
|Well distance from pesticide application sited (P2)
| No pesticides applied on farm
| <100 ft
| 100–200 ft
| >200 ft
| Did not complete P2
|Well ever tested for nitratee (P2–P3)
| Yes, n (%)
| No, n (%)
| Do not know/missing f
| Did not complete P2 or P3
|Safe nitrate levels for bottle-fed infants (<10 mg/L NO3-N)e (P2–P3)
| Yes, n (%)
| No, n (%)
| Do not know/missingf
| Did not complete P2 or P3
Percentages were computed after excluding those with unknown/missing information; Private well characteristics are based on data collected by questionnaires at phases 1–3 (P1–P3).
aInformation only collected from spouse at phase 1 (P1), spouse values applied to private applicator (partner).
bInformation on well depth was collected in P1–P3.
cInformation on distance from pesticide mixing was collected P1–P2.
dInformation on distance from pesticide application collected in P2.
eInformation on nitrate testing and safety was collected in P2 and P3.
fThose who responded do not know and those who skipped the question; Responses were recategorized due to varying response options across phases.
Upon further assessment, we found no major differences in reporting of private well nitrate testing by age, race/ethnicity, sex, educational attainment, smoking status, alcohol consumption, whether children lived at home, or marital status in each state (eTable 1; https://links.lww.com/EE/A189). We also found no differences in testing by whether applicator owned or worked on a farm, reported fertilizer application, age of home, or housing foundation type. No differences in testing were found by well depth, or distances of the well from pesticide application and mixing sites. Last, the medians and IQRs for NO3-N were similar for those with and without well testing in both states, and proportions of wells with predicted levels at or above the ½ MCL and MCL were similar for wells reported to be tested and those not reported to be tested.
Among those on PWS, the medians of participants’ average concentrations (1990–2009) of NO3-N, HAA5, and TTHM were all below the EPA MCLs (10, 60, and 80 μg/L, respectively) in both states (Table 4).27 However, 0.8%, 1.4%, and 0.2% of PWS users in IA and 0.5%, 1.7%, and 0%, in NC had average concentrations at or above the MCLs for HAA5, TTHM, and NO3-N, respectively. Among those with annual average levels at or above ½ MCL for at least 1 year, the median number of years for HAA5, TTHM, and NO3-N, respectively, was 3.0, 6.0, and 2.0 for IA, and 6.0, 10.0, 1.0 for NC. Concentrations of DBPs were higher in NC than in IA (medians: HAA5: 12.3 vs. 5.0 μg/L and TTHM: 35.4 vs. 13.0 μg/L), while PWS nitrate levels were higher in IA (medians: 0.9 vs. 0.1 mg/L).
Table 4. -
Distributions of nitrate and disinfection byproduct analyte measurements among AHS participants, 1990–2009 among those using public water supplies and private wells.
||Analyte or years of exposure
|Public water supply
||Average NO3-N (mg/L)
|Years ≥5 mg/L NO3-N
|Years ≥10 mg/L NO3-N
|Average HAA5 (μg/L)
|Years ≥30 μg/L HAA5
|Years ≥60 μg/L HAA5
|Average TTHM (μg/L)
|Years ≥40 μg/L TTHM
|Years ≥80 μg/L TTHM
|Private well (all)
|Well depth ≤150 ft
|Well depth >150 ftd
aValues in these rows represent number years that participants had measurements at or above half the maximum contaminant level for at least 1 year.
bValues in these rows represent number years that participants had measurements at or above the maximum contaminant level for at least 1 year.
cExcludes commercial applicators.
dWells of the >100 ft category based on phase 2 questionnaire; Analyte measurements below the detection limit in NC were imputed as a single value below the limit of detection; No LODs for private wells since nitrate levels were predicted.
For private well users, the estimated median concentrations of NO3-N were below the EPA’s MCL in both states (Table 4). NO3-N levels were higher in private wells than in PWS in both states (medians: 1.5 vs. 0.9 mg/L in IA and 1.9 vs. 0.1 mg/L in NC). The 90th (11 mg/L) and 95th (15.3 mg/L) percentiles for predicted NO3-N in private wells in IA were above the MCL. More private wells in IA (12%) exceeded the 10 mg/L NO3-N regulatory limit than NC (<1%). Deeper wells (>150 ft) had lower predicted nitrate concentrations, particularly in IA (median: 3.3 mg/L, IQR: 1.1–8.8 vs. 1.0 mg/L, IQR: 0.7–1.8 mg/L for ≤150 ft). In IA, wells with the highest NO3-N concentrations (>2 mg/L) were generally found in the western part of the state and in the northeast (Figure 1A). In NC, wells with average NO3-N concentrations >2 mg/L were more evenly distributed across the state (Figure 1B). There were only a few wells in NC with average concentrations >5 mg/L (1%), and these wells were located in the central area of the state.
We determined drinking water sources at enrollment for approximately 84% of AHS participants. We matched 97% of PWS users to their respective water utility and historical monitoring data and estimated private well nitrate levels for 92% of private well users. The medians for nitrate, TTHM, and HAA5 among PWS and private well users were all below the EPA MCLs. Nitrate concentrations in private wells in both states were higher than those in PWS; 12% of IA private well users had nitrate estimates at or above the MCL compared with <1% in NC. Although concentrations of each contaminant were mostly below the MCLs, it is important to note that MCLs are infrequently updated and that adverse health effects have been reported at exposure levels below MCLs for both nitrate and TTHM.26,28,29 Therefore, exposures below the MCL may not be fully indicative of safe drinking contaminant water levels.
We found that a much higher percentage of participants reported testing their private wells for nitrate in IA (79%) compared with NC (44%). There were minimal differences in demographic, lifestyle, and other characteristics between those who tested and those who did not test their wells. Likewise, there were minimal differences in nitrate levels. Testing differences between the states may be due to differing access to testing. Since 1987, IA has had the statewide Grants to Counties program that provides funding to assist private well users with well testing (nitrate/nitrite, Escherichia coli, and arsenic) and maintenance.30,31 As of 2018, 98 of IA’s 99 counties participated.31 The NC DEQ has provided financial support for private well testing only since 2006 through the Bernard Allen Memorial Emergency Drinking Water Fund; furthermore, businesses, including farms, do not qualify and testing is restricted to anthropogenic contaminants.32 Since 2008, testing is required for newly constructed wells.33 Some local health departments and nonprofit organizations, such as the Southeast Rural Community Assistance Project, offer free or subsidized testing in NC.34 Still, both cost and lack of information have been identified as barriers to testing and maintenance of private wells in NC.35 In a small study in Wake County, NC, all residents (n = 18) who were interviewed about private well use in 2015, thought they could detect water contaminants through appearance, odor, and taste alone.36 However, aesthetic effects in drinking water are typically related to nuisance chemicals, which often pose less health risks compared with other substances, such as nitrate that has no taste, odor, or color.37 In sum, although some programs assist well owners with testing, it is ultimately the owner’s responsibility to seek out the necessary resources and, in many cases, to cover well maintenance and testing costs. In contrast, PWS users’ water is regularly tested and they receive annual drinking water quality reports.38
There are a limited number of epidemiologic studies in agricultural areas with information on drinking water exposures for comparison to our study. The IA Women’s Health Study (IWHS), a population-based cohort of postmenopausal women, collected information on drinking water source for approximately 36,000 women in 1989.26 Because this cohort was population-based, a smaller percentage used private wells (18.5%) compared with the AHS (70%). Based on PWS monitoring data from 1955 to 1988, the median HAA5 concentration was 3.5 µg/L (IQR: 1.9–6.4 µg/L), similar to the IA AHS participants on PWS [4.3 µg/L (IQR: 3.0–10.8 µg/L)]; whereas median TTHM levels were lower [4.6 µg/L (IQR 0.9–14.3 µg/L) vs. 13.0 µg/L (IQR: 3.7–32.3 µg/L)].26,39,40 Differences in the PWS (e.g., size of operation, treatments, and water sources) serving each of these populations would be expected to result in different distributions of water contaminants between the two studies. Median nitrate levels in PWS in the IWHS (1.07 mg/L, IQR: 0.47–2.97 mg/L) were similar to those in the AHS using PWS (0.9 mg/L, IQR: 0.2–3.1 mg/L).
A study by the Environmental Working Group (EWG) reported similar nitrate levels in Iowa private wells and corroborated the role of well depth, which we found to be a major predictor of nitrate levels.41 The EWG analysis using 2002–2017 data found a mean nitrate level of 4.4 mg/L NO3-N (IA AHS mean = 3.9) and that 12% of private wells exceeded the MCL (same as the IA AHS population).41 Shallow wells (<50 ft) had higher average nitrate concentrations compared with deeper wells; average concentrations for wells <50, 50–150, and >150 ft were 8.79, 4.35, and 2.90 mg/L, respectively. We observed similar patterns in the AHS, particularly in IA, where shallower wells had higher median nitrate (3.3 mg/L for wells ≤150 ft vs. 1.0 mg/L >150 ft).
Studies of water quality in agricultural areas of California demonstrated large variation in drinking water nitrate concentrations in PWS and private wells. Balazs et al42 estimated average nitrate levels for 327 active PWS in the San Joaquin Valley of California. In 1999–2001, 15.2% and 0.2% of the population using PWS had nitrate levels ½ MCL-<MCL and ≥MCL, respectively. A study of private wells in the Central Valley of California, which includes the San Joaquin Valley, estimated that percentages of private and public supply wells with nitrate levels >5 mg NO3/L were 21% and 4.7%, respectively.43
Population-based case-control studies in Iowa, Nebraska, and the New England region assessed drinking water nitrate and TTHM exposures.29,44–46 Ward et al45 reported median average nitrate levels among IA PWS users in 1960–1990 [1.4 mg/L (IQR: 0.6–2.9 mg/L)] that were similar to AHS IA PWS users. However, private well nitrate measurements at the time of interviews (1998–2000) were lower with only 5% (vs. 25% in IA AHS) of private wells above 5 mg NO3-N/L. Median PWS nitrate levels in eastern Nebraska in 1947–1979 were higher [1.8 mg/L (IQR: 1.7–3.8 mg/L)] and a higher percentage of PWS users (21% vs. 0.3% in IA AHS) had average levels above the MCL.46 Median private well nitrate levels [2.7 mg/L (IQR: 0.7–6.6 mg/L)] were also higher than our study and about 20% of private well users had levels above the MCL. Cantor et al44 and Beane Freeman et al29 described TTHM exposures in IA (1980s) and New England (2001–2004), respectively. The IA study reported geometric mean values of 38 μg/L for chlorinated surface water and 1.7 μg/L for chlorinated groundwater sources. The median TTHM level for PWS in the New England study was similar to IA AHS (15.7 vs. 13 μg/L in AHS).
Studies of drinking water nitrate have been conducted in other countries with high agricultural production. For example, Schullehner et al47 used a Danish drinking water quality database (JUPITER) to assess the population’s nitrate exposures over a 35-year period. Similar to US studies, they found that private well users were exposed to higher nitrate levels (37% of private wells were above the drinking water standard compared with 2.5% of the PWS in 2012). Generally, average groundwater nitrate levels are reported to have remained stable across Europe over the last two decades (1992–2018) and have largely remained below the drinking water standard.48 Similarly, rates of noncompliance to nitrate standards in PWS are generally low across the EU.49 In summary, most AHS participants had average nitrate and DBP exposures that were generally below the MCLs, and most exposures were below half the MCLs. However, those using private wells in IA had higher nitrate exposures than described for most prior study populations. Therefore, future analyses of drinking water nitrate exposures and health effects in the AHS will include a wider range of nitrate exposures than prior studies.
A limitation of our drinking water exposure assessment is the lack of information on other contaminants in private wells, most notably pesticides. About one-third of the private wells in each state were located within 200 ft of pesticide application sites; however, we were unable to describe drinking water pesticide levels. We did not have pesticide measurement data to estimate pesticide levels in AHS private wells and while some pesticides are measured in PWS, most are not regulated (no MCLs). Monitoring data for our study period and population was scarce. Studies of drinking water quality in agricultural areas have generally not found pesticides at levels above the MCL.50 An IWHS analysis of atrazine in PWS found that 32% of participants’ PWS had detectable atrazine, but median levels were <0.1 mg/L (MCL: 10 mg/L) in 1986–1987.51 Studies of farming communities primarily using well water in Canada found detectable herbicide levels, predominantly 2-methyl-4-chlorophenoxyacetic acid (MCPA) and 2,4-Dichlorophenoxyacetic acid (2,4-D), in less than 4% of tap water samples in the 1990s.52,53 Notably, a 1997 pilot study of six farms in the AHS detected no pesticides in drinking water samples.54 However, more recently, neonicotinoid insecticides have been detected in 36% of private wells sampled in 2017 and 2018 in IA and southern Minnesota.55,56 Although, neonicotinoids have been highly detected in ground and surface water resources across the United States, Canada, and China, these insecticides remain unregulated by the EPA and few studies have assessed human exposure through drinking water.57 On the national scale, the US Geological Survey (USGS) reported that only 1% of the private wells sampled for the National Water Quality Assessment exceeded human-health benchmarks for one or more pesticides from 1992 to 2001 in agricultural areas.58 Furthermore, a recent nationwide USGS study of public supply wells detected pesticides in 41% of the wells surveyed in 2012–2018, with atrazine, atrazine degradates (didealkylatrazine and deethylatrazine), and a metolachlor degradate (metolachlor sulfonic acid) being detected in >10% of the wells.59 Hexazinone, 2-hydroxyatrazine, prometon, and tebuthiuron were also all detected in at least 6% of the wells; however, no detected pesticides were found at levels exceeding MCLs or other health benchmarks. Based on these reports, regulated pesticides would not be likely to occur in private wells at levels exceeding EPA standards. However, many pesticides have yet to be assessed in drinking water, so more research is needed.
Strengths of our study include the estimation of private well nitrate levels and assessment of PWS contaminant measurements over two decades. In future analyses of contaminant exposures and health risks in AHS, we will link PWS measurements to participants by duration on drinking water source. Although we estimated nitrate concentrations in private wells around the time of enrollment, the private well models could be used to provide estimates for later time periods by updating time-sensitive predictor variables such as land use. Additionally, the model could be used to estimate nitrate levels in new locations for participants who moved. One limitation is the lack of information for some applicators since only spouses were asked about their drinking water source at P1 and not all applicators participated at later phases. Additional limitations are that commercial applicators only reported drinking water source at P2 and were only asked if they used a private well or not and that we were unable to match all PWS users to monitoring data. Although we estimated private well nitrate levels through modeling, we did not have measured nitrate levels for these wells. There were some differences between the state-specific nitrate prediction models. Information on well depth, which was an important predictor in the IA models,23 was unavailable for NC private wells. Ultimately, predictions based on the final NC model were less accurate compared with the IA model.23,24
In this study, we describe drinking water sources and contaminant levels for more than 80% of the AHS participants. The development of these metrics will facilitate studies of cancer and non-cancer-related health risks associated with drinking water exposures in the AHS. These future studies will provide an opportunity to replicate results from the limited number of existing studies that have evaluated the health effects of drinking water contaminants in the United States and other countries, and to study associations with health outcomes not previously studied. Additionally, assessments of health risks associated with water contaminants in PWS will inform drinking water regulatory standards.
Conflict of interest statement
The authors declare that they have no conflicts of interest with regard to the content of this report.
We thank Anne Taylor at Information Management Systems for programming support, Abigail Flory at Westat, Inc. for Geographic Information Systems support, the Center for Health Effects of Environmental Contamination at University of Iowa, especially Darrin Thompson, for their assistance in conducting the Iowa public water supply linkages, and the NCDEQ, especially Eric Chai, for providing public water supply monitoring data. Contact corresponding author to inquire about access to data and computer code.
1. Ward MH, Jones RR, Brender JD, et al. Drinking water nitrate and human health: an updated review. Int J Environ Res Public Health. 2018;15:E1557.
2. Cantor KP, Steinmaus CM, Ward MH, Beane Freeman LE. Water contaminants. In: Cancer Epidemiology and Prevention. Oxford Scholarship Online; 2017.
3. Graves CG, Matanoski GM, Tardiff RG. Weight of evidence for an association between adverse reproductive and developmental effects and exposure to disinfection by-products: a critical review. Regul Toxicol Pharmacol. 2001;34:103–124.
4. Tardiff RG, Carson ML, Ginevan ME. Updated weight of evidence for an association between adverse reproductive and developmental effects and exposure to disinfection by-products. Regul Toxicol Pharmacol. 2006;45:185–205.
5. Dieter CA, Maupin MA, Caldwell RR, et al. Estimated Use of Water in the United States in 2015. U.S. Geological Survey Circular 1441, 65. 2018: 1441. https://doi.org/10.3133/cir1441
6. EPA. Information about public water systems. 2021. Available at: https://www.epa.gov/dwreginfo/information-about-public-water-systems
. Accessed 7 February 2022.
7. Dubrovsky NM, Burow KR, Clark GM, et al. The Quality of Our Nation’s Waters- Nutrients in the Nation’s Streams and Groundwater, 1992–2004. National Water-Quality Assessment Program, U.S. Department of the Interior, U.S. Geological Survey; 2010.
8. Davidson EA, David MB, Galloway JN, et al. Excess nitrogen in the U.S. environment: Trends, risks, and solutions. Issues Ecol. Report Number 15. 2011 https://www.esa.org/esa/wp-content/uploads/2013/03/issuesinecology15.pdf
9. Howarth RW. Coastal nitrogen pollution: a review of sources and trends globally and regionally. Harmful Algae. 2008;8:14–20.
10. Vitousek PM, Aber JD, Howarth RW, et al. Human alteration of the global nitrogen cycle: sources and consequences. Ecol Appl. 1997;7:737–750.
11. Zirkle KW, Nolan BT, Jones RR, Weyer PJ, Ward MH, Wheeler DC. Assessing the relationship between groundwater nitrate and animal feeding operations in Iowa (USA). Sci Total Environ. 2016;566-567:1062–1068.
12. IARC. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans: Ingested Nitrate and Nitrite and Cyanobacterial Peptide Toxins. International Agency for Research on Cancer; 2010.
13. García Torres E, Pérez Morales R, González Zamora A, et al. Consumption of water contaminated by nitrate and its deleterious effects on the human thyroid gland: a review and update. Int J Environ Health Res. 2020;32:1–18.
14. EPA. Quality Criteria for Water. Environmental Protection Agency; 1976.
15. Richardson SD, Plewa MJ, Wagner ED, Schoeny R, Demarini DM. Occurrence, genotoxicity, and carcinogenicity of regulated and emerging disinfection by-products in drinking water: a review and roadmap for research. Mutat Res. 2007;636:178–242.
16. DeMarini DM. A review on the 40th anniversary of the first regulation of drinking water disinfection by-products. Environ Mol Mutagen. 2020;61:588–601.
17. Alavanja MC, Sandler DP, McMaster SB, et al. The agricultural health study. Environ Health Perspect. 1996;104:362–369.
18. Fisher JA, Spaur M, Buller ID, et al. Spatial heterogeneity in positional errors: a comparison of two residential geocoding efforts in the Agricultural Health Study. Int J Environ Res Public Health. 2021;18:1637.
19. Jones RR, DellaValle CT, Flory AR, et al. Accuracy of residential geocoding in the Agricultural Health Study. Int J Health Geogr. 2014;13:37.
20. Agricultural Health Study. Questionnaires & study data. 2021. Availble at: https://aghealth.nih.gov/collaboration/questionnaires.html
. Accessed 16 July 2021.
21. EPA. Learn about private water wells. Available at: https://www.epa.gov/privatewells/learn-about-private-water-wells
. Accessed 4 September 2021.
22. Drinking water resilience interactive project. Map your water supply. Available at: http://drip.unctv.org/maps/
. Accessed 11 May 2022.
23. Wheeler DC, Nolan BT, Flory AR, DellaValle CT, Ward MH. Modeling groundwater nitrate concentrations in private wells in Iowa. Sci Total Environ. 2015;536:481–488.
24. Messier KP, Wheeler DC, Flory AR, et al. Modeling groundwater nitrate exposure in private wells of North Carolina for the Agricultural Health Study. Sci Total Environ. 2019;655:512–519.
25. Lubin JH, Colt JS, Camann D, et al. Epidemiologic evaluation of measurement data in the presence of detection limits. Environ Health Perspect. 2004;112:1691–1696.
26. Jones RR, Weyer PJ, DellaValle CT, et al. Nitrate from drinking water and diet and bladder cancer among postmenopausal women in Iowa. Environ Health Perspect. 2016;124:1751–1758.
27. EPA. Chemical Contaminant Rules- Drinking Water Requirements for States and Public Water Systems. Environmental Protection Agency; 2018.
28. Villanueva CM, Fernández F, Malats N, Grimalt JO, Kogevinas M. Meta-analysis of studies on individual consumption of chlorinated drinking water and bladder cancer. J Epidemiol Community Health. 2003;57:166–173.
29. Beane Freeman LE, Cantor KP, Baris D, et al. Bladder cancer and water disinfection by-product exposures through multiple routes: a population-based case-control study (New England, USA). Environ Health Perspect. 2017;125:067010.
30. Iowa Department of Health. Private well testing, reconstruction, and plugging—grants to counties. In: Iowa Administrative Code. Public Health Department Rule 641-24.4. Chapter 24. 2016. https://idph.iowa.gov/Portals/1/userfiles/197/BEHS/PDFs/641_24.pdf
31. Secchi S, Cwiertny D. Iowa’s Grants to Counties Program: A Valuable but Underutilized Program for Protecting the Public Health of Private Well Users. Center for Health Effects of Environmental Contamination; 2019. doi: 10.17077/rep.001117.
32. General assembly of North Carolina. General statute Chapter 87-98: Bernard Allen memorial emergency drinking water fund. https://www.ncleg.gov/EnactedLegislation/Statutes/PDF/BySection/Chapter_87/GS_87-98.pdf
33. General Assembly of North Carolina. House Bill 2873, Sections 3-4. An act to require permitting and inspection of new private drinking water wells and to require water quality testing of private drinking water wells. Session Law 2006-202. https://ehs.ncpublichealth.com/oswp/docs/WellStatutes-GS-87-87.pdf
34. Southeast Rural Community Assistance Project, Inc. Programs and services for individuals. Availble at: http://www.southeastrcap.org/get-assistance/for-individuals
. Accessed 18 July 2021.
35. MacDonald Gibson J, Pieper KJ. Strategies to improve private-well water quality: a north carolina perspective. Environ Health Perspect. 2017;125:076001.
36. Fizer CM. Barriers to Private Well and Septic Management: An Analysis of Homeowner Decision-Making. [Master’s Thesis]: Department of Environmental Sciences and Engineering, Gillings School of Global Public Health, University of North Carolina at Chapel Hill; 2016.
37. EPA. Secondary drinking water standards: guidance for nuisance chemicals. 2021. Available at: https://www.epa.gov/sdwa/secondary-drinking-water-standards-guidance-nuisance-chemicals
. Accessed 8 February 2022.
38. EPA. Consumer Confidence Reports (CCR)- find your local CCR. 2017. Available at: https://ofmpub.epa.gov/apex/safewater/f?p=136:102
. Accessed 18 July 2021.
39. Jones RR, Weyer PJ, DellaValle CT, et al. Ingested nitrate, disinfection by-products, and kidney cancer risk in older women. Epidemiology. 2017;28:703–711.
40. Jones RR, DellaValle CT, Weyer PJ, et al. Ingested nitrate, disinfection by-products, and risk of colon and rectal cancers in the Iowa Women’s Health Study cohort
. Environ Int. 2019;126:242–251.
41. Schechinger A. Contamination of Iowa’s Private Wells: Methods and Detailed Results. Environmental Working Group; 2019.
42. Balazs C, Morello-Frosch R, Hubbard A, Ray I. Social disparities in nitrate-contaminated drinking water in California’s San Joaquin Valley. Environ Health Perspect. 2011;119:1272–1278.
43. Ransom KM, Nolan BT, A Traum J, et al. A hybrid machine learning model to predict and visualize nitrate concentration throughout the Central Valley Aquifer, California, USA. Sci Total Environ. 2017;601-602:1160–1172.
44. Cantor KP, Lynch CF, Hildesheim ME, et al. Drinking water source and chlorination byproducts in Iowa. III. Risk of brain cancer. Am J Epidemiol. 1999;150:552–560.
45. Ward MH, Cerhan JR, Colt JS, Hartge P. Risk of non-Hodgkin lymphoma and nitrate and nitrite from drinking water and diet. Epidemiology. 2006;17:375–382.
46. Ward MH, Mark SD, Cantor KP, Weisenburger DD, Correa-Villaseñor A, Zahm SH. Drinking water nitrate and the risk of non-Hodgkin’s lymphoma. Epidemiology. 1996;7:465–471.
47. Schullehner J, Hansen B. Nitrate exposure from drinking water in Denmark over the last 35 years. Environ Res Lett. 2014;9:095001.
49. Bryan N, Grinsven H. The role of nitrate in human health. Adv Agron. 2013;119:153–182.
50. Deziel NC, Friesen MC, Hoppin JA, Hines CJ, Thomas K, Freeman LE. A review of nonoccupational pathways for pesticide exposure in women living in agricultural areas. Environ Health Perspect. 2015;123:515–524.
51. Inoue-Choi M, Weyer PJ, Jones RR, et al. Atrazine in public water supplies and risk of ovarian cancer among postmenopausal women in the Iowa Women’s Health Study. Occup Environ Med. 2016;73:582–587.
52. Arbuckle TE, Bruce D, Ritter L, Hall JC. Indirect sources of herbicide exposure for families on Ontario farms. J Expo Sci Environ Epidemiol. 2006;16:98–104.
53. Fitzgerald D, Chanasyk DS, Neilson RD, Kiely D, Audette R. Farm well water quality in Alberta. Water Qual Res J. 2001;36:565–588.
54. Melnyk LJ, Berry MR, Sheldon LS. Dietary exposure from pesticide application on farms in the Agricultural Health Pilot Study. J Expo Anal Environ Epidemiol. 1997;7:61–80.
55. Thompson DA, Kolpin DW, Hladik M, Barnes KK, Vargo JD, Field RW. Prevalence of neonicotinoids and sulfoxaflor in alluvial aquifers in a high corn and soybean producing region of the Midwestern United States. Sci Total Environ. 2021;782.
56. Thompson DA, Hruby CE, Vargo JD, Field RW. Occurrence of neonicotinoids and sulfoxaflor in major aquifer groups in Iowa. Chemosphere. 2021;281:130856.
57. Thompson DA, Lehmler HJ, Kolpin DW, et al. A critical review on the potential impacts of neonicotinoid insecticide use: current knowledge of environmental fate, toxicity, and implications for human health. Environ Sci Process Impacts. 2020;22:1315–1346.
58. Gilliom RJ, Barbash JE, Crawford CG, et al. Pesticides in the Nation’s Streams and Ground Water, 1992–2001. U.S. Geological Survey Circular 1291, The Quality of Our Nation's Waters, 184p., https://pubs.er.usgs.gov/publication/cir1291
59. Bexfield LM, Belitz K, Lindsey BD, Toccalino PL, Nowell LH. Pesticides and pesticide degradates in groundwater used for public supply across the United States: occurrence and human-health context. Environ Sci Technol. 2021;55:362–372.