The dominant source of drinking water in rural Nevada, United States, is privately-owned domestic wells. Because the water from these wells is unregulated with respect to government guidelines, it is the owner’s responsibility to test their groundwater for heavy metals and other contaminants. Arsenic, lead, cadmium, and uranium have been previously measured at concentrations above Environmental Protection Agency (EPA) guidelines in Nevada groundwater. This is a public health concern because elevated levels of these metals are known to have negative health effects.

We recruited individuals through a population health study, the Healthy Nevada Project, to submit drinking water samples from domestic wells for testing. Water samples were returned from 174 households with private wells. We found 22% had arsenic concentrations exceeding the EPA maximum contaminant level (MCL) of 10 mg/L. Additionally, federal, state, or health-based guidelines were exceeded for 8% of the households for uranium and iron, 6% for lithium and manganese, 4% for molybdenum, and 1% for lead. The maximum observed concentrations of arsenic, uranium, and lead were ~80, ~5, and ~1.5 times the EPA guideline values, respectively. 41% of households had a treatment system and submitted both pre- and post-treatment water samples from their well. The household treatments were shown to reduce metal concentrations, but concentrations above guideline values were still observed. Many treatment systems cannot reduce the concentration below guideline values because of water chemistry, treatment failure, or improper treatment techniques. These results show the pressing need for continued education and outreach on regular testing of domestic well waters, proper treatment types, and health effects of metal contamination. These findings are potentially applicable to other arid areas where groundwater contamination of naturally occurring heavy metals occurs.

Recruitment and survey of participants

Participants of the Healthy Nevada Project (HNP) were recruited for this research. The HNP is a large (>50,000 participants) all-comers population health study in Nevada that includes cross-referenced electronic health records, socio-demographic data, whole-exome sequences, and social health determinant data. Details about the HNP were previously published (Grzymski et al., 2020; Read et al., 2021; Schlauch et al., 2022, 2020). Specifically, HNP participants who consented to be contacted for future studies were invited via email to complete the HNP private well survey. The HNP private well survey contained 11 questions regarding their well, water treatment, and drinking water habits (Supplementary Material). The private well survey was created and disseminated on the Survey Monkey platform (www.SurveyMonkey.com).

Sampling

Survey respondents who had a private well and wanted to submit a well water sample were emailed an informed consent document for water sampling. Consented individuals were mailed an at-home sampling kit. The kit included sampling instructions, a data collection sheet, pre-cleaned (1% nitric acid washed) sample bottle(s), and a return-shipment box. Sampling protocols (Supplementary Material) requested participants to purge their well casing of stagnant water by allowing the water to run for 2 to 4 hours prior to sampling, which is similar to state well water sampling guidelines (Donaldson et al., 2012). After purging, participants were asked to sample from the tap they used for drinking, cooking, and other household activities (referred to as “household water” herein). Households without a water treatment system were instructed to sample from the tap they most commonly used for household activities (referred to as “households without treatment”). Homeowners with a water treatment system in their home were instructed to sample after the treatment system from the tap they most commonly used for household activities (referred to as “households with treatment”). For the purpose of this study, any method, media, etc. that is used to treat, purify, or filter well water is considered a “treatment.” To study the effectiveness of water treatment, households with treatment were asked to submit an additional pre-treatment sample. The pre-treatment sample represents the groundwater geochemistry and the households do not use the pre-treatment water for drinking, cooking, and other household activities. The post-treatment sample represents the water after undergoing water treatment.

 Additional metadata requested of all participants were sample location, date, time, treatment method prior to sampling, color and smell of the well water, duration of well purging, and any other relevant information pertaining to the participant’s water and sampling experience.

Geochemical analysis

Participant sampling kits were mailed back to the Desert Research Institute (DRI) in Reno, Nevada. Samples were accessioned and de-identified. Related data were manually entered in an electronic database. Using ultra-high purity nitric acid (Aristar Ultra, VWR Chemicals BDH), the water samples were then preserved in 1% nitric acid and analyzed following EPA Method 200.8. All but one sample were received within the sample hold times outlined in EPA Method 200.8. The one sample received outside of the sample hold time was discarded. Sample geochemical analysis was conducted at the University of Nevada Core Analytical Laboratory using a Shimadzu 2030 Inductively Coupled Plasma Mass Spectrometry (Shimadzu Corporation, Columbia, Maryland, US) instrument. Settings for the ICP-MS are provided in Table S1. The targeted elements include As, copper (Cu), Fe, lithium (Li), Mn, Mo, Pb, and U. Calibration standards (Inorganic Ventures, Christiansburg, Virginia, US) for the ICP-MS instrument captured the concentration range of the samples.

Quality assurance (QA) and quality control (QC) steps were taken at DRI and during ICP-MS analysis. DRI laboratory blanks and internal standards were submitted with the samples for analysis. DRI internal standards were made using a standard of known elemental concentration (Inorganic Ventures, Christiansburg, Virginia, US) with the elemental concentration targeting expected concentrations. QC included comparing the calculated internal standard concentrations to measured values as well as ensuring the DRI laboratory blank elemental concentrations were below sample concentrations or below detection limits. The average offset of a ~10 mg/L internal DRI standard (n=6) was <5% for As and Pb and <11% for Cu, Mn, Mo, and U. The average offset of a ~50 mg/L internal DRI standard (n=4) was <5% for Fe, Mo, and Pb and <16% for As, Cu and Mn. Additionally, during ICP-MS analysis precision recovery check standards (Agilent, Santa Clara, California, US) were measured every 10 to 15 samples (depending on sample analysis duration) to ensure the ICP-MS instrument response did not drift > 5% during analysis. Also during ICP-MS analysis, internal standards (Inorganic Ventures, Christiansburg, Virginia, US) were added to further monitor ICP-MS instrument drift.

The detection limit for all analyzed metals was 0.05 mg/L. The detection limits were determined by a signal-to-noise ratio of >3, where the noise is the instrument response to the matrix.  In some cases, the measured detection limit was below 0.05 mg/L, but 0.05 mg/L was used in this study as it is the reported detection limit for the element by the instrument manufacturer. The results were then reported to the homeowners including remediation options, as necessary.

Statistical analysis

Non-detects complicate subsequent data analysis. Discarding values <0.05 mg/L (i.e., below detection limit) or replacing them with zero may introduce a bias (Palarea-Albaladejo and Martín-Fernández, 2015) when computing summary statistics (i.e., mean median, and standard deviation), testing differences among groups, and correlation coefficients and regression equations (Helsel, 2011). Therefore, we chose to impute values <0.05 mg/L using R package (R Core Team, 2021) zCompositions (Palarea-Albaladejo and Martín-Fernández, 2015). We used the robust regression on order statistics (ROS) multiplicative lognormal replacement approach via the R package NADA for the imputation. This approach uses the measured values and assumes a distribution for the imputation of the censored portion (Helsel, 2005; Lee, 2020). To assess differences in metal concentrations in household water between homes without treatment and those with treatment as well as differences between pre- and post-treatment concentrations, we used the cendiff function in R, which is part of the NADA package. This function tests if there is a difference between two empirical cumulative distribution functions. The cendiff function is the Peto and Peto modification of the Gehan-Wilcoxon test (Helsel, 2005). This approach is appropriate for left-censored log-normal data, as is commonly observed in geochemical studies (Helsel, 2005), and was observed in this study.

xlsx files can be opened by most data handling software