Prioritization and Risk Ranking of Regulated and Unregulated Chemicals in US Drinking Water

Abstract

Drinking water constituents were compared using more than six million measurements (USEPA data) to prioritize and risk-rank regulated and unregulated chemicals and classes of chemicals. Hazard indexes were utilized for hazard- and risk-based chemicals, along with observed (nondetects = 0) and censored (nondetects = method detection limit/2) data methods. Chemicals (n = 139) were risk-ranked based on population exposed, resulting in the highest rankings for inorganic compounds (IOCs) and disinfection byproducts (DBPs), followed by semivolatile organic compounds (SOCs), nonvolatile organic compounds (NVOCs), and volatile organic compounds (VOCs) for observed data. The top 50 risk-ranked chemicals included 15 that were unregulated, with at least one chemical from each chemical class (chromium-6 [#1, IOC], chlorate and NDMA [#11 and 12, DBP], 1,4-dioxane [#25, SOC], PFOS, PFOA, PFHxS [#42, 44, and 49, NVOC], and 1,2,3-trichloropropane [#48, VOC]). These results suggest that numerous unregulated chemicals are of higher exposure risk or hazard in US drinking water than many regulated chemicals. These methods could be applied following each Unregulated Contaminant Monitoring Rule (UCMR) data collection phase and compared to retrospective data that highlight what chemicals potentially pose the highest exposure risk or hazard among US drinking water, which could inform regulators, utilities, and researchers alike.

Short abstract

National drinking water data for regulated and unregulated chemicals were risk ranked to inform researchers, utilities, and policy makers which chemicals pose the highest exposure risk or hazard.

Introduction

Access to clean, potable water is a growing area of concern for many US water systems due to aging infrastructure, impaired source waters, social vulnerability, limited budgets, and an increase in unregulated contaminants of concern.¹⁻⁴ This problem is exemplified by well-known, recent events such as lead exposure from drinking water in Flint, Michigan, toxic algal bloom formation in Lake Erie, Ohio, and its subsequent impact on drinking water in Toledo, Ohio, to the frequent detection of per- and polyfluorinated compounds (PFAS) in drinking waters.⁵⁻⁷ Given the vast quantity of chemicals produced annually, with an estimated 1500 new chemicals generated globally each year,⁸ and the paramount importance of preserving access to potable water, identifying and prioritizing chemicals in source and drinking waters is a key area for maintaining public safety. To protect against the risks and hazards of impaired drinking water and sources, it is essential that water utilities have the tools to properly monitor levels of chemicals that represent the highest exposure risk or hazard.

The current approach to regulating and identifying contaminants in US drinking water is a multistep, iterative process, conducted by the Environmental Protection Agency (USEPA). Regulation is directed by the Safe Drinking Water Act (SDWA), in collaboration with drinking water utilities and state and local health agencies that enforce the National Primary Drinking Water Regulations (NPDWR). The NPDWR contains a list of both microbiological and chemical contaminants that have been assigned either a maximum contaminant level (MCL) or a required treatment method to minimize known associated hazards. The evaluation of these MCL chemicals and potentially new MCL chemicals utilizes both national surveillance and desktop studies. Regulated drinking water contaminants are evaluated by collecting MCL data from utilities nationally and compiling these data into what is known as the Six-Year Review of Drinking Water (Six-Year Review). This data from the Six-Year Review are then evaluated along with other factors (e.g., health effects to treatment technologies), to determine if a regulatory revision is necessary to strengthen public health protection. The Unregulated Contaminant Monitoring Rule (UCMR) addresses unregulated contaminants that may impact drinking water. The UCMR utilizes a desktop study to generate a Contaminant Candidate List (CCL) that prioritizes the potential hazard and risk of exposure to thousands of contaminants and selects up to 30 for a multiyear national study. While these two studies together are intended to periodically update the regulated MCL list, only a single chemical, uranium, has been added to the regulated list in over 20 years. Over the past two decades, numerous chemicals (e.g., chromium-6 to perfluorooctanesulfonic acid (PFOS)) have been detected in national drinking water at levels of public health concern, but they are still absent from the list of regulated contaminants.^9,10 This point demonstrates a need for new evaluation methods to ensure that the chemicals representing the highest exposure risk or hazard are identified and prioritized based on their prevalence at levels of concern in public water systems.¹¹

Health reference-normalized evaluation methods are proven techniques used by the USEPA (e.g., CCL) to evaluate and compare chemicals and have recently been used in the proposed regulation of several PFAS chemicals. These approaches, such as the Margin of Safety, are also used in a variety of national and international drinking water guidelines and reports. Two USEPA examples are (1) USEPA’s CCL, which uses a health reference level (HRL) to estimated environmental concentration ratio to narrow a list of contaminants from thousands to roughly 30 to aid in the UCMR selection process,^13,14 and (2) the proposed use of a summed hazard index (HI) for several PFAS (PFNA, PFBS, GenX, and PFHxS) with action being required if this combined level exceeds a threshold of 1.0. To create this HI, the EPA uses a chemical’s environmental concentration divided by its health-based reference level, generating a health reference normalized value. The CCL approach is like a margin of safety, where a smaller value is of greater risk/hazard, while the HI approach for PFAS is the opposite since the HRL is the denominator, resulting in larger values being of greater risk/hazard. Unfortunately, health reference normalization is not used beyond the current UCMR study to compare retrospective occurrence data of regulated contaminants (i.e., Six-Year Review data) and past UCMR studies. Such a comparison would be possible through the application of HIs for data from both the UCMRs and the most recent Six-Year Review. This approach could be applied at the end of the UCMR process, which compares the national prevalence of contaminants at or above their respective HRLs, thereby providing a new tool to evaluate regulated and unregulated contaminants nationally and regionally.

Thus, the goal of this study is to synthesize and organize national drinking water data to prioritize chemical classes and rank individual chemicals (regulated and unregulated) using HIs. HIs are commonly used for hazard-based (noncarcinogenic) and not risk-based (carcinogenic) chemicals; however, this study uses HIs similarly to the National Research Council, which applied a margin of safety approach to compare both hazard- and risk-based chemicals in drinking water.¹⁵ The data set was assembled using USEPA national occurrence data (Six-Year Review, Safe Drinking Water Information System (SDWIS), and UCMRs) along with the HRLs of the USEPA and other authoritative agencies to generate HIs. This data set was then combined with geospatial and population-based information for individual potable water systems, using their public water systems identification numbers (PWSID). The combined data set was then used to achieve the following objectives: (1) to characterize the frequency and national distribution of chemical classes in US drinking water with concentrations greater than their HRLs, (2) risk-rank all 139 chemicals studied (Table 1) based on their individually exposed populations to concentrations greater than their HRL for that chemical’s studied population, and (3) to identify the highest exposure risk or hazard chemicals, particularly unregulated, from each chemical class. As such, this work applies HIs on a scale that has yet to occur to answer key questions around drinking water chemicals while also demonstrating a new tool for identifying these hazard- and risk-based chemicals in US drinking water.

Table 1. List of Chemicals and Their Chemical Classes^a.

chemical class	chemicals
inorganic chemicals (IOCs)	antimonyR; arsenicR; bariumR; berylliumR; cadmiumR; chromium (MCL)R; chromium-6UR; cobaltUR; combined radium (226 and 228; MCL)R; combined uraniumR; cyanideR; germanium (CCL4)UR; gross alpha (MCL; no radon or uranium)R; gross beta (MCL)R; lead (WHO)R; manganeseUR; mercuryR; molybdenumUR; nitrateR; nitriteR; seleniumR; strontiumUR; thalliumR; vanadiumUR
volatile organic chemicals (VOCs)	1,1-dichloroethaneUR; 1,1-dichloroethyleneR; 1,1,1-trichloroethaneR; 1,1,2-trichloroethaneR; 1,2-dichloroethaneR; 1,2-dichloropropaneR; 1,2,3-trichloropropaneUR; 1,2,4-trichlorobenzeneR; 1,3-butadieneUR; benzeneR; bromomethaneUR; carbon tetrachlorideR; chlorobenzeneR; chloromethane (CCL3)UR; (cis and trans)-1,2-dichloroethyleneR; ethylbenzeneR; halon 1011UR; HCFC-22 (CCL3)UR; o-dichlorobenzeneR; p-dichlorobenzeneR; styrene (MCL)R; tetrachloroethyleneR; tolueneR; trichloroethyleneR; vinyl chlorideR; total xylenesR
semivolatile organic chemicals (SOCs)	1-butanolUR; 1,2-dibromo-3-chloropropaneR; 1,3-dinitrobenzeneUR; 1,4-dioxaneUR; 2-methoxyethanolUR; 2-propen-1-olUR; 2,3,7,8-TCDDR; 2,4-DR; 2,4,5-TPR; acetochlorUR; alachlorUR; alpha-hexachlorocyclohexaneUR; atrazineR; bde-153UR; bde-47UR; bde-99UR; benzo(a)pyreneR; bhc-gammaR; butylated hydroxyanisole (CCL4)UR; carbofuranR; chlordaneR; chlorpyrifos (MDH)UR; dalaponR; dimethipinUR; di(2-ethylhexyl)adipateR; di(2-ethylhexyl)phthalateR; dimiethoateUR; endothallR; endrinR; ethoprop (CCL4)UR; ethylene dibromideR; glyphosateR; heptachlorR; heptachlor epoxideR; hexabromobenzeneR; hexachlorobenzeneR; hexachlorocyclopentadieneR; methoxychlorR metolachlorUR; o-toluidine (CCL4)UR; oxamylR; oxyfluorfen (CCL4)UR; pentachlorophenolR; PCBsR; picloramR; profenofos (CCL4)UR; quinoline (CCL3)UR; RDXUR; simazineR; tebuconazole (CCL4)UR; terbufos sulfoneUR; total permethrin (CCL4)UR; tribufosUR; TNTUR; toxapheneR
nonvolatile organic chemicals (NVOCs)	17-alpha-ethynylestradiol (CCL3)UR; 17-beta-estradiol (CCL3)UR; acetochlor esaUR; acetochlor oaUR; alachlor esaUR; alachlor oaUR; equilin (AGWR)UR; estriol (AGWR)UR; estrone (AGWR)UR; metolachlor esaUR; metolachlor oaUR; PFBS (DREPA)DR; PFHxS (DREPA)DR; PFNA (DREPA)DR; PFOA (DREPA)DR; PFOS (DREPA)DR; testosterone (AGWR)UR
disinfection byproducts (DBPs)	bromateR; bromodichloromethaneR; bromoformR; chlorate (CCL3)UR; chloriteR; chloroformR; dibromochloromethaneR; dichloroacetic acidR; monochloroacetic acidR; NDBAUR; NDEAUR; NDMAUR; NDPAUR; NMEAUR; NPYRUR; trichloroacetic acidR

^aThe HRL source used for each chemical is also provided with designations if they were not drawn from the USEPA Regional Screening Tables for Resident Tap Water (n = 109). Superscript “R” means regulated; superscript “UR” means unregulated; superscript “DR” means draft regulations, currently unregulated.

Materials and Methods

Data Acquisition

This study includes national drinking water data from systems serving more than 90% of the US population (PWSID population served) compiled from the Third Six-Year Review, SDWIS for lead from 2010–2020, and UCMRs 2 (2008–2010), 3 (2013–2015), and 4 (2018–2020), utilizing the largest and most current USEPA data sets available to achieve this study’s national-scale goals. Each data source was downloaded from their respective USEPA Web site, cleaned, and organized as detailed in the Supporting Information (SI). The data set consisted of more than six million testing records for 139 contaminants, of which 74 are regulated and 65 are unregulated. Chemical values, PWSID characteristics, and population numbers were obtained from the Six-Year Review, UCMRs, and SDWIS. For Six-Year Review data, to be qualified for inclusion in the data set, a potable water system had to serve over 500 people, be deemed as “finished water”, and either occur at an accepted facility type (clear well, distribution system (DS), sampling station, or treatment plant (e.g., water system facility)) or at one of the accepted sampling points (DS, entry point, first customer, midpoint in DS, or point of maximum residence). The data were geotagged at the county level using Federal Information Processing Standards (FIPS) codes for each PWSID. Several data-handling methods were used in the interpretation and presentation of the data. These included calculating the mean HI value for each chemical in that PWSID (n = 30,362), with that chemical class’s highest HI value being mapped for that county (n = 2,982), in Figure 1 and the mean HI for PWSIDs (n = 24,00) in Figure S2, while Figure 2 used all data to identify PWSIDs that experienced a HI value greater than 1 for each chemical. For measurements that were below a chemical’s method detection limit (MDL), these nondetects were set to zero (observed) or to half of a chemical’s MDL (censored, discussed later and in the SI). See Table S1 for individual chemical’s MDL values (average MDL when a chemical had more than one MDL) and the data acquisition method and source.

Figure 1.

PWSIDs with at least one individual chemical in each chemical class (DBPs, SOCs, NVOCs, IOCs) that has a measured environmental concentration greater than its respective chemical HRL (VOCs; Figure S1). Since multiple PWSIDs (n = 30,362) and samples exist for a single FIPS code (n = 2,982), the mean HI value for each chemical in that PWSID was calculated with that chemical class’s highest HI value being mapped for that county.

Figure 2.

Risk ranking by population exposed for the top 50 chemicals in the study that had a measured environmental concentration greater than its HRL. Population estimates were determined by identifying each PWSID that had a chemical occurrence above its HRL and then summing up the population studied from each of these public water systems. The PFAS chemicals using UCMR5 data are included in the risk ranking counts given in the text.

Individual chemicals were divided into classes based on chemical characteristics. The chemical classes included inorganic compounds (IOCs; n = 24), volatile organic compounds (VOCs; n = 27), semivolatile organic compounds (SOCs; n = 55), nonvolatile organic compounds (NVOCs; n = 17), and disinfection byproducts (DBPs; n = 16). Classes were determined by the chemical nature of the compounds: IOCs and radionuclides were grouped into the IOC class, while organic chemicals were separated by their volatility or designated as a DBP. The full list of chemicals and their classes is summarized in Table 1.

HRLs

Health reference data were drawn from multiple national and state government agencies, including USEPA’s Regional Screening Level Tables for Resident Tap Water (RSL; HI value of 0.1 for noncancer and a carcinogenic target risk of 10^–6); USEPA’s CCLs (CCL3 and CCL4); Michigan Department of Environment, Great Lakes, and Energy (MDEGE); Australia’s Guidelines for Water Recycling (AGWR); Minnesota’s Department of Health (MDH); National Primary Drinking Water Regulations (MCL); USEPA’s Draft Regulations on Drinking Water Health Advisories for PFAS (DREPA); and the World Health Organization (WHO).^{12,13,16−20} Health reference data were drawn from these eight references, yet 109 of the 139 HRLs were drawn from USEPA’s RSL, with the next highest source being the USEPA’s CCL supporting documents (n = 14). This coupled with five MCLs being used for HRLs (e.g., total chromium, gross alpha and beta, and combined radium and styrene) and PFBS, PFHxS, PFNA, PFOA, and PFOS from the DREPA, which resulted in 133 of the 139 chemicals’ HRL values coming from USEPA sources. The other six were selected from these other sources if they were not reported in RSL tables or had more conservative values from national and state government agencies (e.g., lead). Table S2 details the sources for each HRL and how they were derived by the national and state government agencies (e.g., RSL Tables, CCL, WHO). Also, Table 1 details all non-RSL sources, while Table S1 details the HRL values evaluated and selected and their references for all of the chemicals. It should be noted that because fluoride has a known positive impact that could not be evaluated and accurately compared to the other chemicals, it was excluded from this study.²¹ Data management, manipulation, and processing were all done in RStudio (1.2.5042). The code used to clean, organize, and create the figures is included in the SI.

HI

The measured environmental concentrations (MEC) were normalized to their most conservative HRL values to generate individual HIs (eq 1). HIs greater than 1 are not statistical probabilities of a harmful outcome occurring; rather, they merely represent an observation that an MEC exceeded its HRL, like a hazard quotient or margin of safety.

1

Results and Discussion

Chemical Class Results for Chemicals Detected above Their Health Reference Levels

Summary Data

Application of HIs to the national drinking water data consisting of more than 6.2 million measurements, irrespective of sample date or frequency, resulted in 16% of detectable measurements being at or above a chemical’s HRL. Breaking out the 139 chemicals by chemical class^22,23 resulted in DBPs being the most common chemical class detected above their HRLs, followed by IOCs, NVOCs, SOCs, and finally VOCs (Table 2). When comparing PWSIDs impacted and their populations in this data set, IOCs exposed 86% of the 309 million studied and DBPs exposed 79% of the 282 million studied, while SOCs, VOCs, and NVOCs exposed 19% (282 m), 9% (271 m), and 7% (259 m), respectively. This represents the observed data scenario that only considers chemicals above their MDL, while the censored data result in all chemical classes exposing nearly 90% or more of their populations, which is discussed later and in the SI.

Table 2. Summary Table of Chemical Class Data that Details Measurements, Detects Greater than a Chemical’s MDL or HRL, and Percent Detected Greater than an MDL or HRL^a.

	IOCs	DBPs	SOCs	VOCs	NVOCs
# of measurements	1,036,742	1,697,772	1,711,824	1,471,266	310,861
# > MDL	454,867	1,021,621	12,768	12,340	1,043
# > HRL (observed)	112,771	888,718	3,768	2,777	893
% > HRL (observed)	10.9%	52.3%	0.22%	0.19%	0.29%
# > HRL (additional with censored)	125,641	93,270	492,077	193,212	109,070
% > HRL (additional with censored)	12.1%	5.5%	28.7%	13.1%	35.1%

^aThe number > HRL for censored data are the additional incidences that are greater than their HRL when nondetects are set to MDL/2.

The chemical class results for the percentage of chemicals detected above their HRL in observed data, which represents our known–knowns, align with those of Allaire et al., who examined national MCL violations from 1982 to 2015 and found that select IOCs (e.g., arsenic) and DBPs were the most commonly regulated chemicals cited in health-based violations.²⁴ Additionally, recent reconnaissance studies by the USEPA and the United States Geological Survey (USGS) that examined influent and effluent chemical concentration from 29 drinking water plants across the USA found similar results to those presented here with respect to prevalence of NVOCs and SOCs. Twenty-five effluent drinking water samples were tested for 118 individual pharmaceuticals (2950 chemical measurements) in that study’s second phase, with a detection frequency of 1.2%.²⁵ These samples were also analyzed for 35 anthropogenic waste indicators (AWIs, mixtures of NVOCs and SOCs; 900 chemical measurements), which were present in just 3.7% of samples.²² It should be noted that the high detection frequencies of several PFAS in the aforementioned studies were exceptions compared to UCMR3, which is discussed in detail later (NVOCs).⁷ These studies present a snapshot of many AWIs that are regulated or unregulated for a small subset of drinking water locations around the USA, while the results (>MDL) presented are representative of a much larger temporal and geographical distribution of water samples, but with fewer AWIs (e.g., SOCs and NVOCs).

Regional Maps

Figure 1 elucidates where chemicals are detected at concentrations higher than their HRLs. US counties with chemical concentrations greater than their HRLs appear to cluster for specific chemical classes. For example, IOCs have the greatest occurrence above their HRLs in the southwestern states, parts of the west-south central states, and other states scattered around the USA (e.g., Oregon, Illinois, New York). Initially, it was hypothesized that regions using groundwater as their source water were responsible for IOC distribution, but Figure S2 depicts that IOC values greater than their HRLs are similar for surface water and groundwater samples. However, locations with the highest density of HIs greater than 100 are in regions that utilize groundwater (e.g., Arizona, California, and Illinois; Figure S3 and Table S4). It should be noted that the PWSIDs within these counties are not solely groundwater or surface water but can be under the influence of one or the other; however, the data pulled from the SDWIS for this analysis did not consider these influences.

As shown in Figure 1, DBPs had the highest frequency of elevated HIs in the west-south-central states (e.g., Texas, Oklahoma, and Kansas) while parts of the northeast (e.g., New York and Maine) and the southeast (e.g., Virginia and North Carolina) had individual PWSIDs that exist in a county with HIs substantially greater than 1. Research has shown that Oklahoma and Texas are acutely susceptible to elevated DBPs, particularly in summer months with high temperatures requiring elevated levels of disinfectants to maintain residuals in distribution systems.²⁴ Increased chlorination coupled with elevated total organic carbon (TOC) concentrations in source waters (Figure S4) creates an ideal environment for DBP formation.²⁴ Another observation from Figure S2 is that DBPs with HIs greater than one are linked to surface water (65%) sources more than groundwater (41%).

In contrast to IOCs and DBPs, NVOCs, SOCs, and VOCs were infrequently detected at levels greater than their HRLs for the observed data. An NVOC was detected above its HRL in at least one county in half of the states (Figure 1, lower 48 and Hawaii). Figure S3 illustrates that NVOCs had identical occurrences of HI greater than 1 (∼0.8%) for both groundwater and surface water. SOCs were detected above their HRLs in at least one county for 40 states. The SOCs studied were three and a half times more likely to occur in drinking water sources using surface water versus groundwater (Figures S2 and S3). VOCs were rarely detected compared to other chemical classes and occurred equally in surface water and groundwater, although nearly two-thirds of states had at least one county with a mean value greater than its HRL (Figure S1). Iowa, California, and Texas had the highest number of counties with VOC concentrations greater than their HRLs, which could be linked to agricultural or industrial practices in these states (Figure S1). While observed data in Figure 1 show comparatively low levels of NVOCs, SOC, and VOCs, this figure might not capture most of the risk and hazard associated with these chemical classes because many of these chemicals have an MDL greater than their HRL, illustrating the need for lower detection limits with these chemicals to adequately evaluate their risk and hazard in drinking water. In contrast, many IOCs and DBPs have an MDL less than their HRL, and therefore, the risk and hazard are better captured with the observed data. This is explored further with the censored data approach in the Supporting Information.

Risk Ranking Chemicals Based on Population Exposed

The prioritization of chemicals within each class is central to this research, as it identifies which chemicals represent the highest exposure risk or hazard in drinking water regardless of regulation, which was done with a population-based risk ranking. Figure 2 highlights the population exposed as a percentage of the population studied for the top 50 chemicals in the study that was measured above its HRL. Observed data (Figure 2) illustrate that chromium-6 occurred above its HRL for the largest percent of the US population, followed by six DBPs. The top 10 chemicals are either DBPs or IOCs, two of which are unregulated (chromium-6 and chlorate). Of the top 50 chemicals in Figure 2, every chemical class is included and there are 15 unregulated chemicals (Figure 2). This illustrates the importance of reevaluating regulated and unregulated chemicals since four unregulated chemicals occur in just the top 15, and 15 occur in the top 50. The following discussion primarily focuses on the top-ranking chemicals from each class based on observed data, providing a reference to what is currently known among the chemicals and their chemical class (Figure 2)

IOCs

The data presented in Figure 2 identify chromium-6 as the chemical with the highest exposure risk or hazard in the nation’s drinking water because it was detected at levels above its HRL for 94% of the studied population (Figure 2 and Table 3). This carcinogenic constituent is considered primarily anthropogenic in nature, although it has recently been shown to also occur naturally, and is detected across the USA (SI Maps).²⁶ Uranium (combined) and arsenic, which are the 9th and 10th risk-ranked chemicals for the observed data, are both considered carcinogens,^16,27,28 have maximum contaminant level goals (MCLGs) of zero, and are detected throughout the USA (SI Maps); discussion into these two chemicals can be found in the SI. Chromium-6 is present in the environment as a result of naturally eroding chromium deposits.²⁹ Also, it is used in batteries, construction and building materials, and home maintenance materials such as adhesives, paints and stain, and surface sealers.³⁰ Total chromium is a regulated chemical (100 μg/L MCL), which is thought to be primarily chromium-3 and chromium-6 in natural waters.²⁹Figure S5 shows the statistically relevant relationship (p < 0.05) between total chromium and chromium-6 in both groundwater (r² = 0.85) and surface water samples (r² = 0.51) collected for UCMR3. Even though chromium-6 constitutes a substantial proportion of the total chromium in these samples, and the EPA considers chromium-6 to be the dominant form in total chromium measurements, total chromium is regulated at a concentration four orders of magnitude greater than the USEPA RSL screening level value (Table 3). This discrepancy likely results from different exposure pathways having defined or undefined carcinogenic values. Exposure from inhalation routes is considered carcinogenic for chromium-6, based on substantial occupational and epidemiologic studies, animal studies, and in vitro bioassay data that suggest carcinogenesis may result from mutagenic oxidative DNA lesions following intracellular reduction to trivalent chromium.³⁰ Alternatively, an oral route of exposure to chromium-6 is classified as “not yet known” for carcinogenicity.^30,31 Irrespective of this, the USEPA’s Office of Pesticide Program and the State of California Environmental Protection Agency determined that chromium-6 has a mutagenic mode of action for all cells regardless of exposure pathway. This resulted in the more health-protective approach of applying age-dependent adjustment factors for early-life exposure via ingestion for use in the USEPA RSL screening level tables and thus the rationale for the value used here. Overall, the omnipresence of chromium-6 in national drinking water coupled with its carcinogenic nature illustrate the dire need for future toxicological research for an oral route to lend support for its regulatory consideration in drinking water to better reflect the highest exposure risk or hazard chemicals.³²

Table 3. List of the Four Highest Risk-Ranked Chemicals from Each Chemical Class and Their Associated Prevalence Data, Health Reference Levels and Source, Method Detection Limits, Regulated Levels, and Data Sources.

chemical class	chemical	no. of measurements	no. of measurements > MDL	no. of measurements > HRL	percent > HRL	no. of PWSID > HRL	probable carcinogen	health reference level selected (μg/L)	regulated or advisory limit (μg/L)	method detection limit (μg/L)	EPA or regulatory body	health reference level source	data source
IOC	chromium-6	61,609	46,731	44,755	72.6%	4,160	Y	3.50 × 10–2	1.00 × 102	3.00 × 10–2	CWRCB	USEPA RSL	UCMR3
IOC	combined uranium	19,313	10,357	9,569	49.5%	1,079	Y	4.00 × 10−1	3.00 × 101	1.00 × 100	USEPA (MCL)	USEPA RSL	Six Year Rev
IOC	arsenic	47,858	19,415	19,415	40.8%	3,315	Y	5.20 × 10–2	1.00 × 101	5.00 × 10–1	USEPA (MCL)	USEPA RSL	Six Year Rev
IOC	nitrate	97,989	65,377	13,693	13.9%	1,082	Y	3.20 × 102	1.00 × 104	2.00 × 10–3	USEPA(MCL)	USEPA RSL	Six Year Rev
DBP	bromodichloromethane	237,196	194,637	194,637	81.9%	9,025	Y	1.30 × 10–1	8.00 × 101a	3.00 × 10–2b	USEPA (MCL)	USEPA RSL	Six Year Rev
DBP	chloroform	237,699	194,344	194,344	79.2%	8,382	Y	8.70 × 10–1	8.00 × 101a	1.00 × 10–2b	USEPA (MCL)	USEPA RSL	Six Year Rev
DBP	dibromochloromethane	237,333	165,210	151,298	63.7%	8,399	Y	8.70 × 10–1	8.00 × 101a	3.00 × 10–2b	USEPA (MCL)	USEPA RSL	Six Year Rev
DBP	dichloroacetic acid	189,027	158,106	158,106	80.6%	6,019	Y	1.50 × 100	8.00× 101a	4.50 × 10–1c	USEPA (MCL)	USEPA RSL	Six Year Rev
SOC	atrazine	39,965	3,555	2,031	5.1%	381	Y	3.00 × 10–1	3.00 × 100	3.00 × 10–3	USEPA (MCL)	USEPA RSL	Six Year Rev
SOC	1,4-dioxane	36,121	4,161	803	2.2%	258	Y	4.60 × 10–1	1.00 × 100	7.00 × 10–2	CWRCB	USEPA RSL	UCMR3
SOC	benzo(a)pyrene	34,317	63	53	0.2%	45	Y	2.50 × 10–2	2.00 × 10–1	1.60 × 10–2	USEPA (MCL)	USEPA RSL	Six Year Rev
SOC	hexachlorocyclopentadiene	35,138	134	125	0.4%	67	N	4.10 × 10–2	5.00 × 10–1	4.00 × 10–3	USEPA (MCL)	USEPA RSL	Six Year Rev
NVOC	PFOS	36,287	288	288	0.8%	92	UDd	2.00 × 10–5	2.00 × 10–5	4.00 × 10–2	USEPA Health Adv	USEPA Health Adv	UCMR3
NVOC	PFOA	36,287	377	377	1.0%	116	UDd	4.00 × 10–6	4.00 × 10–6	2.00 × 10–2	USEPA Health Adv	USEPA Health Adv	UCMR3
NVOC	alachlor ESA	10,308	5	3	0.03%	2	UDe	1.10 × 100	2.00 × 101	1.00 × 100	USEPA (MCL)	USEPA RSL	UCMR2
NVOC	PFHxS	36,490	204	204	0.6%	53	UDd	2.70 × 10–2	2.70 × 10–2	3.00 × 10–2	USEPA Health Adv	USEPA Health Adv	UCMR3
VOC	carbon tetrachloride	61,560	288	249	0.4%	135	Y	4.60 × 10–1	5.00 × 100	2.10 × 10–1	USEPA (MCL)	USEPA RSL	Six-Year Rev
VOC	trichloroethylene	61,727	1,407	1,168	1.9%	135	Y	4.90 × 10–1	5.00 × 100	1.90 × 10–1	USEPA (MCL)	USEPA RSL	Six-Year Rev
VOC	ethylbenzene	30,108	907	237	0.4%	170	N	1.50 × 100	7.00 × 102	1.00 × 10–2	USEPA (MCL)	USEPA RSL	Six-Year Rev
VOC	1,2,3-trichloropropane	36,164	251	251	0.7%	65	Y	7.50 × 10–4	5.00 × 10–3	3.00 × 10–2	CWRCB	USEPA RSL	UCMR3

^aTotal trihalomethanes (TTHM) and HAA5 (five haloacetic acids) are regulated as a summation of four and five chemicals, respectively.

^bTTHMs for the Six-Year Review had up to five different methods used for their quantification that all had differing MDLs ranging from 0.03 to 0.002 μg/L.

^cHAA5s for the Six-Year Review had up to five different methods used for their quantification that all had different MDLs ranging from 0.45 to 0.054 μg/L.

^dUndetermined (UD) whether the chemicals are carcinogenic (USEPA).

^eUndetermined (UD) whether the chemical is carcinogenic (USEPA); however, alachlor is a probable human carcinogen.

DBPs

Six of the seven highest-exposure risks or hazardous chemicals, in terms of population exposed above their HRLs for observed data, were DBPs (Figure 2). This included bromodichloromethane, the second risk-ranked chemical, which exposed 92% of its studied population to levels greater than its HRL. Chloroform, dibromochloromethane, and dichloroacetic acid were the third, fourth, and fifth risk-ranked chemicals for the observed data, which exposed 90, 86, and 85% of the studied populations to levels greater than their HRLs, respectively. All of these DBPs are considered carcinogenic, and both bromodichloromethane and dichloroacetic acid have MCLGs of zero.³³ Because DBPs are predominantly formed as a byproduct of disinfection, they are produced during water treatment and distribution. Several of the DBPs studied here are nearly ubiquitous in US drinking water at or above their HRL (Figures 1 and 2 and SI Maps). This is compounded by there being over 600 DBPs identified in drinking water, with only a select few being regulated or having national prevalence data.³⁴ To illustrate this point, the recently completed UCMR4 detected bromochloroacetic acid (average value = 3.0 μg/L) and bromodichloroacetic acid (average value = 2.75 μg/L) (method detection limit of 0.3 μg/L) in 92 and 81% of the 51,000 samples across the USA (Table S1). The haloacetic acids (HAAs) are underrepresented in this study, as only dichloroacetic acid, trichloroacetic acid, and monochloroacetic acid (ranked 5th, 7th, and 13th in Figure 2) have HRL values derived from an authoritative agency, meaning that the other six HAAs could not be compared using HIs in this study. With so many potential DBPs present in drinking water, understanding which DBPs present the greatest exposure risk or hazard is critical. To date, only 24 DBPs have received carcinogenic in vivo testing, with 22 testing positive for carcinogenicity, which further illustrates that there is a shortfall of DBP toxicity data.^35,36 Numerous unregulated DBPs (e.g., haloacetaldehydes) have demonstrated levels of toxicity orders of magnitude greater than regulated DBPs when compared using in vitro cytotoxicity and genotoxicity bioassays.^37,38 Additionally, disinfection practices of municipalities in the USA are shifting away from chlorination to alternatives, such as chloramination, that reduce THM4 levels but form nitrosamines, iodinated DBPs, and other DBP classes. Meanwhile, nations like The Netherlands have stopped using chlorine-based disinfection altogether.³⁹ DBPs' presence in US drinking water (Figures 1 and 2) and their frequency at levels 100× their HRL (Figure S7), coupled with the paucity of toxicological data and a changing disinfection regulatory landscape, illustrate the dire need for additional surveillance and toxicological studies into both regulated and particularly unregulated DBPs (e.g., future UCMRs for haloacetonitriles, haloacetaldehydes, haloketones, haloacetamides, halonitromethanes, and iodinated THMs) that could help further evaluate the risks and hazards of a fundamental drinking water treatment process.

NVOCs

The 17 NVOCs investigated in this study were detected infrequently across the USA for observed data (Figure 1 and Table S1), with only six of them having any detection above their HRL (Figure 2 and Table S1). The NVOCs that impacted the most people were PFAS chemicals (PFOS, PFOA, and PFHxS, ranked 42nd, 44th, and 49th, respectively, for UCMR3 data) and alachlor esa (a herbicide, ranked 47th), which occurred above their HRLs and impacted around 2 to 4% of the studied populations (Figure 2). Designation of PFAS as a carcinogen is predominantly classified as “undetermined”, yet the International Agency of Research on Cancer did classify PFOA as a possible human carcinogen in 2017.⁴⁰ In 2022, the US EPA established Drinking Water Health Advisories for PFAS Chemicals and set interim updated lifetime drinking water health advisories at 0.02 ng/L for PFOS and 0.004 ng/L for PFOA. The MDLs in UCMR3 are 2000–5000 times higher than these recent health advisory levels, which are used for this study’s HRLs (Table 3) causing the large differences in what would be their risk ranks for observed versus censored data; in fact, for censored data, PFNA, PFOA, PFOS, and PFBS expose all of their studied population (Figure S6). These MDLs for UCMR3 surely underestimate the prevalence of PFAS in US drinking water, as described by Hu et al.⁴¹ and others that suggest that if the UCMR3 had used a lower MDL, much higher detection frequency would likely be observed (USEPA⁷ and USGS⁴² along with the statewide surveillance studies by Ohio⁴³ and Michigan⁴⁴). As such, it is difficult to make an accurate comparison between the UCMR3 PFAS observed data set and the other chemicals studied.

For PFAS compounds, the most comprehensive national study to date is UCMR3; however, data collection for UCMR5 is currently underway and is focused on 29 PFAS compounds and lithium. The third set of UCMR5 data was released in January 2024 and represents 24% (∼17,000 samples) of the total results, which are expected by 2026. UCMR5 has lower MDLs of 0.004 μg/L for PFOS and PFOA and 0.003 μg/L for PFHxS, which are still orders of magnitude higher than the USEPA's health advisory levels. Nonetheless, the partial UCMR5 data show significant jumps in population exposed compared to UCMR3 data: PFOS 15th instead of 39th, PFOA 18th instead of 41st, and PFHxS ranks 35th instead of 46th (Figure 2). Additionally, UCMR5 showed that PFNA and GENx had exposures of 2 and 0.5%, respectively, for their studied population, while PFHxA, PFBS, and PFBA had no exposures above their HRLs so far.

PFAS are clearly priority constituents that deserve national attention and possible regulation (Figure 2), especially because an estimated 3000 PFAS compounds are produced globally and present-day analytical methods can only explain a portion of PFAS compounds that occur in the environment.^45,46 This is compounded by shorter-chain PFAS compounds (e.g., GenX), which have limited toxicological and occurrence data, replacing legacy PFAS, and recent health studies linking PFAS levels in tap water to elevated blood levels of PFAS (e.g., bioaccumulation).⁴⁷ These studies, coupled with the current UCMR5 data, support numerous PFAS chemicals as the most important anthropogenic drinking water constituents (VOCs, SOCs, and NVOCs), in terms of exposure risk or hazard, yet not the most important drinking water chemicals when compared to select DBPs and IOCs (Figure 2).

SOCs

Among the frequently detected anthropogenic organic compounds measured in this study (NVOC, SOC, and VOC), SOCs were the most prevalent (Figure 2). Atrazine and 1,4-dioxane occurred above their respective HRLs most frequently and were the 17th and 25th risk-ranked chemicals, while benzo(a)pyrene was the 29th for observed data. Atrazine occurred above its HRL for 18% of the studied population, making it the most health-relevant SOC studied. All three chemicals are considered probable human carcinogens,^10,16,33 while just atrazine (MCLG equal to MCL) and benzo(a)pyrene (MCLG of zero) are regulated and 1,4-dioxane is not.

The presence of 1,4-dioxane in US drinking water has been known since at least 1978; yet, after more than 40 years, it still remains a contaminant of emerging concern.¹⁰ 1,4-Dioxane is used as a stabilizer for chlorinated solvents and typically reaches the environment through improper disposal of industrial waste or accidental solvent spills.⁴⁸ This lack of regulation at a national level has resulted in select states taking action. For example, New Hampshire implemented nonenforceable guidance values for drinking water (0.25 μg/L), and California established notification (1 μg/L) and response levels (35 μg/L).⁴⁹ The HRL used in this analysis (0.46 μg/L) is half of California’s notification value (Table 3). This value, taken from USEPA’s RSL residential tap water tables, is rooted in dozens of animal studies since there are inconclusive human epidemiologic data. Kano et al.’s 2-year study of 1,4-dioxane-spiked drinking water in rats and mice was used as the principal study to derive an oral cancer slope factor.^33,50 Adamson et al. studied 1,4-dioxane in the UCMR3 data set and suggested that the loading of 1,4-dioxane in drinking water supplies may be decreasing.⁵¹ Regardless of Adamson et al.’s conclusion, 1,4-dioxane is and likely will continue to be the highest-exposure risk or hazardous SOC and an important US drinking water chemical, demonstrated by its chemical risk ranking (25th; Figure 2). These results presented (Figure 2 and Table 3) for 1,4-dioxane lend support for its regulatory consideration in drinking water, especially when compared to other SOCs that are rarely detected at levels above their HRLs.

VOCs

VOCs were shown to impact the smallest percentage of the US population for the observed data, with carbon tetrachloride, trichloroethylene, ethylbenzene, and 1,2,3-trichloropropane occurring above their HRLs for 5–6% of the studied US populations and had chemical risk-ranks of 31st, 37th, 39th, 48th (Figure 2 and Table 3). Both carbon tetrachloride and trichloroethylene are regulated chemicals, considered probable human carcinogens,³³ and have MCLGs of zero. Ethylbenzene is regulated but is not considered a human carcinogen. 1,2,3-Trichloropropane, a chlorinated solvent that reaches the environment through industrial settings or hazardous waste sites, is currently unregulated. 1,2,3-Trichloropropane is highly mobile in groundwater, resistant to natural attenuation, and considered to be carcinogenic and mutagenic and to have reproductive effects, which is drawn from carcinogenicity observed in several animal (e.g., rat and mice) studies and mutagenic/genotoxic activity for both in vivo and in vitro testing.^33,52 California’s advisory level for 1,2,3-trichloropropane is 5 ng/L, which is six times lower than the UCMR’s MDL (30 ng/L) and more than six times higher than the HRL used (0.75 ng/L; Table 3). Therefore, when it was detected in this study, it was well above its HRL. The MDL of 1,2,3-trichloropropane is 40 times higher than its HRL and its prevalence could have been underestimated in UCMR3. However, it is an industrial solvent that is likely found in select regions that have industries using it, unlike PFAS. Therefore, like several of the other unregulated highly risk-ranked constituents identified here, 1,2,3-trichloropropane should receive additional toxicological and surveillance research to evaluate whether it should be considered for regulation as a VOC, due to it being in the top 3 in terms of population risk ranking for evaluated VOCs (Figure 2).

Limitations and Future Considerations

Data Uncertainties

Uncertainty in this study, and in any study utilizing these USEPA data, is attributed to the frequency of observations occurring below a chemical’s MDL. This was the case for nearly 75% of the 6.2 million measurements, and it is possible that chemicals with zero or few detections could be more important than current data indicate due to inadequate MDLs. How to best manage these nondetects is a common environmental analysis and data management challenge.⁵³⁻⁵⁵ A standard way to approach this issue is with a censored data method, which was employed in this research by setting nondetects equal to MDL/2.⁵⁴⁻⁵⁷ Applying censored data methods here skews results toward chemicals that have an HRL lower than their MDL, irrespective of how many times they were detected above their MDL (Figure S6 and Table S3). This shifts 30 chemicals to having exposed 100% of their studied population, like terbufos sulfone or BDE-99 that were only detected one time out of more than 30,000 measurements (terbufos sulfone) or not at all (BDE-99; Table S3), illustrating a challenge when analyzing the censored data alone (Figure S6). Another way to view this is by comparing between percent above MDL and a chemical’s population exposed risk ranking, which was done in Figure S7, demonstrating a weak correlation (R² = 0.035) between the two variables, while Table S3 provides some insight into how chemicals would risk-rank based on the differences/magnitude between a chemical’s MDL and HRL, as compared to the known data. There is an in-depth discussion on the MDL/2 data, but this is found in the SI. Irrespective of how one manages this missing data, MDL is an important consideration of this study that illustrates significant gaps between the known (observed) and unknown (censored) exposure risks and hazards associated with the HIs for the chemicals studied herein; suggesting the exposure risk and hazard when using HIs is likely somewhere between the observed and censored data maps and rankings depicted in Figures 1 and 2 and Figures S1, S2, and S6.

Limitations

HI approaches are critical to the research presented here and other studies to identify and prioritize chemicals in drinking water that present the greatest exposure risk or hazard to consumers of US drinking water; yet, it is also important to acknowledge their limitations. HI is a simple tool that relies on quality analytical (MECs) and toxicological data (HRLs) to try and provide answers to complex problems (e.g., imprecision of population exposure assessments, the unknown and possibly unknowable shape of the dose–response curves at low environmental levels of exposure, the translation of this information from rodents to humans as well as within the human population,⁵⁸ to mixture effects). In this study, HIs are necessitated by an effort to achieve estimates of risks and hazards for various environmental contaminants that at this time are needed to advance the field. The potential flaws of using different HRLs that have a variety of end points, different exposure pathways considered, and even different risk or hazard methods applied are not unique to this study; health impact assessments, cumulative or otherwise, suffer from our present lack of knowledge and data availability.^42,53,58,59 Yet, other researchers have applied similar approaches to drinking water recently.^21,23,42,59

In this study, ideally, all 139 chemicals would come from the same source (e.g., USEPAs RSL tables) that used the same methods and were also the most conservative. Unfortunately, that is not the case for 29 of the 139 chemicals that came from different sources. Of these 29, only eight had detections greater than 1%, and of these, six are regulated chemicals. These included the following IOCs, chromium (51%), germanium (7%), combined radium (63%), gross alpha (54%), gross beta (50%), and lead (73%), while the others were HCFC-22 (2%, SVOC) and chlorate (55%, DBP). If this was expanded to UCMR5 data, several PFAS would also be included (e.g., PFOS, PFOA, and PFHxS), which would result in 11 of the top 50 risk-ranked chemicals coming from non-RSL sources. The consequence of this for MCL and CCL chemicals is potentially higher population-based risk rankings (due to lower HRL) since MCLs consider more than just health-based data (e.g., associated costs of treatment) while CCL HRL calculations utilized an adult population (70 kg) versus the RSL noncancer chemicals that used a child (15 kg; Table S2). Conversely, both lead (WHO) and PFAS (DREPA) HRLs were more conservative than the RSL table values (e.g., PFOS 6 ng/L [RSL] vs 0.004 ng/L [DREPA]), which would shift their population-based risk ranking to lower risk or hazard chemicals if their RSL values were used. This highlights the challenge with calculating HRLs and the associated considerations (e.g., uncertainty factors), resulting in this research relying on experts at authoritative agencies (e.g., USEPA to WHO) who calculated HRLs, and this study using the values that are the most protective/conservative.

Future Considerations and Applications

Overall, the chemical class results illustrate the importance of IOCs and DBPs in US drinking water, regardless of regulation, especially when compared to NVOCs, VOCs, and SOCs that are less likely to be detected at levels greater than their HRL, except for select PFAS chemicals. Of the 139 chemicals evaluated for observed data, chromium-6 was determined to be the highest exposure risk or hazard among drinking water chemicals. The top 10 chemicals were either DBPs or IOCs, with two of these being unregulated (chromium-6 and chlorate), while if this was expanded to the top 25, there were nine unregulated chemicals (DBPs, IOCs, NVOCs, and SVOCs). These results suggest that, for the known data set, many unregulated chemicals are of higher risk or hazard than numerous regulated chemicals. Unfortunately, several of these unregulated chemicals are recalcitrant to conventional drinking water treatments, reinforcing the need to understand the drinking water chemicals that are of highest risk or hazard to that specific region or utility (e.g., elevated incoming organic carbon levels that can lead to DBP formation to source water high in PFAS that may require specialized media). Although PFAS, at this time, do not appear to be the highest risk or hazard constituents, their potential regulation could have a multitude of cobenefits since mitigation strategies that use activated carbon would also lower the TOC of that treated drinking water, likely resulting in lower DBP formation and mitigation of other organic constituents. Additionally, the results presented here on DBPs coupled with recent studies illustrating regulated DBPs, making up only a small fraction of DBP toxicity in drinking water, suggest that substantially more research is needed around DBP prevalence in US drinking water and that UCMR6 should potentially focus on these unregulated DBPs.⁶⁰

This manuscript presents a new dynamic tool that could be employed following the UCMR data collection phase by which the resulting data (e.g., UCMR5) are compared to retrospective data, allowing for a side-by-side evaluation of regulated and unregulated constituents using population-based risk ranking. This process, like the SDWA, would use updated environmental concentration (2023 Six-Year data) and toxicological data as they are available along with new chemicals to be evaluated versus retrospective data (e.g., UCMR5). As such, this research aims to provide a tool for regulators and researchers alike to inform their assessment, prioritization, and comparison of the chemicals representing the highest exposure risk or hazard among drinking water in the USA.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.3c08745.

• Chemical names and data sources (XLSX)

• Results discussion on risk ranking and censored data, links to data sets, code used in R for data wrangling and figures, and individual chemical maps (PDF)

The authors declare no competing financial interest.