Advertisement

  • Loading metrics

Open Access

Peer-reviewed

Research Article

Bottled water, tap water and household-treated tap water–insight into potential health risks and aesthetic concerns in drinking water

Abstract

Understanding drinking water quality at the point-of-use across a range of consumer options is essential for designing effective public health interventions in the face of deteriorating source waters and complex contaminant mixtures. This is especially pressing as the popularity of tap water alternatives like bottled water and household treatment increases, yet this data is largely missing from the academic literature and policy discussions. This study presents one of the first evaluations of water quality comparing three common consumer drinking water options in the nine county San Francisco Bay Area with a survey of 100 analytes in 100 bottled water samples, 603 tap water samples, and 111 samples of household-treated tap water. Analytes measured included general water quality characteristics, metals, other inorganics, volatile organic compounds (including disinfection byproducts), and three microbial indicator species in bottled water only. Samples were evaluated to assess potential taste, odor, and color issues, as well as potential health risks by calculating cumulative toxicity quotients to reflect the additive toxicity of chemical mixtures. All three drinking water options had potential health risks, primarily driven by the presence of trihalomethanes (contributing from 76.7 to 94.5% of the total cumulative toxicity across the three drinking water options). While tap water had the highest potential toxicity among the three drinking water options, results suggest that household-scale treatment may reduce the potential for aesthetic issues and health risks of tap water.

Citation: Bear SE, Waxenberg T, Schroeder CR, Goddard JJ (2024) Bottled water, tap water and household-treated tap water–insight into potential health risks and aesthetic concerns in drinking water. PLOS Water 3(9): e0000272. https://doi.org/10.1371/journal.pwat.0000272

Editor: Inderjeet Tyagi, ZSI: Zoological Survey of India, INDIA

Received: March 5, 2024; Accepted: July 31, 2024; Published: September 4, 2024

Copyright: © 2024 Bear et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All data required to reproduce the results and statistics of this study are provided in the SI or are readily available for download. Individual sample results were de-identified and anonymized and are available in Supporting Information Table S8.

Funding: The authors received no specific funding for this work.

Competing interests: The authors have declared that no competing interests exist.

Abbreviations: BW, bottled water; HTTW, household-treated tap water; TW, tap water; TQ, toxicity quotient; POU, point-of-use; POE, point-of-entry; EPA, Environmental Protection Agency; FDA, Food and Drug Administration; SOQ, Standards of Quality; PET, polyethylene terephthalate; VOC, volatile organic compound; DBP, disinfection byproduct; THM, trihalomethane

1. Introduction

Tap water quality issues are on the rise in the US. Deteriorating infrastructure [1], climate change pressures on source waters [2, 3], increasingly complex chemical mixtures in the environment [4, 5], governance failures [6], and low technical, financial and managerial capacity have led to compounding pressures on water systems. Lead, arsenic, nitrate, uranium and bacterial contaminants are all federally regulated and frequently found in violation of drinking water standards across the US [7–9]. Where tap water quality is in question or consumers distrust their water system for other reasons, people turn to alternative sources like bottled water [10] and household-scale treatment of their tap water [11–13]. Bottled water use is growing rapidly, with per capita consumption in the US having increased from 27.8 gallons per person per year in 2010 to 45.2 gallons per person per year in 2020 [14]. Household water treatment is a 2.09-billion-dollar market in the US [15]. While these alternatives to tap water are attracting a growing share of tap water consumers, insight into their water quality implications is scant. Lack of adequate monitoring data across consumer options hinders policymakers’ ability to identify effective economic and public health interventions to address tap water quality issues and public health more broadly. This study offers a comparative analysis of bottled water, household-treated tap water, and tap water quality using a unique dataset to evaluate potential organoleptic and toxicity trade-offs among drinking water options for households in the San Francisco Bay Area. The resulting analysis offers insight into multiple toxicity mitigation priorities for improving drinking water quality.

There is a widespread perception that bottled water is “pure” and free of contaminants, due in part to misleading marketing [16–18]. Additionally, aesthetic properties of tap water like taste and odor have been documented to impact people’s perception of tap water safety [19–21], potentially driving people toward alternatives like bottled water [22, 23]. However, a wide range of contaminants have been identified in US bottled water due to contamination from source water, bottle processing, or packaging material leachate [16]. Contaminants detected in bottled water products include bacteria, various heavy metals (including lead), volatile and semi-volatile organic contaminants (including phthalates), disinfection byproducts, radiological elements, microplastics, and various PFAS compounds [24–31].

Households seeking improved tap water also turn to point-of-entry (POE) and point-of-use (POU) treatment of delivered tap water. In the international context, concerns about appropriate technology selection and whether systems are maintained appropriately are central to understanding household-treated tap water quality [32]. In the US, household behaviors with treatment equipment and their impact on water quality are unknown. Evidence suggests that people purchase treatment units in response to publicized water quality contamination events or concerns about the taste, odor and color of their water [23], rather than in response to targeted testing to characterize on-site contamination. While perceptions can correlate with contaminant occurrence, there are many contamination issues that have no organoleptic effects. As such, household-treated tap water poses an exposure risk if treatment is not tailored to a household’s water quality.

The existing literature on drinking water quality is not typically geared toward understanding differences among drinking water options at the household level primarily due to a lack of adequate data. Monitoring and reporting for water system-supplied tap water quality is limited due to a focus on regulated contaminants that are not monitored at the tap (except for lead). Previous studies of tap water quality have been extensive in the contaminants they evaluate but have smaller sample sizes (≤45) [4, 33–35]. In a study of public tap water and private well supplies in North and South Dakota, the authors concluded that public monitoring data beyond water system compliance is needed “to inform consumer POE/POU treatment decisions’’ across the US [35]. However, while household-treated tap water quality has been studied internationally [36], to our knowledge, there are no large-scale studies of household-treated tap water in the US. Also, while numerous bottled water quality analyses exist, they rarely represent the quality of a “typical” bottled water purchase based on the market share of products in a given region. Thus, while these studies indicate potential contamination across the different water sources, a comparative picture of water quality across these three consumer options is lacking.

To our knowledge, this study is one of the first large sample size water quality comparisons among consumer drinking water options and the first U.S. based study of household-treated tap water. One hundred bottled water products (across 89 brands) were collected and tested, and this data was compared to 714 tap water samples across the San Francisco Bay Area collected by households using a consumer test kit product–including 603 direct tap water samples and 111 samples with follow-up household water treatment. The bottled water sample set reflects a “typical” bottled water purchase in the Bay Area, whereas the 714 tap water samples are part of a citizen-science dataset (a convenience sample). One hundred analytes, spanning metals, volatile organic compounds, and common water quality parameters, were surveyed in all three sources, as well as three bacteriological indicators in bottled water.

Potential risks to human health were evaluated using a cumulative toxicity framework in which the concentrations of analytes in a sample were compared to health benchmarks, and potential organoleptic effects were also identified via comparison with aesthetic benchmarks [37, 38]. The findings highlight potential drivers of water source preference with respect to organoleptics, as well as drivers of cumulative toxicity, for all three drinking water options. Tractable recommendations are offered for consumers and policymakers aspiring to improve POU tap water quality in the face of complex environmental and financial trade-offs among drinking water options.

2. Methods

To compare the quality of drinking water options in the Bay Area, this study leverages a primary dataset of bottled water samples and a unique secondary dataset of (unpaired) tap water samples and household-treated tap water samples. Each sample was evaluated for 100 analytes, many with possible organoleptic effects and/or potential toxicity. A dataset of aesthetic and health benchmarks for drinking water analytes was compiled from engineering, toxicology, and public health literature to identify concentrations of concern. These data were combined to analyze levels of exceedance and potential toxicity across all three drinking water options. This section first presents the methods of sample collection and laboratory analysis used in each dataset, followed by the comparative data analysis approach.

2.1. Bottled water sample selection and collection

A sample design was developed to represent a “typical” bottled water (BW) purchase for consumers in the San Francisco Bay Area nine county region (Alameda, Contra Costa, Marin, Napa, San Francisco, San Mateo, Santa Clara, Solano, and Sonoma counties). A representative sample set was considered to be one that approximately mimicked the market share of BW products sold in the US in 2020 by Food and Drug Administration (FDA) category [39]. FDA categories are based on the source water and/or water treatment used and include “spring water”, “artesian water”, “mineral water”, “well water” and “purified water” [40]; see Table A in S1 File for sample details. The final sample included 100 BW products encompassing 89 brands (Table A in S1 File) purchased from 80 stores across the nine-county region (Fig A in S1 File). Forty six percent of products were from groundwater sources with no specific treatment requirements from FDA, while another 46% of products were from various sources (often unspecified) with FDA treatment requirements (referred to as “purified”, including those marketed as “distilled”). The remaining 8% of products did not meet FDA requirements for any official bottled water categories.

All 100 products were purchased between July 12th and July 28th, 2022. They were sent, unopened, to the Microbac Laboratories facility in Dayville, CT for sampling and analysis within one week of purchase.

2.2. Tap water and household-treated tap water sample selection and collection

Tap water and household-treated tap water samples were selected from a secondary data set obtained from SimpleLab, Inc of individual samples taken over time (as opposed to repeat samples at specific locations). All samples in this dataset were from water system supplied households in the nine county San Francisco Bay Area that purchased water quality testing kits from Tap ScoreTM (a product of SimpleLab, Inc.) between September 17th, 2020 and August 16th, 2022 (Fig B in S1 File). Selected tap water (TW) and household-treated tap water (HTTW) samples were limited to those with the complete set of 100 analytes analyzed in the BW samples to enable a direct comparison. Samples were categorized as HTTW if residents reported using in-home water treatment devices (excluding water softeners), and as TW if not. HTTW samples and TW samples are not paired samples and therefore contaminant removal efficiencies of household treatment choices could not be calculated. The final dataset includes 714 samples: 603 TW samples and 111 HTTW samples. Seventy-seven percent of water systems supplying the TW and HTTW samples in this study rely on surface water as the primary water source, though nearly all have some groundwater-reliant facilities (see Table F in S1 File for summary of water system treatment and water source characteristics).

For sample collection, Tap ScoreTM test kits included sample containers and detailed instructions regarding appropriate sampling technique. Two sample containers were provided: one 250 mL HDPE bottle for a first draw sample filled to the shoulder (for analysis of general chemical characteristics, metals and other inorganics), and one 40 mL clear glass VOA vial containing 25 mg of ascorbic acid for a fully flushed sample with no headspace (for volatile organic compound analysis).

2.3. Laboratory analyses

All samples across the two datasets were analyzed at accredited, commercial laboratories using EPA-approved methods for 100 select analytes, including general water quality characteristics (e.g., pH, hardness, total dissolved solids, etc.) and chemical constituents (Tables B and C in S1 File). Analytes were selected based on the contaminants most frequently tested among households that used Tap ScoreTM. BW was further analyzed for three microbiological indicators (total coliform, E. coli, and total HPC); these indicators were not measured in TW or HTTW samples as they are not included in the standard water testing package. Only quantitative data (> method reporting limits) was reported and considered a detection.

2.4. Data analysis

Sample results were assessed individually for potential toxicity and aesthetic concerns, as well as cumulatively within samples and across drinking water options. Anonymized analytical sample results are provided in Table J in S1 File.

2.4.1. Health and aesthetic benchmarks.

To evaluate potential health risks and organoleptic impacts from the analytes measured, a list of health and aesthetic benchmark values was developed for each analyte. If an analyte has no known impact on human health and/or the aesthetic experience of drinking water, or research is insufficient for determining an impact, no benchmark was assigned.

Health benchmarks were aggregated from various public health agencies and governmental bodies. These benchmarks reflect concentrations of contaminants below which no known non-cancer impacts are expected over a lifetime of exposure (typically toxicity to specific organ systems), or concentrations that result in specific cancer risk levels over a lifetime of exposure (from 10−4 to 10−6). Similar to previous studies, the lowest health benchmark available for each analyte was used in order to evaluate all analytes against the most health-protective value available [24, 33–35, 37, 38, 41] (Table D in S1 File).

Similarly, aesthetic benchmarks were derived by creating a database of benchmark concentrations for analytes that have been associated with organoleptic effects like off tastes or odors, or discoloration, and the lowest concentration among benchmarks was selected as the aesthetic benchmark. These benchmarks were gathered from institutional, governmental and academic sources (Table E in S1 File).

2.4.2. Cumulative toxicity quotient calculations.

Frameworks for cumulative health impacts assume health risks from exposures are additive and they guide many public health assessments of drinking water quality [42–44]. Without information on duration of exposure, volume of water consumed, or personal susceptibility, health risks or effects cannot be fully characterized. Instead, detected concentrations were compared to benchmarks that indicate potential toxicity of an exposure given (typically) lifetime exposure duration and common assumptions about relative source contributions across exposure pathways. The resultant ratio is known as a toxicity quotient (TQ).

TQs were calculated by dividing the concentration of a contaminant by its health benchmark: , where Ci is the concentration of contaminant i, and Bi is the benchmark for contaminant i [37, 38]. Any health benchmarks of zero were replaced with the value of the reporting limit in order to calculate a finite toxicity quotient [24, 33–35, 41]. Cumulative toxicity quotients (∑TQs) were calculated by summing the TQs of individual contaminants in a sample: . ∑TQs were used to assess potential toxicity concern using the effects-screening-level threshold of concern, ∑TQ = 0.1, and the benchmark equivalent concentration, ∑TQ = 1. ∑TQ = 0.1 indicates that a particular contaminant may be relevant to the overall risk profile of the sample, whereas ∑TQ = 1 defines the point at which exposure to the sample may present a health risk over the course of a lifetime. This approach has origins in EPA’s hazard quotient and hazard index approach to quantifying potential health risks from environmental exposures [45, 46].

2.4.3. Statistical analyses.

Welch’s one-way analysis of variance (ANOVA) was conducted using the rstatix package in R to evaluate difference of means among ∑TQs in the three sample groups [47, 48]. Welch’s test accounts for unbalanced groups and unequal variances among groups. Because means can be sensitive to outliers, a sensitivity analysis was performed removing outliers defined as ∑TQs greater than 75th percentile plus 1.5 times the interquartile range for all three sample groups. The null hypothesis was that there was no significant difference in mean ∑TQs among the three sample groups. Where Welch’s ANOVA was significant, it was followed up with nonparametric Games-Howell posthoc tests to evaluate differences in means among the three sample groups.

3. Results and discussion

A wide variety of contaminants were detected in samples from all three drinking water options, including many with the potential to impact human health or the odor, taste, or color of drinking water. The presence of contaminant mixtures in these samples is consistent with previous studies of BW [24, 30] and TW quality in California [9, 49–51]. Contaminant level summary statistics by drinking water option are shown in Table G in S1 File.

3.1 Exposure-benchmark comparison: Aesthetic benchmarks

Potential aesthetic issues were identified in samples from all three drinking water options, but higher concentrations of contaminants with aesthetic benchmarks were detected in TW and HTTW compared with BW. Forty-one of the 100 analytes measured have aesthetic benchmarks. Twenty of these 41 analytes were detected in at least one sample, including volatile organic compounds (VOCs), metals, other inorganics, and general water quality characteristics such as hardness and total dissolved solids (Fig 1).

thumbnail
Download:
Fig 1.

Detected concentrations (in μg/L) of analytes with aesthetic benchmarks in bottled water (blue, n = 100), household-treated tap water (green, n = 111) and tap water (purple, n = 603) for contaminants that were detected in more than one water source (a) and those detected in only one water source (b). Percent of samples in which analyte was detected on the right. Concentrations are plotted on a log-scale to allow for legibility in comparing concentrations among BW, TW, and HTTW.

https://doi.org/10.1371/journal.pwat.0000272.g001

Across BW samples, six analytes exceeded their aesthetic benchmarks, as compared with eight and 12 in HTTW and TW samples, respectively (Fig C in S1 File). At least one aesthetic benchmark exceedance was found in 24% (n = 24/100), 44% (n = 49/111), and 41% (n = 246/603) of BW, HTTW, and TW samples, respectively. The concentrations of analytes were generally similar among water sources, with some exceptions where BW concentrations were lower (including zinc, copper, fluoride, and bromoform).

Calcium was the only analyte to exceed aesthetic benchmarks at a higher proportion in BW than in HTTW and TW (4% of results in exceedance in BW versus 0% for both HTTW and TW). Aesthetic exceedances across water sources were largely from analytes that impact taste and staining–such as metals (aluminum, copper, iron, manganese and, for TW only, zinc), magnesium, and sodium (Fig C in S1 File). For example, magnesium and sodium exceeded taste thresholds for bitter or salty flavors in all water sources. The benchmark for magnesium was exceeded in 16% (n = 16/100), 33% (n = 37/111) and 29% (n = 173/603) of BW, HTTW and TW samples, respectively, and the aesthetic benchmark for sodium was exceeded in 10% (n = 10/100), 19% (n = 21/111) and 20% (n = 122/603) of BW, HTTW and TW samples, respectively. These elements are naturally occurring, likely present in all sources due to natural processes, and are not fully removed via many at-home treatment processes–sodium, in fact, may be added to water that is softened via an ion exchange water softener [52, 53].

Iron and copper both exceeded taste thresholds in TW and HTTW in higher proportions than in BW. The aesthetic benchmark for iron was exceeded in 2% (n = 2/100), 4% (n = 4/111) and 10% (n = 62/603) of BW, HTTW and TW samples, respectively, and that of copper was exceeded in 0%, 7% (n = 8/111) and 5% (n = 30/603) of BW, HTTW and TW samples, respectively. These metals can enter drinking water via water distribution lines and household plumbing and fixtures [54, 55], all of which BW mostly avoids. These results are consistent with studies showing that aesthetic impacts like off-tastes are common in TW and a primary reason people turn to BW [19–21].

3.2 Exposure-benchmark comparison: Health benchmarks

Analytes with potential health risks were detected in samples from all three drinking water options in exceedance of health benchmarks. Eighty-four of the 100 contaminants tested in all samples have health benchmarks. Forty of these 84 contaminants, including disinfection byproducts, other VOCs, metals, and other inorganics, were detected in at least one sample (Fig 2). Seventeen health benchmarks were exceeded in both BW and HTTW, and 25 were exceeded in TW (Fig D in S1 File). At least one health benchmark was exceeded in 53% (n = 53/100), 61% (68/111), and 98% (n = 590/603) of BW, HTTW, and TW samples, respectively. These differences likely stem, in large part, from treatment and source water effects across samples.

thumbnail
Download:
Fig 2.

Detected concentrations (in μg/L) of analytes with health benchmarks in bottled water (blue, n = 100), household-treated tap water (green, n = 111) and tap water (purple, n = 603) for contaminants that were detected in more than one water source (a) and those detected in only one water source (b). For panel (b), one very high value for tin obscures the concentrations of other contaminants and therefore an axis break is plotted showing observations between 0 μg/L and 100 μg/L, and then the very high value at 5,110 μg/L. Percent of samples in which analyte was detected are indicated on the right. Concentrations are plotted on a log-scale to allow for legibility in comparing concentrations among BW, TW, and HTTW.

https://doi.org/10.1371/journal.pwat.0000272.g002

Fewer health benchmark exceedances in BW likely reflects both the high degree of treatment of the 46 purified products (44 of which employed either distillation or reverse osmosis, see Table A in S1 File) and the overall high integrity of source waters for the 46 groundwater-sourced products. Specific treatment is not required by the FDA for groundwater-sourced BW products as long as they meet quality standards [40]; only 11 of 46 groundwater-sourced BW products listed any treatment, which included filtration and disinfection (either via UV or ozone). On the other hand, nearly all TW and HTTW samples in this study were taken from water systems that rely on conventional drinking water treatment (coagulation, flocculation, sedimentation, filtration, and disinfection) (Table F in S1 File). Conventional treatment is largely adequate to meet regulatory standards, but not necessarily effective in reducing contaminant concentrations below health benchmarks or removing them entirely.

While 98% of TW samples had at least one health benchmark exceedance, 61% of HTTW samples had an exceedance, indicating that some treatment effect is likely. Specific treatment information for HTTW samples is not comprehensive in the dataset, but 88 out of 111 samples reported using carbon-based filters and/or reverse osmosis systems, which likely contributed to the lower overall exceedances. However, the proportion of health benchmark exceedances remaining in HTTW samples indicates that barriers still persist for achieving health risk reduction using at-home treatment. For example, household treatment technologies are often purchased in the absence of water testing to identify appropriate technologies, and effective contaminant removal can be hindered by technology selection and improper maintenance.

3.2.1 Trihalomethanes.

The trihalomethanes (THMs) chloroform, bromodichloromethane and dibromochloromethane were three of the contaminants with the most frequent health benchmark exceedances across all samples tested (Fig D in S1 File). THMs are formed when chlorine disinfectant reacts with natural organic matter in source waters, which is why mitigation strategies employed by water utilities largely focus on controlling precursors to THMs (i.e. monitoring total organic carbon) [56]. The proportion of samples with THM exceedances was highest in TW, followed by BW and HTTW. There has been growing concern about disinfection byproducts (DBPs) in drinking water [57–59], which is consistent with the detections of these chlorinated DBPs. Health effects of the three THMs include developmental/reproductive effects, liver toxicity, and an increased risk of cancer [60–63]. THMs were the only contaminants analyzed in this study that exceeded the limits guiding their regulation. Four THMs are regulated as a group called total THMs: chloroform, bromoform, bromodichloromethane and dibromochloromethane. Eight BW samples (8%) exceeded the California BW quality regulatory limit of 10 μg/L for total THMs (which is lower than the FDA’s SOQ of 80 μg/L) [64], while 13 TW samples (2%) exceeded the federal maximum contaminant level of 80 μg/L [65]. No HTTW samples exceeded either of these regulatory limits. However, individual THMs exceeded their health protective benchmarks–ranging from 0 to 0.22 μg/L (Table D in S1 File)–in all three drinking water options. For example, chloroform exceeded its health benchmark in 32% of BW samples (n = 32/100), 25% of HTTW samples (n = 28/111) and 89% of TW samples (n = 534/603). Exceedances (equal to detections) in BW were similar to the proportion of detections in Bradley et al. [24]. Most purified BW products employed reverse osmosis as the primary treatment, which has been shown to incompletely remove THMs [66]. Because regulatory limits for THMs are so much higher than health benchmarks, it is likely that all drinking water options will continue to present health risk while remaining in compliance with regulations.

The high proportion of THM detections in TW is consistent with violation trends in California [49] and reported concentrations in Bay Area water systems [51]. Surface water typically has higher levels of organic matter (i.e., THM precursors) than groundwater and, as such, water systems primarily reliant on surface water are at higher risk of THM formation when using chlorine as a disinfectant [63]. The vast majority of TW and HTTW samples (92%) were taken by households in water systems reliant on surface water as their primary source (of these systems, 52% have facilities that also rely on at least one groundwater source). Moreover, 60% of these samples are from households in water systems that use chlorine or chloramine as a disinfectant (Table F in S1 File). Taken together, the impact of THMs on potential toxicity across water sources is unsurprising and reflects a significant challenge for water systems reliant on surface water and chlorination.

3.2.2 Metals.

Contamination due to metals is another well-known drinking water safety challenge, often caused by corrosion of pipes and fixtures and/or inadequate corrosion control in the presence of corrosive source waters [55, 67]. Lead was detected (and exceeded its health benchmark of 0) in samples from all three drinking water options, though only in one BW sample. In contrast, lead was detected in 30% (n = 33/111) of HTTW samples and 51% (n = 306/603) of TW samples. The proportion of detections in BW is lower than that reported by Bradley et al. [24], who found lead in five out of 30 samples (17%) but had lower method detection limits than this study. Lead is a persistent challenge due to aging infrastructure across the US, and exposure to lead can cause adverse neurological, developmental, learning and behavioral effects, especially in children [68–70]. The higher proportion of exceedances found in HTTW and TW is consistent with the potential for contamination from the distribution system and on-premises plumbing.

Other metals that exceeded health benchmarks have a variety of potential sources, including on-site fixtures and faucets subject to less stringent regulations than lead in pipes (e.g. copper and nickel), and source waters (arsenic, uranium, cadmium, cobalt). Copper and nickel were detected in samples from all drinking water options and exceeded their health benchmarks in 7% to 14% of HTTW and TW samples, but neither exceeded their benchmarks in BW. A number of geogenic elements also exceeded health benchmarks in samples from all drinking water options, consistent with previous studies [24, 35, 49, 51]. Uranium exceeded its health benchmark in 6% of BW samples (n = 6/100), as compared with 2% in HTTW (n = 2/111) and TW (n = 13/603). Uranium has been shown to cause kidney and osteo- toxicity in human and animal studies, as well as adverse female reproductive and developmental effects in animal studies [71, 72]. Arsenic was detected in a similar proportion of samples in BW and HTTW samples (7% or n = 7/100, and 8% or n = 9/111, respectively), and in a lower proportion of TW samples (3% or n = 16/603). Exposure to arsenic in drinking water has been shown to increase the risk of cancer, as well as cause adverse dermal and blood system impacts [73, 74].

Source water is likely the primary driver of these geogenic exceedances. All uranium detections in BW were in groundwater-sourced products (13% of groundwater-sourced BW products; n = 6/46). Similarly, arsenic exceeded its health benchmark in groundwater-sourced BW products only (15%, n = 7/46). Bradley et al. [24] similarly found uranium and arsenic in groundwater-sourced BW, but at substantially higher proportions (74% and 87% for uranium and arsenic, respectively) using lower detection limits. The lower proportion of these contaminants in HTTW and TW is consistent with the fact that the majority of tap water samples were sourced from surface water, with a smaller proportion of facilities reliant on groundwater sources (Table F in S1 File).

3.2.3 Other VOCs.

In BW, additional detections and exceedances suggest that bottle production and processing may play a role in water quality. Two petroleum-derived compounds–benzene and toluene were detected in BW only. One benzene detection was in exceedance of the health benchmark, and all 5 toluene detections were below the health benchmark. One possibility is that these contaminants were introduced to bottled water products during processing [75].

3.2.4 Bacteria.

The microbiological indicators total coliform, E. coli, and total heterotrophic bacteria (total HPC) were evaluated in BW samples only. While no E. coli or total coliform were found in BW samples, heterotrophic bacteria were detected in 43% of BW samples (Fig E in S1 File). Heterotrophic bacteria are not considered good indicators of pathogenic bacteria [76], but total HPC serves as an indicator of overall sterility. These findings corroborate previous studies identifying HPC in BW, which found detections in 30% - 71% of samples [24, 27]. While tap water samples are likely to have HPC as well because of its ubiquitous presence in the environment, these findings are particularly interesting in BW because of marketing claims about the purity of BW.

3.3 Cumulative toxicity quotients

A cumulative toxicity quotient (∑TQ) was calculated for each sample to assess overall potential toxicity by summing detected contaminant concentrations divided by their health benchmarks. Fig 3 plots cumulative toxicity quotients for samples of each drinking water option in comparison with the benchmark equivalent concentration (∑TQ = 1).

thumbnail
Download:
Fig 3. Distribution of cumulative toxicity quotient values for each sample for bottled water, household-treated tap water, and tap water.

The red line indicates the benchmark equivalent concentration (∑TQ = 1). Posthoc test results with p-values adjusted for multiple hypothesis testing shown above boxplots to indicate significant difference in group means between pairs at p <0.0001 (****).

https://doi.org/10.1371/journal.pwat.0000272.g003

Eighty percent (n = 80/100), 97% (n = 108/111), and >99% (n = 602/603) of BW, HTTW and TW samples, respectively, exceeded the effects-screening-level of concern, while 54% (n = 54/100), 83% (n = 92/111), and >99% (n = 598/603) of BW, HTTW and TW samples, respectively, exceeded the benchmark equivalent concentration. This indicates a high potential for health risk due to lifetime consumption of water from each of the drinking water options tested, with the highest potential hazard posed by TW and the lowest by BW. These ∑TQ results agree with those of prior studies of both nationwide POU TW and BW [33–35, 41], though Bradley et al. [24] saw a higher proportion of effects-screening-level and benchmark equivalent concentration exceedances for BW. The mean (and standard deviation) of ∑TQ was 12.0 (±23.5), 20.3 (±39.6), and 130.3 (±75.4) in BW, HTTW, and TW, respectively, indicating the disproportionately higher potential toxicity of TW compared with BW and HTTW (Table H in S1 File).

Drinking water option was significantly associated with cumulative toxicity quotient outcomes using Welch’s one-way ANOVA (F = 502.44, p<0.0001) to account for unequal variance [77] (Tables A and B in S1 Text). Effect size as measured by omega-squared was large (0.77; Table C in S1 Text). This finding was robust to a sensitivity analysis removing outliers. Overall, Games-Howell posthoc tests indicated the average TW sample had significantly higher potential toxicity than that of both HTTW and BW samples (p<0.0001), whereas the average HTTW and BW sample had no significant difference in mean toxicity quotients (Fig 3; Table D in S1 Text). While mean differences were insignificant between BW and HTTW, the variability of HTTW ∑TQs was greater than that of BW ∑TQs (Table H in S1 File), as indicated by a wider interquartile range (1.3–18.4) compared with that of BW (0.2–12.2). This suggests that HTTW had slightly more variability in potential toxicity at a sample level. This likely reflects the range of technologies applied to the HTTW samples, which included sediment filters, carbon-based filters, reverse osmosis, ion exchange, distillation, KDF filters, UV disinfection and unspecified filtration.

To identify which contaminants contributed most to the total potential toxicity of each drinking water option, a “percent contribution” to potential toxicity was calculated for each contaminant. First, within a given sample group, individual TQs were grouped by contaminant and summed. The ∑TQ of each contaminant across all samples in a group was then compared to the ∑TQ of all contaminants within that sample group (Eq 1): Eq 1

Individual contaminant TQs across samples indicate that, at the sample level, a diverse range of contaminants may be responsible for the potential toxicity of any individual sample (Figs F-H in S1 File), underscoring the importance of sample-level toxicity assessments to consider exposure risks at a household-level. However, when contaminants are identified for their percent contribution to the overall potential toxicity of a drinking water option (Fig 4; Table I in S1 File), toxicity mitigation strategies for broader public health improvements can be prioritized, as discussed in the following section.

thumbnail
Download:
Fig 4. Proportion of toxicity contributed by contaminants contributing 1% or more of cumulative toxicity to a given water source.

https://doi.org/10.1371/journal.pwat.0000272.g004

Consistent with the exceedance results in Section 3.2 and findings in Bradley et al. [24], THMs–driven primarily by chloroform–were responsible for most of the potential toxicity for all three drinking water categories– 79.4%, 76.7%, and 94.5% for BW, HTTW and TW, respectively (Fig 4; Table I in S1 File). Addressing chloroform and THMs generally is important across all drinking water options, but Fig 4 (and Table I in S1 File) illustrates that 94.5% of potential toxicity as determined by the 100 analytes evaluated in TW samples would be addressed at the tap by treating THMs (0.87% of total potential toxicity is contributed by bromodichloromethane which is displayed in the “Other” group in Fig 4).

The remaining risk for TW and HTTW was primarily driven by lead, with contributions from other metals (including nickel, copper, cadmium and cobalt), geogenic elements (arsenic and uranium) and fluoride for HTTW, while the remaining risk for BW was driven by geogenic elements (including lithium, uranium, boron and arsenic). In the case of HTTW, potential toxicity was composed of a wider range of compounds–likely because many household treatment technologies are mitigating trihalomethanes and thus other unaddressed contaminants drive overall toxicity.

3.4 Implications for toxicity mitigation

Taken together, the absolute exceedances and cumulative toxicity results have a range of implications for consumer choice among BW, HTTW and TW. First, analytes impacting taste thresholds exceeded benchmarks in TW and HTTW samples more than in BW samples, supporting evidence that aesthetic differences between TW and BW may drive consumers toward BW. However, cumulative toxicity results indicate that BW is not free from exposure risk despite emphatic marketing otherwise. Though not addressed in this study, the environmental and financial burden of BW is also substantial when compared with those of HTTW and TW–environmental impacts of BW are orders of magnitude greater than that of TW [78, 79] and the price of BW exceeds that of HTTW or TW by thousands of dollars per year [17]. While BW may be preferable for a temporary period where TW is declared unsafe, HTTW is likely a more affordable and sustainable option where aesthetic issues and potential toxicity risk persist in TW [80]. This is supported by the finding that the difference in mean cumulative potential toxicity was insignificant between HTTW and BW.

While HTTW has the potential to mitigate certain contaminants, potential toxicity was still identified in HTTW samples. This underscores the importance of technology selection tailored to POU water quality and ongoing user maintenance to ensure HTTW efficacy. Fig 4 demonstrates the potential mitigating effect of household treatment on disinfection byproducts relative to TW and illustrates the diverse range of contaminants that contribute to the toxicity of these samples overall. The wide range of contaminants found in HTTW samples would likely be reduced if treatment technology were chosen in accordance with POU water quality. Further research to support user compliance and maintenance of treatment devices, perhaps in partnership with utilities, is warranted.

Third, TW and HTTW samples had higher exceedances and potential toxicity than BW from metals commonly used in distribution systems, plumbing, and fixtures. For example, lead was detected in 51% of TW samples and 30% of HTTW samples (but only one BW sample), and contributed 2.9% and 6.6% of overall toxicity to TW and HTTW, respectively (Fig 4). Lead is a well-studied problem that has garnered significant policy response–the recently proposed Lead & Copper Rule Improvements would require utilities to replace all lead service lines within 10 years [81]. While this should reduce lead levels in TW and HTTW, it has been over 30 years since the original Lead & Copper Rule was established [54] and almost 10 years since the Flint, MI water crisis. Policy can be slow to address even well-understood contaminants, and other metals–like cobalt, nickel, and cadmium detected in TW and HTTW samples–have diverse sources that may be more challenging to mitigate through targeted policies. In such cases, properly designed, targeted household-scale treatment may be an effective intervention to reduce toxicity further than currently possible with drinking water standards. Some cities have begun to implement this, for example Denver Water delivered free pitcher filters to households to support their lead remediation goals [82].

Finally, targeted treatment for THMs would reduce the risk profile for all drinking water options, but would have the greatest impact in TW. Given the prevalence of surface water-supplied water systems in the San Francisco Bay Area, THMs will likely persist above health protective benchmarks, but below regulatory standards, if current water treatment practices continue. At least 68% of treatment systems for HTTW samples included carbon-based filters, which should be effective in reducing THM concentrations, and HTTW samples had much lower THM concentrations than TW samples in this study. Given the gap between regulatory standards and health benchmarks for THMs, treatment at home may be the only near-term strategy for mitigation of THMs to health-protective levels. Further THM mitigation in TW is important even if BW is the primary drinking water source for a household, as most people still use TW or HTTW for other domestic purposes and would be exposed to THMs via inhalation [79, 83].

3.5 Study limitations

This study addresses several limitations of previous studies, with the inclusion of a large sample size of BW products representative of San Francisco Bay Area consumer choices, an unprecedented number of in-home TW quality results, and “in practice” water quality from people treating their water at home. Still, there are known limitations of this study and areas for future research. First, while data was produced using EPA approved methods, commercial laboratories often have higher detection limits than academic or agency laboratories; several differences in these findings from the recent study on BW by Bradley et al. [24] indicate differences in analytical sensitivity.

Second, the cumulative toxicity quotient approach assumes that contaminant impacts are cumulative, though some chemicals may be more or less than additive [84]. The approach is also limited by the analytical scope of the study (conservative with respect to the full range of potential contaminants present) and availability of health benchmarks. For example, various organic contaminants that were not measured here–such as pesticides, PFAS compounds, phthalates and additional DBPs–have been detected in BW in prior studies [24, 28, 31, 85]. The limitation of scope is also salient in the case of DBPs. There are over 700 known DBPs, many of which are more toxic than THMs, and THMs are not necessarily good surrogates for other DBPs, especially iodine- and nitrogen-containing DBPs that can be formed at higher levels when chloramines are used as disinfectants [57, 86]. Thus, it is expected that a broader contaminant panel would indicate further contamination challenges across the three water sources, with unique issues for BW.

The representativeness of the TW and HTTW results is another potential limitation. Samples were taken by individuals in the San Francisco Bay Area but not with explicit representation across water systems, though most water systems had similar water source types and treatment practices. Fig B in S1 File indicates a bias over-representing the five southern counties of the San Francisco Bay Area with respect to the location of samples. The large sample size of results across the nine-county area reflects a range of scenarios, but research into the defining characteristics of people analyzing their tap water would allow for a better characterization of bias in the citizen science data.

Lastly, detailed statistical analyses of source water impacts on toxicity profiles (i.e., surface water versus groundwater versus mixed-sources) were not possible given the available data (see Table F in S1 File), and future research into this area would support more targeted toxicity mitigation efforts.

4. Conclusions

Alternatives to tap water, including household-scale treatment of tap water and bottled water, are increasing in popularity but information regarding their water quality is limited. This is the first large-scale study comparing the water quality of the realistic consumer drinking water options of bottled water, tap water, and household-treated tap water. Potential aesthetic concerns were identified in all drinking water options, but were more common in TW and HTTW samples, supporting previous evidence that people switch to BW in response to aesthetic concerns. Potential toxicity of samples was also identified despite the quality of the municipally supplied water in the study being largely within state and federal drinking water limits, and the BW quality mostly falling within FDA standards.

Overall, TW had significantly higher average potential toxicity than BW and HTTW. Potential toxicity in all three water sources was primarily attributed to THMs, followed by metals for TW and HTTW, likely derived from distribution systems, household plumbing and/or fixtures, and geogenic elements in groundwater-derived BW and HTTW. Improving THM management across all three water sources would have a significant impact on the cumulative potential toxicity of samples. Average potential toxicity of samples was not significantly different between HTTW and BW, suggesting that BW–which has higher environmental and financial costs than HTTW–is not a superior alternative to TW where household treatment is possible. Persistent aesthetic issues and potential toxicity in HTTW could be addressed by designing household-scale treatment to specifically address identified issues, perhaps in partnership with water systems. This approach is not common and represents an area for future research and policy design to achieve safe water goals. Evidence is provided showing higher water quality among households using household-scale treatment as compared with those using TW, which indicates that treatment at POU can address water quality issues and thus may be a tractable complement to improved TW to improve public health.

Acknowledgments

The bottled water sample collection and analysis was funded by SimpleLab, Inc. We thank Noor Brody for technical review of early versions of the analysis, Alea Laidlaw for research assistance at the project’s inception, and eight anonymous experts for their support on study design. Finally, we thank Carsten Prasse for providing us with critical feedback on the original manuscript.

References

  1. 1. Riggs , Hughes J, Irvin D, Leopard K, Riggs E Irvin D, Leopard K HJ. An Overview of Clean Water Access Challenges in the United States. Chapel Hill: US Water Partnership; UNC Environmental Finance Center; 2017. Available: https://www.urbanwaterslearningnetwork.org/wp-content/uploads/2019/05/UNC-Clean-Water-Access-Challenges2017.pdf
    • 2. Barlow PM, Reichard EG. Saltwater intrusion in coastal regions of North America. Hydrogeol J. 2010;18: 247–260.
    • 3. Proctor CR, Lee J, Yu D, Shah AD, Whelton AJ. Wildfire caused widespread drinking water distribution network contamination. AWWA Water Sci. 2020;2: e1183.
    • 4. Bradley PM, Kolpin DW, Romanok KM, Smalling KL, Focazio MJ, Brown JB, et al. Reconnaissance of Mixed Organic and Inorganic Chemicals in Private and Public Supply Tapwaters at Selected Residential and Workplace Sites in the United States. Environ Sci Technol. 2018;52: 13972–13985. pmid:30460851
    • 5. Ferraro PJ, Prasse C. Reimagining safe drinking water on the basis of twenty-first-century science. Nat Sustain. 2021;4: 1032–1037.
    • 6. Hanna-Attisha M, LaChance J, Sadler RC, Schnepp AC. Elevated blood lead levels in children associated with the flint drinking water crisis: A spatial analysis of risk and public health response. Am J Public Health. 2016;106: 283–290. pmid:26691115
    • 7. Allaire M, Wu H, Lall U. National trends in drinking water quality violations. PNAS. 2018/02/15. 2018;115: 2078–2083. pmid:29440421
    • 8. Bae J, Lynch MJ. Ethnicity, Poverty, Race, and the Unequal Distribution of US Safe Drinking Water Act Violations, 2016–2018. Sociol Q. 2023;64: 274–295.
    • 9. 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. 2011/06/07. 2011;119: 1272–1278. pmid:21642046
    • 10. Allaire M, Mackay T, Zheng S, Lall U. Detecting community response to water quality violations using bottled water sales. Proc Natl Acad Sci. 2019;116: 20917–20922. pmid:31570603
    • 11. Brown KW, Gessesse B, Butler LJ, MacIntosh DL. Potential Effectiveness of Point-of-Use Filtration to Address Risks to Drinking Water in the United States. Environ Health Insights. 2017;11: 1178630217746997. pmid:29270018
    • 12. Herkert NJ, Merrill J, Peters C, Bollinger D, Zhang S, Hoffman K, et al. Assessing the Effectiveness of Point-of-Use Residential Drinking Water Filters for Perfluoroalkyl Substances (PFASs). Environ Sci Technol Lett. 2020;7: 178–184.
    • 13. Verhougstraete MP, Gerald JK, Gerba CP, Reynolds KA. Cost-benefit of point-of-use devices for lead reduction. Environ Res. 2019;171: 260–265. pmid:30690272
    • 14. Rodwan JG. Bottled Water 2020: Continued Upward Movement—U.S. and International Developments and Statistics. Bottled Water Reporter. 2020.
    • 15. Varma J. Home Water Filtration Unit Market Analysis, 2017–2030. In: Grand View Research. 2022.
    • 16. Diduch M, Polkowska Ż, Namieśnik J. Factors affecting the quality of bottled water. J Expo Sci Environ Epidemiol. 2013;23: 111–119. pmid:23093103
    • 17. Bottled Gleick P. and Sold: The Story Behind Our Obsession With Bottled Water. Island Press; 2010.
    • 18. Euijin Y A. BD, A. EM, M. FK. Exploring the Relationship between Public Trust toward the Water Sector and the Use of Bottled Water within US Shrinking Cities. J Water Resour Plan Manag. 2023;149: 4022071.
    • 19. Doria M de F, Pidgeon N, Hunter PR. Perceptions of drinking water quality and risk and its effect on behaviour: A cross-national study. Sci Total Environ. 2009;407: 5455–5464. pmid:19664799
    • 20. Doria MF, de França Doria M. Factors influencing public perception of drinking water quality. Water Policy. 2010;12: 1–19.
    • 21. Doria MF. Bottled water versus tap water: understanding consumers’ preferences. J Water Health. 2006;4: 271–276.
    • 22. Javidi A, Pierce G. U.S. Households’ Perception of Drinking Water as Unsafe and its Consequences: Examining Alternative Choices to the Tap. Water Resour Res. 2018;54: 6100–6113.
    • 23. March H, Garcia X, Domene E, Sauri D. Tap Water, Bottled Water or In-Home Water Treatment Systems: Insights on Household Perceptions and Choices. Water. 2020.
    • 24. Bradley PM, Romanok KM, Smalling KL, Focazio MJ, Evans N, Fitzpatrick SC, et al. Bottled water contaminant exposures and potential human effects. Environ Int. 2023;171: 107701. pmid:36542998
    • 25. Diduch M, Polkowska Ż, Namieśnik J. Chemical Quality of Bottled Waters: A Review. J Food Sci. 2011;76: R178–R196. pmid:22416719
    • 26. Krachler M, Shotyk W. Trace and ultratrace metals in bottled waters: Survey of sources worldwide and comparison with refillable metal bottles. Sci Total Environ. 2009;407: 1089–1096. pmid:18990431
    • 27. Lalumandier JA. Fluoride and Bacterial Content of Bottled Water vs Tap Water. Arch Fam Med. 2000;9: 246–250. pmid:10728111
    • 28. Luo Q, Liu Z, Yin H, Dang Z, Wu P, Zhu N, et al. Migration and potential risk of trace phthalates in bottled water: A global situation. Water Res. 2018;147: 362–372. pmid:30326398
    • 29. Mason SA, Welch VG, Neratko J. Synthetic Polymer Contamination in Bottled Water. Frontiers in Chemistry. 2018. Available: https://www.frontiersin.org/articles/10.3389/fchem.2018.00407 pmid:30255015
    • 30. Olson ED, Poling D, Solomon G. Bottled Water: Pure Drink or Pure Hype? 1999.
    • 31. Chow SJ, Ojeda N, Jacangelo JG, Schwab KJ. Detection of ultrashort-chain and other per- and polyfluoroalkyl substances (PFAS) in U.S. bottled water. Water Res. 2021;201: 117292. pmid:34118648
    • 32. Rosa G, Clasen T. Estimating the Scope of Household Water Treatment in Low- and Medium-Income Countries. Am Soc Trop Med Hyg. 2010;82: 289–300. pmid:20134007
    • 33. Bradley PM, LeBlanc DR, Romanok KM, Smalling KL, Focazio MJ, Cardon MC, et al. Public and private tapwater: Comparative analysis of contaminant exposure and potential risk, Cape Cod, Massachusetts, USA. Environ Int. 2021;152: 106487. pmid:33752165
    • 34. Bradley PM, Argos M, Kolpin DW, Meppelink SM, Romanok KM, Smalling KL, et al. Mixed organic and inorganic tapwater exposures and potential effects in greater Chicago area, USA. Sci Total Environ. 2020;719: 137236. pmid:32126404
    • 35. Bradley PM, Romanok KM, Smalling KL, Focazio MJ, Charboneau R, George CM, et al. Tapwater Exposures, Effects Potential, and Residential Risk Management in Northern Plains Nations. ACS ES&T Water. 2022;2: 1772–1788. pmid:36277121
    • 36. Sobsey MD, Stauber CE, Casanova LM, Brown JM, Elliott MA. Point of Use Household Drinking Water Filtration: A Practical, Effective Solution for Providing Sustained Access to Safe Drinking Water in the Developing World. Environ Sci Technol. 2008;42: 4261–4267. pmid:18605542
    • 37. Baldwin AK, Corsi SR, De Cicco LA, Lenaker PL, Lutz MA, Sullivan DJ, et al. Organic contaminants in Great Lakes tributaries: Prevalence and potential aquatic toxicity. Sci Total Environ. 2016;554–555: 42–52. pmid:26950618
    • 38. Corsi SR, De Cicco LA, Villeneuve DL, Blackwell BR, Fay KA, Ankley GT, et al. Prioritizing chemicals of ecological concern in Great Lakes tributaries using high-throughput screening data and adverse outcome pathways. Sci Total Environ. 2019;686: 995–1009. pmid:31412529
    • 39. Varma J. Bottled Water Market Analysis, 2016–2028. In: Grand View Research. 2021.
    • 40. US FDA. 21 CFR §165.110. 2022.
    • 41. Smalling KL, Bradley PM, Romanok KM, Elliot SM, Lambert J de, Focazio M, et al. Exposures and potential health implications of contaminant mixtures in linked source water, finished drinking water, and tapwater from public-supply drinking water systems in Minneapolis/St. Paul area, USA. Environ Sci Water Res Technol. 2023.
    • 42. Evans S, Campbell C, Naidenko O V. Cumulative risk analysis of carcinogenic contaminants in United States drinking water. Heliyon. 2019;5: e02314. pmid:31687532
    • 43. Pace C, Balazs C, Bangia K, Depsky N, Renteria A, Morello-Frosch R, et al. Inequities in Drinking Water Quality Among Domestic Well Communities and Community Water Systems, California, 2011‒2019. Am J Public Health. 2022;112: 88–97. pmid:34936392
    • 44. Stoiber T, Temkin A, Andrews D, Campbell C, Naidenko O V. Applying a cumulative risk framework to drinking water assessment: a commentary. Environ Heal. 2019;18: 37. pmid:31036021
    • 45. Goumenou M, Tsatsakis A. Proposing new approaches for the risk characterisation of single chemicals and chemical mixtures: The source related Hazard Quotient (HQS) and Hazard Index (HIS) and the adversity specific Hazard Index (HIA). Toxicol Reports. 2019;6: 632–636. pmid:31334033
    • 46. US EPA. An Overview of Methods for EPA’s National-Scale Air Toxics Assessment. 2011. Available: https://www.epa.gov/sites/default/files/2015-10/documents/2005-nata-tmd.pdf
    • 47. Kassambra A. rstatix: Pipe-Friendly Framework for Basic Statistical Tests. 2023. Available: https://rpkgs.datanovia.com/rstatix/
    • 48. R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2018. Available: https://www.r-project.org/
    • 49. Acquah S, Allaire M. Disparities in drinking water quality: evidence from California. Water Policy. 2023;25: 69–86.
    • 50. Balazs C, Goddard JJ, Chang C, Zeise L, Faust J. Monitoring the human right to water in California: development and implementation of a framework and data tool. Water Policy. 2021;23: 1189–1210.
    • 51. Bangia K, August L, Slocombe A, Faust J. Assessment of contaminants in California drinking water by region and system size. AWWA Water Sci. 2020;2: e1194.
    • 52. WHO. Hardness in Drinking-water. World Health Organization; 2011. Available: https://www.who.int/teams/environment-climate-change-and-health/water-sanitation-and-health/chemical-hazards-in-drinking-water/hardness
    • 53. WHO. Sodium in Drinking-water. World Health Organization; 2003. Available: https://www.who.int/teams/environment-climate-change-and-health/water-sanitation-and-health/chemical-hazards-in-drinking-water/sodium
    • 54. Schock MR, Lytle DA. Corrosion Control Principles and Strategies for Reducing Lead and Copper in Drinking Water Systems. US Environmental Protection Agency; 1991.
    • 55. US EPA. Optimal Corrosion Control Treatment Evaluation Technical Recommendations for Primacy Agencies and Public Water Systems. 2016.
    • 56. US EPA. Stage 1 and Stage 2 Disinfectants and Disinfection Byproducts Rules. 1 May 2015. Available: https://www.epa.gov/dwreginfo/stage-1-and-stage-2-disinfectants-and-disinfection-byproducts-rules
    • 57. Richardson SD. Invited Perspective: Existing Rules for Disinfection By-Products Are Good, but They Are Not Enough. Environ Health Perspect. 2022;130: 81302. pmid:35913907
    • 58. Richardson SD, Ternes TA. Water Analysis: Emerging Contaminants and Current Issues. Anal Chem. 2022;94: 382–416. pmid:34841857
    • 59. Villanueva CM, Evlampidou I, Ibrahim F, Donat-Vargas C, Valentin A, Tugulea A-M, et al. Global assessment of chemical quality of drinking water: The case of trihalomethanes. Water Res. 2023;230: 119568. pmid:36621278
    • 60. ATSDR. Toxicological profile for chloroform. Atlanta, GA: US Department of Health and Human Services, Public Health Service; 1997 Apr. Available: https://wwwn.cdc.gov/TSP/ToxProfiles/ToxProfiles.aspx?id=53&tid=16
    • 61. ATSDR. Toxicological profile for bromoform and dibromochloromethane. Atlanta, GA: US Department of Health and Human Services, Public Health Service; 2005 Apr. Available: https://www.atsdr.cdc.gov/ToxProfiles/tp130.pdf
    • 62. ATSDR. Toxicological profile for bromodichloromethane. Atlanta, GA: US Department of Health and Human Services, Public Health Service; 2020 Apr. Available: https://www.atsdr.cdc.gov/ToxProfiles/tp129.pdf
    • 63. Weisman RJ, Heinrich A, Letkiewicz F, Messner M, Studer K, Wang L, et al. Estimating National Exposures and Potential Bladder Cancer Cases Associated with Chlorination DBPs in U.S. Drinking Water. Environ Health Perspect. 2022;130: 87002. pmid:35913906
    • 64. California Department of Public Health. Sherman Food Drug Cosmetic Law. Health & Safety Code 2018.
    • 65. US EPA. National Primary Drinking Water Regulations. 2015. Available: https://www.epa.gov/ground-water-and-drinking-water/national-primary-drinking-water-regulations
    • 66. Breitner LN, Howe KJ, Minakata D. Effect of Functional Chemistry on the Rejection of Low-Molecular Weight Neutral Organics through Reverse Osmosis Membranes for Potable Reuse. Environ Sci Technol. 2019;53: 11401–11409. pmid:31513383
    • 67. Triantafyllidou S, Edwards M. Lead (Pb) in Tap Water and in Blood: Implications for Lead Exposure in the United States. Crit Rev Environ Sci Technol. 2012;42: 1297–1352.
    • 68. ATSDR. Toxicological profile for lead. Atlanta, GA: US Department of Health and Human Services, Public Health Service; 2020 Apr. Available: https://wwwn.cdc.gov/TSP/ToxProfiles/ToxProfiles.aspx?id=96&tid=22
    • 69. Lanphear BP, Lowry JA, Ahdoot S, Baum CR, Bernstein AS, Bole A, et al. Prevention of Childhood Lead Toxicity. Pediatrics. 2016;138: e20161493. pmid:27325637
    • 70. (NTP) NTP. NTP Monograph on Health Effects of Low-Level Lead. U.S. Department of Health and Human Services; 2012. Available: https://ntp.niehs.nih.gov/ntp/ohat/lead/final/monographhealtheffectslowlevellead_newissn_508.pdf
    • 71. Arzuaga X, Gehlhaus M, Strong J. Modes of action associated with uranium induced adverse effects in bone function and development. Toxicol Lett. 2015;236: 123–130. pmid:25976116
    • 72. ATSDR. Toxicological profile for uranium. Atlanta, GA: US Department of Health and Human Services, Public Health Service; 2013 Apr. Available: https://www.atsdr.cdc.gov/ToxProfiles/tp150.pdf
    • 73. ATSDR. Addendum to the Toxicological Profile for Arsenic. Atlanta, GA: US Department of Health and Human Services, Public Health Service; 2016. Available: https://www.atsdr.cdc.gov/toxprofiles/Arsenic_addendum.pdf
    • 74. OEHHA. Public Health Goal for Arsenic in Drinking Water. Sacramento, CA: California Environmental Protection Agency; 2004. Available: https://oehha.ca.gov/media/downloads/water/chemicals/phg/asfinal.pdf
    • 75. Allen L, Darby JL. Quality Control of Bottled and Vended Water in California: A review and comparison to tap water. J Environ Health. 1994;56: 17–22. Available: https://www.jstor.org/stable/44534073
    • 76. WHO. Heterotrophic Plate Counts and Drinking-water Safety. IWA Publ; 2003.
    • 77. Blanca MJ, Alarcón R, Arnau J, Bono R, Bendayan R. Non-normal data: Is ANOVA still a valid option? Psicothema. 2017;29: 552–557. pmid:29048317
    • 78. Gleick PH, Cooley HS. Energy implications of bottled water. Environ Res Lett. 2009;4: 014009.
    • 79. Villanueva CM, Garfí M, Milà C, Olmos S, Ferrer I, Tonne C. Health and environmental impacts of drinking water choices in Barcelona, Spain: A modelling study. Sci Total Environ. 2021;795: 148884. pmid:34247071
    • 80. Cohen A, Ray I. The global risks of increasing reliance on bottled water. Nat Sustain. 2018;1: 327–329.
    • 81. US EPA. Proposed Lead and Copper Rule Improvements. 2023. Available: https://www.epa.gov/ground-water-and-drinking-water/proposed-lead-and-copper-rule-improvements
    • 82. Water Denver. Using Filters. Lead Reduction Program. 2023. Available: https://www.denverwater.org/your-water/water-quality/lead/filter-program
    • 83. Villanueva CM, Cantor KP, Grimalt JO, Malats N, Silverman D, Tardon A, et al. Bladder Cancer and Exposure to Water Disinfection By-Products through Ingestion, Bathing, Showering, and Swimming in Pools. Am J Epidemiol. 2007;165: 148–156. pmid:17079692
    • 84. Sexton K, Hattis D. Assessing Cumulative Health Risks from Exposure to Environmental Mixtures—Three Fundamental Questions. Environ Health Perspect. 2007;115: 825–832. pmid:17520074
    • 85. Bach C, Dauchy X, Chagnon M-C, Etienne S. Chemical compounds and toxicological assessments of drinking water stored in polyethylene terephthalate (PET) bottles: A source of controversy reviewed. Water Res. 2012;46: 571–583. pmid:22196043
    • 86. Richardson SD, Plewa MJ. To regulate or not to regulate? What to do with more toxic disinfection by-products? J Environ Chem Eng. 2020;8: 103939.