Assessing Exposure and Health Consequences of Chemicals in Drinking Water: Current State of Knowledge and Research Needs

The safety of water supplies is of paramount public health importance. An estimated 13% of the world population lacked access to improved drinking-water sources in 2008 [UNICEF and World Health Organization (WHO) 2011], and almost 10% of the total burden of disease worldwide could be prevented by improving drinking-water supply, sanitation, hygiene, and the management of water resources (Prüss-Üstün et al. 2008). Microbiological contamination is the largest cause of waterborne disease at a global scale. However, chemicals in water supplies can be related to health risks, generally when associated with long-term exposures (Thompson et al. 2007).

There are uncertainties about the safety of current standards for some regulated chemicals, and the potential health impacts of unregulated or emerging chemical contaminants are largely unknown. In May 2012, a workshop was held at the Centre for Research of Environmental Epidemiology (CREAL), Barcelona, Spain, with the aim of advancing the field of epidemiology and chemical contaminants in water and to make recommendations for future research. Our aspiration was that the proposed suggestions be useful and applicable to any type of chemical contaminant occurring in drinking water. Chemicals that we discuss as examples in this review are substances whose main pathway of human exposure is through drinking water. Although the chemical universe is broad and most chemicals do not occur exclusively in drinking water, water is essential for life, and exposures to chemicals in drinking water, even at low concentrations, may have important consequences across the entire population. Here we focus on cancer as an example and summarize the main discussion points and conclusions of the workshop.

Regulated chemicals. Drinking-water quality is regulated in most countries, and monitoring is conducted routinely. A complete list of chemicals that are currently regulated in drinking water, and the regulatory limits promulgated for each chemical by the WHO (2011), the U.S. Environmental Protection Agency (EPA 2009), and the European Union (EU) Council (1998) are provided in Table 1. These regulatory guidelines require periodic review to be updated according to new evidence. For example, the U.S. EPA reduced its maximum contaminant level (MCL) for arsenic from 50 μg/L in 1942 to the current level of 10 μg/L in response to growing scientific evidence of its adverse health effects (Smith et al. 2002). Epidemiological studies have reported associations of trihalomethane (THM) levels in drinking water (a surrogate measure of the disinfection by-product mixture) and bladder cancer (Villanueva et al. 2004) at THM levels lower than the current regulations in the United States and the European Union (80 and 100 μg/L, respectively; Table 1). The current MCL for nitrate was set based on methemoglobinemia among infants, but there is uncertainty concerning the safety of this MCL for chronic effects over longer exposure periods (e.g., on cancer) (Ward et al. 2005). Manganese is a neurotoxin associated with learning disabilities and deficits in intellectual function in children (Zoni and Lucchini 2013). The WHO manganese guideline has been fluctuating from the initial 500 μg/L in 1958 (Ljung and Vahter 2007) to its discontinuation in the current (fourth) edition of the WHO guidelines (WHO 2011). This has generated controversy in the scientific community because the last guideline before discontinuation (400 μg/L) was questionable according to some authors (Ljung and Vahter 2007) and the discontinuation of the manganese guidelines has received criticisms (Frisbie et al. 2012). Although many contaminants are monitored and regulated, the adequacy of the MCL approach is open to debate, in part because these limits are often based on toxicological studies of animals because human studies are not available or are inconclusive.

Emerging chemical contaminants. Non-regulated chemicals are of particular concern and constitute a main focus of current research (Richardson and Ternes 2011). Wastewater from human activities may contaminate water supply sources with pharmaceuticals, nanoparticles, consumer products (such as sunscreens), and other contaminants (Table 2), and these chemicals have been identified in drinking water (Ternes 2007). For example, iodinated or nitrogenated disinfection by-products (DBPs) [which are unregulated DBPs that are more toxic than their chlorinated and carbonaceous DBP analogs (Plewa et al. 2008b)] may occur in water supplies at very low concentrations (Plewa et al. 2004, 2008a). Degradation by-products of pharmaceuticals, which may be more toxic than their parent compounds, also have been identified in drinking water (Shen and Andrews 2011). The contribution of drinking water as a source of exposure to perfluorinated chemicals may be as important as dietary intake (Ericson et al. 2008), and evidence suggests that continued human exposure to even relatively low concentrations of perfluorooctanoic acid (PFOA) in drinking water results in elevated body burdens that may increase the risk of health effects (Post et al. 2012). Although concentrations are generally low (usually in the range of nanograms per liter) and some individual chemicals may pose no appreciable risks to human health (Schriks et al. 2010), there are concerns about potential risks of exposures to mixtures (Silva et al. 2002). The removal efficiency by drinking-water treatment processes has been evaluated for some substances (WHO 2012) but is poorly known for many emerging pollutants.

Water supplies often include mixtures of chemical contaminants that vary in time and space. In addition, the epidemiological and toxicological evaluation of mixtures involves significant challenges, in many cases beyond the limits of current research methods. In vitro bioassays (or biosensors) developed through toxicological research are promising tools for measuring the global toxicity of chemical mixtures in water samples and may be coupled with more in-depth analysis of specific contaminants when a positive response is detected. For example, Jeong et al. (2012) evaluated in vitro mammalian cell toxicity for a range of DBPs in an attempt to identify specific DBPs responsible for genomic DNA damage. End points that can be measured by in vitro bioassays include mutagenicity (Ames test) (Richardson et al. 2010), genotoxicity (micronuclei, Comet assay) (Plewa et al. 2010), endocrine disruption (DR-CALUX bioassay) (Brand et al. 2013; Sato et al. 2010), and cytotoxicity (Plewa et al. 2010). Although the use of these markers is not without limitations (such as the need for complex and nonstandardized sample pretreatment methods in order to obtain concentrates before laboratory analysis and the uncertain validity for some of the assays, limited throughput development, elevated cost, low sensitivity, and results reflecting only short-term exposure evaluations). Further development of these techniques and their incorporation into epidemiological research may improve our understanding of the effects of mixtures. These efforts will require improved, interdisciplinary communication and collaboration including analytical chemists, toxicologists, and epidemiologists.

Accurate exposure assessment in human observational studies is essential to obtain valid results and constitutes a main methodological challenge, as summarized in Table 3. Difficulties in identifying and measuring contaminants in water supplies at very low concentrations and substances occurring in mixtures hamper the evaluation of human exposure, requiring new methods in health risk analysis (Schwarzenbach et al. 2006).

DBPs are an example of chemicals occurring in complex mixtures, and this has been addressed in part by using a few compounds as surrogates for the DBP mixture as a whole. For example, observational studies of human DBP exposures and health effects have focused on a small subset of the several hundred DBPs that may occur in public water supplies (Richardson et al. 2007), particularly the THMs and haloacetic acids (HAAs) (Hinckley et al. 2005; Hoffman et al. 2008; Righi et al. 2012). However, although these compounds are often used as a surrogate for other DBPs, the assumption that they correlate with other DBPs is not universally supported, and correlations can vary in time and space (Villanueva et al. 2012).

Methods of exposure assessment are influenced by the specific outcome under study. For instance, for end points with a long latency, such as cancer, long time periods of more than several decades need to be evaluated, whereas for reproductive outcomes, it is very important to accurately capture the temporal variation in exposure over a shorter period covering the relevant time windows before and during gestation.

Chemicals or metabolites have been measured in biological samples in epidemiological studies to estimate exposures [e.g., urinary or toenail arsenic measurements in cancer studies (Karagas et al. 2004)]. Urinary trichloroacetic acid is a promising biomarker of DBPs that requires methodological development before a generalized use in epidemiological studies (Savitz 2012). In addition, among the available biomarkers specific for drinking-water contaminants, many have short half-lives (e.g., urinary trichloroacetic acid) and are thus of limited value to associate with health outcomes that require long-term exposures (Savitz 2012). Consequently, exposure assessment in most instances relies on assessing personal behavior (ascertained through questionnaires) and measuring environmental levels of the chemicals (Hoffman et al. 2008; Levallois et al. 2012).

Inhalation and dermal contact may be relevant exposure routes for volatile or skin-permeable chemicals. In such cases, activities involving different water uses at home (e.g., showering, bathing), in recreation (e.g., swimming in pools), and through occupations involving water contact should be considered.

Alternative methods of exposure assessment may involve statistical modeling; for example, modeling based on known geographic distributions of contaminants (Toledano et al. 2005), hydrological modeling of underground plumes of contaminants (Gallagher et al. 2010), and/or the use of surrogate parameters such as land use (Aschebrook-Kilfoy et al. 2012). Several methods can be used in combination, tailored to the availability of data; for example, in a recent study on the long-term exposure to arsenic and cancer, Nuckols et al. (2011) combined arsenic data from their own measurements in water samples collected at homes of the participants, data from public water utilities, and historical data for aquifers.

Exposure estimates with minimal measurement error are necessary to produce valid effect estimates. Misclassification of exposure is of particular concern at the low exposure range because it tends, under most scenarios, to attenuate associations toward the null (Cantor and Lubin 2007; Waller et al. 2001) or to reduce the precision of associations (Wright and Bateson 2004). Strategies to minimize measurement error are necessary from study design to data analysis, and include, for example, the collection of repeated measures of individual water use over the relevant exposure period (Forssen et al. 2009) and assessing reliability of interviews to exclude unreliable questionnaires (Villanueva et al. 2009).

The following is an overview of epidemiological findings from individual-based studies of chemical contaminants in water and cancer. Table 4 displays a summary of the evidence of carcinogenicity as evaluated and concluded by the WHO International Agency for Research on Cancer (IARC).

There is sufficient evidence in humans that arsenic in drinking water causes cancers of the urinary bladder, lung, and skin (IARC 2004). Studies conducted in areas with lower levels of arsenic in drinking water (i.e., at or below the MCL) have reported inconsistent results, and cancer risks associated with exposure to low arsenic levels over decades remain uncertain.

Bladder cancer has been consistently associated with DBP exposure (Cantor 2010), and pooled analyses combining data from studies conducted in different countries have reported associations between bladder cancer and THM at levels below current MCLs (Costet et al. 2011; Villanueva et al. 2004). Some (Cragle et al. 1985; King et al. 2000; Wilkins and Comstock 1981), but not all (Doyle et al. 1997; Hildesheim et al. 1998; Koivusalo et al. 1997), studies of DBP exposure and colon cancer have reported positive associations. Similarly, positive associations for DBP exposure have been found for rectal cancer (Bove et al. 2007; Doyle et al. 1997; Hildesheim et al. 1998) not replicated in other studies (King et al. 2000; Koivusalo et al. 1997; Wilkins and Comstock 1981).

The epidemiological investigation for nitrate and cancer has been challenging. Drinking water may be a primary source of nitrate exposure when drinking-water concentrations are > 50 mg/L (IARC 2010). Below this threshold, diet is the main exposure route, involving complex mechanisms of action through endogenous formation of N-nitroso compounds (IARC 2010). Long-term exposure to nitrate in drinking water has been evaluated in relation to multiple cancer sites including the esophagus, stomach, bladder, and colon (IARC 2010). Although there is inadequate human evidence for carcinogenicity, there is sufficient evidence from experimental animals for the carcinogenicity of nitrite in combination with amines or amides, and ingested nitrate under conditions that result in endogenous nitrosation has been classified as probably carcinogenic to humans (IARC 2010).

Other contaminants have been less extensively investigated in relation to cancer risk. Fluoride is added to drinking water at low concentrations in some countries to prevent dental caries, and naturally occurs in water at higher levels in certain parts of the world such as the Rift Valley in Africa (Malde et al. 2011). The IARC (1987) evaluated fluoride carcinogenicity and concluded that human and animal evidence was inadequate (Table 3). Some epidemiological studies on osteosarcoma have been published after this evaluation (Bassin et al. 2006; Kim et al. 2011), but consistent associations have not been observed.

The liver is a target organ for microcystin-LR (IARC 2010), which are toxins produced by cyanobacteria as a result of algae blooms and the eutrophication of surface waters. Individual-based studies evaluated by IARC (2010) have assessed exposure by comparing water consumed from ponds or ditches versus other sources and no measurements of toxins or bacteria were considered. In consequence, IARC concluded that evidence in humans for the carcinogenicity of microcystin-LR is inadequate (IARC 2010). Other carcinogens such as heavy metals, pesticides, and solvents may occur in drinking water as a consequence of human activities and natural hydrogeochemical processes. However, evidence on the cancer risk on human populations is limited.

The elucidation of mechanisms of action to provide biological plausibility and support causality suggested by epidemiological associations is a priority in current research. Biomarkers of early effect can be used in epidemiological studies to provide evidence about subclinical or intermediate effects of exposures (e.g., cytogenetic changes), and effects of very low exposure levels, and they can be used in experimental studies to evaluate the effect of an intervention. For an intermediate biomarker to be informative, it should be associated with both the disease and exposure of interest and reflect an intermediate step in the pathway between exposure and disease. For example, a suggested mechanism of action for arsenic is through epigenetic dysregulation, although there are limited human studies available (Ren et al. 2011). In addition, the evaluation of genetic variants may be used to identify susceptible populations underlying the biological mechanisms of action. For example, the evaluation of genetic variants of DBP-metabolizing enzymes in an epidemiological study on bladder cancer and THM exposure has shown that polymorphisms in key metabolizing enzymes modified DBP-associated bladder cancer risk (Cantor et al. 2010). In addition, the consistency of these findings with experimental observations of GSTT1 (glutathione S-transferase theta 1), GSTZ1 (glutathione S-transferase zeta 1), and CYP2E1 (cytochrome P450, family 2, subfamily E, polypeptide 1) enzymatic activity strengthens the hypothesis that DBPs cause bladder cancer and suggests possible mechanisms, as well as the classes of compounds likely to be implicated (Cantor et al. 2010).

There are few validated biomarkers specific for chemical contaminants in drinking water. However, the availability of prospective studies with biobanked samples and biotechnological development allowing large numbers of compounds to be measured in small amounts of biological samples (e.g., urine, plasma, serum) is encouraging. These technologies include genomics, epigenomics, transcriptomics, adductomics, proteomics, and metabolomics (Rappaport and Smith 2010; Wild 2005). Application of these techniques will facilitate a comprehensive approach to identify perturbations in biological systems and associated mechanisms of action (Moore et al. 2013). These technologies have not been widely applied in water research but have shown promising results in other areas of environmental research.

A significant and growing body of evidence suggests that climate change will have a detrimental effect on the quality of water available for human consumption in the future. For example, increasing temperatures may enhance conditions for the proliferation of cyanobacteria and algae (Joehnk et al. 2008; Newcombe et al. 2012; Paerl and Huisman 2008). Cyanobacteria are of particular concern for human populations because they can produce cyanotoxins such as microcystin that have carcinogenic effects (IARC 2010). The frequency of extreme weather events is expected to increase as a consequence of climate change, and the concentrations of chemical contaminants may be affected by extreme precipitation events. For example, tests conducted in models of different types of soils showed that certain mobile pharmaceuticals occur at higher concentrations in soil and groundwater during and directly after intense precipitation events (Oppel et al. 2004). Simulation studies have shown that pesticide concentrations fluctuate with changes in precipitation intensity and seasonality (Bloomfield et al. 2006; Probst et al. 2005). Evidence concerning the effect of drought is mixed. For example, concentrations of heavy metals (e.g., chrome, mercury, lead, cadmium) introduced primarily from anthropogenic activities in the Rhine River basin are higher during drought years (Zwolsman and van Bokhoven 2007). In contrast, no significant changes during drought conditions, but significant variability between seasons, has been described in the Dommel River, a tributary of the Meuse river in the Netherlands where increased groundwater flow in winter led to increased metal concentrations (Wilbers et al. 2009). In summary, it is expected that climate change could adversely affect drinking-water quality, but there is limited knowledge concerning the magnitude and distribution of the impact at different scales (global, regional, local).

General aspects. Although microbiological contamination is the largest contribution to waterborne disease and mortality at a global scale, chemical contaminants in water supplies also can cause disease, sometimes after long periods of exposure. The concentrations in drinking water, the prevalence of human exposure in the population, and the level of toxicity can be used to prioritize chemicals for further research. These characteristics may vary geographically and, therefore, further research should be designed to local-, region-, or country-specific circumstances as appropriate. Finally, exposures and risks affecting vulnerable populations (e.g., children and pregnant women) require special attention and are of particular interest.

Arsenic is a unique example of a substance in drinking water with conclusive evidence from human epidemiological studies. There is no doubt that arsenic is a human carcinogen at high concentrations (IARC 2004); however, there is inadequate information to determine the carcinogenic potential of other chemicals that occur in drinking water (Table 4). Arsenic has several unique characteristics—including the fact that drinking water represents the predominant source of exposure in humans; the levels in water, and thus the magnitude of the exposure, is very high in certain areas (e.g., Bangladesh); the availability of measurements in drinking water has allowed the development of epidemiological studies; the wide variability in exposures facilitates the detection of risks; the occurrence as an isolated substance rather than in mixtures allows the direct measurement of the putative agent; the magnitude of the risks are high compared with other chemicals; and the existence of biomarkers—all of which have helped to improve exposure assessment and elucidation of mechanisms of action of arsenic.

Recommendations on occurrence and exposure assessment. Improved exposure assessment to water contaminants is essential to derive valid exposure–response curves and useful knowledge for risk assessment and regulation, and here we provide some suggestions.

Recommendations on epidemiological methods. Epidemiological studies based on rigorous study design are essential to properly evaluate the human health risks associated with chemical contaminants in drinking water. Here we summarize some suggestions in this direction.

General conclusions. Assessing the health impacts of chemical contaminants in drinking water is a challenge that requires improved methodologies and enhanced interdisciplinarity in future epidemiological studies. Useful and valuable knowledge will increase if future studies successfully integrate existing and new developments from analytical chemistry, toxicology, exposure science, molecular epidemiology, statistics, environmental epidemiology, environmental sciences, engineering, and geospatial sciences. Improved cooperation and collaboration with stakeholders such as the water industry, regulatory, and public health agencies and affected communities would serve to produce higher-quality risk analyses, as well as to improve the likelihood of implementing effective and early intervention measures. Institutional support promoting access to reliable routine monitoring data at all levels and collaboration with stakeholders (e.g., water utilities, regulators, and consumer groups) would be beneficial. Finally, research efforts in this area are frequently hampered by the lack of specific funding for this research field, and the availability of stable and substantial financial support is needed, either from governmental or nongovernmental sources.