``No human studies of sufficient statistical power or scope have examined whether consumption of arsenic in drinking water at the current MCL results in an increased incidence of cancer or noncancer effects (NRC, 1999, pg. 7).''

There have only been a few studies of inorganic arsenic exposure via drinking water in the U.S., and most have not considered cancer as an endpoint. People have written EPA asking that the new MCL be set considering that these U.S. studies have not seen increases in cancers at the low levels of arsenic exposure in U.S. drinking water.

A large number of adverse noncarcinogenic effects have been reported in humans after exposure to drinking water highly contaminated with inorganic arsenic. The earliest and most prominent changes are in the skin, e.g., hyper pigmentation and keratoses (callus-like growths). Other effects that have been reported include alterations in gastrointestinal, cardiovascular, hematological (e.g., anemia), pulmonary, neurological, immunological and reproductive/developmental function (ATSDR, 1998).

The most common symptoms of inorganic arsenic exposure appear on the skin and occur after 5-15 years of exposure equivalent to 700 µg/day for a 70 kg adult, or within 6 months to 3 years at exposures equivalent to 2,800 µg/day for a 70 kg adult (pg. 131 NRC, 1999). They include alterations in pigmentation and the development of keratoses which are localized primarily on the palms of the hands, the soles of the feet and the torso. The presence of hyper pigmentation and keratoses on parts of the body not exposed to the sun is characteristic of arsenic exposure (Yeh, 1973, Tseng, 1977). The same alterations have been reported in patients treated with Fowler's solution (1% potassium arsenite; Cuzick et al., 1982), used for asthma, psoriasis, rheumatic fever, leukemia, fever, pain, and as a tonic (WHO 1981 and NRC 1999).

Although peripheral neuropathy (numbness, muscle weakness, tremors, ATSDR 1998) may be present after exposure to short-term, high doses of inorganic arsenic (Buchanan, 1962; Tay and Seah, 1975), there are no studies that definitely document this effect after exposure to levels of less than levels (50 µg/L) of inorganic arsenic in drinking water.

There have been a few, scattered reports in the literature that inorganic arsenic can affect reproduction and development in humans (Borzysonyi et al., 1992; Desi et al., 1992; Tabacova et al., 1994). After reviewing the available literature on arsenic and reproductive effects, the National Research Council panel (NRC 1999) wrote that ``nothing conclusive can be stated from these studies.''

Based on the studies mentioned in this section, it is evident that inorganic arsenic contamination of drinking water can cause dermal and internal cancers, affect the GI system, alter cardiovascular function, and increase risk of diabetes, based on studies of people exposed to drinking water well above the current arsenic MCL. EPA's MCL is chosen to be protective of the general population within an acceptable risk range, not at levels at which adverse health effects are routinely seen (see section III.F.7. on risk considerations).

In terms of implications for the risk assessment, the panel noted that risk per unit dose estimates from human studies can be biased either way. For the Taiwanese study, the ``* * * biases associated with the use of average doses and with the attribution of all increased risk to arsenic would both lead to an overestimation of risk (US EPA, 1997d, page 31).

May 1999 Utah Mortality Study

EPA scientists conducted an epidemiological study of 4,058 Mormons exposed to arsenic in drinking water in seven communities in Millard County, Utah (Lewis et al., 1999). The 151 samples from their public and private drinking water sources had arsenic concentrations ranging from 4 to 620 µg/L with seven mean (arithmetic average) community exposure concentrations of 18 to 191 µg/L and all seven community exposure medians (mid-point of arsenic values) 200 µg/L. Observed causes of death in the study group (numbering 2,203) were compared to those expected from the same causes based upon death rates for the general white male and female population of Utah. Several factors suggest that the study population may not be representative of the rest of the United States. The Mormon church, the predominant religion in Utah, prohibits smoking and consumption of alcohol and caffeine. Utah had the lowest statewide smoking rates in the U.S. from 1984 to 1996, ranging from 13 to 17%. Mormon men had about half the cancers related to smoking (mouth, larynx, lung, esophagus, and bladder cancers) as the U.S. male population from 1971 to 1985 (Lyon et al., 1994). The Utah study population was relatively small (4,000 persons) and primarily English, Scottish, and Scandinavian in ethnic background.

While the study population males had a significantly higher risk of prostate cancer mortality, females had no significant excess risk of cancer mortality at any site. Millard County subjects had higher mortality from kidney cancer, but this was not statistically significant. Both males and females in the study group had less risk of bladder, digestive system and lung cancer mortality than the general Utah population. The Mormon females had lower death rates from breast and female genital cancers than the State rate. These decreased death rates were not statistically significant.

Although deaths due to hypertensive heart disease were roughly twice as high as expected in both sexes, increases in death did not relate to increases in dose, calculated as the years of exposure times the median arsenic concentration. The Utah data indicate that heart disease should be considered in the evaluation of potential benefits of U.S. regulation. Vascular effects have also been reported as an effect of arsenic exposure in studies in the U.S. (Engel et al. 1994), Taiwan (Wu et al., 1989) and Chile (Borgono et al., 1977). The overall evidence indicating an association of various vascular diseases with arsenic exposure supports consideration of this endpoint in evaluation of potential noncancer health benefits of arsenic exposure reduction.

Study of Bladder and Kidney Cancer in Finland Kurttio et al. (1999) conducted a case-cohort design study of 61 bladder and 49 kidney cancer cases and 275 controls to evaluate the risk of these diseases with respect to arsenic drinking water concentrations. In this study the median exposure was 0.1 µg/L, the maximum reported was 64 µg/L, and 1% of the exposure was greater than 10 µg/L. The authors reported that very low concentrations of arsenic in drinking water were significantly associated with being a case of bladder cancer when exposure occurred 2-9 years prior to diagnosis. Arsenic exposure occurring greater than 10 years prior to diagnosis was not associated with bladder cancer risk. Arsenic was not associated with kidney cancer risk even after consideration of a latency period.

The NRC report examined the question of essentiality of arsenic in the human diet. It found no information on essentiality in humans and only data in experimental animals suggesting growth promotion (arsenicals are fed to livestock for this reason). Inorganic arsenic has not been found to be essential for human well-being or involved in any required biochemical pathway. Given this and the fact that arsenic occurs naturally in food, consideration of essentiality is not necessary for public health decisions about water.

The NRC report concluded: ``For arsenic carcinogenicity, the mode of action has not been established, but the several modes of action that are considered plausible (namely, indirect mechanisms of mutagenicity) would lead to a sublinear dose-response curve at some point below the point at which a significant increase in tumors is observed. * * * However, because a specific mode (or modes) of action has not yet been identified, it is prudent not to rule out the possibility of a linear response.''

Given the current outstanding questions about human risk at low levels of exposure, decisions about safe levels are public health policy judgments.

Risk Characterization In 1983 the National Academy of Sciences (NAS, 1983) defined risk assessment as containing four steps: hazard identification, dose- response assessment, exposure assessment, and risk characterization. Risk characterization is the process of estimating the health effects based on evaluating the available research, extrapolating to estimate health effects at exposure levels, and characterizing uncertainties. In risk management, regulatory agencies such as EPA evaluate alternatives and select the regulatory action. Risk management considers ``political, social, economic, and engineering information'' using value judgments to consider ``the acceptability of risk and the reasonableness of the costs of control (NAS, 1983).''

Unlike most chemicals, there is a large data base on the effects of arsenic on humans. Inorganic arsenic is a human poison, and oral or inhalation exposure to the chemical can induce many adverse health conditions in humans. Specifically oral exposure to inorganic arsenic in drinking water has been reported to cause many different human illnesses, including cancer and noncancer effects, as described in Section III. The NRC panel (1999) reviewed the inorganic arsenic health effects data base. The panel members concluded that the studies from Taiwan provided the current best available data for the risk assessment of inorganic arsenic-induced cancer. (There are corroborating studies from Argentina and Chile.) They obtained more detailed Taiwanese internal cancer data and modeled the data using the multistage Weibull model and a Poisson regression model. Three Poisson data analyses showed a 1% response level of male bladder cancer at approximately 400 µg of inorganic arsenic/L. The 1% level was used as a Point of Departure (POD) for extrapolating to exposure levels outside the range of observed data.

For an agent that is either acting by reacting directly with DNA or whose mode of action has not been sufficiently characterized, EPA's public health policy is to assume that dose and response will be proportionate as dose decreases (linearity of the extrapolated dose- response curve). This is a science policy approach to provide a public health conservative assessment of risk. The dose-response relationship is extrapolated by taking a straight line from the POD rather than by attempting to extend the model used for the observed range. This approach was adopted by the NRC report which additionally noted that using this approach for arsenic data provides results with alternative models that are consistent at doses below the observed range whereas extending the alternative models below the observed range gives inconsistent results. Drawing a straight line from the POD to zero gives a risk of 1 to 1.5 per 1,000 at the current MCL of 50 µg/ L. Since some studies show that lung cancer deaths may be 2- to 5-fold higher than bladder cancer deaths, the combined cancer risk could be even greater. The NRC panel also noted that the MCL of 50 µg/L is less than 10-fold lower than the 1% response level for male bladder cancer. Based on its review, the consensus opinion of the NRC panel was that the current MCL of 50 µg/L does not meet the EPA's goal of public-health protection. Their report recommended that EPA lower the MCL as soon as possible.

A factor that could modify the degree of individual response to inorganic arsenic is its metabolism. There is ample evidence (NRC, 1999) that the quantitative patterns of inorganic arsenic methylation vary considerably and that the extent of this variation is unknown. It is certainly possible that the metabolic patterns of people affect their response to inorganic arsenic.

There are studies underway in humans and experimental animals under the EPA research plan and other sponsorships. Over the next several years these will provide better understanding of the mode(s) of carcinogenic action of arsenic, metabolic processes that are important to its toxicity, human variability in metabolic processes, and the specific contributions of various food and other sources to arsenic exposure in the U.S. These are important issues in projecting risk from the observed data range in the epidemiologic studies to lower environmental exposures experienced from U.S. drinking water.

Until further research is completed, questions will remain regarding the dose-response relationship at low environmental levels. The several Taiwan studies have strengths in their long-term observation of exposed persons and coverage of very large populations (>40,000 persons). Additionally, the collection of pathology data was unusually thorough. Moreover, the populations were quite homogeneous in terms of lifestyle. Limitations in exposure information exist that are not unusual in such studies. In ecological epidemiology studies of this kind, the exposure of individuals is difficult to measure because their exposure from water and food is not known. This results in uncertainties in defining a dose-response relationship. The studies in Chile and Argentina are more limited in extent, (e.g., years of coverage, number of persons, or number of arsenic exposure categories analyzed), but provide important findings which corroborate one another and those of the Taiwan studies.

These epidemiological studies provide the basis for assessing potential risk from lower concentrations of inorganic arsenic in drinking water, without having to adjust for cross-species toxicity interpretation. Ordinarily, the characteristics of human carcinogens can be explored and experimentally defined in test animals. Dose-response can be measured, and animal studies may identify internal transport, metabolism, elimination, and subcellular events that explain the carcinogenic process. Arsenic presents unique problems for quantitative risk assessment because there is no test animal species in which to study its carcinogenicity. While such studies have been undertaken, it appears that test animals, unlike humans, do not respond to inorganic arsenic exposure by developing cancer. Their metabolism of inorganic arsenic is also quantitatively different than humans. Inorganic arsenic does not react directly with DNA. If it did, it would be expected to cause similar effects across species and to cause response in a proportionate relationship to dose. Moreover, its metabolism, internal disposition, and excretion are different and vary across animal and plant species and humans--in test studies and in nature.

Until more is known, EPA will take a traditional, public health conservative approach to considering the potential risks of drinking water containing inorganic arsenic. EPA recognizes that the traditional approach may overestimate risk, as explained in the next section.

Most of the 25-States data had reporting limits of less than 2 µg/L. In addition, the database includes multiple samples from the water systems over time and from multiple sources within the systems. The multiple samples provide for a more accurate estimate of the arsenic levels in the systems, than a survey with one sample per system. The arsenic compliance monitoring data provides point-of-entry or well data within systems from eight States, which is used for intrasystem variability analysis (discussed in Section V.G). Intrasystem variability analysis provides an understanding of the variation of arsenic levels within a system, from well to well or entry point to entry point.