Arsenic exposure in US public and domestic drinking water supplies: a comparative risk assessment.

Although approximately 35 million people in the US obtain drinking water from domestic wells, few studies have investigated the risk of arsenic exposure from this source. In this paper arsenic concentrations were modeled for public and domestic wells using a dataset from the US Geological Survey (USGS). Excess lifetime and annual risks for lung and bladder cancer were calculated based on the carcinogenic potency and average arsenic concentrations in public and domestic water supplies. Monte Carlo uncertainty analysis was used to estimate the degree of confidence in these estimations. Results indicated that domestic well users accounted for 12% of the US population, but 23% of overall arsenic exposure from drinking water. Assuming that the new and more restrictive arsenic maximum contaminant limit (MCL) is implemented for public water supplies, it is anticipated that the proportion of people experiencing excess annual fatalities from drinking water from domestic wells will increase to 29% unless corresponding efforts are made to reduce exposures among domestic well users. Differences between public and domestic wells were not consistent across the nation. Public wells tend to tap deeper aquifers than domestic wells, and as a result local arsenic-depth trends can contribute to differences between public and domestic wells. Domestic wells and public wells in the western US have the highest arsenic levels with excess fatality risks estimated to be in the range of 1 per 9300 to 1 per 6600 in these regions. Uncertainty distributions of excess fatalities were developed and resultant uncertainties were propagated in arsenic exposure and potency factor. Uncertainty in the carcinogenic potency of arsenic was the dominant source of uncertainty in most regions, but for domestic wells in the New England and Southeast regions uncertainty in arsenic exposure was dominant, indicating that additional data on arsenic concentrations in these areas would substantially improve regional risk estimates.