EPA Model Will Assess Safety Of Less-Strict Arsenic Standard
RESEARCH TRIANGLE PARK, NC

reported by
Matt Shipman

EPA has recently begun developing a potentially precedent-setting arsenic risk assessment model that will help the agency decide whether it can implement a less-stringent arsenic drinking water standard without sacrificing human health protection, according to an agency presentation.

Rory Conolly, of EPA's National Center for Computational Toxicology (NCCT), outlined the agency's plans to develop a biologically based dose-response (BBDR) model to assess the carcinogenicity of inorganic arsenic at a meeting of the agency's Board of Scientific Counselors (BOSC) Computational Toxicology Subcommittee here Dec. 18. The goal, Conolly said, is for NCCT to have a functional BBDR model in place by 2012 -- so that EPA's Office of Water can use it when performing its next six-year review of the arsenic drinking water standard. Conolly said that NCCT is working with EPA's National Health & Environmental Effects Research Laboratory (NHEERL) to develop a research proposal for generating the data necessary to create the BBDR model. An EPA source says a draft of the proposal will likely be ready in January for internal review by agency staff, followed by external review involving an invited panel of experts. The EPA source says the research plan is important because the "modeling [EPA] would like to do depends on getting more data" than currently available. For example, the source says, developing a BBDR model would require a great deal of additional dose-response and "time-course" data. Most current studies examine the effects of one or two doses of a chemical at one point in time, the source explains. For a BBDR model, EPA will need studies looking at a wide array of doses and measuring their effects over time, in an attempt to mimic drinking water exposures to arsenic as closely as possible. The goal for the BBDR model would be to use it to determine the carcinogenic toxic effects of inorganic arsenic at very low levels, Conolly told the BOSC panel. A BBDR model would incorporate a physiologically based pharmacokinetic (PBPK) component to assess the relationship between the concentration of arsenic in drinking water and the dose received by the target organ, the EPA source says. PBPK models are regarded as a more sophisticated approach to understanding how chemicals -- once they enter the body -- are transported and transform as physiological processes break them down, potentially reducing their toxicity. But the model would also describe the series of biological and chemical events that link the dose received by that target organ's tissue to those physical changes that lead to cancer, the source explains.

The question, the EPA source says, is "how much of that tissue response data can we develop?" The source says it is "still early," but "I'm optimistic we can do a lot." The EPA source adds that "in my personal opinion" the BBDR model for arsenic, if successful, will provide a "prototype or blueprint for how this could be done for other chemicals." A second EPA official agrees, noting that "the insights gained from this effort will help inform the further development of PBPK and BBDR" models. The source says an additional benefit of the BBDR model is that it could be used to look at "time-limited" risks. For example, the source says, the BBDR model should allow EPA to examine the risks associated with childhood exposure to arsenic -- even if an individual moves away when they get older and are not exposed to arsenic in drinking water as an adult. In his presentation, Conolly explained that EPA relied on "linear extrapolation" to support its 2001 decision to lower the drinking water standard for arsenic from 50 parts per billion (ppb) to 10 ppb. But Conolly noted that linear extrapolation, which effectively draws a straight line from data on higher doses through zero to estimate the risks posed by a substance at lower doses, is a default assumption that is used in the absence of data on low-dose toxicity. Linear approaches are considered by critics to be overly "conservative," or tending to overstate risk. The BBDR model will hopefully allow EPA to minimize its uncertainty on the low-dose toxic effects of arsenic, and allow it to avoid using the linear default.

Conolly noted that the agency's decision to use linear extrapolation to set the arsenic standard resulted in a drinking water standard that is "probably health protective" but is "unlikely to be accurate." The EPA source says "it is possible [the 10 ppb standard] is more stringent than it needs to be," and that the BBDR model may show that it is possible to continue to protect human health with a less stringent standard that has cheaper compliance costs for drinking water treatment providers. Conolly told the BOSC panel that EPA is scheduled to complete its current review of the arsenic drinking water standard in 2009, and that it will then begin a second six-year review cycle that would end in 2015. It is too late for NCCT and NHEERL to develop a BBDR model that could influence the 2009 review, Conolly said.

And in order for the BBDR model to be used by water office personnel in the 2015 review, Conolly said, the model would have to be available for use by 2012. The first EPA source says the agency hopes to meet that deadline. However, it is unclear whether the model would be sufficient to warrant revisions to the arsenic standard, regardless of its results. The Safe Drinking Water Act requires EPA to review each of its drinking water standards at least once every six years and to revise them if the agency deems necessary, but also mandates that "any revision must maintain or increase public health protection." Given this statutory language, it may be difficult for EPA to overcome the findings of its Science Advisory Board, which said in a June 2007 report that exposure to inorganic arsenic poses cancer risks at any level above zero. The finding was viewed as supporting the 10 ppb standard, particularly because it backed EPA's use of a linear model when setting the standard. -