body weight of 70 kg and divided by an assumed daily water consumption of 2 L to yield a drinking water equivalent level. That level is multiplied by a percentage of the total daily exposure contributed by drinking water (usually 20%) to calculate the MCLG. EPA then uses the MCLG to set an enforceable standard (the MCL). The MCL is set as close to the MCLG as feasible.

EPA sets MCLGs of zero for contaminants that are known or probable human carcinogens. For chemicals judged to be possibly carcinogenic to humans, EPA has recently begun applying an uncertainty factor between 1 and 10 to the reference dose derived from noncancer health effects to determine some exposure standards, such as certain ambient water-quality criteria (EPA 2000d). EPA stipulates that the water concentrations estimated to result in 1 × 10⁻⁶ to 1 × 10⁻⁵ excess cancer risks should also be compared with the reference dose.

Since the fluoride MCLG and SMCL were originally issued, there have been a number of developments in risk assessment. A few of those issues were described above in the discussion of current risk assessment practices (e.g., use of benchmark dose). Below, a few specific issues relevant to the committee’s review of the drinking water standards for fluoride are discussed, including advances in carcinogenicity assessment, relative source contribution, special considerations for children, and explicit treatment of uncertainty and variability.

In 2005, EPA issued its new Guidelines for Carcinogen Risk Assessment (EPA 2005a) as a replacement for its 1986 guidelines (EPA 1986). The revised guidelines were issued partly to address changes in the understanding of the variety of ways in which carcinogens can operate. For example, the guidelines provide a framework that allows all relevant biological information to be incorporated and the flexibility to consider future scientific advances.

The guidelines provide several options for constructing the dose-response relationship, in contrast to the single default dose-response relationship of the 1986 cancer guidelines. Biologically based extrapolation is the preferred approach for quantifying risk. It involves extrapolating from animals to humans based on a similar underlying mode of action. However,