Abstract
The development of scientifically sound risk assessments based on mechanistic data will enable society to better allocate scarce resources. Inadequate risk assessments may result in potentially dangerous levels of hazardous chemicals, whereas overly conservative estimates can result in unnecessary loss of products or industries and waste limited resources. Risk models are used to extrapolate from high-dose rodent studies to estimate potential effects in humans at low environmental exposures and determine a virtually safe dose (VSD). When information to the contrary is not available, the linearized multistage (LMS) model, a conservative model that assumes some risk of cancer at any dose, is traditionally employed. In the case of airborne chloroform, the dose at which an increased lifetime cancer risk of 10-6 could be calculated was chosen as the target VSD. Applying the LMS model to the mouse liver tumor data from a corn-oil gavage bioassay yields a VSD of 0.000008 ppm chloroform in the air. The weight of evidence indicates that chloroform is not directly mutagenic but, rather, acts through a nongenotoxic-cytotoxic mode of action. In this case, tumor formation results from events secondary to induced cytolethality and regenerative cell proliferation. Toxicity is not observed in rodents when chloroform is not converted to toxic metabolites at a rate sufficient to kill cells. Thus, tumors would not be anticipated at doses that do not induce cytolethality, contrary to the predictions of the LMS model. Inhalation studies in rodents show no cytolethality or regenerative cell proliferation in mouse liver at a chloroform concentration of 10 ppm as the no observed effect level (NOEL) or below. Using that NOEL and a safety factor approach, one can develop a VSD of 0.01 ppm. Integrating these data into the risk assessment process will yield risk estimates that are appropriate to the route of administration and consistent with the mode of action.