Trichloroethylene and cancer: epidemiologic evidence

adopt a traditional review of the mutagenicity data on TCE and its metabolites but instead raise several issues regarding the interpretation of mutagenicity and genetic toxicity tests in shedding light on whether these processes are key events in tumor initiation. As discussed in the U.S. EPA proposed cancer guidelines, a salient question is whether TCE or its metabolites interacts directly with and mutates DNA to bring about changes in gene expression or whether DNA mutation is achieved through some other process. The Moore and Harrington-Brock article examines this question. Bull ((italic)10(/italic)), Lash et al. ((italic)11(/italic)), and Green ((italic)12(/italic)) present the experimental support for several modes of action for tumor development in rodents. These articles discuss a number of hypotheses including the influence on tumor development from mutagenesis, cytotoxicity, cell proliferation, *(subscript)2u(/subscript)-globin, peroxisome proliferation, oxidative stress, receptor binding, and perturbation of cell-signaling pathways. Quantitative dose-response issues important to the statistical modeling of both noncarcinogenic and carcinogenic effects are discussed in articles by Fisher ((italic)5(/italic)), Bois ((italic)13(/italic)), Clewell et al. ((italic)14(/italic)), Bois ((italic)15(/italic)), Boyes et al. ((italic)16(/italic)), Barton and Clewell ((italic)17(/italic)), Chen ((italic)18(/italic)), and Rhomberg ((italic)19(/italic)). Because pharmacokinetic data are available for the TCE assessment, dose metrics other than applied dose may be evaluated in benchmark and other dose-response analyses. Fisher's article ((italic)5(/italic)) describes modeling liver concentration of TCE and its oxidative metabolites, while Clewell et al. ((italic)14(/italic)) model plasma concentrations of the oxidative metabolites and flux through the kidney for metabolites of the glutathione-(italic)S(/italic)-transferase pathway. Both models are scaled from mice or rats to humans and provide estimates of human equivalent doses simulating inhalation and oral exposure routes. Parameters from these models have been further subjected to an uncertainty analysis in the articles by Bois ((italic)13,15(/italic)). The application of Baysian statistical methods is increasingly used for updating estimates of pharmacokinetic model parameters. Moreover, these analyses can provide an additional set of dosimetric estimates that in some instances are very different from those obtained with the original model. These findings make the risk assessor's job more complex. Boyes and others ((italic)16(/italic)) test whether Haber's Law or a dose metric that integrates time and concentration best describes neurologic effects with high-level TCE exposure. Barton and Clewell ((italic)17(/italic)) examine both experimental and pharmacokinetically derived dose metrics in their analysis of neurologic and systemic organ toxicity seen in the rodent studies. The article further applies benchmark dose methodology to the quantitative analysis of these effects. The U.S. EPA proposed cancer guidelines recommend that dose-response modeling be carried out in two parts: analysis of the curve shape within the range of the data and extrapolation below the observable data. Application of a biologically based model is preferred for evaluating the dose-response relationship for carcinogenic effects. Such an approach is described in the article by Chen ((italic)18(/italic)), which explores the relationship between TCE and two of its oxidative metabolites, dichloroacetic acid and trichloroacetic acid, under the hypothesis that these chemicals induce liver tumors in mice through promotion of preexisting initiated cells. Unfortunately, the data ne