Evaluation of models for predicting terrestrial food chain behavior of xenobiotics.

The authors evaluated the adequacy and capabilities of proposed univariate regression models for predicting the bioaccumulation of xenobiotic compounds in terrestrial vertebrates for physicochemical properties (structure-activity relationships) or from bioaccumulation by fish. Their findings raised doubts about the generic application of such statistical models, even for purposes for screening large numbers of compounds for potentially high bioaccumulation. They developed a screening level model based on lipophilic tendency of xenobiotic compounds which identified all compounds, but one, with a high bioaccumulation potential in fish, mammals, and birds, from large numbers of compounds. The one compound with a high bioaccumulation potential, which was not identified as such, represents an important class of toxic organic materials not accounted for in existing structure-activity relationships. Furthermore, their model incorrectly placed some compounds into the high bioaccumulation category when existing experimental data demonstrate low actual bioaccumulation. Problems such as those identified suggest that structure-activity relationships are not an adequate substitute for laboratory feeding trials in predicting the bioaccumulation potential of a xenobiotic compound. 15 references, 5 tables.