Goals and Experimental Design:The goal of this study is to develop a general empirical model to predict responses following exposures to mixtures of compounds. The strategy for the model, described below, was originally developed to model mixtures of dioxin-like compounds. A critical need in developing the model was that it be independent of biologically based assumptions. For example, in calculating a toxic equivalency factor using the ratio of ED50s, it is assumed that one compound behaves biologically as a dilution of the other, i.e., that the compounds act similarly, but with different potencies.

In developing the basic model for the dioxin-like compounds, three properties became apparent. First, because of the empirical, rather than biological, basis of the model, it is generally applicable to any mixture of compounds. Second, the model provides information on the nature of the differences in mechanisms of action for different compounds. Potentially this information can be used in designing experiments to study the effects of mixtures of compound and in improving biologically based dosimetry models. And third, because the model consists of empirical curve fits, estimates of inter-subject variability may be made. These would most appropriately be done using Monte Carlo techniques.

The work to be performed in this study will expand the basic model to include different functional forms of the curves. For example, the current functional form used in fitting the calibration curve is appropriate for saturable processes but not for a biological process that could increase continually. This will address and expand upon the first property of the model. By increasing the family of curves that can be used in the model, different kinds of biological endpoints can be incorporated easily. The other focus of the work to be performed in this study will be a detailed examination of the nature of the mechanism differences that the model can provide. This addresses the second property of the model. Although not part of the current project, future plans include addressing the incorporation of Monte Carlo simulation techniques to provide robust estimates of inter-subject variability. This study serves as a pilot project to develop and explore statistical approaches to model responses observed in future Center research.

Schedule/Status: This study was completed during the year, with the exception that a manuscript for submission to a journal is still under development. The study was nearing completion at the time of the 1999 Annual Report, and no significant new findings or conclusions were reached since the report. In summary, the study compared in statistical terms, the relative values of investing in the number of treatment groups vs. the number of subjects per group. The conclusion was that, for programs like NERC, the number of different treatment groups (dose levels) should be as large as possible within the constraints of cost and generating reasonable statistical power by group size. This information assisted the decision to use four dose levels plus a negative control in the NERC core studies.