Symposium VI: New Statistical Approaches to the Qualitative Interpretation of Toxicology Data

Abstract

This symposium describes some new statistical methods of use in the analysis of toxicology data for purposes of hazard evaluation, particularly as related to the qualitative phase of carcinogen risk assessment. C.J. Portier discusses life table analysis of cancer bioassays. Current assumptions about the lethality of treatment and tumor can lead to incorrect interpretation of results. D.W. North provides an introduction to the concepts of Bayesian statistics and shows how the application of judgmental probabilities can assist in reaching broad conclusions. L.B. Lave and G.S. Omenn apply a Bayesian method to estimate the likelihood of bioassay outcomes given short-term test data. The best screening test is not necessarily the most accurate or the least expensive. W.P. Darby and J.K. Gohagan use a bayesian method to integrate data from different bioassays of the same substance. In a case study of saccharin, they show that the totality of results can lead to conclusions that differ from those achieved with a single bioassay.

References

1. Hitchcock, M. , Walter, S.D. , Freeman, D.H. , D'Atri, C. , and Charpertier, P. (1979). Evaluation of the NIEHS Environmental Toxicology Training Program (Contract N01-ES-9005). Research Triangle Park, NC: National Institute of Environmental Health Sciences, p. 169.

2. Environmental Protection Agency. (1986). Guidelines for carcinogen risk assessment. Federal Register 51, 33992–34003.

3. Raiffa, H. (1968). An overview of statistics. In: Decision Analysis. H. Raiffa (ed.). Reading, MA: Addison-Wesley, Chap. 10.2, p. 278.

4. Ralston, A. (1986). Discrete mathematics: The new mathematics of science. Am. Scientist 74: 611–618.