Document jg25gw0RpaNjLNw3MXqaR18gk

participation in a , AND FRAUMENI, sarcoma. J. Amer. lion and processing ie polymer produc- nmental epidemioon. Health 8, 250- lyrene-polystyrene ng styrene-exposed LE, J. R. (1980). A ne-based products. 9-1311. . J. Med. 304,600- ARD, J. K. (1984). v Zealand. J. Natl. Can^^mevention: l. BeoKnd M. J. iltural and forestry . (1987). Soft tissue nated phenol expo- REGULATORY TOXICOLOGY AND PHARMACOLOGY 9, 53-55 (1989) Copyrighted Material Do Not Reproduce Without Publisher's Permission Are Negative Toxicological Data Suspect: An Epilog1 Perry J. Gehring The Dow Chemical Company, P.O. Box J 706, Midland, Michigan 48640 Received May 2,1988 Curiously, the public and their leaders are requesting scientists to tell them what chemicals under what conditions are safe, while the scientists persist in equivocating whatever data orjudgments they provide with very, if not overly, conservative cave ats. A positive study is a priori deemed creditable; a negative is deemed suspect. While we complain that journals do not publish negative studies, the truth is none of us-- 37 R&S 142745 toxicologists, epidemiologists, statisticians, etc.--feel comfortable with negative studies. We are driven to find evidence for disease caused by treatment. This is as it should be. However, the science offinding no untoward effects or defining under what condi- tions none will occur is in the end more important. If these issues are not decided in M a scientific manner, they will be decided in a less desirable forum. * If one accepts the responsibility of our science to elucidate negative results, we * 01 must then accept the responsibility to communicate those results. Do not blame the journals for not publishing; journals publish what we want to read or hear. The speakers today have addressed some of the problems inherent to current test ing methods in animals and surveillance of human populations. As Dr. Clayson demonstrated, use of the MTD in animal studies renders interpre tation fuzzy at best. Using data from many ofthese studies requires extrapolation not only from animals to man but also from sick animals to man. There is good reason to believe that the latter is less creditable than the former. Shown in Table 1 are some risk assessments for a few compounds, using the linear ized multistage model. For perchloroethylene, trichloroethylene, acrylonitrile, and butadiene, exposure to {the threshold limit value (TLV) for 20 years is predicted to cause an 8 to 26% increase in cancer. Although definitive epidemiological studies are not available, it is highly unlikely that such increases have occurred without recogni tion. For perchloroethylene and trichloroethylene, at least 8-10 small studies have been published, which indicate no increase in cancer of the liver. While not one of ` Presented at the symposium "Significance of Negative Data in Evaluating Environmental Toxicologi cal Hazards," February 18, 1988, Society ofToxicology meeting, Dallas, TX. 53 0273-2300/89 $3.00 Copyright 1989 by Academic Press, Inc, All nghts of reproduction in any form reserved. I 54 PERRY 1. GEHRING TABLE 1 Risk Assessment Based on Extrapolation of Animal Data" Chemical Exposure Risk Perchloroethylene Trichloroethylene Acrylonitrile Butadiene Vinyl chloride Bischlormethylether Ethylene dibromide Aflatoxin 60 ppm for 20 years 60 ppm for 20 years 10 ppm for 20 years 500 ppm for 20 years 200 ppm for 20 years 0.01 ppm for 10 years 1 year 3 ppm for 20 years 10 years 4.2 years Average U.S consumption 0.23 0.08 0.13 0.26 0.16 1.0 0.45 0.65 0.41 0.20 789/100,000 " Risk calculated from cancer assessment group potency estimates published in methylene chloride health assessment document EPA/600/8-82/004F, February' 1985. R&S 142746 these studies is in itselfindicative, their consistency is an indication that the incidence of liver cancer is well below that predicted. Vinyl chloride has been predicted to produce a 0.16 risk of cancer. However, two large interindustry studies indicate no increase in cancer in vinyl chloride monomer or PVC polymer workers other than angiosarcoma ofthe liver (Fox and Collier, 1977; Tabershaw and Gaffey, 1974), The worldwide incidence of angiosarcoma, 108 cases from 1955 to 1983 (Forman et ai, unpublished data, 1985), is far less than that pre dicted by the risk models, considering that tens of thousands of workers have been exposed and that, historically, exposure to several hundred parts per million were common. Exposure to concentrations of bischlormethylether reportedly sufficient to cause marked respiratory irritation (1 ppm and greater) caused cases of respiratory tract cancer; however, the incidence was not 100% or even 45% as predicted from linear models. For ethylene dibromide, no increase in cancer was seen in 156 employees exposed to an average of 3 ppm for 4.2 years, while the predicted incidence is 20% (Ramsey etal., 1978). The probability of dying from liver cancer in the United States is 210/100,000, yet the risk predicted from the average exposure to aflatoxin alone is 789/100,000. The lack or inadequacy ofepidemiological evaluation weakens the foregoing super ficial analysis. However, it is not likely that the magnitude and consistency of overprediction will be discounted by more intense evaluation. More rigorous comparison ofpredicted risk with epidemiological results is needed. Such comparisons are needed to define the boundaries for the preconceptions inherent to risk assessment models. In addition to using risk assessment models to predict the incidence of cancer in man from chronic bioassay data, similar predictions of the incidence in the same or other species ofanimals exposed to lower doses in other experiments need to be made when feasible. For chloroform, such an assessment shows that the risk assessment . V* Risk 0.23 0.08 0.13 0.26 0.16 1.0 0.45 0.65 0.41 0.20 789/100.000 i methylene chloride that the incidence :er. imwever, two hlo^Bmonomer and^Ilier, 1977; trcoma, 108 cases less than that pre-orkers have been per million were efficient to cause f respiratory tract iicted from linear mployees exposed e is 20% (Ramsey s 210/100,000, yet 39/100,000. e foregoing supernsistency of overarous comparison arisons are needed essment models, .ence of cancer in ice in the same or s need to be made e risk assessment are negative toxicological data suspect? 55 models fail miserably (Reitz et ai, 1980). At the very least, it is reasonable to expect such predictions to be more accurate than predictions for man. Assessments such as the foregoing indicate toxicologists better get about assessing the credibility of their science. Many, including epidemiologists, criticize the lack of power in epidemiological studies. It is high time we critically assess and admit the lack of power in animal toxicology studies with respect to interpretations currently being made. As a start, extrapolation ofcarcinogenicity data using the linearized mul tistage model should be considered only for materials found positive in mutagenicity evaluations. Biological rationale for using this model for nongenetic carcinogens does not exist. REFERENCES Fox, A. J,, AND COLLIER, P. F. (1977). Mortality experience of workers in the manufacture of polyvinyl chloride in Great Britain. Brit. J. Ind. Med. 34,1-10. Ramsey, J. C., Park, C. N,, Ott, M. G,, and Gehring, P. J. (1978). Carcinogenic risk assessment: Ethylene dibromide. Toxicol. Appl. Pharmacol 47,411-414. Reitz, R. H,, Quast, J. F,, Stott, W. T, Watanabe, P. G., and Gehring, P. J. (1980). Pharmacokinet ics and macromolecular effects ofchloroform in rats and mice: Implications forcarcinogenic risk estima tion. In Water Chlorination: Environmental Impact and Health Effects (R. L. Jolley, W. A. Brunge, and R. B. Cummings, Eds.), Vol. 2, pp. 983-993. Ann Arbor Press, Ann Arbor, MI. Tabershaw, I. R., and Gaffey, W. R. (1974). Mortality study of workers in the manufacture of vinyl chloride and its polymers. J. Occup. Med. 16,509-518. Zdeb, M. S. (1977). The probability ofdeveloping cancer. Amer. J. Epidemiol 106,6,