Reducing uncertainty in risk assessment by using specific knowledge to replace default options.

Abstract

This paper has advocated the development of specific scientific information, especially information on the mechanisms of action of chemicals, to use in place of default options in assessing human cancer risks. Four examples have been discussed that build largely on information from the CIIT research program. These four examples are worthy of consideration as a group, with a view to developing insights for increasing the effectiveness and efficiency of obtaining such data in the future and, most of all, to increase their acceptance for use instead of default options. In my view, key features of all four examples are that the data are framed within an exposure-dose-response paradigm and that there is a clear linkage to the end point of concern-cancer. As the number of techniques available for making observations at the cellular and molecular levels continues to increase at a rapid pace, linking these observations to the health end points of concern such as cancer is going to be increasingly important, especially in enhancing the value of the observations for risk assessment purposes. Equally as important, the mechanistic observations must be linked to realistic exposures and associated tissue dose that can be related to realistic human exposure scenarios. In my opinion, the likelihood of obtaining information of value for risk assessment purposes using the most sophisticated of molecular and cellular techniques will be of limited value if the exposures or doses are not realistically linked to those likely to be encountered by humans. The mechanism of alpha 2u-globulin nephropathy and its association with kidney tumors in male rats and the conclusion that the male rat kidney tumor findings are not applicable to assessing human cancer risk is an example of a qualitative decision. I suspect this may be a somewhat unusual case. As one looks across the various mammalian species used for experimentation and makes comparisons with humans, a unifying theme is the relative abundance of similarities. Indeed, this is a major argument for the use of laboratory animals to obtain information relevant to humans. Nonetheless, vigilance to differences among species is important. When differences are observed, we must capitalize on them to better understand the underlying biological mechanisms that mediate the differences. If, as I have suggested, laboratory animal species are more like than different from humans in their basic biological characteristics, there is a rationale for continuing to use laboratory animals as sources of data to help assess human risks of exposure to chemicals. It follows from this that quantitative differences among species such as observed with both formaldehyde and 1,3-butadiene assume major importance for assessing human risks. In my opinion, quantitation of the likely human carcinogenic potency of chemicals is of major importance. It is not sufficient to simply classify chemicals with regard to the likelihood of their being human carcinogens, as done by IARC (1994) and U.S. EPA (1986). IARC has placed more than 60 chemicals or processes (such as coke production) in group 1, carcinogenic to humans; more than 50 in group 2a, probably carcinogenic to humans; and 250 in group 2b, possibly carcinogenic to humans. This rank order implies differing levels of concern for three categories. However, even this rough three-bin system does not convey a very clear picture as to the degree of concern that should be accorded a given chemical for producing cancer. For example, the chemicals categorized as group 1, human carcinogens, using potency estimates developed by the U.S. EPA differ in potency by roughly 4 orders of magnitude. For example, a lifetime cancer risk is 6.2 x 10(-2) per micrograms/m3 for bischloromethyl ether and 8.3 x 10(-6) for benzene (NRC, 1994). Differences such as this offer strong arguments for complementing simplistic hazard identification schemes such as the IARC and EPA carcinogen classification systems w