a2u-Globulin Nephropathy and Ravens: Do Ravens of a Different Feather Flock Together?

differentiate significant and trivial chemical hazards and to assess likely risks to humans. This latter justification can be illustrated by the following statement: if all chemicals that are mutagenic in vitro were mutagenic and carcinogenic to humans, and if all rodent carcinogens posed an equal carcinogenic hazard to humans, then there would be no role for mechanistic studies. The fact that these statements are self-evidently untrue therefore provides the main justification for the conduct of mechanistic studies. The value of the results of mechanistic studies aimed at differentiating hazards should be judged solely in terms of the extent to which they either increase or decrease the level of concern over the potential hazard posed to humans by exposure to the substance under study. Within this context, simple experiments, such as confirming that an in vitro mutagen is also mutagenic in vivo or establishing the effect that changing the route of exposure, or the test species, has on the carcinogenicity of a substance, qualify as mechanistic studies. However, the term is usually reserved for investigations that provide a framework within which the rodent toxicity of a substance can be explained and extrapolated to humans. Loose use of the term mechanistic studies can lead to confusion when the derived results offer no guidance on human hazard assessment. For example, it may be interesting to establish that nitrogen mustard is clastogenic by virtue of its ability to bind covalently with DNA, while the similar activity of etoposide is due to its ability to inhibit topoisomerase II enzymes, but such knowledge is of no value for human hazard assessment in the absence of data indicating that one of these mechanisms of clastogenesis is more relevant to human hazard assessment than the other. This single example indicates that when mechanistic data are invoked to qualify the hazard of a rodent carcinogen, there is a need to discuss the context of those data and to make clear how they should alter our perception of the human hazard implied by the initial test data. To date, there are no examples of a mechanism of cancer induction in rodents that is rigorously established and uniformly accepted as being of no relevance to humans. Further, given the potential for occasional high accidental exposures to chemicals and the existence of human polymorphisms among the enzymes that activate or detoxify chemicals, and among the enzymes and processes that maintain genetic integrity, such an example is unlikely ever to be documented. Nonetheless, there are several proposed mechanisms of cancer induction (rodent tumor etiologies) and tumor types (e.g., B6C3F1 mouse liver tumors) that carry a high level of certainty regarding their probable irrelevance to humans. The central issue therefore becomes whether a given level of certainty is acceptable or not. Whichever conclusion is reached in such cases, a compromise will have been made in the process. So, for example, if a chemical induces specifically renal or thyroid gland tumors in a lifetime rat carcinogenicity bioassay and if those tumors are preceded by the induction of ax2u-globulin-mediated nephropathy or thyroid-stimulating hormone-mediated thyroid follicular cell proliferation, respectively, then a decision will have to be made regarding the relevance of those tumor etiologies to humans. The differences evident in the viewpoints of Huff (1) and Dietrich (3) are due to both different starting premises regarding the value of attempting to separate mechanisms of chemical carcinogenesis and differences in their use and interpretation of the available mechanistic data for the cx2u-globulin-mediated nephropathy mechanism. Those differences are of interest at two levels. First, they lead to identification of the key data by which to assess whether or not humans will be subject to this particular rodent tumor etiology. Uncertainties over the sufficiency and the significance of those data remain a valid subject for future research. Second, the different viewpoints expressed confirm that we have some way to go before we can all share a common motivation for undertaking mechanistic studies. Put simply, we will make no useful progress until we can share the motivation of attempting to improve human hazard assessments within the limited confines of the resources available for the task, mindful that compromises will have to be made if we are to distinguish, for example, the human hazards posed by exposure to the rodent kidney carcinogens limonene and diethyl nitrosamine. That motivation will lead only to rational and balanced discussion of the relevant peerreviewed data. As summarized in the old Chinese adage, "if you want to find the truth about an issue, don't take either side." Attempts to guard the scientific high ground and to search for certainty, ostensibly on behalf of less well-informed colleagues, or attempts to dismiss rodent carcinogenicity data by reference to insecure and preliminary mechanistic constructs must be equally resisted. The success or failure of the margin of exposure (MOE) approach to nongenotoxic carcinogen hazard assessment proposed by the U.S. EPA will depend largely on the extent to which such common ground can be found and secured in the future. John Ashby Zeneca Central Toxicology Laboratory Alderley Park, Cheshire, United Kingdom