An empirical examination of factors influencing prediction of carcinogenic hazard across species.

This study was stimulated by a recent U.S. Environmental Protection Agency (EPA, 1994) statement in draft environmental carcinogen risk assessment guidelines: "Several kinds of observations from animal studies can contribute to the judgment whether animal responses indicate a significant carcinogenic hazard to humans." We have investigated each of these kinds of observation using the cancer bioassay data system database. We obtained concordances from rat to mouse (and vice versa) for various subgroups of chemicals as follows: chemicals that induced tumors at multiple sites, chemicals that induce cancer in both sexes, chemicals that display reduced latency, and chemicals increasing the rates of rare tumors. The concordances are much higher for these chemical subgroups than the chemical groups that induce tumor at a single site, in only one sex, or without reduced latency, respectively. Thus, our findings support some of the EPA's suggested factors.