ABSTRACT
It is a well recognised property of chemicals that they are toxic and that the division of toxin from non-toxin depends upon the dose. The manner in which this toxicity is expressed is also dose-dependent, so that, for example, the division of carcinogens from non-carcinogens is dependent upon their carcinogenic potency and whether the expression of some lethal, non-neoplastic property of the substance occurs at a dose level that is higher or lower than that required to induce the expression of neoplastic processes. This idea is wellaccepted in experimental reproductive toxicology, where an agent may induce malformations, but is not described as a teratogen unless this activity is expressed at a dose at which embryotoxicity or foetotoxicity are not the dominant effects. It is also readily accepted in mutagenesis that dead cells do not mutate. The definition of a carcinogen also depends upon the statistical power of the study to detect differences between an exposed and a nonexposed group, after confounding and other sources of bias have been eliminated or minimized. Such control is much more readily achieved in experimental than in epidemiological studies, although the latter possess the property of greater relevance.
