Statistical Issues in Toxicology

Abstract

Toxicology is study of the nature and mechanism of toxic effects of substances on living organisms and other biologic systems (Lu 1996). Sometimes, data from human populations serve as the sentinel event indicating adverse health effects associated with environmental exposures. For example, the serious developmental effects associated with prenatal methyl mercury exposure were discovered only after some rural Japanese women ate fish contaminated by effluent from a nearby factory. Usually, however, the complexity and inherent variability of human populations complicates the evaluation of adverse environmental effects. For this reason, the field of toxicology has traditionally relied heavily on controlled studies in laboratory animals. Conducting studies in animals also allows researchers to explore questions that are difficult or unethical to address in a human population. For example, the pharmaceutical industry assesses safety in controlled animal experiments before products are used in humans. This article reviews just a few of the interesting statistical problems in toxicology. After presenting a short history, I turn to some problems of current interest, followed by some emerging problems in the field. Many important topics (e.g., the use of toxicological information to inform risk assessment decisions and policy making) are touched on only briefly or not at all.