The value of epidemiolgical studies

The ongoing Air Force Health Study, the U.S. Air Force investigation of health effects in Ranch Hand veterans exposed to Agent Orange and its contaminant, 2, 3, 7, 8-tetrachlorodibenzo-p-dioxin (TCDD), is presented as a model epidemiologic study of occupational exposure to a toxic chemical. Three points are discussed: 1) The interpretation of the many statistical associations that can arise in an epidemiologic study requires careful considerations of the multiple testing artifact and esthablished causal criteria. 2) Recently published work indicates that epidemiologic studies designed to demonstrate safety are, in practice, not feasible, and an observed relative risk of 1.0 in a study designed to detect hazard is not a valid basis for assurances for safety. 3) Work history indices of exposure effect are subject to error when the exposure is weak or the period of exposure is short; this error can lead to a strong bias towards finding no effect.