Abstract
The classification variables related to exposure in epidemiological studies of occupational hazards are reviewed, and the criteria for choosing from among different indicators of exposure to obtain maximal information from the association(s) under study are outlined. A valid classification of exposure requires that, in the planning stage of the study, the population chosen is such that the contrast between exposure and the lack of exposure is clear, good quality data or other types of reliable information on exposure are available, there is no unresolvable confounding, and the population is large enough to allow a powerful study and analysis to be carried out. Also, the study hypotheses need to be accurately formulated in terms of specific agents, their characteristics, and the possible modifying or confounding factors. Model calculations to illustrate the relationships between several commonly used indicators of exposure are presented, and some problems of a classification strategy are briefly discussed. The importance of indicators predicting the occurrence of a disease accurately, its time-dependence and exposure-response relationship are emphasized. In studies of chronic diseases, a useful set of exposure variables often includes qualitative exposure (e.g., definitely/possibly/not exposed) to a well-specified agent, cumulative exposure, and time since the onset of exposure. Indicators adjusted for the main confounders and modifiers, as well as those addressing the mode or pattern of exposure, are likely to be predictive of the outcome if the effect of these factors is strong. Whenever proxies of qualitative exposure or target dose are used, one should note that they may seriously bias the observed risk and the exposure-response relationship due to misclassification of exposure or through other mechanisms.