sometimes be wrong. These authors acknowledge that the problem they identify is not common, but they do suggest that caution is needed in interpreting results in the presence of nondifferential misclassification. (See Sexton et al., 1995, for a discussion of problems of misclassification associated with exposure measurement.)
Better measures of exposure can improve the ability of a study to assess adverse effects from environmental agents. Such improvements lead to an increase in the power of the study and reduction in bias, but also to increased cost. The health outcomes and exposure analysis must be considered together to arrive at a balanced prioritization of study requirements.
A wide array of exposure-assessment tools is available to the epidemiologic investigator, ranging from personal monitoring to the use of diaries or other indirect means. All the tools have potential value when used logically and reasonably. A continuing dialogue is necessary between scientists whose emphasis is on exposure assessment and epidemiologic investigators. In this regard, Lioy (1991a) has stressed the need for continued dialogue to focus on critical questions that will reduce ambiguity in terminology and conceptual design and improve the experimental design of both health and exposure studies. The development of the EPA guidelines on exposure assessment (EPA, 1992) and NHEXAS (Burke et al., 1992) should stimulate further discussion on these issues.
Landrigan (1983) has commented on the advantages of improved exposure characterization by using individual versus grouped data in a study of the health effects of arsenic in drinking water. The average concentration of arsenic in well water was a poor indicator of individual exposure because some of the persons studied had supplemented their consumption or changed completely to drinking bottled water. When estimates of bottled water consumption were incorporated into individual-exposure assessment, the dose-effect relationship was strengthened.
Kennedy et al. (1991), in a cross-sectional study of pulp and paper workers exposed to chlorine gas, found no differences from workers in other industries. However, when those pulp and paper workers who had an acute gas exposure were considered, symptom and forced-expiratory-volume (FEV1 and FEV25-75) differences were found. The authors concluded that accidental chlorine or chlorine dioxide exposures in pulp mills are associated with increased respiratory symptoms and airflow obstruction, particularly among nonsmokers and former smokers.
Monster and Smolders (1984) studied teachers and pupils at a kindergarten near a factory with emissions of tetrachloroethane. The levels of