Abstract
Epidemiologic information and analysis are essential for the process of health risk assessment and management in pollution episodes. Recognition of causal associations between exposure and disease requires an understanding of the nature of data upon which such associations are based and of the limitations that affect such data. Data concerned with disease detection are often affected by long and variable latencies, clinical non-specificity, low frequencies (small population sizes), and reporting biases. Data concerned with measuring exposures must take into account pathway uncertainties, probable low dose levels, inability to develop dose-response information, and the frequent necessity to rely on indirect surrogates for dose estimation. Where such difficulties can be overcome, epidemiologic analysis can be decisive in identifying causal relationships. More often, data limitations require more limited conclusions. Case studies from New York, Michigan, South Carolina, and Massachusetts illustrate the impact of these principles on the process of health risk assessment.