established. It is possible to validate a true elevation in a rate only if it can be demonstrated that an event is unusual (or improbable); this implies that the researcher drawing the inference has a good grasp of what is usual (Rothenberg et al., 1990). For instance, the assessment of time trends in birth defects or reproductive health must contend with the lack of well-established national and regional information about rates of major birth defects and spontaneous abortion. Moreover, the system used to code and classify minor and major birth defects can differ from one place to the next. Evaluating the occurrence of spontaneous abortions requires information about regional and cultural variations in rates and kinds of contraceptives used, rates of elective abortion, and genetic-screening tests that can provide the basis for such procedures. In areas where proportionally more pregnancies are voluntarily terminated, reported rates of spontaneous abortion might be lower.

For multiply caused diseases the strength of association measured depends on many factors, including the power of the overall study to detect an effect. Power is a statistical measure of the potential of the study to find an association. It varies with the inverse of the square root of size of the population studied and the expected relative risk of the disease. In order to detect significant patterns, rare diseases are best studied in larger populations. More common diseases can be studied in smaller populations. However, to the extent that multiple causes are involved, as they are with most chronic diseases, larger populations are generally required in order to obtain significant results in studies of more common diseases as well. Refining the measures of diseases and the assessment of exposure can improve the power of a study to detect an association. “Strong” associations are not more biologically correct then “weak” associations. They may be less readily dismissed as confounding, however, and are more readily detected.

Cancer clusters and spontaneous abortion clusters are among the most commonly reported events linked to exposure to hazardous-waste sites. These clusters also rank as among the most difficult outcome for which causation can be inferred. In part, this is because both outcomes reflect multiple causes and because it is difficult to determine the relevant regional baseline rate. Also, for cancer, the latent period (between exposure and onset of disease) is often long.

Neutra (1990) notes that because of the small populations exposed at many hazardous-waste sites, the observed rates of occurrences for diseases studied in a given cluster often must be at least 20 times greater than expected to support an inference of causation. Today,



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement