Childhood leukemia near nuclear plants in the United Kingdom: the evolution of a systematic approach to studying rare disease in small geographic areas.

A cluster of childhood leukemia in a village near a nuclear plant in northern England prompted further studies of cancer in the vicinity of other nuclear plants in the United Kingdom. These studies demonstrated that the risk of childhood leukemia was increased near certain other nuclear plants. Although the reasons for the increase are still unclear, the scientific debate stimulated by these findings has clarified some of the special methodological problems encountered when studying rare diseases in small areas. Firstly, unless a specific hypothesis is defined in advance, the relevance of a single geographic cluster of disease can rarely be interpreted. Even when a prior hypothesis exists, the small number of cases which generally occur in a small area make the findings highly sensitive to reporting, diagnostic, or classification errors. The statistical power of such investigations is also usually low and only marked increases in risk can be detected. Furthermore, conventional statistical tests may be inappropriate if the underlying spatial distribution of the disease is not random; and little is known about the background distribution of disease in small areas. Investigations of specific hypotheses about defined sources of environmental contamination, especially if they can be replicated, are more likely to result in conclusive findings that are in-depth studies of individual clusters.