Cancer clusters: the importance of monitoring multiple geographic scales.

Analysis of cancer incidence data at a variety of geographic scales provides surveillance information that can allay fears of the general public, prevent costly and unwarranted epidemiologic studies driven by political pressures, and target appropriate cases for further investigation. We systematically examined New Jersey Cancer Registry data (1979-1985) for childhood and young adult (0-24 years) cancers at multiple geographic scales--at the state level, then by degree of urbanization, county boundaries, and minor civil divisions. The state had increased rates for some cancers when compared to four other SEER (Surveillance, Epidemiology and End Results) states. No meaningful patterns at either the most urban/suburban/most rural scales or at the county level of analysis were found. At the minor civil division level, the Ederer, Myers and Mantel method found evidence of clustering of pediatric and young adult cases statewide. Stratification of cases by race yielded even stronger findings and indicated that whites had clustering of cases for several cancer types. In-depth analysis of individual cases yielded hypotheses for investigating identified clusters.