Application of the double kernel density approach to the analysis of cancer incidence in a major metropolitan area.

Rationale: Although cancer is a main cause of human morbidity worldwide, relatively small numbers of new cancer cases are recorded annually in single urban areas. This makes the association between cancer morbidity and environmental risk factors, such as ambient air pollution, difficult to detect using traditional methods of analysis based on age standardized rates and zonal estimates.

Study Goal: The present study investigates the association between air pollution and cancer morbidity in the Greater Haifa Metropolitan Area in Israel by comparing two analytical techniques: the traditional zonal approach and more recently developed Double Kernel Density (DKD) tools. While the first approach uses age adjusted Standardized Incidence Ratios (SIRs) for small census areas, the second approach estimates the areal density of cancer cases, normalized by the areal density of background population in which cancer events occurred. Both analyses control for several potential confounders, including air pollution, proximities to main industrial facilities and socio-demographic attributes.

Results: Air pollution variables and distances to industrial facilities emerged as statistically significant predictors of lung and NHL cancer morbidity in the DKD-based models (p<0.05) but not in the models based on SIRs estimates (p>0.2).

Conclusion: DKD models appear to be a more sensitive tool for assessing potential environmental risks than traditional SIR-based models, because DKD estimates do not depend on a priory geographic delineations of statistical zones and produce a smooth and continuous disease 'risk surface' covering the entire study area. We suggest using the DKD method in similar studies of the effect of ambient air pollution on chronic morbidity, especially in cases in which the number of statistical areas available for aggregation and comparison is small and recorded morbidity events are relatively rare.