An analysis of the geographic variation in cancer incidence and its determinants in Ontario

Cancer incidence data for the Ontario Public Health Units in 1980-91 were examined to investigate regional patterns and the existence of outlier values after adjustment for known risk factors. Candidate risk factors were derived from the Ontario Health Survey and the census. Weighted regressions were fit to the data, and the spatial pattern of the residuals was analyzed. The number of outlier data points with significant elevations or reductions in risk was close to that expected by chance. They were dispersed geographically, and occurred in a variety of cancer types. We conclude that, in general, most of the geographic variation in cancer risk can be associated with variation in known risk factors, and that there appear to be no broad regional effects remaining after adjustment for these factors. A few cancer sites provide limited evidence of regional effects that may warrant further investigation.