Register analysis of measures of urbanization and cancer incidence in Sweden

Abstract

The Swedish Cancer Registry and Cause-of-Death Registry are being studied with different statistical models (regression and Poisson models) in order to characterize the structure of the “urban factor”, i.e., the causes of higher cancer incidence in urbanized areas. Lacking epidemiologically useful variables for air pollution, this component of the urban factor must be described in terms of a component that is correlated with urbanization but remains unexplained by other known factors. Correlations of degree of urbanization with effectiveness of diagnostic routines and with intensity of tobacco smoking are confounding, and smoking also acts synergistically in presumptive effects of air pollutants. Some 40% of the Swedish lung cancer incidence in males and some 20% of that in females are statistically explainable in terms of urbanization variables after subtraction of the effects of diagnostic routines and smoking. These figures are higher than expected from previous estimates, indicating that these are underestimates (e.g., because of strong synergisms with smoking) or that life style factors such as food habits contribute. Similar conclusions are reached concerning incidences of cancer at other sites. Incidences have been referred to Swedish municipalities. This leads to analytical difficulties due to internal heterogeneity of the “units” and a great variation in size that causes a variation in variances. These disadvantages could be counteracted by the construction of “pseudo-municipalities” from parishes. It may be concluded that epidemiological “register analysis” is an instrument that is complementary to risk estimations based on exposure doses determined by chemical methods. Register analysis may further generate hypotheses to be tested by more refined epidemiological and chemical methods.