Epidemiological Studies of Geographic Variations of Cancer Incidence in Sweden: Choice of variables and statistical units

Abstract

The Swedish Cancer-Environment Registry has been used for studies of the parts of cancer incidences and of their variations between subpopulations, which, in a statistical sense, can be ‘explained’ by environmental variables. In previous studies the dependence of age-standardized incidences in municipalities (279 in Sweden) on population density, socio-economic variables, smoking habits and a variable for ‘diagnostic intensity’, assumed to allow for variations in under-diagnosing and under-reporting, was studied in, i.a., multiple-regression analyses. Due to intrinsic variation in potentially etiologic factors and in size, municipalities are less suitable as geographic units. Therefore a system was developed to build up, from parishes (about 2600), pseudo-municipalities standardized with respect to size and some environmental variable(s), in the present study population density. This paper shows, for a number of cancer diseases and for total cancer, the superiority of pseudo-municipalities to municipalities with respect to power of explaining variation and, preliminarily, also incidence.