[Problems in the extrapolation of attributable risk estimates]

The problematic nature of the generalization of epidemiologic results from the study population to other populations has been controversially discussed for a long time in the epidemiologic literature. There is no methodological justification of this inference beyond the study population, it constitutes a conclusion by analogy which has to be substantiated by medical arguments. In the special case of the extrapolation of attributable risk estimates from a study population to other populations two different methods of extrapolation have been proposed. In the paper they are critically reviewed. In addition the concept of attributable risk is briefly introduced. Data from the Leubeck Blood Pressure Study, a cross-sectional study in the field of cardiovascular epidemiology, is used to illustrate the practical application and the limitations of the methods.