The misuse of attributable and prevented fractions in the evaluation of screening in case-control studies

Methodological problems of evaluating screening programs in case-control studies have been the subjects of controversial discussion in recent years. The increasing number of published papers presenting results of case-control studies on the evaluation of cancer screening during the last few years demonstrate that this design has rapidly gained popularity in cancer epidemiology, although the methodological criticism has not been solved yet and is usually not reflected in the interpretation of the study results. A new development in this area has been the incorporation into the evaluation of some measure of the population impact of screening programs. This paper discusses the application of these measures (attributable and prevented fractions) to screening situations. The implicit assumptions required for their valid interpretation are delineated. Primarily, it is our intention to raise the flag of caution against the naive practical application and overoptimistic interpretation of attributable and prevented fraction estimates derived from case-control studies on the evaluation of screening.