This study reviews articles on case-control studies in which the cases were classified in two or more types. Application of multinomial models and their adequacy for case-control studies are discussed. Among the available multinomial adjustments, we argue that the polytomous logistic model is the most suitable for obtaining epidemiological measures of risk and association in case-control studies. By way of illustration, we present an application of this model in a population-based case-control study, comparing the results with those obtained in a binomial logistic model. The multinomial approach allows investigating, in a single analysis, the occurrence of associations between covariates and more or more subclasses of cases, thus providing the epidemiologically relevant possibility of identifying individualized risk and protective factors for each subclass.
Case-Control Studies; Logistic Models; Statistical Analysis
