[Case-control design in cardiovascular disease epidemiology. II--Analysis of data]

The wide spread distribution of statistical software recommended for multivariate analysis as well as the ease in handling it can lead the users into adopting wrong measures if they do not pay attention to the theoretical principles behind those methods. With a view to bringing out some of these principles some steps for the data analysis of a case-control study undertaken in the city of S. Paulo-Brazil from March, 1993 to February, 1994 in order to test the association between diabetes mellitus and ischaemic heart disease after adjusting for potential confounders and/or modifiers of effect are presented. Methodologic issues are emphasized in the development of four steps: a) the data bank structure: b) the calculation of statistical power; c) the definition of variables strata and codification and d) the choice of the logistic regression method.