The aim of the present study was to evaluate the relative usefulness of some of the diagnostic methods used to estimate the prevalence of diabetes mellitus in epidemiological studies. A cross-sectional descriptive study with multistage stratified sampling was designed. A personal interview, a capillary baseline glycemia and a OGTT were carried out. The medical records and the use of antidiabetic drugs and self-control products were also evaluated. The prevalences obtained with these methods were: Medical records = 1.31%, personal interview = 7.91%, use of drugs = 1.93%, self-control products = 1.05%, capillary baseline glycemia greater than or equal to 120 mg/dl = 4.04%, OGTT with WHO criteria = 5.62%. Using OGTT as a reference, the diagnostic test with the greatest overall value was capillary baseline glycemia equal to or higher than 120 mg/dl. We think that, as a whole, the election of one method or another will depend on both the precise aim of the study and the type of design and the cost that can be sustained.