A population of 495 volunteer subjects who applied to the screening diabetes Centre of Hôtel-Dieu hospital in Paris has been studied and divided into two samples. In the first 300 subjects sample the sensitivity and the specificity of different fasting blood glucose threshold values, of different glycosylated hemoglobin threshold levels and of various combinations of the above mentioned parameter have been evaluated as a test for diabetes diagnosis defined by a 2-hr OGTT value greater than or equal to 200 mg/dl. The predictive values for both positive and negative diagnosis were also evaluated. The combination of a fasting blood glucose greater than or equal to 120 mg/dl with a glycosylated hemoglobin greater than 5.8% was found to be the best association with a good relative specificity (97.9%), a fair sensitivity (64.7%) and a fair predictive value for a positive diagnosis (64.7%). A second sample includes 195 subjects. These were classified according to either the above criterion or either the 2-hr OGTT value greater than or equal to 200 mg/dl. A great majority of subjects (95.4%) was classified in concordance by the two methods. The characteristics of the subjects that were identically classified as diabetic subjects (n = 5) were compared to those of differently classified subjects (n = 9). Longitudinal surveys are needed in order to assess the validity of HbA1c measurement as a tool for diabetes screening by reference to unquestionable criteria of the disease.