In order to evaluate the prevalence of gestational diabetes mellitus (GDM) and the presence of risk factors for GDM, we conducted a retrospective study of a cohort of Italian women. In addition, we compared universal versus selective screening to validate the ADA's recommendations in our population. From June 1st, 1995 to December 31st, 2001, universal screening for GDM was performed in 3950 women. The glucose challenge test (GCT) was positive (GCT+) in 1389 cases (35.2%). The 1-h glucose level after GCT enabled us to diagnose GDM directly in 24 pregnant women. Oral glucose tolerance test (OGTT) was performed in 1221 GCT+ women (144 cases with GCT+ dropped out) and GDM was diagnosed in 284 (23.2%) of them. OGTT was also performed in 391 randomly chosen, women from the GCT negative (GCT-) group. In this last group 25 (6.3%) women had GDM. Thus, the total number of subjects with GDM was 333 out of 3806 with a prevalence of 8.74% in the entire cohort. Assuming that the rate of GDM observed in the random sample of GCT- women is applicable to the whole group of 2561 GCT- women, then 161 GCT- patients could also have GDM. This will further increase the estimated prevalence for the whole cohort up to 12.3% (i.e. 469 out of 3806 pregnant women). There were 236 (5.6%) women with a low risk for GDM (normal weight, age less than 25 years and without a family history of diabetes). In this group we found 34 cases and five cases with positive screening test and GDM, respectively. Thus, if we excluded low risk women from the screening test, as suggested by ADA recommendations, only five women with GDM would have been missed. However, about 95% of our population were at medium or high risk for GDM and, therefore, would have been screened. The rate of GDM was significantly higher in women with a positive history of diabetes, increasing age, previous pregnancies, pre-pregnancy overweight and short stature. After logistic regression analysis, GDM diagnosis was significantly correlated with age (P<0.0001), pre-pregnancy BMI (P<0.0001), weight gain (P<0.0001) and family history of diabetes (P<0.01).