The aim of the study was to assess gamma-glutamyl transpeptidase (gamma-GT), alanine aminotransferase, and aspartate aminotransferase (AST) in the prediction of diabetes and cardiovascular disease (CVD) in subjects free from hepatic diseases other than nonalcoholic fatty liver disease. The present analysis was performed on the cohort of subjects enrolled in the Firenze Bagno a Ripoli (FIBAR) study, a screening program for diabetes performed between 1 March 2001 and 31 December 2003 in the city of Florence on 3124 subjects who underwent an oral glucose tolerance test. Incident cases of diabetes in nondiabetic subjects (n = 2662) were obtained through databases of drug prescriptions, hospital admissions, and lists of subjects eligible for reimbursement. Incident CVD in subjects free of diabetes and CVD at enrollment (n = 2617) was identified through hospital admissions and through the register of causes of death. Mean follow-up was 39.6 +/- 12.0 months and 39.8 +/- 11.4 months for diabetes and CVD, respectively. Yearly incidence of diabetes and CVD was 0.4% and 0.2%, respectively. After adjustment for age and sex, gamma-GT >40 U/L was associated with increased incidence of diabetes and CVD (hazard ratio [95% confidence interval]: 2.54 [1.26-5.11], P < .05 and 2.21 [0.98-5.43], P < .10, respectively). Risk of diabetes, but not of CVD, was increased in patients with gamma-GT in the 25- to 40-U/L range. After adjustment for confounders, AST >40 U/L predicted CVD (hazard ratio, 6.5 [95% confidence interval, 1.5-28.1]), but not diabetes. Elevated gamma-GT or AST is an independent predictor of CVD. An increase of gamma-GT levels above the reference range, or also in the upper reference range, is an independent predictor of incident diabetes.