OBJECTIVE An increase in the serum gamma-glutamyltransferase (GGT) concentration has been regarded as a marker of alcohol drinking or liver disease. Some reports, however, have suggested that the serum GGT may be a sensitive and early biomarker for the development of prediabetes and diabetes. In this study we investigated whether serum GGT is a reliable predictor of the incident impaired fasting glucose (IFG), including diabetes. METHODS We performed a prospective study for two years (2002-2004). We analyzed the periodic health examination data from a total of 4,711 men. The examinations were done in the years 2002 and 2004. The analyzed data included a self-questionnaire, a physical examination and the laboratory results. Both IFG and diabetes were defined as a serum fasting glucose concentration of more than 100 mg/dL and 126 mg/dL, respectively. RESULTS A total of 738 cases (15.7%) of incident IFG and 13 cases (0.3%) of diabetes occurred. The mean serum GGT concentrations were quite different between the normal (38.0 IU) and incident IFG groups (50.3 IU), and the incident diabetes group (66.0 IU) (p <0.001). After multivariable adjustment, the relative risks for incident IFG or diabetes across the baseline GGT categories (<10th, 10th-20th, 30th-40th, 50th-60th, 70th-80th and >90th percentile) were 1.0, 1.172 (0.769-1.785), 1.107 (0.725-1.689), 1.444 (0.934-2.232), 2.061 (1.401-3.031) and 2.545 (1.784-3.631) (p-value for trend: <0.001). The risks significantly increased with increasing levels of GGT for 2 years; when comparing the increased groups (<10%, .10-20%, >20%) versus the decreased over 20% group of GGT, the risks for IFG or diabetes were 1.334 (1.002-1.776), 1.613 (1.183-2.199) and 1.399 (1.092-1.794). CONCLUSIONS Our findings suggest that serum GGT concentrations within its normal range may be an early predictor of the development of IFG and diabetes. As serum GGT is a relatively inexpensive test and a reliable marker, it might have important implications in public health promotion.