The aims of this study were to determine if impaired fasting glucose should be redefined as a fasting plasma glucose (FPG) of 100 to 125 mg/dL (5.6-6.9 mmol/L) in Korea. A prospective cohort study was undertaken involving 13189 male workers aged 30 to 59 years who did not have medication for diabetes, a history of any cancer, or a fasting glucose level of 126 mg/dL or higher at the initial examination between January 1999 to December 2000. Subjects were reexamined at periodic annual health examination over a 5-year period. The receiver operating characteristic curve for predicting the future onset of diabetes was derived by plotting the sensitivity against 1 - specificity for a baseline FPG of less than 126 mg/dL. The age- and body mass index-adjusted incidence density of type 2 diabetes mellitus was examined according to the percentile of the distribution for the baseline FPG. The baseline FPG for predicting the future onset of diabetes at a point on the receiver operating characteristic curve that was closest to the ideal 100% sensitivity and 100% specificity was 92 mg/dL. There was a threshold for the age- and body mass index-adjusted incidence density of diabetes in the group with FPG of 93 to 95 mg/dL, at a mean of 93.9 mg/dL. Lowering the lower limit of impaired fasting glucose to 100 mg/dL (5.6 mmol/L) would optimize its sensitivity and specificity for predicting the future onset of diabetes in Korea.