To determine the optimal fasting plasma glucose (FPG) cut-off value which effectively identifies high risk subjects for type 2 diabetes in Japanese, we conducted a population-based prospective study on diabetes as part of the Japan Public Health Center-based Prospective Study and estimated the 5-year incidence of diabetes. The subjects of the analysis of this study were 2,207 Japanese aged 51-70 at baseline from whom a fasting blood sample was collected in both the baseline and the 5-year follow-up surveys and who completed the questionnaires at both times. Diabetes was defined as an FPG value > or = 126 mg/dL (7.0 mmol/L) and/or self-reported diabetes. A total of 125 subjects developed diabetes during the 5 years after the baseline survey, and the incidence rate for a baseline FPG value of 95-99, 100-104, 105-109, 110-114, 115- 119, and 120-125 mg/dL was 6.1, 11.5, 30.3, 52.6, 86.4, and 115.2 per 1,000 person-years, respectively. The results of receiver operating characteristic curve analysis suggested that an FPG value of 102 mg/dL (5.67 mmol/L) was optimal for predicting diabetes during the next 5-years. The cut-off value was similar in both genders and in the 51- to 60-year-old group and 61- to 70-year-old group. Use of hemoglobin A(1c) level > or = 6.1% for an additional diagnostic criterion resulted in a small increment in incidence, but the cut-off value for predicting diabetes was almost the same (101 mg/dL). The results of this study suggested that the cut-off FPG value should be lowered in terms of prediction of type 2 diabetes among Japanese populations.