To determine whether the new category of impaired fasting glucose (IFG) recently proposed by the Expert Committee of the American Diabetes Association is a risk factor for cardiovascular disease. Death certificates and residence transfer documents from the cohort population consisting of participants of the diabetes prevalence study in Funagata, Yamagata prefecture, Japan, 1990-1992, were analyzed up through the end of 1996. First, the cohort population was classified into three groups: normal glucose tolerance (NGT) (n = 2,016), impaired glucose tolerance (IGT) (n = 382), and diabetic (n = 253). Then the same population was reclassified into normal fasting glucose (NFG), IFG, and diabetic. The cumulative survival rates among the groups were compared using the classical life-table method, and age-adjusted analyses, the person-year method, and Cox's proportional hazard model were adopted. At the end of seven observed years, the cumulative survival rates from cardiovascular disease of IGT and diabetes were 0.962 and 0.954, respectively, both significantly lower than that of NGT (0.988). The Cox's proportional hazard model analysis showed that the hazard ratio of IGT to NGT on death from cardiovascular disease was 2.219 (95% CI 1.076-4.577). However, the cumulative survival rate of IFG from cardiovascular disease was 0.977, not significantly lower than that of NFG (0.985). The Cox's hazard ratio of IFG to NFG on death from cardiovascular disease was 1.136 (0.345-3.734), which was not significant either. IGT was a risk factor for cardiovascular disease, but IFG was not.