The validity of the new diagnostic criteria for diabetes mellitus for Japanese subjects was assessed by the long-term prognosis of 501 patients, found in an epidemiological survey of 6,681 people in a Japanese town. Ten-year relative survival rates were examined in relation to the 50 g oral glucose tolerance test (serum glucose determined by a ferricyanide method). Subjects were classified using the new criteria. The prognosis of normal subjects (fasting glucose less than 140 mg/dl (7.8 mmol/l) and 2-hour glucose level less than 140 mg/dl) and of subjects with fasting glucose less than 140 mg/dl and 2-h 140-200 mg/dl (7.8-11.1 mmol/l) did not differ from that of the general population. Diabetics (fasting greater than or equal to 140 mg/dl and 2-h greater than or equal to 200 mg/dl) showed a significantly reduced survival rate, corresponding to an excess mortality. Subjects with a fasting level less than 140 mg/dl and 2-h greater than or equal to 200 mg/dl had a relative survival rate similar to that of normal subjects. As a group, fasting levels greater than or equal to 140 mg/dl indicated a reduced survival rate, regardless of the 2-h level, while those with a 2-h level greater than or equal to 200 mg/dl alone failed to show a significantly reduced survival rate. Thus, the fasting glucose values appeared to be more closely related to death rate than the 2-hour glucose values.