We have analyzed data from 3370 OGTTs performed during epidemiologic studies in three different ethnic groups (Micronesian, Polynesian, and Melanesian) in various Pacific countries to examine the value of a single fasting plasma glucose greater than or equal to 140 mg/dl as a diagnostic test for diabetes (defined as 2-h plasma glucose greater than or equal to 200 mg/dl). A fasting plasma glucose greater than or equal to 140 mg/dl is a highly specific test for diabetes, specificity in the various populations ranging from 98.1% to 99.7%. On the other hand, the sensitivity of fasting plasma glucose was not high and varied greatly between the populations (46.2%-79.0%). The predictive value of fasting plasma glucose for the diagnosis of diabetes was lowest in populations with a low diabetes prevalence and improved in higher prevalence groups. These data indicate that a fasting plasma glucose greater than or equal to 140 mg/dl is not a good screening test, apart from populations with a high prevalence of diabetes mellitus, and the 2-h postload plasma glucose is preferable.