OBJECTIVE - For epidemiological purposes, it has now been recommended that a fasting plasma glucose value of 7.0 mmol/l can be used to diagnose diabetes, instead of a 2-h value of 11.1 mmol/l. This study assesses the impact of making this change on the prevalence of diabetes and on the phenotype of individuals identified. RESEARCH DESIGN AND METHODS - Data were collated from nine population- based southern hemisphere studies in which a 75-g oral glucose tolerance test was performed. Comparisons were made between the prevalence derived from fasting values only and the prevalence derived from 2-h values only. Cardiovascular risk was assessed in all individuals. RESULTS - There were 20,624 subjects in the nine surveys, of whom 1,036 had previously diagnosed diabetes and 1,714 had newly diagnosed diabetes, according to either fasting or 2-h glucose. The differences in prevalence within each population resulting from changing the diagnostic criteria ranged from +30 to - 19% (relative difference) and +4.1 percentage points to -2.8 percentage points (absolute difference). BMI was the most important determinant of disagreement in classification. A total of 31% of those individuals who were diabetic on the fasting value were not diabetic on the 2-h value, and 32% of those with diabetes on the 2-h value were not diabetic on the fasting value. Apart from obesity, there were no differences in cardiovascular risk between those identified by the fasting and the 2-h values. CONCLUSIONS - Changing the diagnostic criteria is likely to have variable and some- times quite large effects on the prevalence of diabetes in different populations. Furthermore, the fasting criterion identifies different people as being diabetic than those identified by the 2-h criterion.