The usefulness of fasting plasma glucose (FPG) in the diagnosis of diabetes mellitus was assessed in Asian Indians in South India. Oral GTT values in 570 newly screened adults were studied. Taking the WHO criteria of 2 h plasma glucose (PG) of > or = 11.1 mmol l-1 for diagnosis of diabetes, the validity of a FPG of > or = 7.8 mmol l-1 cut off value for diabetes was assessed. Using the regression analysis, the correlations of the FPG to 2 h PG were examined. Among the 268 with 2h PG value of > or = 11.1 mmol l-1, 205 (76.5%) had FPG > or = 7.8 mmol l-1. Sensitivity of FPG was 76.5% for diagnosis of diabetes and its specificity was 99%. An exponential regression model gave the best fit for FPG vs 2 h PG and using the regression equation, the predicted FPG for a 2 h PG of 11.1 mmol l-1 was 7.05 mmol l-1. Sensitivity increased to 90.3% with FPG of 7.05 mmol l-1 while the specificity remained at 92%. It is concluded that in the south Indian subjects, the sensitivity for diagnosis of diabetes with FPG of > or = 7.8 mmol l-1 was 76.5%; and it increased to 90.3% with FPG of > or = 7.05 mmol l-1.