Assessment of glucose tolerance test criteria for diagnosis of diabetes in Chinese subjects.

Abstract

To examine and compare WHO diagnostic criteria for diabetes mellitus. The relationship between FPG and 2-h glucose are examined in 680 OGTTs with a quadratic regression model and ROC analysis. Simultaneous measurements of HbA1 and fructosamine are also compared with multiple linear regression. Two hundred eighteen subjects (32%) had 2-h glucose greater than or equal to 11.1 mM, of which only 86 had FPG greater than or equal to 7.8 mM. Only 2 subjects had FPG greater than 7.8 mM and 2-h glucose less than 7.8 mM. Of subjects with 2-h glucose less than 7.8 mM (n = 332), only 9 had FPG greater than 6.0 mM. From the quadratic model, the predicted FPG corresponding to 2-h glucose = 11.1 mM was 5.7 mM, whereas the predicted 2-h glucose corresponding to FPG = 7.8 mM was 15.2 mM. ROC analysis showed that, with 2-h glucose greater than or equal to 11.1 mM as indicating diabetes, an FPG of 5.6 mM gave an intersect for sensitivity and specificity of 87%. HbA1 and fructosamine correlated more closely with 2-h glucose and area under the OGTT curve than with FPG. Given that a 2-h glucose cutoff of 11.1 mM can be justified from other studies, our results suggest that the FPG cutoff of 7.8 mM when used for screening purposes should be reduced. At a suggested value of 7.0 mM, specificity remains 98.5%, whereas sensitivity increases to 57%.