Prognostic value of haemoglobin A1c and fasting plasma glucose for incident diabetes and implications for screening

The aim of this analysis is to compare screening strategies with haemoglobin A_1c (HbA_1c), fasting plasma glucose (FPG) or combined measures in the identification of individuals at high risk for diabetes. Applying American Diabetes Association thresholds for FPG and HbA_1c screening, 6,803 subjects free of diabetes were classified as non-diabetic, pre-diabetic and possibly diabetic by FPG (<100, 100–125 and >125 mg/dl) and HbA_1c (<5.7, 5.7–6.4 and >6.4%). Hazard ratios, sensitivity and specificity were estimated for individuals with pre-diabetes with respect to incident diabetes in the following 5 years. Areas under the receiver operating characteristic curves (AUC) were estimated for levels of FPG ≤ 125 mg/dl and HbA_1c ≤ 6.4% in diabetes prediction. Although FPG and HbA_1c screenings poorly agreed in classifying individuals as pre-diabetic, hazard ratios [95% confidence interval] for incident diabetes were similarly increased in univariate models in the two pre-diabetic groups: FPG 100–125 mg/dl, 4.72 [3.69; 6.05]; HbA_1c 5.7–6.4%, 3.97 [3.05; 5.23]. HbA_1c and FPG had comparable AUCs (FPG, 0.732; HbA_1c, 0.725) and consequently similar 5-year sensitivities and specificities for their pre-diabetes definitions (when the lower cut-off for HbA_1c-defined pre-diabetes was increased to a level between 5.8 and 5.9%). Combining HbA_1c and FPG increased the AUC to 0.778, and a further increase to 0.817 was seen with additional inclusion of conventional risk factors. FPG and HbA_1c have comparable (yet insufficient) abilities in identifying individuals at high risk for diabetes. Effectiveness of a diabetes screening program could be improved by a risk score including FPG and HbA_1c.