An Assessment of HbA1c in Diabetes Mellitus and Pre-diabetes Diagnosis: a Multi-centered Data Mining Study

HbA1c test has been widely used to evaluate glycemic control in patients with diabetes. However, there are controversial results regarding the value of HbA1c in the diagnosis of diabetes mellitus (DM). The present study investigates the diagnostic effectiveness of HbA1c in a large patient group. The oral glucose tolerance test and HbA1c results of 6551 patients (4704 healthy, 1345 pre-diabetes, 502 DM) in 12 different medical centers in Turkey between 2010 and 2016 were examined to understand the effectiveness of HbA1c in the diagnosis of DM. Different Roche systems were used for measuring HbA1c via the immunoturbidimetric method. The DM ROC curves revealed the diagnostic sensitivity, specificity, and AUC of 74.5%, 87.1%, and 0.866 (CI 95% 0.858–0.875), respectively, for HbA1c (at the cut-off 41 mmol/mol, 5.9%). For HbA1c at the universal diagnostic decision value of 48 mmol/mol (6.5%), the sensitivity and specificity were determined as 32.4% and 99.9%, respectively. The ROC curves for fasting plasma glucose (FPG) revealed the diagnostic sensitivity, specificity, and AUC of 71.3%, 85.3%, and 0.853 (CI 95% 0.844–0.861), respectively. However, the ROC curve results for pre-diabetes (HbA1c at the cut-off value of 39 mmol/mol, 5.7%) revealed the diagnostic sensitivity, specificity, and AUC of 45.7%, 76.4%, and 0.641, respectively. Furthermore, it was shown that the changes in HbA1c values due to gender and age had no clinical effect on the diagnosis. According to our results, it remains challenging to suggest HbA1c measurements can have a significant contribution to the FPG measurements. It was found that the sensitivity is specifically low in the assessment of the pre-diabetes data. Additionally, considering the problems associated with Hb1Ac measurements, further studies conducted in different regions by using different methods are required.