Diagnostic performance of HbA1c for detecting prediabetes and diabetes in Turkish adults

Abstract

We evaluated the diagnostic performance of HbA1c in diabetes and prediabetes and compared its efficiency in a model where HbA1c is used as a screening test. Records of 945 patients who had undergone simultaneous OGTT and HbA1c were evaluated. American Diabetes Association (ADA) criteria were used for diagnosis. Agreement between OGTT and HbA1c was evaluated. Using OGTT as the gold standard, the diagnostic performance of HbA1c was evaluated with ROC analysis and optimum cutoffs were determined. Then, the group was rearranged as if HbA1c was performed first as a screening test with new cutoffs and if OGTT was performed for diagnosis among suspected individuals. According to OGTT, 53 (5.6%) patients were diabetic and 247 were prediabetic; 18 of diabetics (34%) and 94 of prediabetics (38%) were diagnosed by HbA1c with present ADA cutoffs. The kappa coefficient for agreement between OGTT and HbA1c was 0.756 for diabetes and 0.336 for prediabetes. In ROC analysis, optimal HbA1c cutoff points were 38 mmol/mol (AUC 0.953) for diabetes and 35.5 mmol/mol (AUC 0.673) for prediabetes. In a new model after initial screening with HbA1c, 398 patients with HbA1c > 35.5 mmol/mol were taken for further OGTT. Fifty-three of diabetics (100%) and 199 of prediabetics (81%) were diagnosed. Prediabetic 48 patients (19%) were missed. HbA1c showed poor performance in diagnosis of diabetes and prediabetes. Instead, using HbA1c as a screening test with well-defined cutoff may offer practical and economical advantages greatly reducing the number of OGTTs required.