1533-P: Detecting Diabetes in Africans Is Improved by Both Modifying the A1C Threshold and Adding Glycated Albumin

A1C at the standard threshold of 6.5% has been a poor predictor of diabetes (DM) in Africans. Reasons for the poor performance of A1C are myriad and include hemoglobinopathies and anemia. To begin to address challenges with A1C as a diagnostic test for Africans, there needs to be a determination that the threshold for A1C of 6.5% is appropriate. In addition, there could be value in determining if glycated albumin (GA), a non-fasting marker of glycemia not affected by hemoglobinopathies and anemia, would improve detection of DM. Therefore, we tested in Africans (a) the ability of A1C and GA independently and combined to predict DM and (b) the optimal cut-points for A1C and GA. OGTT were performed in 366 African-born blacks (age 38±10 (mean± SD); BMI 27.5 ±4.4 kg/m²) who self-identified as healthy and currently living in the Washington, DC area. Diagnosis of DM was based on glucose criteria for the OGTT. Ability of A1C and GA to predict DM was estimated by area under the receiver operating characteristic curve (AROC). Youden index established optimal cut-points for A1C and GA. Diabetes was newly diagnosed in 6% (23/366). For the prediction of DM by A1C the AROC was 0.78 (95% CI 0.68-0.88), and the optimal cut point was ≥5.8%. For the prediction of DM by GA the AROC was 0.73 (95% CI 0.61-0.85), and the optimal cut point was ≥14.05%. At these cut-points, DM was identified by A1C in 61% (14/23) (sensitivity 61%, specificity of 82%), and by GA in 65% (15/23) (sensitivity 65%, specificity 72%). The combination of A1C and GA identified DM in 83% (19/23) (sensitivity 83%, specificity 60%). Overall, the combination of A1C ≥5.8% and GA ≥14.05% markedly improved the detection of undiagnosed diabetes in Africans. Combining two non-fasting markers of glycemia at African-specific thresholds may be an important innovation in healthcare delivery.