Performance of glycated albumin for type 2 diabetes and prediabetes diagnosis in a South African population

Abstract

ObjectiveTo assess the utility of glycated albumin (GA%) as a diagnostic marker of type 2 diabetes and prediabetes in an African population. MethodsGA% levels were determined in a sample of 1294 mixed ancestry adults (74.2% women) residing in Cape Town using an enzymatic method. The participants' glycemic status was based on oral glucose tolerance test (OGTT). ResultsThe mean age was 47.8 years with a mean body mass index (BMI) of 28.7 kg/m2. Obesity was more pronounced in the screen-detected diabetes and prediabetes groups with mean BMI's of 32.5 kg/m2 and 31.5 kg/m2 respectively. The optimal thresholds of GA% to diagnose screen-detected diabetes and prediabetes, were 14.90% and 12.75% respectively. For screen-detected diabetes, the C-statistic was higher for HbA1c than GA% (p = .034) with values of 0.899 (95% CI 0.855–0.943) and 0.873 (0.782–0.892) respectively. The agreement between GA% and HbA1c at their optimal thresholds for diagnosing screen-detected diabetes, was kappa = 0.33 (95% CI 0.26–0.40) and was higher than the agreement for prediabetes, kappa = 0.16 (0.11–0.21). The performance of GA% to identify screen-detected diabetes at the optimal threshold of 14.90%, was 64.8% (95% CI 54.1%–74.6%) for sensitivity and 93.5% (92.0%–94.9%) for specificity. GA% was significantly less sensitive, but more specific than HbA1c (at the optimal threshold of 6.15%) for screen-detected diabetes diagnosis (both p ≤ .002 from McNemar tests for sensitivity and specificity comparisons). ConclusionsGA% performed less well than HbA1c to identify participants with OGTT-diagnosed type 2 diabetes or prediabetes (HbA1c cut-off of 6.15% and 5.95% respectively) in this population.