Sensitivity and Specificity of Glycated Hemoglobin as a Diagnostic Test for Diabetes and Prediabetes in Arabs

Abstract

Glycated hemoglobin (A1C) has been recommended by the American Diabetes Association for the diagnosis of diabetes and prediabetes. The diagnostic utility of A1C has not been evaluated in Arabs, a population at increased risk for developing diabetes. Our objective was to examine the sensitivity and specificity of A1C for the diagnosis of diabetes and prediabetes in Arabs. In this cross-sectional study, glucose tolerance was classified by the American Diabetes Association diagnostic criteria specified for A1C, fasting plasma glucose, and 75-g oral glucose tolerance test. A population-based representative sample of 482 randomly selected adult Arabs without known diabetes was studied. Sensitivity, specificity, and area under the receiver operating characteristic curve of A1C diagnostic cutpoints for diabetes and prediabetes were calculated. κ Coefficients were used to test for agreement between A1C categorization and glucose-based diagnoses. A1C testing correctly identified 5% of individuals diagnosed with diabetes by oral glucose tolerance test, 13% by fasting plasma glucose, and 41% by both criteria. A1C alone identified 14% of individuals diagnosed with impaired glucose tolerance, 9% with impaired fasting glucose, and 33% with both abnormalities. Sensitivity, specificity, and area under the receiver operating characteristic curve were 19% (16-23%), 100% (99-100%), and 77% (69-85%) for diabetes A1C cutpoint and 14% (11-17%), 91% (89-94%), and 57% (52-62%) for prediabetes A1C range. A1C cutpoint of 6.2% for diabetes and 5.1% for prediabetes yielded the highest accuracy but still missed 73% of those with diabetes and 31% with prediabetes. Agreement between A1C and diabetes (κ = 0.2835) or prediabetes (κ = 0.0530) was low. A1C-based criteria yield a high proportion of false-negative tests for diabetes and prediabetes in Arabs. Racial/ethnic differences in A1C performance for diagnosis and prediction of diabetes exist. This paper examines its utility against glucose measurements in an at-risk Arab population.