The HbA1c threshold for diagnosis of type 2 diabetes mellitus in an adult Black South African population: A cross-sectional study from 2016 to 2020

Abstract

Background Diabetes mellitus is a growing public health concern in Africa. The HbA1c threshold of 6.5% used for diagnosing diabetes is based on populations of European ancestry and may not be optimal for African populations. This study aimed to identify an optimal HbA1c threshold for the diagnosis of type 2 diabetes mellitus in a Black South African population and to compare its diagnostic performance using fasting plasma glucose (FPG) and two hour postprandial glucose (2hrPG) criteria. Methods A cross-sectional study (2016–2020) in Agincourt, Mpumalanga, was conducted among Black adults aged ≥40 years. HbA1c diagnostic performance was assessed using FPG, 2hrPG, and diabetic retinopathy as the reference standards. A receiver operating characteristic (ROC) curve analysis was used to determine optimal thresholds, sensitivity, specificity, and area under the curve (AUC). We used logistic regression to explore how HbA1c levels are associated with FPG, 2hrPG, hypertension, age, gender, and body mass index (BMI). Results Using FPG (standard diagnostic threshold ≥7.0 mmol/L) as the reference, an HbA1c threshold of 6.0% achieved a sensitivity of 100%, specificity of 83%, and AUC of 0.97.The optimal HbA1c threshold was 6.1% when compared to 2hrPG (standard diagnostic threshold ≥11.1 mmol/L), with a sensitivity of 85.0%, specificity of 85.9%, and AUC of 0.89. Diabetic retinopathy could not be meaningfully analysed as only one of 300 participants with interpretable images had diabetic retinopathy, with 157 images uninterpretable or missing. Conclusion Our findings indicate that using an HbA1c cut-off of 6.1% presents the best balance between sensitivity and specificity when compared with the results of the oral glucose tolerance test. A cut-off of 6.0% also performed very well against fasting glucose, making it useful in settings where fasting tests are more practical. This highlights the need to adapt HbA1c thresholds to local populations.