Derivation and validation of an HbA1c optimal cutoff for diagnosing prediabetes in a South African mixed ancestry population

Abstract

Introduction: Prediabetes compromises impaired fasting glucose and impaired glucose tolerance and is a high risk for future diabetes mellitus and cardiovascular disease. Traditional diagnostic methods involve a fasting sample or oral glucose tolerance test, which is cumbersome, time-consuming and inconvenient. An HbA1c-based approach has been incorporated into new guidelines, but cut-offs may vary and have not been defined for all population groups. We derived and validated HbA1c cut-offs to diagnose prediabetes in mixed ancestry South Africans. Methods: Participants were 667 (derivation sample), 234 (validation sample 1) and 674 (validation sample 2) diabetes-free individuals. They underwent standard 2-hour OGTT with HbA1c test. Receiver–operator charac- teristic curves were used to determine optimal HbA1c cut-off to predict prediabetes. Results: A total of 27.7% participants in the derivation sample had prediabetes versus 17.5% (validation sample 1) and 15.4% (validation sample 2). The optimal cut-off was 5.75% in all three cohorts with sensitivity and specificity of 64.8% and 60.4% in combined derivation and validation sample 1, and 59.6% and 69.8% in validation sample 2. Conclusion: The discriminatory capacity of HbA1c for predicting prediabetes in this population is modest at the derived cut-off. The use of HbA1c alone in this setting may result in an inaccurate diagnosis