AbstractAimTo investigate the performance of HbA1c in predicting incident diabetes among Korean adults with normal fasting glucose and impaired fasting glucose levels.MethodsThis study used data from the Korean Genome Epidemiology Study‐Kangwha Study. A prospective analysis was carried out on 2079 people (820 men and 1259 women) who completed follow‐up examinations up until 2013. Diabetes was defined as fasting blood glucose level ≥ 7.0 mmol/l, HbA1c level ≥ 48 mmol/mol (6.5%), or current treatment for diabetes. Areas under the receiver‐operating characteristic curves were used to assess the different performances of HbA1c, glucose and insulin in predicting diabetes.ResultsThe median follow‐up time was 3.97 years, during which 7.7% of men and 6.3% of women developed incident diabetes. The areas under the receiver‐operating curves (95% CI) for diabetes prediction were 0.740 (0.692–0.787) for HbA1c, 0.716 (0.667–0.764) for glucose and 0.598 (0.549–0.648) for insulin. HbA1c showed better predictive power in people with impaired fasting glucose (area under the curve 0.753, 95% CI 0.685–0.821) than in those with normal glucose (area under the curve 0.648, 95% CI 0.577–0.719). An HbA1c threshold of 40 mmol/mol (5.8%) was found to have the highest predictive value for diabetes, with a relative risk of 6.30 (95% CI 3.49–11.35) in men and 3.52 (95% CI 2.06–6.03) in women after adjusting for age, waist circumference, triglycerides, hypertension, family history of diabetes, smoking, alcohol intake, exercise and baseline glucose level.ConclusionsHbA1c can be used to identify people at high risk for the development of diabetes, especially in those with impaired fasting glucose levels.