Utility of A1C for Diabetes Screening in the 1999–2004 NHANES Population

Thirty percent of people with diabetes are undiagnosed (1), and up to 25% already have microvascular complications at diagnosis (2). The American Diabetes Association recommends screening adults ≥45 years of age, especially those with a BMI ≥25 kg/m2, at 3-year intervals with a fasting plasma glucose (FPG) measurement (3). However, physicians infrequently use an FPG for screening. For example, in a large health maintenance organization within the University of Michigan health system, 184 physicians at 22 separate locations screened 5,752 (69%) of 8,286 people without diabetes over a 3-year period. Ninety-five percent of the screening tests were random glucose measurements, 3% were FPGs, and 2% were A1Cs (4). Random glucose levels depend on the length of time after the previous meal and the carbohydrate content of that meal. There is no agreement on what random glucose values should lead to further investigations. Because A1C levels reflect average glycemia during the preceding 3–4 months, meal issues are not a factor. To determine what A1C level should lead to further tests to diagnose diabetes, we examined the 1999–2004 National Health and Nutrition Examination Survey (NHANES) population of those not known to have diabetes to determine the sensitivity and specificity with which various A1C levels identified people with diabetes. NHANES oversampled African-American and Hispanic individuals to obtain enough data for analysis in this minority subset. Accordingly, results were weighted to reflect the …