The relative merits of haemoglobin A1c and fasting plasma glucose as first-line diagnostic tests for diabetes mellitus in non-pregnant subjects.

HbA1c was measured by high-performance ion-exchange chromatography in 401 non-pregnant patients undergoing oral glucose tolerance tests (OGTT). All those with HbA1c>6.2% (reference range 3.8-5.5%) had diabetic OGTT (sensitivity 41%, specificity 100%). Although a fasting plasma glucose (FPG) cut-off > or =7.0 mmol l(-1), as recommended by the American Diabetes Association (ADA), had greater sensitivity (78%), false positives (12%) limited its usefulness, so more diagnostic confidence could be placed in a positive HbA1c. In agreement with the ADA, we found FPG gave only slightly lower diabetes prevalence than the OGTT, but this masked a significant number of individual discrepancies (false positives and negatives) cancelling out each other. The new ADA category of impaired fasting glucose did not correlate well with impaired glucose tolerance. HbA1c is insufficiently sensitive as a direct substitute for the OGTT. A third of subjects diabetic on OGTT had normal HbA1c values, so it cannot exclude diabetes as currently defined, but HbA1c screening could make sufficient positive diagnoses to reduce our non-pregnant OGTTs by one-fifth. If a 'risk threshold' for diabetic complications could be applied to HbA1c, it could replace the OGTT as a more pragmatic diagnostic/prognostic test.