Screening for diabetes and pre-diabetes with proposed A1C-based diagnostic criteria.
Level 3 - non-randomized controlled study
Cross-sectional diagnostic accuracy study comparing screening thresholds across three population datasets
PubMed 20639452 · doi:10.2337/dc10-0433
What was done
The authors evaluated the diagnostic accuracy of proposed A1C-based screening thresholds for diabetes (A1C ≥ 6.5%) and pre-diabetes (International Expert Committee [IEC]: 6.0–6.4%; American Diabetes Association [ADA]: 5.7–6.4%) compared with standard 75-g oral glucose tolerance tests (OGTT). Data were pooled from non-Hispanic white and black adults without known diabetes across three cohorts: the Screening for Impaired Glucose Tolerance study (n = 1,581), NHANES III (n = 2,014), and NHANES 2005–2006 (n = 1,111), totaling 4,706 participants.
What was found
By OGTT criteria, 35.8% of participants had pre-diabetes and 5.2% had diabetes. A1C yielded receiver operating characteristic (ROC) curve areas of 0.79–0.83 for diabetes, but ≤ 0.70 for dysglycemia or pre-diabetes. The proposed A1C criteria missed 70% of OGTT-diagnosed diabetes, 71–84% of dysglycemia, and 82–94% of pre-diabetes. Compared to IEC criteria, ADA criteria produced fewer false negatives but more false positives. False-positive results were more frequent in black participants, whereas false negatives were more frequent in white participants. Applying these estimates to NHANES 2005–2006 data suggested that A1C screening alone would miss approximately 5.9 million non-Hispanic US adults with unrecognized diabetes and 43–52 million with pre-diabetes.
Why it matters
This study demonstrates that relying exclusively on A1C cutoffs for screening misses a large majority of individuals with glycemic abnormalities identified by OGTT and introduces systematic racial discrepancies.
Limits
The study used a single OGTT as the diagnostic gold standard without repeat confirmatory testing. Clinical outcomes (such as long-term microvascular or macrovascular complications) were not evaluated to determine which testing modality better predicts clinical risk. The sample was limited to non-Hispanic white and black adults, precluding generalization to other racial and ethnic groups.
Cited by
- supports According to US diagnostic thresholds, an HbA1c between 5.7% and 6.4% indicates prediabetes, and an HbA1c above 6.5% indicates type 2 diabetes.