Impact of mismatches in HbA 1c vs glucose values on the diagnostic classification of diabetes and prediabetes.

Aims: To determine whether HbA _1c mismatches (HbA _1c levels that are higher or lower than expected for the average glucose levels in different individuals) could lead to errors if diagnostic classification is based only on HbA _1c levels.
Methods: In a cross-sectional study, 3106 participants without known diabetes underwent a 75-g oral glucose tolerance test (fasting glucose and 2-h glucose) and a 50-g glucose challenge test (1-h glucose) on separate days. They were classified by oral glucose tolerance test results as having: normal glucose metabolism; prediabetes; or diabetes. Predicted HbA _1c was determined from the linear regression modelling the relationship between observed HbA _1c and average glucose (mean of fasting glucose and 2-h glucose from the oral glucose tolerance test, and 1-h glucose from the glucose challenge test) within oral glucose tolerance test groups. The haemoglobin glycation index was calculated as [observed - predicted HbA _1c ], and divided into low, intermediate and high haemoglobin glycation index mismatch tertiles.
Results: Those participants with higher mismatches were more likely to be black, to be men, to be older, and to have higher BMI (all P<0.001). Using oral glucose tolerance test criteria, the distribution of normal glucose metabolism, prediabetes and diabetes was similar across mismatch tertiles; however, using HbA _1c criteria, the participants with low mismatches were classified as 97% normal glucose metabolism, 3% prediabetes and 0% diabetes, i.e. mostly normal, while those with high mismatches were classified as 13% normal glucose metabolism, 77% prediabetes and 10% diabetes, i.e. mostly abnormal (P<0.001).
Conclusions: Measuring only HbA _1c could lead to under-diagnosis in people with low mismatches and over-diagnosis in those with high mismatches. Additional oral glucose tolerance tests and/or fasting glucose testing to complement HbA _1c in diagnostic classification should be performed in most individuals.
(Published 2019. This article is a U.S. Government work and is in the public domain in the USA.)