Pitfalls in the use of HbA_1c as a diagnostic test

In their correspondence (A conundrum addressed: The prognostic value of HbA_1c. Nat. Rev. Endocrinol. 7, doi:10.1038/nrendo.2010.126-c1), Selvin and Brancati raise four issues regarding my Perspectives article (Pitfalls in the use of HbA_1c as a diagnostic test: the ethnic conundrum. Nat. Rev. Endocrinol. 6, 589–593 (2010))¹ on the diagnostic use of HbA_1c.

First, they question the clinical significance of genetic contribution to the variation in HbA_1c levels. In addition to the paper by Snieder et al.,² several reports document differential genetic influences on HbA_1c and fasting glucose values that would make it clinically precarious to use HbA_1c alone for the diagnosis of diabetes mellitus or prediabetes.^3,4,5,6,7 Among patients with type 1 diabetes mellitus enrolled in the Diabetes Control and Complications Trial, genome-wide association studies identified a major locus (near SORCS1) that was associated with both HbA_1c and glucose (P = 7 × 10⁻¹⁰ for the association with HbA_1c, P = 2 × 10⁻⁵ for the association with mean glucose).³

Studies in nondiabetic individuals provide a less perturbed internal environment for assessing independent genetic contributions to the variations in HbA_1c and fasting glucose values.^4,5,6,7 Such studies have indicated that heritable factors⁴ and genetic loci^5,6,7 for HbA_1c exist that are not necessarily shared by fasting and nonfasting blood glucose values. The clinical significance of the nonoverlapping genetic loci for HbA_1c and blood glucose may be moot in a given patient receiving treatment for established diabetes mellitus. However, the genetic contribution to the discordance between HbA_1c and mean blood glucose can be clinically significant if HbA_1c is used as the sole diagnostic test for diabetes mellitus or prediabetes.^1,8 The use of HbA_1c as the primary criterion for diagnosis assumes a degree of universal concordance between HbA_1c and mean blood glucose values that is unsupported by current evidence.

Second, Selvin and Brancati suggest that the effects of age and ethnicity on the relationship between HbA_1c and blood glucose values could be owing to inadequate sampling of postprandial glucose levels and differences in dietary practices among older individuals and those from ethnic minorities—an interesting suggestion. However, dietary practices in the US overlap to such a degree⁹ that it is doubtful whether any systematic differences can account for the ethnic disparities in HbA_1c.

Third, the standardization of HbA_1c methodology does represent progress in the right direction. Unfortunately, such standardization does not obviate the genetic or many of the biological concerns about the diagnostic use of HbA_1c.¹ Finally, without question, the collection of multiple glucose data points would increase the validity of HbA_1c comparisons. In this regard, data from continuous glucose monitoring also showed wide variability in HbA_1c–glucose correlation among individuals.¹⁰

To be clear, I do not dispute the role of HbA_1c as a 'gold' standard for monitoring glycemic control and predicting microvascular events among patients with diabetes mellitus. However, the sole or preferential use of HbA_1c for diagnosis is fraught with biological and demographic pitfalls that create genuine clinical concern^1,8 regarding underdiagnosis or overdiagnosis of a substantial number of individuals. In recognition of these issues, clinicians should utilize direct glucose measurement to validate abnormal HbA_1c results when screening patients for diabetes mellitus or prediabetes.