HbA1c has been recommended as a test option for the diagnosis of diabetes by both the American Diabetes Association (1) and the World Health Organization (WHO) (2). Both organizations advised that after a result consistent with the diagnosis of diabetes (≥6.5% [48 mmol/mol]), HbA1c levels should be repeated in asymptomatic patients within 2 weeks to rule out the rare occurrence of a sample being mislabelled. Consensus statements from expert bodies in the U.K. endorsed these recommendations, which were implemented locally in 2012 (3). Since adoption of HbA1c as the principal test for diabetes, we have received a steady trickle of queries about discordant HbA1c results from clinicians. In some cases, the discrepancy between two results obtained within the short repeat interval was striking.
To address this issue, we aimed to determine the magnitude and nature of variation seen in HbA1c results repeated within 14 days of an initial diagnosis of diabetes. We extracted anonymized HbA1c results (analyzed on the Tosoh G8 HPLC Analyzer; Tosoh Bioscience, South San Francisco, CA; coefficient of variation = 2.3% at 10% [86 mmol/mol]) from our laboratory information management system between April and June 2013. We identified cases where HbA1c was used to screen for diabetes, with results of ≥6.5% (48 mmol/mol), and a subsequent repeat HbA1c requested within 14 days (n = 188). We limited our analysis to uncomplicated cases of newly diagnosed type 2 diabetes, excluding patients with rapid-onset type 1 diabetes, drug-associated hyperglycemia (e.g., glucocorticoids), and acute illness.
The average interval between repeat testing was 7 days (range 1–14 days). In 63% of patients, the repeat HbA1c was lower than the first; in 20%, it was higher. The mean change in HbA1c was a fall of 0.2% (1.64 mmol/mol), with 12% of patients having a fall of at least 0.5% (5 mmol/mol) (Fig. 1). In 39% (73/188) of cases, the second HbA1c result was below the diagnostic threshold of 6.5% (48 mmol/mol). This change in HbA1c cannot be explained by analytical variation alone, at the group mean of 6.9% (52 mmol/mol) the analytical variation would account for a maximum of ±0.08% (0.6 mmol/mol) in an individual patient. There was no association between change in HbA1c and initial HbA1c (r = −0.047, P = 0.52). Similarly, there was no association between change in HbA1c and interval between testing (r = 0.01, P = 0.991). Inspection of clinical records in these cases did not reveal any factors that might explain the greater variability.
The results of this study indicate there is significant short-term variability of HbA1c. It is common for a repeat HbA1c to be below the threshold for diagnosis of diabetes after a first apparently diagnostic test, potentially leading to diagnostic confusion and uncertainty. The tendency for HbA1c to fall on repeat testing may reflect regression to the mean and/or genuine glycemic change over the preceding 14 days. This study cannot determine the reason for the fall in mean HbA1c. However, if regression to the mean had been the main mechanism, then we would expect higher initial HbA1c to be associated with greater fall on retesting; no such association was seen. Similarly, if true change in blood glucose levels was the main mechanism, we would expect a greater fall in HbA1c with longer interval between testing; again, no such association was seen. This association may be revealed in a study of larger numbers.
We believe a policy of retesting HbA1c after a single apparently diagnostic result seems to be justified, with nearly 40% of cases having a repeat test below the diagnostic threshold. Clinicians need to be aware that significant test-retest variation exists in HbA1c and is sometimes quite large.
Funding. T.J.M. is supported by the National Institute for Health Research CSO Postdoctoral Fellowship.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. T.J.M. and R.W. researched data and wrote and edited the manuscript. T.J.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
The views expressed are those of the author(s) and not necessarily those of the National Health Service, the National Institute for Health Research, or the Department of Health.