ORLANDO – Significant discordance exists between average glucose and hemoglobin A1c (HbA1c) measures in patients with certain comorbidities, according to findings from a retrospective chart review.
For example, there was a complete lack of correlation between average glucose (AG) and A1c measures in patients with advanced renal dysfunction and non-alcoholic fatty liver disease, (NAFLD) Jordan E. Perlman, MD, reported at the annual scientific sessions of the American Diabetes Association.
KatarzynaBialasiewicz/Thinkstock
Unweighted averages of self-monitored blood glucose (SMBG) and continuous glucose monitor (CGM) readings were calculated based on downloads from 1,039 patients who had been prescribed insulin for diabetes mellitus between January 2011 and October 2016 and who had a comorbid condition proven or hypothesized to invalidate A1c, including anemia, chronic kidney disease (CKD), abnormal liver function tests (LFTs), and NAFLD. Predicted AG was also derived from paired A1c using the equation established by the A1c Derived Average Glucose (ADAG) Study Group in a 2013 re-analysis of its 2008 report, which excluded patients with comorbidities.
The averages calculated using downloads were then compared with the averages derived using the ADAG equation to assess concordance.
“The term ‘discordant’ refers to averages that differ by more than 15%,” Dr. Perlman explained.
She and her colleagues found that CGM, compared with SMBG, decreased the odds of discordance after controlling for diabetes type (odds ratio, 0.39).
Additionally, having type 2 vs. type 1 diabetes mellitus increased the odds of discordance, as did renal dysfunction.
“Having CKD stage 3b or worse increases the odds of ADAG discordance (OR, 2.04),” she said. “The relationship demonstrates statistical significance at a P value of 0.004. Unfortunately, we did not have enough patients to analyze stage 4 or 5 CKD alone.”
Poor linear correlation was clearly seen between AG and A1c in patients with NAFLD, she noted.
“The relationship doesn’t reach statistical significance, but the odds ratio of 1.6 is difficult to ignore. The wide confidence interval (0.67-3.58) leads us to believe that this particular analysis is probably underpowered,” Dr. Perlman said.
Factors assessed and found to have no significant effect on ADAG discordance included abnormal LFTs, age, body mass index, and hemoglobin, including by gender.
“These important data suggest that any patient on insulin who comes to diabetes clinic has an automatic 33.5% chance of mismatch between their A1c and average glucose, and this is before you know anything else about them. To go a step further, it seems excluding comorbidities doesn’t really improve the percent discordance,” she said, adding that this suggests comorbidities have less impact than previously thought. “This makes us wonder if maybe there is a problem with our test and not the person having the test.”
It remains unclear what is acceptable in terms of discordance, Dr. Perlman said, noting that using ADAG to interpret A1c yields a wide range of estimated AG.
“Comorbidities alone do not explain this variation,” she said. “Clinicians should not rely on A1c alone to make treatment decisions because it is unclear when discordance gains clinical relevance.”
This study is limited by the retrospective study design and a number of factors, such as the difficulties of confirming or excluding comorbidities based on a single encounter and the limitless potential for unestablished confounders of A1c and AG, Dr. Perlman noted.