June 2009
Feature Story

Karen Titus

A quick quiz: A hemoglobin A1c value of nine percent represents which one of the following estimated average glucoses?

a) 126 mg/dL
b) 140 mg/dL
c) 154 mg/dL
d) 183 mg/dL
e) 212 mg/dL

Calculating the answer to that question is taking on new significance. A study published last year in Diabetes Care provides the strongest evidence yet of a linear relationship between HbA1c and average blood glucose levels. Numerous groups, including the American Diabetes Association and the AACC, are urging labs to provide eAG calculations alongside A1c measurements, arguing that it will make life easier for patients and clinicians and, they hope, improve diabetes management. Not since Garrison Keillor began touting the superior quality of Lake Wobegon’s children has such attention been settled on the word “average.”

The recommendation could well be one of the few no-brainers in medicine a simple, practically cost-free step laboratories can take to help patients and physicians. If labs are reporting A1cs, they ought to be reporting eAGs, says Gary Horowitz, MD, director of clinical chemistry at Beth Israel Deaconess Medical Center, Boston, and associate professor of pathology, Harvard Medical School. “The CAP can be heroes here,” he says. “Laboratories can start reporting it. Pathologists can start explaining it to their clinical colleagues and take the lead in this. Because a lot of clinicians don’t know about it yet.”

Providing eAGs regularly will put clinicians and patients on the same page, give them a common language—an Esperanto for diabetes—for what are arguably confusing numbers.

Estimated average glucose offers a novel way to describe the relationship between glucose and the degree and duration of hyperglycemia, says William Winter, MD. “For patients, it could very well be an easier way to understand that relationship,” says Dr. Winter, professor, Department of Pathology, Immunology and Laboratory Medicine, and professor, Department of Pediatrics, University of Florida College of Medicine, Gainesville.

Patients typically have to think about their glucose monitoring using two different scales. It’s an odd divide. Day to day, they’re guided by self-monitoring of capillary glucose concentrations, measured in the familiar milligrams per deciliter or millimoles per liter. “They know 150 or 200 [mg/dL] is bad,” says Dr. Horowitz, who chairs the CAP Chemistry Resource Committee.

Self-monitoring is supplemented with measurements of hemoglobin A1c, roughly two to four times a year. Those measurements are reported in percentages. And while many patients may understand that seven percent or below is the target, they may not fully appreciate that a bump up of one or two percentage points is cause for concern. The goal is an A1c value of less than seven percent. Nondiabetic patients typically have values between four and six percent, while those with diabetes range widely. Values between seven and eight percent are considered a gray area, while anything above eight percent calls for more aggressive management.

Are patients confused? David Sacks, MBChB, points to a study published in Annals of Clinical Biochemistry (Iqbal N, et al. 2008;45:504–507) revealing just how little patients know about A1c measurements. The majority of patients in the study (done at a U.K. hospital diabetes clinic) who were familiar with the term A1c thought an A1c of nine percent reflected an average blood glucose of 9 mmol/L (162 mg/dL) for the preceding three months; a small percentage thought the nine percent correlated with a blood glucose reading of 9 mmol/L on the same day.

In his own clinical practice, Dr. Winter found that telling diabetic children and their parents of changes in A1c values, in percentages, carried little weight. “And I’m sure it’s no different for adult patients,” he says. Tell a patient that his or her A1c moved from seven to eight percent, for example, and the typical response will be, “Gee, that’s not much of a change.” To someone not trained in numbers, one percent sounds small.

With the new equation, clinicians can provide eAGs in the same units everyone is already familiar with: mg/dL or mmol/L, making the discussion between patients, clinicians, and diabetes educators more straightforward.

If this is a no-brainer, who are the brains behind it? Plenty of credit goes to those who devised the most recent study, the A1c-Derived Average Glucose, or ADAG, study (Nathan DM, et al. Diabetes Care. 2008;31 1473–1478).

“It’s an amazing study,” says Dr. Horowitz, who was in the audience when the study was presented at an endocrine conference last fall. The delivery, by an endocrine fellow, was “depressing,” Dr. Horowitz recalls. “They had no clue what an incredible paper it was. So I felt compelled to put my two cents in.” Speaking up after the presentation, he told the audience, “They did everything right.”

As it turned out, one of the audience members was one of the study’s authors, who thanked Dr. Horowitz for his favorable blurb. “He said, ‘I was worried that people didn’t appreciate what went into this,’” Dr. Horowitz says.

The study looked at 507 subjects from 10 international centers: 268 with type 1 diabetes, 159 with type 2 diabetes, and 80 without diabetes. Over three months, some 2,700 glucose values were obtained by each subject. Average glucose was calculated by combining weighted results from a minimum of two days of continuous glucose monitoring done four times, with seven-point self-monitoring of finger-stick glucose done at least three times weekly.

Linear regression analysis between the A1c and average glucose values provided the tightest correlations, the researchers reported, allowing them to calculate an estimated average glucose for A1c values.

The equation is simple: 28.7 x A1c – 46.7 = eAG.

Dr. Horowitz says the study “established, without any doubt whatsoever, that the equation they have, which is only marginally different from the original one, holds up under all kinds of patient populations. So based on that data, it’s easy now to make the leap of faith that you can take the A1c and provide physicians and patients with a number that represents the mean blood glucose.”

So strongly has the ADA promoted this that there are those outside the United States who thought the ADAG moniker might be interpreted as “ADA glucoses.” “This is true,” says Dr. Sacks, laughing. “This was actually raised” when the study was first proposed, says Dr. Sacks, associate professor of pathology, Harvard Medical School, and director, clinical chemistry, Brigham and Women’s Hospital, Boston.

The equation is not the first to calculate eAG, as Dr. Horowitz notes. Until now, however, clinicians mostly have done their own conversions, relying on cards they carry around in their pockets that show estimated average glucoses for a range of A1cs.

And while most clinicians understand better than their patients that a change in A1c from seven to eight percent is significant, something may still get lost in translation, says Dr. Horowitz. “I think if you took 10 clinicians today and asked them what they think the estimated average glucose is for an A1c of eight or nine percent, they’ll be nowhere near the truth,” says Dr. Horowitz. Most, he suggests, will drift, erroneously, toward the low end of the scale. (Remember that quiz at the start of this story?)

“The truth is, most physicians are not confused by it so much, but we don’t have a good idea of what nine percent A1c means,” Dr. Horowitz says. “If you say to a physician that the A1c is nine percent, he’s going to say, ‘That’s elevated.’ But he doesn’t have a good sense of how bad that is.” Give him the number in mg/dL units, though, says Dr. Horowitz, and “He’s going to say, ‘Oh my God, that’s terrible.’”

The A1c-Derived Average Glucose study was ambitious from the start—a little on the order of world domination. “We had hoped to have an international cohort, so that when we established this regression equation, it would apply to the vast majority of people in the world,” says lead author David M. Nathan, MD, director of the Diabetes Center at Massachusetts General Hospital, Boston, and professor of medicine, Harvard Medical School.

As it turned out, several clinical centers in the study were unable to store samples properly. “We ended up having to exclude populations of Asians,” Dr. Nathan says. The researchers also had limited numbers of Africans and African-Americans, compared with initial plans for the study.

“So therefore we have somewhat less confidence in whether the equations we established apply to all people with diabetes,” says Dr. Nathan. “We don’t have any reason to think they don’t. But we didn’t have enough people of some of those racial groups to be absolutely confident that they apply to everyone.”

Dr. Winter reframes the matter as a question: Is there any scientific reason to believe that the relationship between glycated hemoglobin and estimated average glucose differs among ethnic groups or age groups? Glycation is a nonenzymatic process, he notes. “This is not going to be like variable metabolism of warfarin,” he says, with its variations in target enzymes and metabolizing enzymes, as well as other possible genetic variations. The nonenzymatic addition of glucose to hemoglobin, by comparison, is a pretty simple biological relationship.

Even so, Dr. Winter says he would like to see the relationship between A1c and eAG confirmed in other populations, including children, who were not part of the ADAG study. “It would be shortsighted to say we know all the story,” he says.

Other groups were left out of the study on purpose, including pregnant women and patients with unstable glucose control. “You can’t do everything all at once,” says Dr. Sacks. “You’re stuck between idealism and reality.”

“We wanted to maximize our ability to establish a stable regression equation,” adds Dr. Nathan.

A similar, though more limited, equation emerged from another study, the well-known Diabetes Control and Complications Trial, published in 1993, though that was not the main purpose of the study. One important difference between the DCCT equation and the ADAG one, says Dr. Sacks, is that the subjects in the more recent study had their glucoses measured through the night (2,400 measurements were done with continuous interstitial glucose monitoring; the other 300 measurements were done with finger-sticks). Since the nadir of glucose usually occurs early in the morning, that may explain why the eAG values obtained in the ADAG study were lower than those in the DCCT study.

These particulars of the ADAG study’s design and execution are worth noting—something to appreciate and perhaps further explore, says Darryl Erik Palmer-Toy, MD, PhD, director of chemistry for Regional Reference Laboratories, Kaiser Permanente, Southern California, North Hollywood. “These are, I think, important caveats in the interpretation of the average blood glucose result.”

Dr. Palmer-Toy says he supports adding eAG to laboratory reports, but wants to draw providers’ and patients’ attention to two matters. One is that monitoring glucose with continuous interstitial glucose monitoring or very frequent capillary blood glucose measurements doesn’t reflect the typical schedule of self-monitoring. Patients may look at the numbers derived from the study and wonder why they differ from the numbers they obtain from their own, less-frequent monitoring. “For their benefit, it’s worth pointing out,” suggests Dr. Palmer-Toy.

The second relates to the study’s reliance on patients with stable (though not necessarily good) glucose control—again, something not all patients might appreciate, says Dr. Palmer-Toy. “Patients who have not been in stable glucose control may find that their self-measurements of glucose differ significantly from those predicted by the estimated average glucose.”

The study was not designed to mimic how patients do self-monitoring. Rather, it was meant to capture average glucose levels as best as possible to establish the strongest equation. Earlier studies, says Dr. Nathan, suffered from too few glucose measurements, making it difficult for researchers to know for certain if mean blood glucose levels were being measured. The sheer volume of glucose measurements in the ADAG study should erase such doubts. “We had remarkable capture of mean glucose,” Dr. Nathan says. “This is about as good as you can get.”

Dr. Palmer-Toy’s observations—one could hardly call them criticisms—reflect a certain undirected spirit that wafts through discussions of reporting eAGs. No one is making rigid demands. At times this feels more like a conversation among Montessori teachers than medical professionals. Pathologists make it clear that it’s up to the clinicians to use the information—or not. And while there’s little hard data to show that reporting eAGs will improve patient care, the general feeling seems to be that it probably can’t hurt, either.

“I can’t see a downside to reporting this,” says Dr. Horowitz.

There are some limitations of the A1c assay, which are well recognized but need to be kept in mind, says Dr. Nathan.

An eAG is only as good as the underlying A1c measurements, as the ADAG study notes. If the A1c is misleading, the eAG will be, too. Anything that alters red cell life span will render the reference intervals for A1c inappropriate, Dr. Horowitz says.

Shorter-than-normal red cell survival can artifactually shorten the time of exposure of red cells to glucose, adds Dr. Winter, creating a lower A1c. Patients who have chronic hemolytic states, for example, may have a negative bias in their A1cs. On the other hand, patients may have a prolonged red cell life span. “That’s not very common, but if somebody’s had a splenectomy, their red cell life span could go up, giving you a positive bias in the A1c,” says Dr. Winter.

And a patient who lacks hemoglobin A—those with sickle cell disease, for example—won’t have A1c. “You have to have hemoglobin A to have A1c,” says Dr. Winter.

It rests with clinicians to know about these issues, he continues, though laboratorians need to understand what can interfere with A1c measurements. These issues should be familiar terrain for clinical endocrinologists and diabetologists, though general practitioners may be on rockier ground, says Dr. Winter.

Buried in all this is an interesting bit of lab history. “It turns out that the A1c values we’ve been reporting have been wrong,” says Dr. Horo­witz. “Totally wrong.”

About a dozen years ago, the National Glycohemoglobin Standardization Program standardized A1c measurements using a high-pressure liquid chromatography method; this was the method used in the DCCT study. Though it appeared stable, it was not a true reference method because it did not use pure HbA1c.

The International Federation of Clinical Chemistry subsequently developed a new method, which uses an enzyme to cut off the terminal six amino acids of the beta chain; the six amino acid fragments are then separated into nonglycated and glycated hemoglobin.

As it turns out, results related to the IFCC standard are two percent lower than those related to the DCCT method.

That news caused a stir. Dr. Horowitz recounts it: “So the scientists in the world said, ‘We’ve got to change everything. We can’t be reporting that the cut point is seven percent for good control; it’s going to be five percent.’” European experts were poised to make the shift about two years ago. “And the Americans said, ‘You’re nuts. You know, it’s taken us 20 years to get people on this page. You can’t do this. So there was a big fight, and then they agreed they’d delay it a little bit.”

Dr. Sacks was at the meeting where the ADAG study was proposed; he chairs the National Glycohemoglobin Standardization Program steering committee and is a member of the IFCC Working Group on HbA1c Standardisation. A compromise suggestion emerged: Why not report average blood glucose? Many clinicians, especially in the United States, were already translating hemoglobin A1c values into mean blood glucose on their own. Relying solely on data from the DCCT trial wasn’t feasible, however, since it was limited to patients with type 1 diabetes and U.S. patients. Hence ADAG.

A broad consensus statement from multiple diabetes groups encourages the reporting of eAG as well as the IFCC and NGSP numbers. Informally, there have also been loose allusions to giving eAG a broader role, which has caused some to wonder about the future role of A1c measurements. Dr. Nathan says, “The thing that has been most confusing to clinicians, and aroused some controversy, is the idea that we wanted to substitute the estimated average glucose calculation for the hemoglobin A1c.”

Read his lips. “That was never the intent,” Dr. Nathan says. “The intent was always to report both measures.” The eAG calculation, he says, is meant to bring “clinical sensibility to this topic.”

Dr. Horowitz, as his enthusiasm might suggest, plans to start reporting eAGs soon. He says he’ll do so without much, if any, discussion. “People don’t listen to you until they see something they don’t understand,” he explains, quickly adding that this approach is what works best for his particular situation; he’s not promoting this as an elastic waistband solution.

Dr. Winter suggests that pathologists meet with their endocrinologist colleagues. The vast majority, he suspects, will want labs to report eAGs, which will make it easier to gain the support of other clinicians who’ll also be receiving eAG numbers.

Beyond that, it’s not much more complicated. “It’s just a calculation,” Dr. Winter says. “We’re not talking about doing CT scans. This would have zero impact monetarily.”

Dr. Nathan’s laboratory has been reporting eAGs for more than a quarter of a century. Dr. Nathan developed his own regression equation years ago, though the lab is now using the new equation. His sense is that the information has been looked at as “quite helpful and hardly ever confusing.”

No one has translated those impressions into harder evidence, however, by comparing the reporting of A1c alone versus reporting A1c and eAG together. Dr. Nathan’s lab reports both. “It never occurred to us to report it any other way, frankly,” he says. “But it would be a reasonable study to do. As far as I know, no one has done it.”

“The feedback we get from patients—we run a large diabetes center—has always been positive,” Dr. Nathan says. “But, in fact, I can’t overplay that, because, in fact, we’ve never studied it.”

Dr. Sacks is aware of only one such study—the aforementioned Iqbal paper, which showed that teaching patients about the relationship between A1c and mean plasma glucose improves glycemic control, especially in those with poorly controlled diabetes and no prior knowledge of A1c.

That lack of evidence may be one reason some countries, including in Europe, are not encouraging the reporting of eAG values, says Randie Little, PhD, associate professor, University of Missouri School of Medicine, Columbia, and codirector of the Diabetes Diagnostic Laboratory. “The acceptance worldwide hasn’t been very good,” says Dr. Little, who is also coordinator of the NGSP network. To her thinking, “It’s not going to hurt anything. If some patients understand it better that way, it seems it can only be helpful and not harmful.”

Dr. Little’s lab has begun implementing eAG reporting. She sent her clinical colleagues information letting them know the lab wanted to start reporting it. “There was not a lot of big discussion,” she says, “probably because it’s going to be their choice whether to use it.”

William Roberts, MD, PhD, medical director of the automated core laboratory at ARUP Laboratories, Salt Lake City, is another experienced hand at reporting eAGs. “We’ve been reporting estimated average glucose at least five years, before that was even the accepted name,” says Dr. Roberts, who’s also professor of pathology, University of Utah. The lab started using an equation derived from the DCCT study, but recently switched to the ADAG one. He considers the differences between the two to be minor, and says clinicians have not commented on the change.

The lab started reporting the eAGs at the request of one client. That led to conversations with ARUP endocrinologists and the decision that reporting them more widely would be helpful. “If clinicians would like to use it, they can. If they choose not to use it, they can just ignore it.”

He’s not aware of any downsides. “I would think after five years, that if there were major problems, if we caused any problems for patients in particular, we would have heard about it. And that has not happened,” says Dr. Roberts.

If the eAG story sounds a little familiar, it’s because it’s been told before, though with a different cast of characters.

Pathologists may be reminded of the push to report estimated glomerular filtration rate, or eGFR. That’s the other parameter frequently invoked as labs think about reporting eAGs, and it may have some fleeing for the exits.

“That was a horrific equation to implement,” says Dr. Horowitz. It took its toll on LISs because of the multiple variables involved. By comparison, eAGs could be done by Shakers. Every A1c gets converted the same way, with a simple, linear equation.

Nor will the information deluge overwhelm labs the way eGFRs have, Dr. Horowitz says. Some patients have multiple creatinines measured daily. “Some labs, like ours, didn’t want to report an eGFR with every one of them.” Patients with diabetes will need eAGs reported only two to four times a year.

What about from an IS standpoint? “They’ll make a big deal out of it, but this is as easy as it gets,” says Dr. Horowitz.

“It’s trivial,” says one pathologist, before backing down slightly.

“Don’t use ‘trivial’—that makes the IT people crazy.”

Dr. Horowitz says the eGFR, with its complex programming requirements, may have caught IT people unawares. “That’s why it may be worth saying this is nothing like the eGFR. The only thing it shares with estimated GFR is the little ‘e,’” says Dr. Horowitz.

Here’s where the answer to the opening quiz would normally be provided. But Dr. Horowitz argues against doing so.

“Give them the equation instead,” he says. “That’s how we’ll need to educate our clinicians.”

Related Links

• Glucose Calculator
  (PDF, 690 KB)