College of American Pathologists
Printable Version

  Too far, too fast on ICU TGC?


CAP Today




March 2011
Feature Story

Anne Paxton

Ten years ago, a study conducted in Leuven, Belgium, took U.S. hospital intensive care units by storm. Reported in the New England Journal of Medicine, the trial, led by Greet Van den Berghe, MD, PhD, found that when adult surgical ICU patients’ blood glucose was maintained between 80 and 110 mg/dL by means of an insulin infusion, it reduced ICU mortality by 42 percent and in-hospital mortality by 34 percent (2001;345:1359–1367). Implementing an ICU protocol of tight glycemic control, the study concluded, also decreased the incidence of acute renal failure, septicemia, and critical illness polyneuropathy.

These conclusions galvanized hospital ICUs. In the wake of the study, “Intensive care units in the U.S. went from almost nobody using insulin to using it on almost every patient,” says Stanley A. Nasraway, MD, director of the surgical intensive care units at Tufts Medical Center, Boston. As one indicator of the impact on laboratory testing, Barnes Jewish Hospital in St Louis, Mo., doubled the number of glucose strips it consumes. “In 2001, we used 250,000 strips a year and we’re at almost half a million strips a year now, just with essentially the same number of beds,” says Mitchell G. Scott, PhD, professor of pathology and immunology and co-medical director of clinical chemistry at Washington University School of Medicine.

But as tight glycemic control (TGC) became more and more the standard for adult patients across the critical care landscape in the U.S., there were ripples of concern. Efforts to duplicate the Leuven study did not all pan out. One of Dr. Van den Berghe’s own later studies in medical ICU patients showed only morbidity benefits. Then, two large-scale studies in 2008 and 2009 suggested that, in fact, very tight glycemic control may actually put patients at greater risk of death and complications from hypoglycemia.

Alarmed, many hospitals, clinicians, laboratories, and standard-setting organizations decided to dial back on tight glycemic control and take stock. Did the medical community embrace the protocol too quickly based on a single trial? Or could there have been methodological or design flaws that kept later research from confirming the initial study? Questions like these have turned a spotlight on a central concern of the clinical laboratory: the critical importance of accurate blood glucose testing.

Few protocols have been adopted so widely and so rapidly as tight glycemic control. “When Dr. Van den Berghe published her landmark paper in 2001, showing intensive insulin therapy in the surgical ICU significantly improved outcomes, the results were so dramatic that almost immediately, without guidelines or anything, practices changed and this became the standard of care,” says David B. Sacks, MBChB, chief of clinical chemistry in the National Institutes of Health Department of Laboratory Medicine and a member of the CAP Chemistry Resource Committee. ”It was implemented in most hospitals in the U.S. and in much of the world.”

Although a 2006 Van den Berghe study of the protocol in the medical ICU (2006;354:449–461) showed no overall mortality benefit from intensive glucose control, it still reported that TGC brought improvement in morbidity and, in a subgroup requiring critical care for three or more days, also mortality. That year, TGC was touted in a pharmacoeconomic analysis as “one of the most cost-effective inpatient interventions ever studied.” But later studies started to cast doubt.

The protocol’s fortunes began to falter, first, with publication of a meta-analysis in the Journal of the American Medical Association in 2008 (Wiener RS, et al. 2008;300:933–944). Looking at 29 randomized trials, the researchers found no difference in mortality between patients assigned to tight glucose control and those assigned to standard glucose control.

Even more sobering was a large multinational study called the Normoglycemia in Intensive Care Evaluation–Survival Using Glucose Algorithm Regulation (NICE-SUGAR) trial (N Engl J Med. 2009; 360:1283–1297). This study tested the hypothesis that intensive glucose control reduces mortality at 90 days. But unexpectedly, it showed that people with intensive glycemic control had more frequent severe hypoglycemia as well as higher mortality.

It was a disquieting finding, because while hyperglycemia is not good, “it’s also not life-threatening—and hypoglycemia is life-threatening,” Dr. Scott points out. The meta-analysis and the NICE-SUGAR trial “really made people stop and look,” Dr. Sacks says. Based on the 2001 paper, the target cutoffs for hospitalized patients had been recommended to be 80 to 110, and there was consensus in favor of intensive gly­ce­mic control for inpatients (Diabetes Care. 2006; 29:1955–1962). But by 2009, in the hope of avoiding severe hypoglycemia, the American College of Endocrinology and the American Diabetes Association increased the target range to 140 to 180.

Disagreement among clinicians continues. Early this year, the Endocrine Society announced that it concurs with new clinical practice guidelines issued the same month by the American College of Physicians, which recommends against tight glycemic control both in the ICU and outside the ICU. But the Endocrine Society thought the ACP-recommended target glucose range for ICU patients of 140–200 was too high. It supports a lower upper target of 180, and said a “more stringent target of 110–140” may be appropriate in some special populations such as cardiac surgery patients. It also called for further research using new blood glucose monitoring technology in hospitalized patients.

As the standards have been revised, many hospitals have backed away from their earlier embrace of TGC. As Dr. Nasraway puts it, “Everybody’s voting with their feet.”

For example, the Yale-New Haven Hospital initially brought an infusion protocol to the hospital with a range of 100 to 140, says Silvio E. Inzucchi, MD, clinical director, Section of Endocrinology, and director of the Yale Diabetes Center. “We were a little dubious that we needed to get as low as 80 to 110 to incur the benefit, but we later tightened the range to 90 to 120 because it seemed that’s where the guidelines were evolving toward.” More recently, however, responding to the findings of the NICE-SUGAR study, new consensus statements, and smaller studies indicating the initial Leuven findings weren’t reproducible at different centers, Yale-New Haven cut back to a TGC range of 120 to 160.

Nevertheless, during the protocol’s heyday many hospitals across the country had started to extend TGC to subcritical patients, and have continued to do so, Dr. Sacks says. “The 2001 paper really did raise everybody’s general awareness that hyperglycemia is bad even if you don’t have diabetes,” he points out. As a result, many clinicians are monitoring people even outside the ICU a lot more carefully and controlling their glucose where necessary.

“Just about everyone in the field feels we can’t go back to the old days, where hyperglycemia was essentially ignored,” Dr. Inzucchi says. “But the pendulum may indeed have swung too far with overly aggressive insulin therapy. That may not necessarily be good for some patients.” NICE-SUGAR would suggest that once you’re under 180 you don’t incur added risk. “We know the Belgian studies done five or 10 years ago showed that taking patients from a very high glucose to a normal glucose has benefits in terms of lower morbidity and mortality. What we don’t know is whether a reasonable glucose in the mid-100s is necessarily any worse than super-stringent control.” The answer, he suspects, lies somewhere between the two schools of thought.

The authors of the JAMA meta-analysis undertook their study to address research findings that were failing to support those of the Belgian study. As a pulmonary and critical care physician practicing in the ICU, says lead author Renda Soylemez Wiener, MD, MPH, “I was aware of the Van den Berghe study which was conducted specifically in the surgical ICU, and I was struck by how quickly it became the standard for all critical care adults. But other studies had conflicting results.”

The key conclusion that came out of the meta-analysis, as well as subsequent randomized trials, says Dr. Wiener, who is now assistant professor of medicine, Boston University School of Medicine, Center for Health Quality, Outcomes and Economic Research, is this: “We really haven’t been able to show a benefit from TGC down to the level of 150—and even less down to the level of 110. At this point the bulk of the evidence suggests that the harm outweighs any type of benefit that was initially observed in the Van den Berghe study.” It’s not clear, she says, that TGC is really worth taking the risk of hypoglycemia.

Dr. Wiener’s institution does not now have a TGC protocol in use for all ICU patients. “We do it where the glucose is kind of out of control, over 180 or 200. Then the goal would be 140 to 180.” From her perspective, the controversy over TGC reached its height in 2008 and has now largely passed. “The questions that remain are about what really is the optimal goal for glycemic control in the ICU, and would there be a benefit if we could really be sure we had less variability in blood glucose.”

Without a doubt, the findings of the JAMA meta-analysis and the NICE-SUGAR study have reshaped the standard of care. But many researchers and clinicians still find the original Belgian study persuasive. How can the differences be explained, and which studies’ findings are closer to the truth?

Dr. Van den Berghe, for one, says that the doubts raised have not been enough to change her thinking on tight glycemic control. “We have shown a consistent decrease in complications in three different ICU populations, and we’ve shown cost savings, and that’s why we are continuing to follow this protocol.” She questions whether the later studies are repeat studies. “At the time of the NICE-SUGAR study, people were treating glucose at 140 to 180 [compared to tolerating much higher levels in the control arm of the Belgian study] because they were thinking about glucose much more. Therefore, the control group started off with this level,” unlike her original study when people were not under a protocol, says Dr. Van den Berghe, who is director of the Department of Intensive Care Medicine, Catholic University of Leuven.

With the multi-center NICE-SUGAR trial, she notes, that was not the only issue. “It appeared that the intermediate group was safer than the strict group, which made people conclude, well, let’s go for the intermediate range.” Partly as a result, the range accepted now is 140 to 180 for ICU patients. “That’s probably okay in the real world, but scientifically speaking, there is not one randomized controlled trial saying this intermediate range is good for anything—except exposing patients to the risks that come with insulin treatment. Our studies have shown a major benefit that comes from perfectly controlled glucose levels, and if people are not able to do this, then whatever target they set will impose a risk. So 140 to 180 is only an expert opinion, an educated guess, as compared to no treatment.”

Of the JAMA meta-analysis, Dr. Van den Berghe says all the data were pooled without enough consideration for why findings might vary. “They did a meta-analysis comparing studies as if the control arm doesn’t matter.” Moreover, she believes, the meta-analysis made a possibly more basic error by leaving out differences in how blood glucose is measured in Europe as compared with the United States, “as if the meters didn’t matter.” In this view, she is joined by many U.S. researchers.

William E. Winter, MD, professor in the Department of Pathology, Immunology, and Laboratory Medicine at the University of Florida College of Medicine, agrees with Dr. Van den Berghe that the untreated group in the Belgian study may have had higher glucoses than the untreated group in the NICE-SUGAR study. He is also skeptical that the NICE-SUGAR trial and JAMA meta-analysis undercut Dr. Van den Berghe’s findings as much as is claimed. “If you look at the field, an important question is, did the other studies duplicate the Belgian study in either design or execution? I don’t think they did,” says Dr. Winter. “Clearly the other analyses don’t show the same benefit. But is it fair to put together studies as dissimilar as these were?”

One difference between the original Van den Bergh study and subsequent ones was consistency of practice, Dr. Scott points out. Researchers from the Belgian study have said that all of the nurses who worked in the Leuven ICUs were devoted to the TGC protocol, understood it well, and made decisions on their own because they had been carrying out the protocol for so long. “It’s not obvious from the study, but I don’t know that other groups have that consistent a routine.”

Another criticism of Dr. Van den Berghe’s initial study was that her hospital “overfed” her patients in a way that American ICUs do not. The idea is that perhaps her findings were not about insulin bringing high sugars down into the normal range but instead a function of overfeeding people and causing hyperglycemia, Tufts’ Dr. Nasraway says. As Dr. Van den Berghe explains it, “Scandinavian pioneers advocated early parenteral feeding for ICU patients not tolerating enteral feeding and it starts within 48 hours in Europe, but under U.S. guidelines patients can tolerate virtual fasting for one week.”

Feeding practices vary in the U.S., but Dr. Nasraway notes that at Tufts, if the patient is catabolic and has suffered burns or massive trauma, “you’re probably going to get fed in the first 24 hours. If we can do that enterally, great, but it’s more likely we’ll start you on TPN [total parenteral nutrition]. A lot of people in the ICU have had operations on the gut, and their gut’s not working anyways, so we have to use another modality.”

But he notes that while feeding can aggravate hyperglycemia, patients even if they don’t get a single calorie can be severely hyperglycemic from stress illness. In addition, “Since her 2001 study, Dr. Van den Berghe has done another large ICU study, published in the NEJM in 2006, in which the glucose intake was better regulated. Other people have since duplicated those findings and confirmed that this is not a feeding question.”

Dr. Nasraway thinks the Van den Berghe study is sound research. “What we’ve learned is not that Dr. Van den Berghe was wrong or off-balance. We’ve learned it’s very hard to export what she conducted in Leuven, Belgium, to everyone else.”

Along those lines, the most serious objection raised to the JAMA 2008 meta-analysis and the NICE-SUGAR trial relates to the glucose measurement instruments employed in the respective studies. As is commonly done in Europe, Dr. Van den Berghe used blood gas analyzers to measure glucose in patients’ arterial blood. By contrast, in the JAMA meta-analysis, Dr. Scott notes, “In only seven or eight of the 29 studies could you even tell how the glucose was measured. You don’t know if they were using home meters, blood gas analyzers—you just don’t know. And the NICE-SUGAR study basically says that glucose was measured either with blood gas instruments or point-of-care meters using arterial, venous, or capillary blood.” In other words, pretty much the gamut of devices.

It’s a feeling of false safety, Dr. Van den Berghe believes, to say, “I gave insulin to keep the patient’s glucose under control, but I don’t have to worry about the accuracy of the measurement.” Dr. Scott does not consider the use of different measurement instruments a trivial matter either. “Keep in mind that glucose meters are not cleared for diagnoses of diabetes, so if something is not good enough to diagnose diabetes, how can it be good enough to make dosing decisions to administer a dangerous drug?”

It was not a matter of deliberate planning that POC glucose meters came to be so prevalent in U.S. ICUs, Dr. Scott points out. “In the late 1980s, diabetes patients started bringing their own meters into the hospitals, people found it convenient that they didn’t have to wait for lab results, manufacturers found a new market, and it just sort of happened.” But the meters have evolved into doing things they were never meant to do and were never cleared to do, Dr. Scott says.

Current regulations allow variations of as much as 20 percent in high glucose values, he says. “This variation will be increased due to hematocrit effects, and when a patient’s glucose is measured with a glucose meter and they’re anemic, their glucose is overestimated. Well, most patients in the ICU are anemic, so using the meters may contribute to an increase of hypoglycemia with tight glycemic control.” Newer meters can overcome this flaw, but most of the meters in the field cannot, he says.

The instruments’ potential bias means, Dr. Sacks explains, that if the true glucose level is 100, under the current guidelines the meter can read anywhere from 80–120. “And that’s only 95 percent of the time. The other five percent of the time it could be three—or 3,000.” With that kind of range, it’s difficult to coordinate with a tight glycemic control protocol, he adds.

Another important issue should be taken into account, says Dr. Sacks, who characterizes the use of glucose meters in the ICU as analogous to off-label uses: It’s not only the instrument itself but also the source of the blood sample that affect the result. “Patients in the ICU are really sick. So it’s not the same as somebody with diabetes who sticks his or her finger, because ICU patients have lots of other things going on that affect their peripheral circulation. When you have low blood pressure, hypotensive shock, the first way the body adapts to that is by shutting down peripheral circulation so there is less blood to the extremities. When you do a fingerstick, there is reduced oxygenation of the blood. So it’s not as accurate, period.”

There were also important differences in the way the protocols were executed, Dr. Winter notes. For example, in the Belgian study, the ICU used basal insulin infusion, whereas in the NICE-SUGAR study, the centers used basal insulin plus a bolus through injection. “It’s not that the quality of research is not as good, but we have to ask what about the quality of the study design? Can we lump together the Belgian and NICE-SUGAR studies and say these really can be put into a meta-analysis?”

Given the questions raised about variations in blood glucose measurement practices, are multi-center trials like NICE-SUGAR useful in answering such important treatment questions? “They are worth doing, but one has to be very, very careful in the design of the study,” Dr. Sacks cautions. As one example of careful design, he cites the HAPO (Hyperglycemia and Adverse Pregnancy Outcome) study of gestational diabetes. It followed women’s pregnancy outcomes prospectively, then looked at glucose results to identify new cutoffs for diagnosing gestational diabetes.

“The samples were handled very, very carefully and all the glucose measurements were done in one place. So even if you have a study done in lots and lots of different centers, by careful design one can eliminate many of the problems. But it’s absolutely critical that you have somebody who understands the lab issues and can eliminate them before the study starts.”

In Dr. Scott’s view, that doesn’t mean a randomized trial to compare TGC with glucose meters and TGC with blood gas analyzers is in the cards. “They have blood gas analyzers in their ICUs in Belgium,” he notes. While it wouldn’t be impossible to do that here, it would be too expensive to do it overnight. And it will be a very long wait before proof is in that it would actually solve the measurement problem, he believes. “You have small effects in very large numbers of people. So you’d literally have to randomize the patients in your ICUs to have half get testing done with meters versus half with blood gas analyzers. Imagine trying to get three shifts of nurses to always make sure that was done correctly for thousands of patients.”

It’s for that reason that most of the data questioning the glucose meters are coming out of modeling studies, he adds. Researchers generally agree that an authentic randomized controlled trial would take way too many patients. “It might also be considered unethical to use the glucose meters if the blood gas analyzers were also there.” However, he notes that buried deep within the NICE-SUGAR study, the measurement method for each ICU is probably specified. “In the database I would guess it could be found out how all those glucoses were measured.” So that might be an avenue for continued research, he suggests.

In the meantime, the FDA has indicated it has concerns about the accuracy of POC glucose meters and plans to tighten standards. Courtney Harper, PhD, director of the FDA’s Division of Chemistry and Toxicology Devices, Center for Devices and Radiological Health, says that considerable feedback from the clinical community indicates that the meters are just not as accurate as they need to be to achieve tight glycemic control.

“We are trying to figure out what is the right standard, and the community is working really hard on that right now.” The FDA is part of the International Organization for Standardization, or ISO, and is working in conjunction with it as part of its normal standards development process, she says.

In answer to concerns that hospitals would have a hard time replacing glucose meters, Dr. Harper says, “We’re not aiming to have glucose meters disappear. They are essential products in all hospitals. The FDA is definitely not interested in disrupting access of hospitals to glucose meters. We are interested in making sure labeling is accurate in terms of how they work.” While the agency does not have a timeline for new glucose meter standards, it is aiming to issue them “as soon as possible.”

However, Dr. Winter is concerned that the quality of POC instruments, especially in the ICU patient settings, is quite different from that of the central lab instruments. “My personal opinion is, why should we provide a quality of testing, even to noncritical care patients at the point of care, that has less quality than the central lab? A 20 percent positive or negative error—we wouldn’t tolerate anywhere near that amount of error in a central lab result.”

Of course, it’s not only quality but also timeliness that is necessary in glycemic control, and this is the source of the push for POC devices at the bedside, he agrees. “But why don’t we take the technology from the central lab that could be placed at the point of care, whether it’s a blood gas analyzer or a different device?” He thinks this could be achievable with modest amounts of expenditures in the ICU. As Dr. Sacks puts it, “Giving an accurate result in 10 minutes is better than an inaccurate one in two.”

Nothing has yet been made official, but Dr. Sacks, who is on the ISO committee and is chairing the CLSI committee considering new standards, says that both organizations will probably tighten the accuracy requirements for glucose meters. “The FDA will likely use these new criteria for determining whether meters should be approved or not.”

Could the FDA theoretically disapprove of the meters that are being used now for glycemic control? Dr. Scott thinks the FDA will adopt a go-slow strategy. “I think there will have to be some sort of phase-in, simply because you just can’t take away the meters from every institution. If all of a sudden 75 percent of the meters on the market just disappeared, I don’t know that the health care world could adjust. Maybe they shouldn’t be allowed in critical care settings. But they still have their place.”

A few institutions implementing a TGC protocol have opted to accept the limitations of POC glucose meters, while trying to compensate for those limitations. Tufts University Medical Center is one example. The intensive insulin therapy program there pre-dates Dr. Van den Berghe’s study. In 1998, Tufts began the program simply to protocolize and make uniform the administration of insulin, says Dr. Nasraway. “We had nothing to guide us at that time as to what would be an appropriate glucose for patients with stress hyperglycemia for critical illness or injury, and we arbitrarily chose 180 as our limit.”

After the Van den Berghe study came out in 2001, “we decided to ratchet that down, and we’ve been at 95 to 135 for this decade.” But the program uses somewhat unusual glucometry. “We now have a glucometer hardwired at each bedside and linked to the computer system for the hospital. Additionally, for the last two and a half years the dosing of insulin has been determined by insulin dosing software, the GlucoStabilizer program.”

Tufts is one of very few institutions routinely using software for this purpose, but he considers it a way to compensate for the inaccuracy of the handheld glucose meters. “We know the handheld meters aren’t accurate enough. The software makes up for some of that the way a thermostat works to control heat by constantly sensing its environment.”

Under the typical “paper protocol,” he explains, a blood sugar of X means an insulin dose of Y. “With the software, you input the glucose and the computer says, ‘When I look at the last 20 glucoses and combine them with the last 20 insulin maneuvers, this is what I want to do now for insulin, and this is exactly when I want you to draw the next blood sugar.’ So it’s much more precise at integrating information.”

A 2009 paper in Critical Care (Juneja R, et al. 2009;13:R163), Dr. Nasraway says, reported that the Tufts software-guided insulin dosing, compared with all other facilities’ dosing based on paper protocols, registered the least amount of documented hypoglycemia. The program forces more blood sugar checks because it sounds an alarm if a scheduled glucose is not inputted. “It’s very clear that the more you measure, the safer it is, the more likely you’ll get into your target range, and the more likely you’ll stay there.”

In fact, Dr. Nasraway’s view is that extending similar glycemic control to subcritical patients would make sense. It should be achieved throughout the continuum of illness, he says, adding: “The degree of control can be loosened given the resources available on surgical and medical wards.”

Dr. Van den Berghe says there has been no study on the use of the TGC protocol on subcritical patients. ”There are a few small studies where people have tested different protocols such as sliding scales and different types of insulins in patients outside the ICU, but there are no outcome studies at all. From my point of view, the nursing staff is not prepared for that and we have not advocated it. We need better tools, and they don’t exist.”

But Dr. Nasraway sees current protocols as just the tip of the iceberg. “In my ICU, the average blood sugar is about 119, but when I send patients out to the surgical floor, the average blood sugar is around 165, and on the medical floor, 185 or 198.” These degrees of severe hyperglycemia are well outside the ADA and the Endocrine Society recommended ranges of glycemic control for patients on the wards in hospitals, he says. He thinks intensive insulin monitoring has a definite place in the treatment of continued illness. “We need better and more effective glucometry, and more effective monitoring, because we’re going to be doing more of this, not less of this.”

That doesn’t mean there’s any consensus on whether glucose is a marker or causes increased morbidity and mortality, he says. “I don’t think that’s been decided. This is a very nascent and early field. But what has been decided is that severe hyperglycemia is bad and severe hypoglycemia is bad.”

“Going forward, we need to argue about whether it’s everybody with a heartbeat in the hospital that we need to worry about—or are some selected cases more at risk than others?” Much more research on hyperglycemia and treatment of it based on outcome determinants is needed, Dr. Nasraway believes. “And I think, with that, the field will evolve.”

Dr. Van den Berghe hopes that researchers and clinicians can continue to find common ground on the issue of TGC protocols in the ICU. She was invited to participate in a debate with a researcher representing the NICE-SUGAR trial. “He’s considered an opponent, but he’s contacted me saying we should not debate this. We should highlight how we are doing our protocols, rather than whether we should or should not do them.” Education on the limitations of the glucose meters, she adds, should be a particular focus for everyone.

Medicine abounds with examples of standards of care that fell out of favor, says Dr. Inzucchi. “We never seem to get it right the first time.” But with glycemic control, he cautions, “We shouldn’t throw the baby out with the bathwater. The importance of good glucose control is doubted by few experts. The main controversy is just how tight it should be.”

Anne Paxton is a writer in Seattle. Dr. Van den Berghe and other researchers will address “Glycemic Control in the Hospital: Evidence, Issues, and Future Directions” at the Arnold O. Beckman Conference on April 12 in San Diego. CAP members can register at the AACC member rate. For details, see