ICUs tighten belts on
  blood glucose levels

February 2005
Cover Story

As George and Ira Gershwin famously wrote, "Love is sweeping the country!" including "All the sexes from Maine to Texas." True, "tight glycemic control in critically ill patients" doesn’t have the instant emotional appeal that love evokes, but it too has been sweeping the country, becoming routine in intensive care units from Oregon to Connecticut and many places in between. And whereas love is often blamed in song with making people feel blue and brokenhearted, tight glycemic control actually makes people feel better and live longer, with some of its strongest health benefits seen in patients with cardiac conditions. Most important, growing evidence indicates that tight glycemic control improves morbidity and mortality not just in diabetics, but in all critically ill patients, setting the stage for comprehensive blood glucose protocols to be adopted in ICUs.

"Right now research shows that survival increases substantially with tight glucose control in the ICU," says Mark D. Siegel, MD, associate professor of medicine in the pulmonary and critical care section at Yale University School of Medicine and director of the medical intensive care unit, Yale New Haven Hospital. "In addition, there seem to be fewer complications." Most interesting of all, he says, is that the majority of patients requiring tight glucose control are not diabetics. Non-diabetics can have trouble with glucose control when they become critically ill. "This represents a dramatic shift in how we practice in just a couple of years," Dr. Siegel says. He credits several publications for this shift, most notably a 2001 article that, he says, has already become a "classic."

The lead author on that article, and principal investigator on the study that it reported, was a Belgian physician, Greet Van den Berghe, MD, PhD, professor of medicine at Catholic University of Leuven and director of the Department of Intensive Care Medicine at University Hospital Gasthuisberg. Dr. Van den Berghe’s study remains the only prospective randomized trial to evaluate tight glycemic control in all comers—on admission to the surgical ICU, only 13 percent of these patients, all of whom were on mechanical ventilation, had a history of diabetes and only five percent were receiving insulin. Dr. Van den Berghe showed that maintaining blood glucose between 80 and 110 mg/dL with insulin infusions reduced ICU mortality by more than 40 percent and in-hospital mortality by 34 percent. It also decreased the incidence of acute renal failure, episodes of septicemia, and critical illness polyneuropathy (Van den Berghe G, et al. N Engl J Med. 2001;345:1359-1367).

Consistent with her dedication to rigorous evidence, Dr. Van den Berghe is hesitant to extrapolate the benefits of tight glycemic control achieved in the controlled study to hospitalized patients outside the surgical ICU, as some have suggested. "I don’t think there is evidence for that," she told CAP TODAY. "I am quite convinced that it would be good to have tight glycemic control even on regular wards, but there is no strict proof." She and others are now conducting controlled trials of tight glycemic control in the medical ICU.

Rapid adoption of tight glycemic control in critically ill patients springs largely from the fact that the ICU is a controlled setting, says Philip Goldberg, MD, fellow in endocrinology in the internal medicine department, Yale University School of Medicine. In the medical ICU at Yale New Haven Hospital, he notes, there are only about 40 nurses, as opposed to hundreds on other wards. "You can get 40 people to do this," he says. A physician places the initial order, and nurses initiate all subsequent actions according to a protocol.

Silvio Inzucchi, MD, professor of medicine at Yale University School of Medicine and director of the Yale Diabetes Center, says physicians are primed to adopt tight glucose control. "It has bothered people for a long time that we worked hard to get control of glucose in outpatients, then it went to pot when they were hospitalized," he says. "Now recent data show that blood glucose is modifiable and has immediate benefits to patients." Still, Dr. Inzucchi says, he’s not sure how widely it’s being adopted. "A lot of hospitals are calling us to find out about our inpatient glucose control projects. There seems to be a lot of new interest in this area, though I’m aware of many hospitals that haven’t yet begun any such interventions."

James Krinsley, MD, director of critical care at Stamford (Conn.) Hospital and associate clinical professor of medicine at Columbia University College of Physicians and Surgeons, says intensive glucose management is an evolving standard of care in ICUs. "Previously we would tolerate glucose values even up to 225 mg/dL. People ignored them. We just accepted that hyperglycemia is something that happens in critically ill patients." Now that has changed, he says. "If a hospital is not onboard with glucose control," Dr. Krinsley advises, "it is missing an opportunity to have a major impact on morbidity and at really low cost compared to other interventions we use in the ICU." His reference point is a drug called Xigris (activated protein C), which is used in patients with severe sepsis and costs $10,000. "It has a dramatic effect in seriously ill patients," he says. In one study Xigris achieved a 20 percent reduction in mortality. However, he points out, "We achieved at least that much with intensive insulin control, which we applied to an entire population of medical/surgical ICU patients."

"Fingersticks and intravenous insulin are not expensive," Dr. Gold berg concurs.

Acceptance of the value of tight glycemic control in critically ill non-diabetic patients is the culmination of a series of steps. It was known for many years that diabetic patients had an excess risk of death after a myocardial infarction. Results of the DIGAMI study, conducted during the 1990s in diabetic patients who had an MI, showed that using a glucose-insulin infusion to keep blood sugar <200 mg/dL during the acute phase reduced mortality over 3.5 years. In the Portland Diabetic Project, a multicenter study in patients with diabetes undergoing coronary artery bypass grafting, or CABG, mortality was significantly reduced with continuous insulin infusion compared with that of a historical control group.

More recently, Harold L. Lazar, MD, professor of cardiothoracic surgery at Boston Medical Center, and colleagues demonstrated similar results in a randomized trial. Keeping blood sugar between 125 and 200 mg/dL with a glucose-insulin-potassium infusion in diabetic patients undergoing CABG reduced length of stay, recurrent wound infections, and two-year mortality compared with patients whose blood sugar was held under 250 mg/dL with conventional subcutaneous sliding-scale insulin administration (Lazar HL, et al. Circulation. 2004;109:1497-1502).

These three trials were all done in diabetics. But observational trials had found that hyperglycemia was a risk factor for poor outcomes in critically ill non-diabetic patients as well. Dr. Van den Berghe’s randomized trial in an overall surgical ICU population grew out of this background and her own research into endocrine changes in critically ill patients. "One marker we found that predicted mortality three to four weeks before any clinical scoring system was a high level of serum insulin-like growth factor binding protein-I [ILGFBP-I], which is produced by the liver and suppressed by insulin," she says. Combined with information on high blood sugar as a predictor, this suggested that critically ill patients need more insulin to maintain normal blood glucose.

Dr. Van den Berghe’s work produced positive results, though her reasoning was faulty. "What I know now from all our subsequent research is that it is not the effect of insulin on the liver that helps, but insulin as a preventive of high blood sugar," she says. "So research sometimes needs a bit of luck." Further analysis of her data showed that the mortality benefit was related not directly to insulin dose but to tight glycemic control. "In older work with cardiac patients, giving insulin plus glucose didn’t help," Dr. Van den Berghe notes, "because they didn’t maintain glycemic control."

Stimulated by Dr. Van den Berghe’s work, Dr. Siegel decided to set up a glucose-control regimen in the medical ICU at Yale New Haven Hospital. "The challenge is that ICU patients are so sick that they are subject to a lot of factors that make it difficult to control glucose levels," he says. These patients have high levels of stress hormones, such as cortisol, growth hormone and adrenaline, which raise blood sugar, and the medicines they’re prescribed, such as steroids and epinephrine, also tend to raise blood sugar. On the other hand, the ICU has an advantage for controlling blood sugar: It is possible to do extensive monitoring, checking blood sugar hourly, and to control glycemia with insulin infusion pumps.

Dr. Siegel sought help from endocrinologists Drs. Inzucchi and Goldberg, who created a protocol for nurses to follow. "This was relatively low-tech," Dr. Siegel says. It required having a blood glucose goal, doing fingersticks to monitor blood glucose, and adjusting the insulin infusion rate accordingly.

"Most studies [of tight glycemic control] reported a specific investigator going into a unit to get data and coming up with a protocol to use in the study," Dr. Inzucchi says. "It was not clear whether this could be applied to practice in a general hospital or whether it only worked under study conditions." He and Dr. Goldberg devised a protocol that nursing staff could implement without much input from physicians. "You have to take decision making out of it," Dr. Inzucchi says. "Unfortunately, physicians will always find a reason not to begin therapy if it involves risk or considerable additional cost or added work or takes more time. So we designed a strategy that was more or less automatic."

Since they had already "bought into the principle," Dr. Goldberg says, they didn’t think it was ethical to randomize patients into a clinical trial. "Our goal was more to teach people how to do it in detail." He obtained protocols from several groups, but found that all had deficiencies. "They were either too loose or left too much up to the user," he says, noting that a protocol has to provide a level of comfort for nurses. "Our protocol tells you what to do in all circumstances and doesn’t leave too much up to experience." It also had to render nurses self-sufficient, so they wouldn’t have to page housestaff frequently.

At the heart of the Yale protocol is a nomogram that incorporates three components needed to regulate insulin delivery: current and previous blood sugar readings, which allow direction and velocity of change of glucose concentration to be calculated; and current insulin delivery rate. Adjustments are given as multiples of the current delivery rate. "It’s complicated," Dr. Goldberg says. "But most nurses, after they had used it on a few patients, found it easy to use."

He was in the ICU for much of the time during training. He taught the nurses how to carry out the protocol and was on hand to answer questions and address concerns, mostly how infusion change was executed. "There was an initial resistance," he acknowledges, "which was definitely an experience-related phenomenon."

He also found that nurses were overly concerned about possible hypoglycemia: If blood glucose went below 100 mg/dL, they started a sugar infusion, particularly in patients who had received insulin. To counter this tendency, Dr. Goldberg asked the nurses, "What do you think your blood sugar is right now?’ Some nurses thought it was as high as 150. But fingersticks showed it was most often in the 70s or 80s, without any symptoms of hypoglycemia. "So this is an educational barrier that we had to overcome," he says. "It was fairly easy to convince people their blood sugars were rather low with just a few fingersticks." Dr. Siegel concedes that hypoglycemic coma was one of his initial concerns as well. But it rarely occurred. "We monitor them so closely that symptomatic low blood sugar rarely happens," he says. "It would be unusual with a good protocol."

A more real problem is the workload increase for the nurses, which Dr. Goldberg estimates as five to 10 minutes per hour per patient.

Dr. Goldberg observed 69 insulin infusions in 52 patients. He found the protocol feasible and effective. Once target blood glucose levels were achieved with intravenous insulin, 52 percent of subsequent hourly fingersticks were in the target range of 100-139 mg/dL, and 66 percent were in a "desirable" range of 80-139 mg/dL (Goldberg PA, et al. Diabetes Care. 2004;27:461-467). Their detailed protocol is in the article.

Dr. Krinsley, too, was motivated by Dr. Van den Berghe’s randomized trial, which he calls "a big eye opener." His response to her "dramatic results," he says, was to analyze the relation between glucose levels during ICU stay and in-hospital mortality among 1,800 consecutive admissions to the ICU at Stamford Hospital. "Since 1998 I had kept a database to track outcomes for every ICU patient, with a detailed core data set," he says. Hyperglycemia turned out to be a strong independent predictor of mortality, over and above illness severity score. There was a gradient of mortality even within the normal range, 80-110 mg/dL. Mortality was lowest, 9.6 percent, among patients with mean glucose values between 80 and 99 mg/dL, and it increased progressively with rising glucose values to a maximum of 42.5 percent among patients with glucose concentrations over 300 (Krinsley JS. Mayo Clin Proc. 2003:78:1471-1478).

These data led Dr. Krinsley to do an intervention in the ICU using intensive monitoring and continuous insulin to maintain blood glucose under 140 mg/dL. From the retrospective review of 1,800 patients, he had learned that mean glucose level was significantly lower among hospital survivors, 138, than among nonsurvivors, 172. The new protocol led to a decrease in mean glucose level in the ICU from 153 mg/dL during the baseline period to 128 mg/dL in the treatment period. Among 800 consecutive patients treated with the intensive protocol, mortality was reduced by almost 30 percent compared with the 800 patients treated immediately before the intervention was introduced. A 10.8 percent reduction in length of stay was also seen (Krinsley JS. Mayo Clin Proc. 2004;79:992-1000). For this work, the Joint Commission on Accreditation of Healthcare Organizations awarded Dr. Krinsley its 2004 Codman Award.

Like Dr. Goldberg at Yale, Dr. Krinsley saw reluctance from nurses initially. "They had been trained to not treat high blood glucose," he notes. Also, there is concern about the bump-up in workload. "The degree of monitoring has increased," he admits. "Nurses and assistants are doing more work." Nevertheless, the number of nursing hours did not increase as a result of the protocol, a finding reported in the 2004 article.

Procedural benefit from a glucose control program was also demonstrated by the diabetes inpatient initiative carried out in the cardiac care unit at Methodist Hospital of Clarian Health Partners, Indianapolis. Diabetes nurse case managers assisted in educating staff nurses and developed order sets and care pathways, says Joni Carroll, RN CDE, of Methodist Diabetes Center. Here, the initial goal was to keep blood sugar <200 mg/dL, a target that was quickly lowered to <180. A reduction in length of stay was achieved that correlated directly with glycemic control.

Instituting a successful tight glycemic control program takes more than a protocol. "Many hospitals want to implement these programs," Dr. Van den Berghe says, "but not all do it well. It requires a very motivated team." In some hospitals nurses are reluctant to do jobs that they believe are the responsibility of physicians, so that even if a policy is established, it is not practiced. "Motivation depends on an enthusiastic leader," Dr. Van den Berghe says. At a recent conference on tight glycemic control sponsored by the National Institutes of Health, one topic was implementation. Participants agreed on this principle: The person in charge of the ICU needs to explain and show how easy it is to follow the protocol. "In our experience what makes it work is to trust the nurses to be responsible," Dr. Van den Berghe says. "You have to hand it over to them. They take pride in excellence and outcome." As Dr. Van den Berghe travels giving lectures, she sees units where nurses have to ask physicians about everything. In such an environment, a program is unlikely to succeed.

Dr. Krinsley takes a similar view. "We have empowered our nurses to follow protocols and make bedside decisions using the protocol as a guide, but being more proactive and using their own initiative. Some nurses like this," he says, "but not all are comfortable with this increased level of interaction."

With growing acceptance of the benefits of tight glycemic control in critically ill patients, both diabetic and non-diabetic, two main questions remain: How far can these benefits be extrapolated? What blood glucose level is optimal? A meeting of major investigators in this area, sponsored by the American College of Endocrinology and the American Association of Clinical Endocrinologists, was held in December 2003 in Washington, DC. From this conference a position paper was published that recommends extending tight glycemic control to all hospitalized patients (Garber AJ, et al. Endocr Pract. 2004;10 suppl 2:4-9). In ICUs, the guideline suggests an upper limit of blood glucose of 110 mg/dL; for noncritical care unit patients, suggested threshold values are 110 mg/dL preprandial and 180 maximal.

Reaction to the recommendation to apply tight glycemic control to all hospitalized patients ranges from complete acceptance to frank disagreement. "[The recommendations] weren’t necessarily evidence-based or rational," Dr. Inzucchi demurs. He views extending tight glycemic control to all hospitalized patients as "a bit extreme" and has co-written an editorial that will soon be published, criticizing the position paper for going beyond ICU patients.

At Yale New Haven Hospital, tight glycemic control has become standard practice in the medical ICU and is being extended to other ICUs. Nursing administrators led the move to apply it to the cardiothoracic ICU, Dr. Goldberg says. Medical ICU nurses talked about the protocol at nursing meetings, which stimulated administrators to do research, turning up published data on cardiothoracic patients. Nursing personnel then approached Dr. Goldberg. "It was definitely harder getting cardiothoracic surgeons to buy into it," he says. "A couple were quite resistant in the beginning, but once we got over the initial hump, it has been well received. Now it’s becoming standard in all ICUs in our hospital."

As for the glucose target, Dr. Inzucchi says, "It is hard to select what the target should be. The optimal glucose level is being hotly debated."

Dr. Siegel agrees that the best glucose threshold is "a bit controversial."

"One article suggested you didn’t have to have perfect control; just sending blood glucose down to the mid-100s is probably sufficient [Finney SJ, et al. JAMA. 2003;290:2041-2047]. But I think the majority of critical care physicians now think the more tightly controlled the better, and that normal blood glucose will lead to the best outcomes. More people believe that the 100-110 range is optimal."

Dr. Van den Berghe has instituted tight glycemic control as standard practice in her 56-bed surgical ICU. Asked about the recommendations of the two endocrinology associations, she replies, "This is a difficult discussion for me. I’m a very evidence-based person. Except for diabetics, I find it difficult to extrapolate these principles outside the ICU, where they have been proven." Some people argue that not adopting universal tight glycemic control implies that accepting hyperglycemia is state of the art. Dr. Van den Berghe acknowledges the force of that argument. Still, she objects, "To implement a new therapy without proof ..." In the end, she says, "I don’t argue with either way."

On optimal glucose levels, she is more certain. In her view the trend toward normoglycemia as a target makes sense. "Our ongoing work shows that if you want to prevent complications of hyperglycemia in critically ill patients, it really matters what level of control you achieve." In her study, outcomes were better among patients with normoglycemia (<110 mg/dL) than in those with an intermediate blood glucose level (110-150 mg/dL).

Clarian Health Partners is adopting the endocrinology groups’ guidelines, Carroll says. She acknowledges that the value of tight glycemic control has not been documented in noncritically ill patients. "But it makes sense that it should cross over to all patients," she says. "Those in the diabetes world have known for a long time that hyperglycemia reduces resistance to infection and reduces healing. I think the onus is on the naysayers to prove that controlling blood sugar is not beneficial."

Implementing tight glycemic control with insulin infusions on general wards runs into a logistical problem: Many patients don’t have intravenous lines or the nursing and management staff is reluctant to institute the use of insulin infusions on their unit. Carroll’s approach is to educate physicians and nursing staff on the basal bolus insulin concept. "This provides a more physiological insulin replacement than the roller coaster sliding-scale insulin effect," she says. "It also allows easier care and smoother control for patients who are NPO, nauseated, anorexic, etc."

Dr. Krinsley has made tight glycemic control standard in the ICU at Stamford Hospital and is now extending it to special care units, starting with an intermediate care unit for subcritical patients. "Designing a hospitalwide glucose management program for less ill patients is a challenge and a goal," he says. "Out on the regular med/surg floors in our hospital, it is the Wild West, with many sliding-scale insulin regimens, no standardization, and typically poor control."

While Dr. Krinsley admits that the value of tight glycemic control hasn’t been proved beyond the ICU, he points out that the relationship between high glucose concentration and poor outcomes has been demonstrated in many different patient groups. "Glucose control leads to less organ system failure in ICU patients," he says. "There is no reason to think that glucose control shouldn’t be strongly beneficial in regular med/surg patients as well." Accepting that better glucose control on the floors is a worthy goal is justifiable, he says. He is more concerned with a practical issue—how to standardize and improve glucose control in a heterogeneous population cared for by so many physicians.

Dr. Krinsley is trending toward normoglycemia as his glycemic goal. Initially, when he set up the intervention study, he says, "we were uncomfortable, especially the nurses, treating to such an intense level as 110. So we chose 140 as our treatment threshold." Having found that 140 mg/dL was attainable, the Stamford Hospital ICU now has adopted 125 mg/dL as the treatment threshold—a change prompted in part by nurses who believed the lower goal would be easy to achieve. A new treatment protocol was written, giving even greater responsibility to the bedside nurses to reach the targeted goal.

Several abstracts on the implementation of glycemic management protocols were presented recently at the national meeting of the Society of Critical Care Medicine. "There is a clear trend to give more responsibility to the nurses and make the protocols less proscriptive as units become more comfortable implementing tight glucose management," Dr. Krinsley says.

But Dr. Lazar calls extending tight glycemic control to patients outside the surgical ICU "a mistake."

"It is really in patients undergoing metabolic derangement of acute coronary syndromes where the combination of insulin and glucose will have benefits for the myocardium," he says. "We need to try to keep it confined." Dr. Lazar likens the spread of tight glycemic control outside the surgical ICU to "a diet fad."

He even disagrees with Dr. Van den Berghe’s use of tight glycemic control in young trauma patients. "To take young kids in trauma and put them on insulin drips-the trauma docs are starting to do that. People will get in trouble," he predicts. "There is a difference between young trauma patients and critically ill septic ICU patients."

For a glycemic goal, Dr. Lazar favors 120-180 mg/dL. "Future study will determine whether patients do better with lower blood glucose," he says. "I am not so sure that will be the case. You reach a finite level with clinical endpoints." Point-of-care blood glucose testing in ICUs is typically a self-contained activity, requiring little input from clinical laboratorians. At Henry Ford Hospital in Detroit, Karen M. Bourlier, MS, MT(ASCP), CLS(NCA), alternate site testing coordinator in the pathology department, makes sure the meters are functioning and that users are properly trained. She also watches the computer interface to make sure meters are uploaded periodically and results are credible. The nursing staff does quality control on the meters.

When the surgical ICU at Henry Ford Hospital set up tight glycemic control, the laboratory was not brought into the activity until Bourlier received a request for additional glucose meters. "We asked them, ’Who are you and why do you need these additional meters?’" she says. It turned out that the surgical ICU was one site in a multicenter study to prevent surgical wound infections; it was sponsored by the Centers for Medicare and Medicaid Services and Centers for Disease Control and Prevention and included a tight glycemic control component.

Fortunately for the people doing the surgical ICU study, the laboratory had six months earlier set up an interface through which glucose results could be transmitted to the electronic medical record. "We work very hard to make sure patient results get onto the medical record with correct ID numbers," Bourlier says. "We try to keep our nonrecorded percentages very low." Currently this figure is about three percent.

As tight glycemic control spreads and critical decisions are based on the results of glucose meter readings, POC testing will be stressed by high volume and demands for rapid feedback. "The critical thing is to have accurate and fairly easy-to-use meters and to have a good interface," Bourlier says. One needs a glucose meter that accepts a patient identification number and from which results can be uploaded to an electronic database. "If it’s all on paper, I can’t imagine how you would implement and monitor a tight glycemic control protocol," Bourlier says. Fortunately, she adds, a variety of glucose meters and interfaces are available that meet these criteria.

While much of the benefit of tight glycemic control protocols appears to derive directly from lowered blood glucose, Dr. Van den Berghe’s research suggests that insulin’s benefit is also mediated through alterations in lipid concentrations (Mesotten DJ, et al. Clin Endocrinol Metab. 2004;89:219-226). In fact, Dr. Van den Berghe has written, analysis of the trial results suggests that normalization of dyslipidemia, along with immunologic effects, "seem to surpass the level of glycemic control in explaining the clinical benefits of intensive insulin therapy on sepsis, organ failure, and death" (Van den Berghe GH. Endocr Pract. 2004;10 suppl2:17-20).

Two corollaries follow from these observations. First, if you are going to do glycemic control, do it with insulin. "To achieve proper metabolic control, it’s insulin that’s crucial rather than reducing glucose intake," Dr. Van den Berghe says. Some have suggested achieving normoglycemia by not feeding, but she doesn’t think this would accomplish the same benefits.

Second, she suggests, "Maybe if we would do something on top of blood glucose control with insulin, it would help even more." In critically ill patients, low levels of the lipoproteins HDL and LDL are predictive of poor outcomes. Dr. Van den Berghe would like to do experiments in animals to see if raising lipoprotein levels with supplements would make a difference.

Improved instrumentation will help achieve tight glycemic control in the future. "The standard way of monitoring blood glucose values by fingersticks and glucose meters is not a very good tool," Dr. Krinsley says. "Many companies are interested in developing continuous glucose monitors, which would facilitate glycemic control, decrease nurse workload, and avoid hypoglycemia."

When Dr. Van den Berghe gives talks, many people tell her they are waiting for a system that provides continuous reading and instructs the user in how to adjust insulin. Currently, she notes, continuous online display of blood pressure or cardiac output, or both, allows for optimal titration of inotropes and vasopressors. In the same way, she says, "Continuous display of blood glucose levels is mandatory for optimal titration of insulin therapy in the ICU." Dr. Van den Berghe is working on developing such an instrument. "But these instruments are not near clinical use yet," she cautions. "Making one is not all that easy."

William Check is a medical writer in Wilmette, Ill.