Evidence and Intuition in the ICU
Updated: Aug 25
Patients will occasionally email me a link to a recent medical study and ask for my opinion. Before even reading the article, my immediate response is that I don’t get that excited about any individual study. It may be of very high quality, published in an excellent peer-reviewed journal, but I will still bring a healthy dose of skepticism. I learned early on in my surgical career that one study – however impressive – should not dictate your practice.
As an inexperienced doctor in the intensive care unit, I was pretty sure that my decisions were life-or-death, and I was determined to have the outcomes skew toward life. Though heroic actions are occasionally necessary, much of critical care is mundane, even boring. One of the essential tasks is collating and interpreting all of the numerous data points produced hourly by every ICU patient. These data points range from basics like heart rate and blood pressure, numbers checked at a normal physical check-up, to more complicated measurements like ins and outs (how much fluid is going in and how much is coming out), mechanical ventilation pressures and volumes, oxygen saturation, medication dosages, and dozens of blood test values.
Though all of these data points are important, some naturally gain more attention than others. Blood pressure, so essential for ensuring that your brain gets enough oxygen, is obviously one of the most consequential and remains one of the most important. But others have had their moments. In 2006, when I started surgical residency, one lab value had obtained an outsized importance: glucose. Glucose is a form of sugar. Sugar may have developed a poor reputation due to its connection with obesity and related health issues, but it is an absolutely necessary source of energy. The brain requires a continuous supply of glucose, and the body carefully regulates blood levels. With too little, one can become shaky, sweaty, and confused, which can progress to a loss of consciousness and seizures if the glucose is not brought to a normal level. People with diabetes mellitus have problems with insulin, the hormone that regulates glucose, and they develop elevated glucose levels, which are also harmful.
Not all of my ICU patients had diabetes, but I closely tracked glucose levels in all of them throughout the day. We were practicing evidence-based medicine, and the best evidence at the time showed that very tight glucose control saved lives. We had always checked glucose levels daily and treated any level that was way out of normal range, but this was different. We wanted the glucose to stay within a very small range, not going below 80 or above 110 mg/dl.
Most people will keep their glucose in a range between 70 and 140 mg/dl. Untreated high glucose levels in someone with diabetes can lead to increased infections, kidney failure, cardiovascular disease, and blindness. But there are situations in which glucose levels can spike temporarily in otherwise healthy people. Eating a sugary dessert can do it. Any kind of stress on the body can do it as well. Stress in this case could be caused by an operation, trauma, or acute illness. Patients who are sick enough to be in the ICU will have some stress response and are likely to have at least a temporary bump in glucose. If the goal is keeping glucose under 110, many patients will need insulin drips, continuous infusions of the hormone. This requires care, since too much insulin causes hypoglycemia, with resulting confusion and seizures. As a result, glucose level checks, done with a drop of blood from the fingertip, have to be performed much more often.
But all of that was worth it if we were saving lives. And we had reason to think we were. In 2001, a landmark trial from Belgium was published in the New England Journal of Medicine, one of our most prestigious medical journals[i]. In one center in Leuven, Belgium, researchers performed a randomized controlled trial (RCT) in a surgical ICU. Any medical student can tell you that an RCT is the best form of research study, the form from which we can draw the most definitive conclusions. In this type of study, the patients are randomly assigned to a treatment group, and the results for each group are then compared. Since the assignment is random, the patients in each group should be largely similar in all ways other than the treatment being studied. As long as there are enough patients in each group, whatever differences you find in the outcomes should be attributable to the treatment. In the Belgium RCT, over 1500 patients were enrolled, with half receiving very tight glucose control (called “Intensive Insulin Therapy”, or IIT, requiring an insulin drip for any glucose levels above 110 mg/dl) and the others receiving more conventional glucose control (glucose was allowed to rise to 215 mg/dl before an insulin drip was started).
The results were clear and striking. Those receiving standard therapy had more infections, higher rates of kidney failure, and higher death rates. For patients in the ICU more than 5 days, mortality within the ICU dropped from 20.2% for those receiving standard therapy to 10.6% for those receiving IIT. There are surprisingly few things doctors can do to improve survival by that much. Physicians and medical societies took note and began changing ICU protocols. The Institute for Healthcare Improvement promoted intensive glucose control for patients with severe sepsis. The Joint Commission on Accreditation of Healthcare Organization proposed tight glucose control as a core measure for quality, which would have allowed Medicare to determine payments based on whether intensive insulin therapy was used. Scientists studied the physiologic explanation for this survival benefit, and published convincing arguments for why insulin therapy produced such incredible results.
As a young physician, I joined the worldwide movement for tight glucose control in the ICU. We all want to do what is best for our patients and we all want to practice the most scientifically sound medicine. The point of “evidence-based medicine” is to improve patient care by moving away from “gut feeling” or intuitive practices and moving toward those based on quality scientific evidence. My gut might tell me that mildly elevated glucose levels in very sick people might be ok, since it is the body’s natural response. But intuition is often wrong, and the evidence was there: allowing glucose to rise above normal was costing lives.
Except that it wasn’t. In the years following the 2001 Leuven study, more studies emerged. Researchers attempted to replicate the Leuven study, but failed to find any benefit to tight glucose control.[ii] Rather than helping, it appeared to harm. Patients who were aggressively treated for mildly elevated glucose levels did not have improved survival, and the insulin frequently caused dangerously low glucose levels. Even the Leuven researchers, when repeating their study in the medical ICU rather than the surgical ICU, failed to show improved survival.
The death blow to intensive insulin therapy came in 2009, when the New England Journal of Medicine published the NICE-SUGAR trial results.[iii] This was also a randomized controlled trial, but with more patients and at multiple hospitals: they studied over 6000 patients in the ICUs of 42 hospitals throughout Australia, New Zealand, and Canada. Critically ill patients were assigned to one of two groups: one received very tight glucose control, as in the Leuven trial, and the other received more moderate glucose control, only treating glucose levels above 180. The results were similar to many smaller trials: tight glucose control did not improve survival. With a 90-day follow-up, patients in the intensive insulin therapy group had a slightly higher mortality: 27.5% died, versus 24.9% in the conventional therapy group. There was no benefit to tight glucose control.
Medical societies and physicians quickly and quietly moved away from intensive insulin therapy. Glucose was still closely monitored, but mildly elevated glucose levels in critically ill patients were acceptable, requiring no aggressive interventions. In retrospect, the Leuven study was not perfect, and the findings likely never should have been applied as broadly as they had been. While we all want to practice scientifically sound medicine, the evidence we have may not be enough for an “evidence-based” approach.
Although the worldwide overreaction to the Leuven study may represent a failure of evidence-based medicine, ultimately the story became one of success. After the 2001 trial, many others tried to replicate the study. The NICE-SUGAR investigators knew of the Leuven results, and were careful to design their study in a way that would answer the most important questions. After many smaller studies and one large trial, it was clear that intensive insulin therapy was not the best treatment for critically ill patients. The evidence is strong enough now to form treatment recommendations.
What went wrong with the 2001 Leuven trial? No one has accused the investigators of committing fraud, so we may assume that the data presented were correct. The patients in that trial really did have better survival with intensive insulin therapy. But looking closer at the patients and the other treatments they received, we can find other explanations for why the results might not apply to all critical care patients. Most of the Leuven patients came to the ICU after cardiac surgery, which could perhaps explain part of the difference in results. Buried in the “Methods” section was a practice that likely had an even larger effect, but was overlooked by many early readers of the study: patients were given high doses of IV glucose on arrival to the ICU, and given total IV nutrition the next day. This is a highly unusual practice.[iv] I normally would not pump patients full of glucose right after surgery. If I did, my gut might tell me that I would need to then control that extra glucose with extra insulin. Sometimes, intuition may be right.
[i] van den Berghe G, Wouters P, Weekers F, et al. Intensive Insulin Therapy in Critically Ill Patients. N Engl J Med. 2001 Nov 8; 345(19):1359-67. [ii] Preiser J, Devos, P. Clinical experience with tight glucose control by intensive insulin therapy. Critical Care Medicine: September 2007 - Volume 35 - Issue 9 - p S503-S507 doi: 10.1097/01.CCM.0000278046.24345.C7 [iii] NICE-SUGAR Study Investigators. Intensive versus conventional glucose control in critically ill patients. N Engl J Med. 2009. March 26;360(13):1283-1297. [iv] Bellomo R, Egi M. What is a NICE-SUGAR for patients in the intensive care unit?. Mayo Clin Proc. 2009;84(5):400-402. doi:10.1016/S0025-6196(11)60557-0