© 2006 American Thoracic Society doi: 10.1164/rccm.2510006
The Highs and Lows of Intensive Insulin TherapySunnybrook and Women's College Health Sciences Centre, University of Toronto, Toronto, Ontario, Canada
Hopital Raymond Poincare, University of Versailles, Garches, France In the past decade, we have witnessed important advances in the care of critically ill patients, from both traditional areas of investigation, such as acute lung injury and sepsis (1–4), but also in relation to the support of systems other than the heart and lungs. One such example is the mounting appreciation for endocrinologic and metabolic issues in the intensive care unit (ICU). Recently, it has been shown that patients with septic shock and relative adrenal insufficiency have improved survival when treated with moderate doses of steroids (5). In 2001, van den Berghe and colleagues published a landmark prospective, randomized, controlled trial of intensive insulin therapy (IIT) for postoperative surgical patients in the ICU, which demonstrated a 42% relative reduction in mortality (from 8.0 to 4.6%) when IIT was used (6). The largest effect was found among patients who stayed in the ICU for more than 5 days, and the greatest reduction in deaths occurred among patients with a proven septic focus (6). Prior work demonstrating that improved glycemic control is associated with reduced incidence of infection lends biological plausibility to the IIT strategy (7–9). Using retrospective observational methodologies, others have shown that glycemic control, as opposed to anabolic effects of insulin, is likely to be most responsible for improved outcomes (10). In the years following van den Berghe's trial, ICUs around the world have instituted insulin protocols. However, many questions have arisen. Can the results from a single-center study that enrolled predominantly postoperative cardiac surgery patients be generalized to all critically ill patients? What is the risk of potential hypoglycemia and harm from IIT among a broad spectrum of critically ill patients? What are the resource implications and costs for ICUs that institute such a program? Finally, how does a higher or lower baseline mortality rate affect the number of patients needed to treat, and number of patients potentially harmed by IIT? In this issue of the Journal (pp. 407–413), Egi and colleagues attempt to address these questions and the generalizability of IIT (11). The authors present a multicenter retrospective cohort study of postoperative patients admitted to the ICUs of four hospitals. Using the Australian and New Zealand Intensive Care Society Adult Patient Database, 783 of 10,125 consecutively admitted postsurgical patients were matched by age, sex, severity of illness, and the diagnostic categories to patients participating in the original van den Berghe IIT trial (6, 11). On the basis of the IIT study, the authors then applied the expected relative risk reduction to the observed mortality for patients in their selection cohort at each participating center. From this calculation, they derived estimates of the number of patients needed to be treated to prevent one death, and also the number of patients needed to be treated to cause one episode of harm (severe hypoglycemia, defined as blood glucose < 2.2 mmol/L) (11). Among all patients in the Australia–New Zealand cohort, 102 would require IIT to prevent a death, whereas harm would occur after treatment of only 13 (11). Importantly, the authors found that there were large variations in these estimates when the IIT strategy and expected effects were applied to various clinical settings and varying patient case mix. The number needed to treat to prevent an individual death ranged from 38 to 125 (as compared with 29 in the IIT trial), whereas the number needed to treat to cause harm varied from 7 to 13 (as compared with 23 in the IIT trial) (6, 12). These results call into question the generalizability of IIT. However, there are important limitations to consider when interpreting Egi and colleagues' study. First, despite extensive attempts to match patients to the IIT trial, there were differences in baseline characteristics, severity of illness (higher in the selection cohort), morning glucose levels (lower in the selection cohort), and baseline mortality rate (lower in the selection cohort). In addition, the selection cohort had a lower incidence of sepsis and proportion of patients with prolonged lengths of stay—the subpopulations deriving most benefit in the IIT trial. Hence, application of a constant relative mortality reduction from the IIT trial may not be appropriate. There were also differences in the methods of nutrition support and glucose control. Patients in the conventional therapy group of the IIT trial were treated with glucose infusions and, by study design, their glucose values were targeted to be higher than that observed in the Australian–New Zealand cohort. There may be yet other unrecognized and unaccounted-for differences between the groups enrolled into the two studies that influence mortality and risk of harm in unanticipated ways. The authors do state that none of the patients in the selection cohort received other therapies subsequently found to improve outcomes, such as activated protein C or steroids for sepsis. Importantly, none of the ICUs contributing patients used an intensive insulin protocol, even though a proportion of the selection cohort was derived after publication of the IIT trial. Finally, and perhaps most important, patients and physicians would likely agree that comparing the number needed to treat to prevent a death and the number needed to treat to cause hypoglycemia is akin to comparing apples and oranges. Recognized and transient hypoglycemia would surely be tolerated when the therapy ultimately saves a life. However, the balance between these competing outcomes is less clear in the Australian–New Zealand study than in the IIT trial (6, 11). Intensive insulin control is an appealing strategy. We have data from a well-designed trial demonstrating improved survival, the strategy has biologic plausibility, and the drug costs are not prohibitive. However, concerns remain about the generalizability of this trial to other critically ill medical and surgical patient populations (11). The possibility of increasing frequency of hypoglycemia has prompted most protocols to direct frequent blood glucose determinations, which has nursing resource implications, laboratory processing demands and costs, and, depending on the mode of glucose monitoring, a substantial increase in amount of phlebotomy volumes. Further investigation on IIT is underway. Van den Berghe and colleagues followed their original trial with a similar study of 1,200 subjects recruited from the medical ICU at the same institution. These investigators have reported a significant improvement in mortality among long-stay patients in an oral session at this year's European Society of Intensive Care Medicine scientific meeting. At the same meeting, Reinhart and colleagues reported that a multicenter German study of intensive insulin therapy in patients with septic shock has been suspended by the data safety monitoring board due to significant excess in the risk of severe hypoglycemia in patients treated with IIT without any evidence of improved survival (12). Other trials continue in Europe, Australia, and North America, which will aid in addressing questions of generalizability and safety of IIT to other patient populations (13, 14). The findings of Egi and coworkers serve to remind us that although IIT has been demonstrated to be effective in reducing mortality in one well-designed trial of predominantly post–cardiac surgery patients, its use may not produce equivalent effects in all patient populations and also there may be important safety concerns, resource implications, and costs to consider before there is widespread acceptance of this promising treatment for all of our critically ill patients. FOOTNOTES Conflict of Interest Statement: Neither author has a financial relationship with a commercial entity that has an interest in the subject of this manuscript. REFERENCES
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
