|Home | About | Journals | Submit | Contact Us | Français|
Tight glycemic control in critically ill patients became a therapeutic paradigm after the 2001 publication of a landmark single-center interventional trial in Leuven, Belgium, (ie, “Leuven 1”) that targeted euglycemia in a population of mechanically ventilated patients in a surgical intensive care unit (ICU).1 A subsequent randomized trial from the same institution in medical patients2 (ie, “Leuven 2”) was not as strongly positive for the benefits of intensive insulin therapy as the first study. Two later interventional studies were terminated prematurely by their respective institutional review boards and were underpowered to show any effect on mortality.3,4 Meta-analysis failed to demonstrate a survival benefit for tight glycemic control.5 Finally, in the large multicenter NICE-SUGAR (Normoglycemia in Intensive Care Evaluation—Survival Using Glucose Algorithm Regulation) trial, the control group with a moderate glycemic target range of 140 to 180 mg/dL had lower 90-day mortality than did the interventional group with a glycemic target range of 81 to 108 mg/dL.6 What went wrong?
Hypoglycemia is a likely culprit. Although the authors of Leuven 1 stated that “no adverse consequences” resulted from the higher rate of severe hypoglycemia (blood glucose concentration <40 mg/dL), the relative risk of this complication among patients in the interventional cohort was 6.65.1,7 Similarly, the hazard ratios for the development of severe hypoglycemia among patients in the Leuven 2,2 Glucontrol,3 VISEP (Efficacy of Volume Substitution and Insulin Therapy in Severe Sepsis),4 and NICE-SUGAR6 trials were 5.94, 3.61, 4.11, and 13.72, respectively.7 Of note, there was no difference in the rate of severe hypoglycemia when control and interventional patients were compared in the 1600-patient “before and after” Stamford, CT, trial, a study that demonstrated significant mortality and morbidity benefit associated with intensive glycemic monitoring and treatment of hyperglycemia.8
There will never be a randomized controlled trial targeting severe hypoglycemia in populations of critically ill patients. Nevertheless, available data suggest that severe hypoglycemia is harmful. In a heterogeneous medical-surgical cohort of 5365 patients admitted to a single ICU, the odds ratio for mortality associated with a single episode of severe hypoglycemia was 2.28 (P=.0008).9 A 1:3 case:control study within this cohort, with matching based on age, severity of illness at presentation, diagnostic category, and diabetic status, confirmed the increased risk of mortality associated with a single episode of severe hypoglycemia (P=.0057). Indeed, investigators of Leuven 2, in which 25.0% of patients in the ICU for at least 5 days sustained at least one episode of severe hypoglycemia, reported that the occurrence of severe hypoglycemia was an independent predictor of death and that “hypoglycemia induced by intensive insulin therapy may have reduced a portion of the potential benefit.”2
See also page 217
The investigation by Egi et al10 in the current issue of Mayo Clinic Proceedings extends these observations. From a cohort of 4946 patients admitted to 2 teaching hospital ICUs in Australia between 2000 and 2004, 1109 (22.4%) had an episode of hypoglycemia, defined as a blood glucose concentration lower than 81 mg/dL (to convert to mmol/L, multiply by 0.0555). The patients were stratified by the severity of hypoglycemia into 6 bands ranging from 72 to 81 mg/dL to less than 36 mg/dL. Strengths of the study include the heterogeneous patient population; use of the glucose oxidase method for all glucose measurements; avoidance of any analytic inconsistencies associated with the use of point-of-care devices11; use of time-weighted glucose concentrations to avoid surveillance bias due to hypoglycemia; creation of an embedded case-control study comparing survivors and nonsurvivors to investigate in detail nutritional support and insulin therapy at the time of hypoglycemia; and use of robust multivariable models to account for potential confounding influences on mortality.
Patients with hypoglycemia had nearly twice the rate of mortality (36.6% vs 19.7%) as did those without. Although this certainly cannot prove causality, a multivariable model was constructed to account for numerous potentially confounding demographic and clinical factors. Increasing severity of hypoglycemia was independently associated with risk of mortality. Compared with patients who had the mildest hypo glycemia, 72 to 81 mg/dL, those experiencing nadir glucose values between 54 and 63 mg/dL, 45 and 54 mg/dL, 36 and 45 mg/dL, and less than 36 mg/dL had statistically significant adjusted odds ratios of mortality of 1.93, 2.13, 2.14, and 2.99, respectively. Of note, the glycemic target was 108 to 180 mg/dL during the period of this retrospective study; only 32.7% of the patients were receiving insulin at the time of the event, and 67.3% were not. Risk factors for insulin-associated hypoglycemia included diabetes, concomitant use of enteral nutrition, and neurologic diagnosis, and insulin-treated hypoglycemic patients were more likely to die of a neurologic cause. In contrast, patients with spontaneous hypoglycemia were more likely not to be receiving nutritional support at the time of the episode and were more likely to die of a cardiovascular cause. Multivariable logistic regression analysis suggested that the degree of hypoglycemia, and not insulin therapy per se, was associated with increased risk of mortality.
It is biologically plausible that hypoglycemia contributes to mortality. Neurons are obligate users of glucose, and a large body of evidence suggests that hypoglycemia, especially if severe or prolonged, can cause irreversible neuronal damage by a variety of mechanisms. Furthermore, hypoglycemia-induced sympathetic stimulation may lead to cardiac arrhythmias and/or myocardial compromise. However, the specific mechanisms of any hypoglycemia-related increases in mortality seen in recent trials in critically ill patients, if they exist, remain to be elucidated.12
Glycemic variability has recently been shown to be an independent predictor of mortality among various populations of critically ill patients.13,14 Oxidative stress induced by glycemic variability may be an important contributor to the risk of mortality among vulnerable patients, such as those with vascular disease.15 Is it possible that the increased risk of mortality observed among patients with even mild degrees of hypoglycemia in the current study is really a reflection of increased glycemic variability? Although the authors do not address this question directly, there is a suggestion that the impact of hypoglycemia is independent of the potential effect of glycemic variability in this cohort; among patients with hypoglycemia, survivors and nonsurvivors had similar mean and maximum glucose values.
What are the clinical implications of this study? The recently published NICE-SUGAR trial demonstrated that patients with a moderate glycemic goal (mean morning glucose, 144 mg/dL) and an admirably low rate of severe hypoglycemia (0.5%) had lower 90-day mortality than did those with more intensive glucose control (mean morning glucose, 118 mg/dL) and a high rate of severe hypoglycemia (6.8%).5 The investigation by Egi et al underlines the important role that hypoglycemia played in determining the negative result of this large randomized trial.
It is highly likely that there will not be another large multicenter randomized trial of insulin therapy in critically ill patients until there is a fundamental change in the manner by which we monitor and control glucose values in the ICU. Specifically, we await the development and clinical implementation of continuous or near-continuous glucose monitoring devices and “closed-loop” glycemic control systems. With the use of these new technologies, coupled with algorithm-driven treatment protocols, the rate of hypoglycemia should plummet; moreover, studies can be completed to prospectively evaluate the potential benefit in targeting a reduction in glycemic variability as an additional therapeutic goal. Until then, we think that the study by Egi et al confirms the deleterious effect of hypoglycemia, especially severe hypoglycemia, in critically ill patients; highlights the complexity of this clinical problem; and reinforces the principle that clinicians practicing glycemic control must do so safely.16,17