Strict glucose control in the critically illBMJ 2006; 332 doi: https://doi.org/10.1136/bmj.332.7546.865 (Published 13 April 2006) Cite this as: BMJ 2006;332:865
- Peter Watkinson, specialist registrar (, )
- Vicki S Barber, research manager,
- J Duncan Young, consultant
In 2001 Van den Berghe et al reported the results of a randomised controlled trial comparing the mortality of critically ill surgical patients receiving insulin infusions to achieve “tight glycaemic control” (target blood glucose 4.4-6.1 mmol/l) with that of patients receiving conventional treatment, where insulin was infused only if the blood glucose exceeded 11.9 mmol/l and was adjusted to maintain values of 10-11.1 mmol/l.1 The trial was stopped after 1548 patients had been enrolled because the mortality in the tight control group was 4.6% compared with 8% in the control group (32% corrected relative reduction; P = 0.04). Ever since, tight glycaemic control has been standard practice, but there are now good reasons to question it.
It always seemed surprising that a simple change in blood glucose management reduced mortality more than other far more costly and complex interventions tested through randomised trials in the critically ill. The only corroborating evidence came from studies of glucose-insulin-potassium treatment in acute myocardial infarction outside a critical care setting2 and an observational study of tight glycaemic control in a general intensive care setting.3 The 2001 study was conducted on a relatively restricted population consisting mainly of post-surgical patients (63% after cardiac surgery) with low admission APACHE II scores and used an unusual feeding regimen. In spite of these limitations, tight glycaemic control rapidly became standard practice in critically ill medical as well as surgical patients in Britain4 and an internationally recommended standard of care in all patients with severe sepsis.5
However, confidence in the benefits of strict glucose control for all critically ill patients is being eroded. Last year the German SepNet group suspended a multicentre randomised controlled trial in medical and surgical patients with severe sepsis.6 Tight glycaemic control produced no reduction in mortality, but it did cause a higher incidence of hypoglycaemia (12.1% v 2.1%). Also last year the CREATE-ECLA trial, a study of insulin-glucose-potassium therapy in 20 000 patients with acute myocardial infarction,7 showed no benefits, removing some of the indirect support for tight glycaemic control. Finally, early this year Van den Berghe and colleagues reported another study of tight glycaemic control, this time in critically ill medical patients,8 and showed no overall benefit on mortality (37.3% v 40% in the control group; P = 0.33).
Why might tight glycaemic control reduce mortality in critically ill patients after surgery but not in critically ill medical patients, where the much higher mortality might be associated with a larger effect? Firstly, the prevention of nosocomial infection, thought to be a major reason for the mortality reduction in the original Van den Berghe study in uninfected surgical patients, did not occur in the medical trial, where established infection was commonly the reason for admission. Secondly, the treatments tested were different. The original surgical study used an unusual feeding regime with 9 g intravenous glucose per hour9 overnight usually followed by parenteral nutrition. This was not repeated in the medical study, where patients were started on enteral feeding once stable. Possibly both insulin and intravenous supplementary glucose are required for benefit. Finally, recent work suggesting that hyperglycaemia on admission is an independent risk factor in patients undergoing intensive care after cardiothoracic surgical procedures but not in those with medical conditions is compatible with blood glucose control showing benefit only in the surgical patients.10
The continuing use of tight glycaemic control in all critically ill patients presupposes that this treatment is harmless. But this may not be the case. In the original surgical study 5.1% of the patients receiving tight glycaemic control had one or more episodes of hypoglycaemia (< 2.2 mmol/l), but they were all also receiving intravenous glucose, and a dedicated doctor cared for the study group. In the subsequent medical study, without glucose infusions or medical support, the figure was 3.7 times this (18.7%), exceeding even the 12.1% seen in the SepNet study. As in the SepNet study, there were more deaths in patients who had hypoglycaemic episodes in the tight glycaemic control group, though this may simply reflect hypoglycaemia as a marker of disease severity.
Trials are underway both to determine the mechanism for the benefit of tight glycaemic control seen in the surgical patients, and to see if the benefit extends to the mixed medical-surgical populations treated in British, European, Canadian, and Australasian intensive care units. In the meantime how should glucose control in the critically ill be managed?
For now tight control (target blood glucose 4.4-6.1 mmol/l) is probably best reserved for critically ill patients after elective surgery in intensive care units with aggressive feeding policies and a high staffing ratio. In the remaining heterogeneous group of critically ill patients, where tight glycaemic control has no proved benefit but the potential of harm through hypoglycaemia, this treatment should be avoided. Instead we should use the common, vigorously contested11 but reasonably supported,12 practice of using less aggressive regimens4 until trials on appropriate populations have reported.
Competing interests None declared.