Opinions regarding the control of blood glucose in critically ill patients have fluctuated dramatically over the past decade, and while our knowledge is far greater than in 2001, many questions remain unanswered.
Hyperglycaemia is highly prevalent during critical illness, with 90% of patients treated in intensive care units (ICU) experiencing blood glucose concentrations greater than 6.1 mmol/L.1 The hypermetabolic ‘stress’ state is associated with both increased hepatic glucose production and peripheral insulin resistance, which is compounded by the hyperglycaemic effects of treatments, such as corticosteroids, sympathomimetic agents and glucose containing infusions. During the 1990s, the prevailing opinion was that hyperglycaemia was a normal and potentially beneficial physiological response to critical illness that promoted cellular glucose uptake.2 Standard care was to tolerate hyperglycaemia as an adaptive response and to start insulin only when blood glucose concentration exceeded the renal threshold (12 mmol/L or 215 mg/dL), with the goal of preventing glycosuria and hypovolaemia.3
Opinions began to alter after the DIGAMI (Diabetes Mellitus Insulin-Glucose Infusion in Acute Myocardial Infarction) trial was published in 1995.4 In that trial, more intensive glucose control (IGC) to maintain a blood glucose concentration below 11.9 mmol/L resulted in improved long-term outcome in patients with diabetes who had suffered acute myocardial infarction. A far greater shift in opinion occurred following the publication of Van den Berghe's landmark paper in 2001.1 Van den Berghe compared IGC (target blood glucose concentration between 4.4 and 6.1 mmol/L (80–110 mg/dL)) with conventional control (blood glucose treated when it exceeded 12.0 mmol/L (216 mg/dL)) with subsequent target of 10.0–11.1 mmol/L (180–200 mg/dL) in patients treated in a surgical ICU in Leuven, Belgium. The trial recruited 1548 patients and reported that intensive glucose control resulted in a 34% reduction in the risk of in-hospital death. The external validity of the trial was questioned because of the high mortality rate in both arms,5 being a single centre study, lack of blinding, excessive administration of intravenous glucose (200–300 g/day) and early termination after multiple looks at the data. Despite these concerns, IGC became a recommended if not mandated treatment.6
Studies designed to confirm Van den Berghe's findings reported less encouraging results. In 2006, Van den Bergh reported a near identical trial in the medical ICU at the same hospital in Leuven. In the second trial, in-hospital mortality was not significantly reduced although there was evidence of reduced morbidity in the form of reduced time to weaning from mechanical ventilation, reduced acute kidney injury and reduced time to discharge from both the ICU and the hospital.7
The first multicentre trial examined the role of IGC in patients with severe sepsis treated in 18 academic tertiary centres in Germany.8 The trial was stopped early for safety reasons because patients randomised to IGC suffered a significantly increased incidence of severe hypoglycaemia, without any evidence of a beneficial effect on either 28- or 90-day mortality. Despite this, guidelines continued to recommend IGC.9
The Glucontrol study was another multicentre trial in Europe that compared the effect of IGC and conventional glucose control (target blood glucose, 7.8–10.0 mmol/L (140–180 mg/dL)) on the mortality of patients admitted to 21 medical-surgical ICU.10 Although designed to recruit 3500 patients, the trial was stopped early because of a high number of protocol violations after the recruitment of 1078 patients. The trial also reported no significant difference in ICU mortality (17.2% for IGC vs 15.3% for conventional treatment, P= 0.41) and again an increased rate of hypoglycaemia in the IGC group (8.7% vs 2.7%, P < 0.0001).
Despite several trials and meta-analyses11 concluding there was no significant benefit from IGC, it took the publication of The Normoglycemia in Intensive Care Evaluation – Survival Using Glucose Algorithm Regulation (NICE-SUGAR) Study12 to change practice recommendations. The NICE-SUGAR study examined the effect of IGC on 90-day all-cause mortality in 6104 critically ill patients drawn from 42 hospitals in Australia, New Zealand, Canada and the United States. The primary outcome measure was available for 6022 patients, and those assigned to IGC had an increased risk of death (27.5% vs 24.9%, P= 0.02). This increased risk of death occurred despite a lower rate of hypoglycaemia in the IGC group than that reported in previous studies of combined surgical and medical patients. Meta-analyses have subsequently included the NICE-SUGAR data and concluded more strongly that IGC is not beneficial in critically ill patients and that it does increase the risk of severe hypoglycaemia.1314
Although the totality of the evidence indicates that intensive glucose control is not beneficial in critically ill patients, we are still left trying to explain and understand why the results of the trials conducted in Leuven differ so markedly from the trials conducted elsewhere. While much attention has been focused on the blood glucose target and mean blood glucose concentration, it has become apparent that various features of blood glucose control are associated with mortality; these include the incidence of hypoglycaemia, measures of blood glucose variability and other measures of the degree of hyperglycaemia.15 While not definitively causal, there is a clear association between the occurrence of hypoglycaemia and subsequent death. A recent retrospective observational study reported the relationship between mild (<4.5 mmol/L (<81 mg/dL)) and severe (<2.1 mmol/L (40 mg/dL)) hypoglycaemic episodes and death.16 Of 4946 patients studied, 1109 (22.4%) experienced hypoglycaemia and, in these patients, hospital mortality was 36.6% compared with 19.7% in those who did not experience hypoglycaemia (P < 0.001). Whether hypoglycaemia is simply a marker of severity of illness or whether it is in itself harmful remains uncertain, nevertheless available data suggest that avoiding even mild hypoglycaemia should be a prominent goal when managing blood glucose in critically ill patients.17
Increased blood glucose variability is also independently associated with increased risk of death.18 Whether this is because reduced blood glucose variability reflects more meticulous medical and nursing care, or is a marker of less severe illness, or that increased variability has a deleterious biological effect in critically ill patients, is currently unknown. Van den Berghe concluded that IGC did not reduce blood glucose variability in the trials in Leuven19 and this makes it an unlikely explanation for mortality differences between the Leuven trials and others.
With increased interest in glycaemic control, the systems currently used to measure blood glucose have been subjected to great scrutiny. Point-of-care (POC) systems were not developed to guide the administration of insulin in critically ill patients, and current standards allow a measurement error of up to 20%.20 As a result even when used meticulously, they are not considered accurate enough to guide therapy aimed at keeping blood glucose within a 1.7 mmol/L (30 mg/dL) range.21 Many POC systems do not account for the patient's haematocrit or degree of oxygenation both of which may cause errors in blood glucose measurement; in anaemic patients, the standard correction factor may result in falsely high blood glucose readings,22 and in hypoxic patients, POC devices may also report a falsely elevated result.23 Presently, for optimal blood glucose control in the ICU, blood glucose should be measured using an accurate glucose electrode, such as those found in many blood gas analysers or in hospital laboratory systems. Blood should be sampled from arterial or central venous catheter rather than finger-prick capillary samples, which increase inaccuracy because of tissue oedema, impaired peripheral perfusion and inadequate sample volume.24
Patients with pre-existing diabetes mellitus may benefit from a higher glucose concentration than that considered safe and desirable in critically ill patients without diabetes. Identifying patients with diabetes after admission to an ICU may be difficult as conventional diagnostic criteria do not apply and hyperglycaemia may be attributable to critical illness alone. The HbA1c level (>6.5%) measured at admission to ICU seems the most useful way to identify pre-existing diabetes although it is insensitive and not applicable to all racial groups.25 A large, two-centre, retrospective observational study suggested that unlike other patients, patients with diabetes show no clear association between hyperglycaemia during ICU stay and mortality; patients with diabetes have markedly lower odds ratios of death at all levels of hyperglycaemia.26 This is consistent with published data as even in Van den Berghe's trials patients with diabetes did not benefit from IGC.27 The optimal target range in critically ill adults with diabetes remains unclear.
The last decade has seen a dramatic evolution of our understanding of glucose control in critically ill patients. While the topic is the most intensively investigated area of critical care practice, the answers to fundamental questions remain unclear. With future promise of more accurate and continuous blood glucose monitoring systems, computerised decision support and closed-loop control of blood glucose,28 tighter glucose control may yet be shown to be safe and beneficial. However, until phase III studies provide sufficient evidence to the contrary, a cautious approach seems prudent. Plasma glucose concentrations should be monitored closely in critically ill adults using an accurate glucose electrode. If the blood glucose concentration exceeds 10 mmol/L, insulin therapy should be commenced, with the goal of keeping blood glucose below 10 mmol/L while avoiding even moderate hypoglycaemia.