As reported in Critical Care, in a very detailed investigation of a set of 37 patients scheduled for cardiac surgery, Lehrke and colleagues 5 sought to better understand the determinants of insulin resistance, the major mechanism of stress hyperglycaemia 67. These investigators hypothesised that both inflammatory mediators (the cytokines interleukin-6 [IL-6] and tumour necrosis factor) and hormones (cortisol, resistin, leptin and adiponectin) independently increased the insulin resistance. This hypothesis is appealing as it is based on robust recent evidence 8 from data recorded in obese patients and in inflammatory situations. The resistance to insulin was assessed by both the insulin requirements to maintain 'euglycaemia' and an insulin glycaemic index deduced from the result of the multiplication of blood glucose by the circulating insulin concentration (the sum of exogenous and endogenous insulin). Using a likelihood ratio test, these authors could elegantly show that cortisol was the best predictor of insulin resistance, followed by IL-6, leptin and adiponectin. Daily clinical practice could be influenced in the future as these data can serve to identify additional risk factors for the development of insulin resistance, namely before surgery is scheduled 9.

Another interesting physiological approach was published last year by Otto and colleagues 10. These investigators sought to discriminate the effects of insulin from those of hyperglycaemia in a cell culture system of human polymorphonuclear cells stimulated with a low dose of lipopolysaccharide. The outcome variable recorded included the release of IL-1 and IL-6, the phagocytosis activity, the oxidative burst and a flow cytometric analysis. The most salient finding was the induction of IL-6 release after incubation of cells, already with 250 mg/dL of glucose, a concentration that can be found in clinical situations. The release of IL-1 needed higher glucose concentrations and the hyperglycaemia-induced production of both cytokines was partially inhibited by insulin. Both the oxidative burst and the phagocytic activity were decreased by hyperglycaemia and by hyperosmolarity. From a clinical standpoint, these data can be viewed as a partial explanation of the susceptibility of hyperglycaemic patients or poorly controlled diabetics to infections and ischaemia.

Data reported from clinical investigations in Critical Care last year can be regarded as either 'observational' or 'interventional'. In the first category, two different teams investigated particular aspects of glucose metabolism during moderately strict glucose control (target blood glucose of 4.4 to 8.0 mmol/L) in selected populations of critically ill patients. The first study was performed in 31 patients with sub-arachnoid haemorrhage 11. In these patients, several indices of brain metabolism and damage (glucose, lactate, pyruvate, glycerol and glutamate) were continuously measured by microdialysis, and correlations with the blood glucose levels were followed. Basically, the authors found that the glucose concentration measured in the cerebral tissue was much lower than the plasma concentration. In spite of the low brain glucose levels, there was no evidence of brain energy failure, as the lactate and pyruvate concentrations remained unchanged. However, the glycerol level tended to increase, consistent with cell membrane damage. These important findings in a selected group of patients are consistent with earlier work from Vespa and colleagues 12, who performed a similar study in a broader population of brain-injured patients. The data presented by Schlenk and colleagues 11 are consistent with the recent report of Oddo and colleagues 13, who reported a risk of brain energy deficit (defined by an elevated lactate-to-pyruvate ratio) in a population of 20 brain-injured patients when blood glucose fell to below 6.7 mmol/L. The data presented by Schlenk and colleagues 11 will definitely contribute to the development of recommendations for blood glucose targets in the specific population of 'neurocritical care patients' in whom both hyperglycaemia and hypoglycaemia, even moderate, can induce detrimental effects, as recently reviewed by Bilotta and colleagues 14.

In another setting, but using the same intermediate glucose target, Waeschle and colleagues 15 compared the severity of sepsis with the time course of glycaemia, including the incidence of hyperglycaemia and hypoglycaemia and glucose variability, in a population of 191 patients. This thorough analysis was clearly needed to confirm the long-held belief that the incidences of both hyperglycaemia and hypo-glycaemia, and hence glucose variability, increased as a function of the severity of sepsis. Interestingly, the multivariate analysis indicated that the frequency of hypoglycaemia and the amplitude of glucose variability were correlated and represented predictors of poor prognosis. These important findings corroborate other data recorded in septic patients (that is, the association between high glucose variability and poor outcome in septic patients) 16. The correlation between hypoglycaemia and higher mortality was reported in each of the prospective trials on glucose control 17 and was found as an independent predictor of mortality in a large cohort of critically ill patients 18. However, this correlation may not imply a direct toxicity of insulin treatment, as recently suggested in patients after myocardial infarction 19.

Two important interventional studies and a detailed analysis of the indices of the quality of glucose control were also published last year in Critical Care. The first interventional study 2, performed in Colombia as a single-centre trial in a mixed (medico-surgical) intensive care unit (ICU), was designed to confirm the external validity of the Leuven I study 1. Twenty-eight-day mortality was the primary outcome variable, and 504 patients were enrolled and randomly assigned to intensive insulin therapy (target of 4.4 to 6.1 mmol/L) versus a 'standard' target of 10.0 to 11.1 mmol/L. Even though there was an overlap between blood glucose achieved in both groups, no advantage in favour of intensive insulin therapy was found. This study, like several other prospective randomised controlled trials 342021, including one study in very-low-birth-weight infants 22, was unable to confirm the survival benefit afforded by tight glucose control by intensive insulin therapy reported in the pioneering study 1. The results of two recent meta-analyses 2324 indeed show no survival benefit but a fivefold increase in the rate of hypoglycaemia in the patients randomly assigned to intensive insulin therapy. However, these trials should not discourage the intensive care community from searching for improvements in glucose control, possibly using different glucose targets 25. In this regard, the data provided by Chase and colleagues 26 are of major importance. Using a combined strategy of adaptation of insulin infusion and feeding rates in a cohort of 371 patients, these investigators reported remarkable results in terms of percentage of blood glucose values within the desired range. This represents a major improvement as compared with the previous performance of this centre, assessed in a retrospective cohort of 413 matched patients. Very importantly, this improvement in glucose control was associated with decreased hospital mortality in patients who stayed in the ICU for more than 3 days. These very impressive data suggest that the degree of achievement of a pre-defined glucose target is the key factor of any strategy used to control glycaemia. In fact, there are several indicators of the success and 'quality' of glucose control. The long list of published quality indicators was systematically reviewed by Eslami and colleagues 27. As stated by the authors, there is a clear need for an unambiguous indicator reference subset rather than a jungle of potential indicators. This important article will probably be very helpful for many investigators in the field of glucose control when designing interventional studies and selecting end points and indices of quality of any intervention.