Abstract
The largest and perhaps most important trial was the multicenter randomized controlled NICE-SUGAR study in 2009. 10 Contrary to expectations, these investigators did not confirm the previous findings of the Leuven trials; moreover, they showed an increase in 90-day mortality from 24.9% to 27.5% (P=0.02) in the intensive insulin therapy group (glucose target, 4.5–6.0 mmol/L) compared with standard therapy (glucose target, <10 mmol/L) in a mixed ICU population. In addition, there was an increase in the incidence of hypoglycemia in the intensive insulin therapy group compared with the conventional-control group (6.8% vs. 0.5%, P<0.001).
Why all these different outcomes from different studies? Several possibly important methodological differences (i.e., different target ranges of blood glucose in the intervention group and the control group, differences in methods used for blood glucose measurements, and different sampling sites) may have contributed to the different outcomes. 14 The more attractive explanation for the different outcomes, however, is the differences in feeding strategies in the major clinical trials. 15 The high amount of parenteral nutrition used in the Leuven trials, compromising a higher total glucose load compared with enteral nutrition, may have increased the severity of stress-induced hyperglycemia, and thus intensive insulin treatment may merely have contravened a side effect of parenteral nutrition. In contrast, tight glycemic control in patients receiving enteral nutrition, as was the feeding strategy used almost exclusively in the NICE-SUGAR study, 10 seems to be harmful, possibly because there is an increased risk of hypoglycemia. The results of the meta-regression analysis of Marik and Preiser 15 suggest that intensive insulin treatment is only of benefit in patients receiving parenteral nutrition, and particularly so in those with a low severity of illness.
Recently, the EPaNIC study 16 compared two feeding strategies in the context of normoglycemia in critically ill patients. They compared early initiation of parental feeding, as supplement to enteral nutrition, thereby preventing a caloric deficit in critical illness (the early-initiation group), with withholding parental feeding during 1 week (the late-initiation group). The late initiation of parental feeding appeared to be superior; specifically, the study found a significantly shorter median duration of hospital stay (relative increase of 6.3% in the likelihood of being discharged earlier from hospital [95% confidence interval, 1.00–1.13; P=0.04]) and significantly fewer complications (22.8% vs. 26.2%, P=0.008). It is important to mention that this study targeted a blood glucose level of 4.4–6.1 mmol/L and that the glucose target level achieved was almost the same in the two groups (5.7±0.8 mmol/L in the late-initiation group vs. 5.9±1.0 mmol/L in the early-initiation group). It remains unclear whether an intermediate blood glucose target, as recommended nowadays, would have affected the outcome of this trial.
Returning to how to translate the contradictory outcomes of the different trials into clinical practice, consensus grew that an intermediate blood glucose target range would be beneficial. Siegelaar et al. 11 provided epidemiological support for this and showed in a retrospective database cohort study that there is a U-shaped curve relationship between mean glucose and mortality during ICU stay in a mixed ICU population. Hence, a “safe range” for mean glucose between 7.0 and 9.0 mmol/L could be defined. Such an increase in the blood glucose target range is very likely to diminish the incidence of hypoglycemia at the same time. Of note is that Hermanides et al. 17 showed that hypoglycemia, even with adjustment for severity of disease using the daily assessed Sequential Organ Failure Assessment score, is related to an increase in ICU death. The defined “safe” blood glucose range by Siegelaar et al. 11 and the importance of diminishing hypoglycemia are in line with the present recommendations for in-hospital glucose targets from the American College of Physicians 18 and the American Association of Clinical Endocrinologists and American Diabetes Association 19 to maintain a medium-high glucose level (7.8–11.1 and 7.8–10.0 mmol/L, respectively) and to avoid hypoglycemia.
Unfortunately, contradiction in the outcomes of clinical trials, as described above, is not uncommon. It generates substantial uncertainty for clinical practitioners and for experts who develop guidelines. Professional societies immediately issued guidelines on target glucose levels in critically ill patients after the report on the initial Leuven study in 2001. 1 However, with publication of the results of subsequent studies, most professional societies have adjusted their guidelines by increasing the glucose target range. With hindsight, professional societies were too hasty to integrate the results of an early, single-center, nonblinded study, such as the Leuven trial, 1 into their clinical guidelines. It is important to mention, and also as a cautionary note to the authors and the readers of these guidelines, that even very highly cited randomized controlled trials may be refuted over the time. 20 Of note is that the workload associated with establishing tight glycemic control is substantial.
However, defining the glucose target range by ending up in the middle between the Leuven and NICE-SUGAR criteria may be somewhat of a poor man's answer. How do we know that higher targets aren't even better? Or can we even be sure that stress hyperglycemia should be treated at all? Presently, these questions are still unanswered. The only solution would be a three-armed randomized trial comparing strict, less strict, and minimal glycemic control in different critically ill populations. The number of patients needed in this trial would be immense, and so will be the cost. Therefore, it may well be that such a trial will never be performed.
Although many researchers have focused on glycemic control in critically ill patient populations in an ICU setting, another interesting question is whether tight glycemic control in the coronary care unit would be beneficial. The initial concept—that metabolic modulation of acute myocardial infarction (AMI) by energy supply with glucose–insulin–potassium (GIK) infusion would improve the outcomes in patients suffering from acute coronary disease 21 —was ultimately studied by the large multicenter randomized controlled CREATE-ECLA 22 trial. These researchers found that high-dose GIK infusion in patients with acute ST-elevation myocardial infarction did not have a beneficial effect on mortality rate (10% in the GIK infusion group vs. 9.7% in the control group, P=0.45). This trial and a subsequent meta-analysis by Puskarich et al. 23 rejected the thought that GIK infusion is beneficial in the treatment of AMI. However, this trial was not designed for glucose control as a primary target, which is important because GIK generally increases glucose concentrations.
Other studies focused on achieving normoglycemia by glucose control with intensive insulin therapy. The DIGAMI study, 24 published in 1995, showed a 1-year mortality rate reduction of 29% (P=0.027) in diabetes patients after an AMI using intensive insulin treatment. However, subsequent studies as the DIGAMI-2 trial 25 and the HI-5 study 8 failed to replicate this result, possibly because these studies were hampered by limited sample sizes. Moreover, since the development of improved reperfusion treatments (i.e., percutaneous coronary intervention) has contributed to an impressive decrease in mortality in patients with AMI 26 and because of the substantial variability in insulin treatment rates across medical centers in a coronary care unit setting, 27 there is a need to confirm the possible benefits of glucose lowering in populations both with and without diabetes treated with state-of-the-art interventions.
Few clinical trials have focused on the optimal management of glycemic control at the ward. Recently, the RABBIT 2 surgery trial 28 investigated the efficacy and safety of a basal–bolus insulin regimen compared with sliding-scale regular insulin in patients with type 2 diabetes admitted to general surgery wards. They showed a significant improvement in glycemic control in the basal–bolus insulin group compared with the sliding-scale regular insulin group (P<0.01), with a lower mean fasting glucose (8.6±2.1 vs. 9.2±2.2 mmol/L, P=0.037) and a lower mean daily glucose during the hospital stay (8.7±1.8 vs. 9.8±2.4 mmol/L, P<0.001). Also, perioperative complications, defined as a composite end point including wound infection, pneumonia, bacteremia, respiratory failure, and acute renal failure, was significantly reduced in the basal–bolus treatment group (8.6% vs. 24.3%, P=0.003). These results are in accordance with the recommendation of the American Association of Clinical Endocrinologists and American Diabetes Association guideline, 19 that scheduled subcutaneous administration of insulin is the preferred method for achieving and maintaining glucose control in noncritically ill patients. Prolonged therapy with sliding-scale regular insulin is discouraged. The blood glucose targets recommended by the American Association of Clinical Endocrinologists and American Diabetes Association guideline 19 on medical or surgical wards in noncritically ill patients treated with insulin are premeal blood glucose values <7.8 mmol/L in conjunction with random blood glucose values <10 mmol/L, as long as these targets can be safely achieved. As mentioned in the guideline, these targets are based on clinical experience and judgment. The blood glucose values achieved in the RABBIT 2 surgery trial 28 indicate that these targets are not easily met, and there is a lack of evidence to support the treatment of patients without diabetes on the ward with insulin.
In conclusion, a decade of intensive clinical research did not provide us with a clear and simple answer to the complex problem of in-hospital glycemic control. Although the current consensus to maintain a medium-high glucose level in an ICU setting is reasonable, we emphasize that further clinical research toward in-hospital glycemic control in noncritically ill patients is warranted.