The question of glycaemic control in the critically ill- unlike many questions in ICU - can be answered with a single paper. However, the path we took to get to this answer is worth reviewing. LITFL has an excellent summary of the relevant literature, all together under one roof in their Glucose Control Literature Summaries. In short, any normal BSL below 10mmol/L seems to be fine, and trying to control things more finely than that will lead to an increased rate of complications.
The DIGAMI study, published in 1995, is discussed briefly but effectively in LITFL. In short, it was never an ICU trial. It was a study of diabetics with MI, irrespective of the significance of their illness. The study found a mortality benefit for the tightly controlled BSl group, and for the first time turned human eyes to the idea that we may need to pay greater attention to stress induced hyperglycaemia of sick patients.
The idea that tight glycaemic control benefits the ICU population had arisen from the notion that we should somehow try to control the normal stress hyperglycaemia response. This response is discussed in greater detail elsewhere; in brief it is the result of cortisol, catecholamines, fasting, and cytokine release. It is supposed to make more glucose available to the active fight-or-flight tissues, but in fact it promotes clotting, impairs leucocyte function, promotes the inflammatory response, and does a number of other counterproductive things. One might be tempted to say that hyperglycaemia is bad for the critically ill population.
Van Den Berghe et al found that intensive glucose control in the ICU had greatly reduced mortality for surgical patients, as well as medical patients (5 years later) and for a period this influenced public opinion.The goalposts for what should be an acceptable BSl had shifed worldwide; everybody was obsessively micromanaging their insulin pumps within a narrow range of 4.4-6.1 mmol/L.
However, not all was well. European trials of this protocol did not find the expected benefit, and in fact had to stop prematurely because too many patients were having hypoglycaemic episodes.
Why was the Leuven trial so far off the mark? It could have something to do with the fact that 87% of their study patients were on TPN. So, of course the control group (whose sugars were left to drift) experienced all the mortality-increasing disadvantages of TPN-associated hyperglycaemia, whereas the ightly controlled group was protected from them. So really, the Van Den Berghe study demonstrated that in patients on TPN, the morbidity and mortality of hypoglycaemia from such tight BSL targets is less of a threat than the morbidity and mortalityfrom hyperglycaemia-related complications of TPN.
In 2009, Finfer et al published the outcomes of their huge multicenter trial, which tested the effects of relaxing our knuckle-whitening grip on glycaemic control. In the control arm, BSL was kept at a more civilized 10.0mml/L. This relaxed population did better; there was a 2.6% increase in mortality associated with tight BSL control. Incorporating this data into a meta-analysis confirmed that overall, there was no mortality benefit with tight sugar control, perhaps with the exception of cardiac surgical ICU patients.
A post hoc analysis of the NICE-SUGAR data was published in 2012, digging deeper and unearthing the prevalence of hypoglycaemia. Patients who had episodes of hypoglycaemia were twice as likely to die, and patients with "intensive" BSL control were ten times as likely to have a hypoglycaemic episode compared to "relaxed" controls. The authors conclude that hypoglycaemia and death had a dose-response relationship, and that this relationship was strongest for patients with "distributive shock".
In spite of the firm and unambiguous "no" to tight glycaemic control, the interest in this practice failed to subside, and further trials were done. The most recent of these was a French attempt to hand the responsibility of a patient's BSL over to a computerised algorithm. Again, with a 4.4-6.1mmol/L BSL target there was no improvement in mortality, and a slight increase in the risk of hypoglycaemia.
However, other people's experience of computerised BSL algorithms has been different (i.e. positive). Van de Berghe has published a recent editorial piece in which he laments the current contempt of strict glycaemic control, and suggests that the brief episodes of hypglycaemia are probably harmless, and hyperglyaemia is uniformly harmful. Furthemore, the overall trend seems to be towards the abolition of human control over the BSL, and the introduction of closed-loop automatic systems, armed with insulin and dextrose, which vigilantly maintain a tight hypoglycaemia-free BSL control.
A recent review by Mesotten et al (2015) makes recommendation which is "not based on findings from randomised controlled trials, but merely represents a very common, pragmatic approach by physicians at the bedside".
Why do we bother?
Why is it bad to have a high BSL in the ICU?