Comment on “Strict versus moderate glucose control after resuscitation from ventricular fibrillation” by Oksanen et al.

Abstract

Sir: The trial published recently in Intensive Care Medicine by Oksanen [1] investigated a very important clinical question in the management of patients treated with therapeutic hypothermia following successful resuscitation from out-of-hospital cardiac arrest due to ventricular fibrillation: whether there is any benefit in using a strict glucose control regime (4–6mmol/l) compared to a less strict regime (6–8mmol/l) during the 48 h post resuscitation. The authors randomised 90 patients from two ICUs, but unfortunately this number is too small to answer the question posed. The authors calculated their sample size based on the ability to detect a 15% reduction in risk of 30-day mortality between treatment groups. This is a far greater reduction than we see with almost any other intervention in ICU; for example, recombinant human activated protein C provided an absolute risk reduction of only 6.1% in 28-day mortality in the PROWESS trial. To expect that resuscitated survivors of out-of-hospital cardiac arrest would gain a further 15% risk reduction from strict glucose control in addition to the approximately 14% risk reduction in hospital mortality gained from therapeutic hypothermia [2] is extremely optimistic. In a combined analysis of the data from their trials of strict glucose control in medical and surgical ICU [3], Van den Berghe’s group found a 3.2% reduction in hospital mortality in the intention-to-treat group, who received strict glucose control. In the group of patients with ICU stay of at least 3 days, there was a reduction of 7.8% in hospital mortality. It has been argued that as strict glucose control appears to prevent secondary complications in critically ill patients, the absolute risk reduction would be expected to be similar in different trials, regardless of the baseline risk of death [4]. Any further trials of strict versus moderate glucose control must be adequately powered to detect an appropriate mortality reduction, which currently appears to be approximately 3–8%. If the baseline mortality risk is 50% and a two-tailed α of 0.05 is used, a total of approximately 9,000 patients would be needed to have an 80% power to show a difference of 3%, and 1,300 would be needed if an 8% difference were considered. Oksanen and colleagues suggested that their 90-patient study was sufficient to exclude any clinically significant differences between the study groups. However, power calculations suggest it only had approximately 10% power to detect an 8% mortality difference. A large trial of strict glucose control in critically ill patients is currently ongoing, a collaboration between the Australia and New Zealand Intensive Care Society Clinical Trials Group and the Canadian Critical Care Trials Group [5]. It has been powered to detect a 3.8% difference in 90-day all-cause mortality (S. Finfer, personal communication) between a strict glucose control group (4.5–6.0mmol/l) and a less strict regime (8.0–10.0mmol/l). By October 2007 it had recruited 4,500 patients towards its target of 6,100. This will provide further evidence of the generalisability of Van den Berghe’s findings, though it is unlikely that post-cardiac arrest patients will form a large sub-group. Until results of this study are reported, we should not be influenced by the results of underpowered trials.