Jean-Charles Preiser Reply to Schultz et al. Accepted: 6 August 2009 Published online: 23 September 2009 Ó Copyright jointly hold by Springer and ESICM 2009 This reply refers to the comment available at: doi:10.1007/s00134-009-1666-2. On behalf of the co-authors of the ‘‘Glucontrol’’ study. We thank Schultz et al. for their thoughtful and constructive com- ments on the Glucontrol study [1]. We agree that changes in the practice of glucose control have likely occur- red in the recent years since the study was designed (in 2002–2003), during the enrolment of patients (2004– 2006) and from the analysis to the publication of the results (2006– 2009). In addition to differences in blood glucose (BG) targets and a more frequent use of insulin, other changes including an increase in the frequency of BG checks, technologi- cal improvements in BG analysers and increased attention and commit- ment to glucose control have occurred as well. At the time of designing the study, we were aware that such a trial mainly aimed to address practical issues of glucose control. Specifically the key question raised by the inten- sive care community was: ‘‘Would my patients benefit from tight glucose control, as compared to the usual practice?’’ Therefore we conducted a survey on the usual practice of potential participants in 2003 and found that the commonest BG threshold to start intravenous insulin was ‘‘intermediate’’ (8.3 mmol/L [ 2]) (see also Table 1 of manuscript [1]). This was the reason for the choice of an intermediate target for glucose control in the ‘‘usual care/control group’’. In fact,the adherence to the target BG in the control group, esti- mated from the means or medians of morning BG values (Table 1), was in fact better in Glucontrol [1] and in NICE-SUGAR [3] than in the trials using a higher BG target for the control group [4–8]. This suggests that the intermediate target range is a more accurate reflection of the usual care than a higher target range. Inevitably, the choice of closer targets than in the pioneering study [4] resulted in an increase in the required size of the study sample. The power calculation, based on the data available in 2002–2003, i.e.a 4% decrease in absolute ICU mortality, and on the mortality data of the participating centres [2], concluded that the sample size needed to detect 4% decrease in mortality with an alpha error rate of 5% and beta error rate of 20% was 1,496 patients in each group. As specified in the discussion, with the final set of 1,078 patients, the power to detect a 4% difference in ICU mortality with an alpha error rate of 5% and beta error rate of 20% was eventually 32%. We are sorry about a potential confusion for the reasons for stopping the recruitment, and the exact word- ing of the DSMB report has been included in the paper [ 1]: ‘‘Following the first interim analysis performed after the 100th death, the DSMB recommended to stop the inclusion of patients. Specifically, the proportion of BG values in the assigned range calculated from the BG readings available at the time of the interim analysis [54.8% (group 1) and 27.8% (group 2) (39.2% after exclusion of day 1 data)] were deemed as a high rate of unintended protocol violation rate.The increased rate of hypogly- caemia in the IIT group was not considered as a safety concern by the DSMB.’’ Of note,there was no mention in this report of an excessive overlap of BG values between groups, but only a too low proportion of BG values in the assigned range. Importantly the proportion of values in range was calculated from all BG values and not Table 1 Mean or median of morning blood glucose (BG) values reported from the prospective controlled trials, in the ‘‘in the ‘‘control’’ groups (corresponding BG targets are given in the column headings) Study Intensive group (target 4.4–6.1 mmol/L) Control group (target 7.8–10.0 mmol/L) Control group (target 10.0–11.1 mmol/L) Glucontrol (median and IQR) [1] 6.1 (5.5–6.8) 7.7 (6.7–8.8) NICE-SUGAR (mean ± SD)[3] 6.6 ± 1.4 8.1 ± 1.4 Leuven I (mean ± SD) [4] 5.7 ± 1.0 8.5 ± 1.8 Leuven II (mean ± SD) [5] 6.2 ± 1.6 8.5 ± 1.7 Brunkhorst (median and IQR) [6] 6.2 (6.1–6.3) 8.4 (8.2–8.6) De la Rosa (median and IQR) [7] 6.5 (5.6–7.8) 8.2 (6.8–10.0) Arabi (mean ± SD) [8] 6.4 ± 1.0 9.5 ± 1.9 The corresponding BG targets are mentioned in the title of the columns Intensive Care Med (2010) 36:175–176 DOI 10.1007/s00134-009-1669-z CORRESPONDENCE