Intensive versus Conventional Glucose Control in Critically Ill Patients

Study Design

We conducted a parallel-group, randomized, controlled trial involving adult medical and surgical patients admitted to the ICUs of 42 hospitals: 38 academic tertiary care hospitals and 4 community hospitals. Eligible patients were those expected to require treatment in the ICU on 3 or more consecutive days (see Appendix A in the Supplementary Appendix, available with the full text of this article at NEJM.org). A detailed description of the study was published previously.²⁵

The study was approved by the ethics committees of the University of Sydney, the University of British Columbia, and each participating institution. Written informed consent, obtained before randomization, or delayed consent was obtained from each patient or from a legal surrogate.

Study participants were randomly assigned to glucose control with one of two target ranges: the intensive (i.e., tight) control target of 81 to 108 mg per deciliter (4.5 to 6.0 mmol per liter), based on that used in previous studies,^12,13 or a conventional-control target of 180 mg or less per deciliter (10.0 mmol or less per liter), based on practice surveys in Australia, New Zealand, and Canada.^23,25 Randomization was stratified according to type of admission (operative or nonoperative) and region (Australia and New Zealand or North America). Patients were randomly assigned to a treatment group by the clinicians treating them or by local study coordinators, with the use of a minimization algorithm²⁶ accessed through a secure Web site. The treatment assignments were concealed before randomization, but subsequently, clinical staff were aware of them.

Control of blood glucose was achieved with the use of an intravenous infusion of insulin in saline. In the group of patients assigned to undergo conventional glucose control, insulin was administered if the blood glucose level exceeded 180 mg per deciliter; insulin administration was reduced and then discontinued if the blood glucose level dropped below 144 mg per deciliter (8.0 mmol per liter). Blood glucose levels in each patient were managed as part of the normal duties of the clinical staff at the participating center. In both groups, this management was guided by treatment algorithms accessed through a secure Web site (for details of the treatment algorithm, see https://studies.thegeorgeinstitute.org/nice/).

The trial intervention was discontinued once the patient was eating or was discharged from the ICU but was resumed if the patient was readmitted to the ICU within 90 days. It was discontinued permanently at the time of death or 90 days after randomization, whichever occurred first.

Blood samples for glucose measurement were obtained by means of arterial catheters whenever possible; the use of capillary samples was discouraged. Blood glucose levels were measured with the use of point-of-care or arterial blood gas analyzers or laboratory analyzers in routine use at each center. All other aspects of patient care, including nutritional management, were carried out at the discretion of the treating clinicians.

Assessments and Data Collection at Baseline

Local study coordinators at each institution collected the data; source data were verified by study monitors from regional coordinating centers. At baseline, demographic and clinical characteristics, including the Acute Physiology and Chronic Health Evaluation II (APACHE II) score²⁷ (which can range from 0 to 71, with higher scores indicating more severe illness) and the diagnostic criteria for severe sepsis,²⁸ were collected. Admissions to the ICU directly from the operating or recovery room were classified as operative admissions. Patients were classified as having diabetes on the basis of their medical history and were classified as having trauma if the ICU admission occurred within 48 hours after admission to the hospital for trauma. Previous treatment with corticosteroids was defined as treatment with systemic corticosteroids for 72 hours or more immediately before randomization.

From the time of randomization to the time of discharge from the ICU or 90 days after randomization (whichever came first), we recorded all blood glucose measurements, insulin administration, red-cell administration, blood cultures that were positive for pathogenic organisms, type and volume of all enteral and parenteral nutrition and additional intravenous glucose administered, and corticosteroid administration. Also recorded were the cardiovascular, respiratory, renal, hepatic, and hematologic components of the Sequential Organ Failure Assessment (SOFA, for which scores can range from 0 to 4 for each organ system, with higher scores indicating more severe dysfunction)²⁹ and the use of mechanical ventilation and renal-replacement therapy.

Outcome Measures

Outcome measures and statistical analyses were defined in a prespecified statistical-analysis plan.³⁰ The primary outcome measure was death from any cause within 90 days after randomization, in an analysis that was not adjusted for baseline characteristics. Secondary outcome measures were survival time during the first 90 days, cause-specific death (see Appendix C in the Supplementary Appendix for more information), and durations of mechanical ventilation, renal-replacement therapy, and stays in the ICU and hospital. Tertiary outcomes were death from any cause within 28 days after randomization, place of death (ICU, hospital ward, or other), incidence of new organ failure, positive blood culture, receipt of red-cell transfusion, and volume of the transfusion.

The primary outcome was also examined in six predefined pairs of subgroups: operative patients and nonoperative patients, patients with and those without diabetes, patients with and those without trauma, patients with and those without severe sepsis, patients treated and those not treated with corticosteroids, and patients whose APACHE II score was 25 or more and those whose score was less than 25.³⁰

Serious Adverse Events

A blood glucose level of 40 mg per deciliter (2.2 mmol per liter) or less was considered a serious adverse event. When the blood glucose level was measured with a bedside point-of-care analyzer, we requested that the treating clinician obtain a blood sample for laboratory confirmation before treating the presumed hypoglycemia. The details of each event were reviewed by the two study management committees and submitted to the research ethics committees of all participating centers and to the independent data and safety monitoring committee.

Statistical Analysis

The study was originally designed to enroll 4000 patients. On the basis of data reported by Van den Berghe et al. in 2006,¹³ the sample size was increased to 6100, thereby providing a statistical power of 90% to detect an absolute difference in mortality between the two groups of 3.8 percentage points, assuming a baseline mortality of 30% at a two-sided alpha level of less than 0.05. All data were analyzed according to the intention-to-treat principle, with no imputation for missing values. The primary analysis for death at 90 days was performed with the use of an unadjusted chi-square test. A secondary analysis based on logistic regression was also conducted, with the strata used for randomization (type of admission and geographic region) as covariates, as well as age, location before ICU admission, APACHE II score, and use or nonuse of mechanical ventilation at baseline. Other binary end points were analyzed by means of a chi-square test or Fisher's exact test. Continuous variables were compared with the use of unpaired t-tests, Welch's tests, or Wilcoxon rank-sum tests. All odds ratios and their corresponding 95% confidence intervals were calculated according to the profile-likelihood method. The time from randomization to death in the two treatment groups was compared with the use of the log-rank test, and the results are presented as Kaplan–Meier curves. Hazard ratios were obtained from Cox models. The time-weighted blood glucose level (with weighting based on the time difference between two consecutive measurements applied to the average of the two consecutive measurements) was computed for all patients for whom data were available.

Subgroup analyses for the primary outcome were based on an unadjusted test of interaction in a logistic model. Estimated distributions of individual patients' average time-weighted blood glucose levels, according to treatment group, were obtained by fitting generalized lambda distributions with the use of the method of maximum likelihood.^31,32 All analyses were conducted with the use of S-PLUS software (version 8.0) and R software (version 2.7.0), and the results were verified independently with SAS software (version 9.1). The data were analyzed by the Statistical Services Division of the George Institute for International Health (University of Sydney).

Two preplanned interim analyses were performed by an independent statistician when 1500 and 4000 of the 6100 patients (25% and 66%, respectively) reached the 90th day of follow-up. The analyses were reviewed by the independent data and safety monitoring committee, which was charged with recommending that the trial be stopped if there was evidence beyond a reasonable doubt of a difference in the rate of death from any cause between the two treatment groups. Since a difference of 3 SE was used as a guideline to recommend early stopping, the final mortality analysis was conducted with an alpha of 0.048.

Study Participants

Figure 1. Figure 1. Assessment, Randomization, and Follow-up of the Study Patients.

A total of 14 of the expected 1132 monthly screening logs (1.2%) were not received at the coordinating center. ICU denotes intensive care unit.

Participants were recruited and had follow-up during the period from December 2004 through November 2008; 6104 were randomly assigned to one of the two treatment groups: 3054 to intensive glucose control and 3050 to conventional glucose control (Figure 1). Where delayed consent was permitted, every patient or a legal surrogate was approached for consent. Delayed consent to the use of study-related data was withheld, or previously obtained consent was withdrawn, for 38 of the 3054 patients (1.2%) assigned to intensive control and 36 of the 3050 patients (1.2%) assigned to conventional control; thus, study data were available for 3016 and 3014 patients, respectively. At 90 days, an additional six patients (0.2%) in the intensive-control group and two (0.1%) in the conventional-control group were lost to follow-up. Of the 6030 patients for whom study data were available, 5275 (87.5%) were recruited in Australia or New Zealand.

Table 1. Table 1. Baseline Characteristics of the Study Patients.

The baseline characteristics of the treatment groups were similar (Table 1). The mean (±SD) age was 60.4±17.2 and 59.9±17.1 years in the intensive-control group and the conventional-control group, respectively; the percentage of male patients, 62.6% and 64.2%; the mean APACHE II score, 21.1±7.9 and 21.1±8.3; and the percentage of operative admissions, 36.9% and 37.2%.

The assigned study treatment, intensive or conventional blood glucose management according to the study-treatment algorithm, was administered to 5997 of 6030 patients (99.5%): 2998 of 3016 (99.4%) in the intensive-control group and 2999 of 3014 (99.5%) in the conventional-control group. The median duration of study treatment was 4.2 days (interquartile range, 1.9 to 8.7) in the intensive-control group and 4.3 days (interquartile range, 2.0 to 9.0) in the conventional-control group (P=0.69).

Study treatment was discontinued prematurely in 304 of 3054 patients (10.0%) in the intensive-control group and 225 of 3050 patients (7.4%) in the conventional-control group. Reasons for discontinuation were withdrawal because of a request by the patient or surrogate (26 patients [0.9%] assigned to intensive control and 22 patients [0.7%] assigned to conventional control) or by the treating physician (115 patients [3.8%] and 48 patients [1.6%], respectively), because of serious adverse events (13 patients [0.4%] and 1 patient [<0.1%], respectively), because of a change in the focus of treatment to palliative care (116 patients [3.8%] and 115 patients [3.8%], respectively), and miscellaneous reasons (34 patients [1.1%] and 39 patients [1.3%], respectively).

At the completion of the trial, data on vital status 90 days after randomization were unavailable for 82 of 6014 patients (1.4%), 44 in the intensive-control group and 38 in the conventional-control group. For 74 of these patients, the vital-status data were missing because consent had been withheld or withdrawn (Figure 1).

Insulin Administration and Treatment Effects

Table 2. Table 2. Blood Glucose Management and Levels, Calorie Administration, and Corticosteroid Treatment, According to Treatment Group.

Figure 2. Figure 2. Data on Blood Glucose Level, According to Treatment Group.

Panel A shows mean blood glucose levels. Baseline data are the averages of the last blood glucose measurement obtained before randomization; day 1 data are the average levels from the time of randomization to the end of the day of randomization. The bars indicate the 95% confidence intervals. The dashed line indicates 108 mg per deciliter, the upper limit of the target range for intensive glucose control. Panel B shows the density plot for the mean time-weighted blood glucose levels for individual patients. The dashed lines indicate the modes (most frequent values) in the intensive-control group (blue) and the conventional-control group (red), as well as the upper threshold for severe hypoglycemia (black). To convert the values for blood glucose to millimoles per liter, multiply by 0.05551.

Patients undergoing intensive glucose control were more likely than those undergoing conventional control to have received insulin (2931 of 3014 patients [97.2%] vs. 2080 of 3014 [69.0%], P<0.001), and they received a larger mean insulin dose (50.2±38.1 units per day, vs. 16.9±29.0 with conventional control; P<0.001) (Table 2). The mean time-weighted blood glucose level was significantly lower in the intensive-control group than in the conventional-control group (115±18 vs. 144±23 mg per deciliter [6.4±1.0 vs. 8.0±1.3 mmol per liter], P<0.001). Additional measures of glycemic control are shown in Table 2 and Figure 2A and 2B.

Nutrition and Concomitant Treatment

During the first 14 days after randomization, the mean daily amount of nonprotein calories administered was 891±490 kcal in the intensive-control group and 872±500 kcal in the conventional-control group (P=0.14). Of this amount, 624±496 kcal (70.0%) and 623±496 kcal (71.4%), respectively, were given as enteral nutrition; 173±359 kcal (19.4%) and 162±345 kcal (18.6%) as parenteral nutrition; and 93.4±88.8 kcal (10.5%) and 87.2±93.5 kcal (10.0%) as intravenous glucose (Table 2, and Appendix B in the Supplementary Appendix).

After randomization, more patients in the intensive-control group than in the conventional-control group were treated with corticosteroids (1042 of 3010 [34.6%] vs. 955 of 3009 [31.7%], P=0.02). The median time from randomization to commencement of corticosteroid treatment was 0 days (interquartile range, 0 to 1) in both groups (P=0.34). The most common indication for corticosteroid administration in both groups was the treatment of septic shock, occurring in 376 of the 1042 patients in the intensive-control group (36.1%) who received corticosteroids, as compared with 328 of the 955 patients in the conventional-control group (34.3%) (absolute difference, 1.7 percentage points; 95% confidence interval [CI], −2.5 to 5.9; P=0.42) (Table 2).

Outcomes

Table 3. Table 3. Outcomes and Adverse Events.

Figure 3. Figure 3. Probability of Survival and Odds Ratios for Death, According to Treatment Group.

Panel A shows Kaplan–Meier estimates for the probability of survival, which at 90 days was greater in the conventional-control group than in the intensive-control group (hazard ratio, 1.11; 95% confidence interval, 1.01 to 1.23; P=0.03). Panel B shows the odds ratios (and 95% confidence intervals) for death from any cause in the intensive-control group as compared with the conventional-control group, among all patients and in six predefined pairs of subgroups. The size of the symbols indicates the relative numbers of deaths. The Acute Physiology and Chronic Health Evaluation II (APACHE II) score can range from 0 to 71, with higher scores indicating more severe organ dysfunction.

Ninety days after randomization, 829 of 3010 patients (27.5%) in the intensive-control group had died, as compared with 751 of 3012 patients (24.9%) in the conventional-control group (Table 3). The absolute difference in mortality was 2.6 percentage points (95% CI, 0.4 to 4.8), and the odds ratio for death with intensive control was 1.14 (95% CI, 1.02 to 1.28; P=0.02). The difference in mortality between the two treatment groups was still significant after adjustment for the predefined baseline risk factors (adjusted odds ratio, 1.14; 95% CI, 1.01 to 1.29; P=0.04). The median survival time was lower in the intensive-control group than in the conventional-control group (hazard ratio, 1.11; 95% CI, 1.01 to 1.23; P=0.03) (Figure 3A).

Overall, the distributions of proximate causes of death were similar in the two groups (P=0.12) (Table 3). However, deaths from cardiovascular causes were more common in the intensive-control group (345 of 829 patients [41.6%]) than in the conventional-control group (269 of 751 patients [35.8%]) (absolute difference, 5.8 percentage points; P=0.02). In the intensive-control group and the conventional-control group, the majority of deaths occurred in the ICU (546 of 829 patients [65.9%] and 498 of 751 patients [66.3%], respectively) or in the hospital after discharge from the ICU (220 of 829 patients [26.5%] and 197 of 751 patients [26.2%], respectively). The remaining deaths (63 of 829 patients [7.6%] undergoing intensive control and 56 of 751 patients [7.5%] undergoing conventional control) occurred after hospital discharge. In both groups, potentially life-sustaining treatments were withheld or withdrawn in more than 90% of the patients who died (see Appendix D in the Supplementary Appendix).

During the 90-day study period, there was no significant difference between the two groups in the median length of stay in the ICU or hospital (Table 3). At 90 days, 7 of the 3016 patients (0.2%) in the intensive-control group and 6 of the 3014 patients (0.2%) in the conventional-control group were still in the ICU (P=0.78), and 174 patients (5.8%) and 166 patients (5.5%), respectively, were still in the hospital (P=0.66).

The number of patients in whom new single or multiple organ failures developed were similar with intensive and conventional glucose control (P=0.11) (Table 3). There was no significant difference between the two groups in the numbers of days of mechanical ventilation and renal-replacement therapy or in the rates of positive blood cultures and red-cell transfusion (Table 3).

With respect to 90-day mortality, subgroup analyses suggested no significant difference in the treatment effect for the comparisons of operative and nonoperative patients (P=0.10), patients with and those without diabetes (P=0.60), patients with and those without severe sepsis (P=0.93), and patients with an APACHE II score of 25 or more and those with a score of less than 25 (P=0.84) (Figure 3B). Tests for interaction indicated a possible trend toward subgroup-specific treatment effects for patients with trauma as compared with those without trauma (P=0.07) and for patients receiving corticosteroids at baseline as compared with those not receiving corticosteroids (P=0.06).

Severe hypoglycemia (defined as a blood glucose level ≤40 mg per deciliter [2.2 mmol per liter]) was recorded in 206 of 3016 patients (6.8%) undergoing intensive glucose control, as compared with 15 of 3014 patients (0.5%) undergoing conventional control (odds ratio, 14.7; 95% CI, 9.0 to 25.9; P<0.001). The recorded number of episodes of severe hypoglycemia was 272 in the intensive-control group, as compared with 16 in the conventional-control group; 173 of all 288 episodes (60.1%) were confirmed by a laboratory measurement, 112 (38.9%) were unconfirmed bedside readings, and 3 (1.0%) were of unknown confirmation status. No long-term sequelae of severe hypoglycemia were reported.

In this large, international, randomized trial involving adults in the ICU, we found that intensive glucose control, as compared with conventional glucose control, increased the absolute risk of death at 90 days by 2.6 percentage points; this represents a number needed to harm of 38. The difference in mortality remained significant after adjustment for potential confounders. Severe hypoglycemia was significantly more common with intensive glucose control.

In conducting our trial, we sought to ensure a high degree of internal and external validity by concealing treatment assignments before randomization, selecting a long-term outcome that is not subject to biased ascertainment, evaluating a number of clinically important outcomes, achieving nearly complete follow-up, and following a predefined statistical-analysis plan.³⁰ The management of blood glucose levels was standardized — nearly all patients received their assigned treatment, the mean blood glucose levels differed significantly between the two treatment groups during the 90-day study period, and the rate of severe hypoglycemia was low in comparison with the rates in other trials.

Limitations of our trial include the use of a subjective criterion — expected length of stay in the ICU — for inclusion, the inability to make treating staff and study personnel unaware of the treatment-group assignments, and achievement of a glucose level modestly above the target range in a substantial proportion of patients in the intensive-control group. We did not collect specific data to address potential biologic mechanisms of the trial interventions or their costs. On the basis of the results in the predefined pairs of subgroups, we cannot exclude the possibility that intensive glucose control may benefit some patients.

Our findings differ from those of a recent meta-analysis showing that intensive glucose control did not significantly alter mortality among critically ill adults.¹⁴ In keeping with the trials included in the meta-analysis, patients in our trial who were assigned to intensive glucose control, as compared with those assigned to conventional control, had lower blood glucose levels, received more insulin, and had more episodes of severe hypoglycemia.¹⁴ A unique feature of our trial was the standardized, complex management of blood glucose made possible at multiple centers through a computerized treatment algorithm accessible on centralized servers. In addition, our patients received predominantly enteral nutrition, consonant with current evidence-based feeding guidelines,³³ whereas a substantial proportion of the patients included in the meta-analysis received predominantly parenteral nutrition.^14,34

Our trial had greater statistical power than previous trials, as well as a longer follow-up period than all but two trials in the meta-analysis. Thus, our results may be due to a specific effect of our treatment algorithm, may be most generalizable to patients receiving predominantly enteral nutrition, or may reflect harm not apparent in trials with shorter follow-up and lower statistical power.

In our trial, more patients in the intensive-control group than in the conventional-control group were treated with corticosteroids, and the excess deaths in the intensive-control group were predominantly from cardiovascular causes. These differences might suggest that reducing blood glucose levels by the administration of insulin has adverse effects on the cardiovascular system.^35,36 However, our trial was not designed to examine such mechanisms; further research is needed to understand the increased mortality in our trial.

Since the original study by Van den Berghe et al.,¹² intensive glucose control has been widely recommended^16,17 on the assumption that treatment aimed at normoglycemia will benefit patients. As noted in other fields of medicine,³⁷ a clinical trial targeting a perceived risk factor (in this case, hyperglycemia) is a test of a complex strategy that may have profound effects beyond its effect on the risk factor (here, the blood glucose level). Our findings suggest that a goal of normoglycemia for glucose control does not necessarily benefit critically ill patients and may be harmful. Whether the harm we observed resulted from the reduced blood glucose level, increased administration of insulin, occurrence of hypoglycemia, methodologic factors specific to our trial, or other factors is unclear.

In conclusion, our trial showed that a blood glucose target of less than 180 mg per deciliter resulted in lower mortality than a target of 81 to 108 mg per deciliter. On the basis of our results, we do not recommend use of the lower target in critically ill adults.

The affiliations of the NICE-SUGAR writing committee members are as follows: Royal North Shore Hospital (S.F.) and the George Institute for International Health, University of Sydney, Sydney (S.F., S.Y.-S.S., D.B, S.H., R.N.); Vancouver Coastal Health (D.R.C.), University of British Columbia (D.F.), and Vancouver General Hospital, University of British Columbia (V.D.) — all in Vancouver, BC, Canada; Austin Hospital, Melbourne University, Melbourne, VIC, Australia (R.B.); McMaster University, Hamilton, ON, Canada (D.C., E.M.); St. Paul's Hospital and University of British Columbia, Vancouver, BC, Canada (P.D.); Vancouver General Hospital, Vancouver Coastal Health Research Institute, University of British Columbia, Vancouver, BC, Canada (W.R.H.); Ottawa Health Research Institute, the University of Ottawa, Ottawa (P.C.H.); Queen's University, Kingston, ON, Canada (D.K.H.); Auckland City Hospital, Auckland, New Zealand (C.M.); Canberra Hospital and Australian National University, Canberra, ACT, Australia (I.M.); St. George Hospital and the George Institute for International Health, University of New South Wales, Sydney (J.M.); Royal North Shore Hospital, University of Sydney, Sydney (J.P.); Kolling Institute of Medical Research, Royal North Shore Hospital and the University of Sydney, Sydney (B.G.R.); Vancouver General Hospital and University of British Columbia, Vancouver, BC, Canada (J.J.R.). The NICE-SUGAR study investigators were as follows: NICE (Australia–New Zealand) Management Committee — S. Finfer (chair), D. Blair, (project manager), R. Bellomo, C. McArthur (lead investigator, New Zealand), I. Mitchell, J. Myburgh, R. Norton, J. Potter; SUGAR (North American) Management Committee — D. Chittock (chair), V. Dhingra (previous chair), D. Foster (senior project manager), D. Cook, P. Dodek, P. Hébert, W. Henderson, D. Heyland, E. McDonald, J. Ronco, I. Schweitzer (ex officio member, at Canadian Institutes for Health Research); Independent Data and Safety Monitoring Committee — R. Peto (chair), P. Sandercock, C. Sprung, J.D. Young; Study Statistical Centre — the George Institute for International Health, University of Sydney, Sydney — S. Su, S. Heritier, Q. Li, S. Bompoint, L. Billot; Study Coordinating Centre — the George Institute for International Health, University of Sydney, Sydney — L. Crampton, F. Darcy, K. Jayne, V. Kumarasinghe, L. Little, S. McEvoy, S. MacMahon, S. Pandey, S. Ryan, R. Shukla, B. Vijayan; Australia–New Zealand site investigators (alphabetically by institution) — Auckland City Hospital (Department of Critical Care Medicine), Auckland, New Zealand — S. Atherton, J. Bell, L. Hadfield, C. Hourigan, C. McArthur, L. Newby, C. Simmonds; Auckland City Hospital (Cardiovascular Intensive Care Unit), Auckland, New Zealand — H. Buhr, M. Eccleston, S. McGuinness, R. Parke; the Austin Hospital, Melbourne, VIC, Australia — R. Bellomo, S. Bates, D. Goldsmith, I. Mercer, K. O'Sullivan; Ballarat Base Hospital, Ballarat, VIC, Australia — R. Gazzard, D. Hill, C. Tauschke; Blacktown Hospital, Blacktown, NSW, Australia — D. Ghelani, K. Nand, G. Reece, T. Sara; Box Hill Hospital, Box Hill, VIC, Australia — S. Elliott, D. Ernest, A. Hamilton; the Canberra Hospital, Canberra, ACT, Australia — R. Ashley, A. Bailey, E. Crowfoot, J. Gissane, I. Mitchell, J. Ranse, J. Whiting; Concord Repatriation Hospital, Concord, NSW, Australia — K. Douglas, D. Milliss, J. Tan, H. Wong; Fremantle Hospital, Fremantle, WA, Australia — D. Blythe, A. Palermo; John Hunter Hospital, Newcastle, NSW, Australia — M. Hardie, P. Harrigan, B. McFadyen; Liverpool Hospital, Liverpool, NSW, Australia — S. Micallef, M. Parr; Middlemore Hospital, Auckland, New Zealand — A. Boase, J. Tai, A. Williams; Nepean Hospital, Nepean, NSW, Australia — L. Cole, I. Seppelt, L. Weisbrodt, S. Whereat; North Shore Hospital, Auckland, New Zealand — A. Flanagan, J. Liang; Prince of Wales Hospital, Sydney — F. Bass, M. Campbell, N. Hammond, L. Nicholson, Y. Shehabi; Queen Elizabeth Hospital, Adelaide, SA, Australia — J. Foote, S. Peake, P. Williams; Royal Brisbane Hospital, Brisbane, QLD, Australia — R. Deans, C. Fourie, M. Lassig-Smith, J. Lipman, J. Stuart; Royal Hobart Hospital, Hobart, TAS, Australia — A. Bell, T. Field, R. McAllister, K. Marsden, A. Turner; Royal North Shore Hospital, Sydney — S. Ankers, S. Bird, S. Finfer, R. Lee, A. O'Connor, J. Potter, N. Ramakrishnan, R. Raper; St. George Hospital, Sydney — V. Dhiacou, K. Girling, A. Jovanovska, J. Myburgh; St. Vincent's Hospital, Melbourne, VIC, Australia — N. Groves, J. Holmes, J. Santamaria, R. Smith; Sir Charles Gairdner Hospital, Perth, WA, Australia — S. Baker, B. Roberts; Wellington Hospital, Wellington, New Zealand — L. Andrews, R. Dinsdale, R. Fenton, D. Mackle, S. Mortimer; Western Hospital, Melbourne, VIC, Australia — C. French, L. Little, H. Raunow; Wollongong Hospital, Wollongong, NSW, Australia — M. Gales, F. Hill, S. Rachakonda, D. Rogan; NSW Institute of Trauma and Injury Management, Sydney — C. Allsop; Australian and New Zealand Intensive Care Research Centre, Melbourne, VIC, Australia — L. Higgins; University of Sydney (Faculty of Medicine), Kolling Institute, and Department of Endocrinology, Royal North Shore Hospital, Sydney — B. Robinson; North American site investigators (alphabetically by institution) — Calgary Health Region, Calgary, AB, Canada — K. Champagne, C. Doig, L. Knox, P. Taylor, C. Wilson; Kingston General Hospital, Kingston, ON, Canada — J. Drover, S. Hammond, E. Mann, M. Myers, A. Robinson; Maisonneuve Rosemont Hospital, Montreal — J. Harvey, Y. Skrobik; Mayo Clinic, Rochester, MN — A. Baumgartner, L. Meade, N. Vlahakis; Mount Sinai Hospital, Toronto — C. Ethier, M. Kramer-Kile, S. Mehta; Ottawa General Hospital, Ottawa — C. Gaudert, S. Kanji, T. McArdle, I. Watpool; St Joseph's Healthcare, Hamilton, ON, Canada — F. Clarke, D. Cook, E. McDonald, A. Tkaczyk; St. Michael's Hospital, Toronto — J. Marshall, J. Morrissey, K. Porretta, O. Smith, V. Wen; St. Paul's Hospital, Vancouver, BC, Canada — B.J. Ashley, P. Dodek, S. Mans; Sunnybrook Health Sciences Centre, Toronto — B. Bojilov, K. Code, R. Fowler, N. Marinoff; Toronto General Hospital, Toronto — L. Chu, J. Granton, M. McGrath-Chong, M. Steinberg; Toronto Western Hospital, Toronto — N. Ferguson, S. Go, A. Matte, J. Rosenberg, J. Stevenson; University of Alberta Hospital, Edmonton, AB, Canada — M. Jacka, L. Sonnema; Vancouver General Hospital, Vancouver, BC, Canada — R. Autio, D. Chittock, D. Davies, P. Ganz, M. Gardner, S. Logie, L. Smith; Vancouver Island Health Authority, VIC, BC, Canada — L. Atkins, F. Auld, M. Stewart, G. Wood.