Normoglycaemia in Intensive Care Evaluation and Survival Using Glucose Algorithm Regulation (NICE - SUGAR STUDY)

The NICE-SUGAR study is a multi-centre, open label, randomised controlled trial of blood glucose management with an intensive insulin regimen to maintain blood glucose between 4.5 - 6.0 mmol/L versus an insulin regimen maintaining blood glucose less than 10.0 mmol/L with insulin being infused if blood glucose exceeds 10.0 mmol/L, and adjusted when needed to maintain blood glucose between 8.0 - 10.0 mmol/L.

To ensure patient safety, the target blood glucose concentration must be closely monitored and the results known to the clinical staff treating the patients. As patient safety is paramount, it is not possible to blind the clinical staff to treatment allocation. In this open-label trial, bias will be minimised by ensuring concealment of treatment allocation prior to randomisation. The unblinded design risks introducing a systematic difference in some other treatment between the two groups, this is unlikely in the ICU setting where many other interventions will be administered simultaneously. The primary outcome measure is mortality and therefore not subject to ascertainment bias.

Patient recruitment

The treatment effect in Van den Berghe's study was limited to patients who stayed in Intensive Care for five days or longer, but all ventilated patients were admitted to the study at the time of ICU admission. Identifying patients who will stay in the ICU for five days or longer is problematic whereas it is relatively easy to identify patients, particularly patients admitted for routine post-operative monitoring, who will be discharged alive from the ICU after the day following admission. For that reason we propose to consider all patients but exclude those expected to be discharged alive or dead before the end of the day following admission.

The attending Intensive Care physician will make this assessment. In addition it is essential that patients who will stay in the ICU for greater than the eligible criteria time frame but who have a very low risk of death are excluded. For this reason we will exclude patients who are able to eat (or who are tube fed due to pre-existing bulbar or laryngeal dysfunction) and patients who do not merit an arterial line as part of their normal management.

Patients who are moribund and at imminent risk of death (brain death or cardiac standstill) will be excluded. This exclusion is on the basis that treatment allocation can not alter the patient's outcome.

Sample size and power calculations

The ANZICS adult patient database contains information on mortality of patients staying more than 48 hours in ICUs. In the financial years 2000, 2001 and 2002, of 43,760 patients treated in intensive care for greater than 48 hours for whom complete data were available, 9476 died prior to hospital discharge. The hospital mortality rate was 22%. As we will exclude less sick patients who stay greater than 48 hours and 90-day mortality is the primary outcome measure, the study has assumed a 90-day mortality rate of 26% in the control group.

A total of 6100 patients will be recruited, approximately 5100 of these patients will be recruited from the ANZICS CTG centres and approximately 1000 patients will be recruited in Canada.

The George Institute for International Health will take responsibility for the web-based randomisation. This will be available 24 hours a day. A minimisation program will stratify treatment allocation by type of critical illness (medical vs. surgical) and by country. Randomisation will be achieved via a password protected fully secure study website.

Study treatments

In the lower range group, a continuous infusion of insulin administered by syringe pump will be commenced if the blood glucose concentration exceeds 6.0 mmol/L and the infusion rate will be adjusted to maintain the blood glucose concentration between 4.5 - 6.0 mmol/L.

In the higher range group, a continuous infusion of insulin administered by syringe pump will be started if the blood glucose concentration exceeds 10.0 mmol/L and the infusion rate adjusted to keep the blood glucose concentration to less than 10.0 mmol/L and titrated when needed to maintain the blood glucose concentration between 8 - 10 mmol/L.

Adjustments to the insulin dose will be made based initially on the measurement of whole blood glucose in undiluted arterial blood performed initially at hourly intervals. Sampling of arterial blood will require the presence of an intra-arterial catheter in situ for routine clinical management at the time of enrolment. The frequency of blood glucose measurement may be reduced to two-hourly and then four hourly once the insulin regimen, blood glucose concentration and calorie intake are sufficiently stable.

Clinical staff (both doctors and nurses) in the study ICUs will undergo formal training and familiarisation with the insulin regimens by local study coordinators assisted by staff from the appropriate national study coordinating centre. Subsequently the administration of insulin will be adjusted by the intensive care doctors and nurses using the study algorithm accessed via the secure, password protected, encrypted study website. The study algorithm recommends insulin infusion rates whilst allowing clinician discretion, ultimate responsibility for the safe and effective use of insulin infusions remains with the treating clinicians

Patients being discharged from the ICU will receive conventional blood glucose management subsequent to discharge.

Reducing the Incidence of Hypoglycaemia

A major focus of the research coordinators and principal investigators will be to educate staff on the safe use of both insulin regimens in the study. All episodes of biochemical hypoglycaemia will be considered serious adverse events and be reported to the coordinating centre within 24 hours. These data will also be reported to the independent data and safety monitoring committee. If it apparent that there is an unacceptable incidence of hypoglycaemia, either in the study overall or in any particular centre or centres, then the study committees will take appropriate steps to reduce the incidence. Depending on the timing and cause of the episodes, this may include any or all of altering the blood glucose control algorithm, altering the nutrition guidelines, instituting routine IV glucose supplementation, increased education at one or more centres or suspending the study at one or more centres.

Discontinuation of randomised treatment

Study treatment will continue until the patient is eating and not requiring supplementary enteral or parenteral nutrition, or until the earlier of ICU discharge or death or 90 days after randomisation. If during the 90-day follow up period the study treatment is ceased and the patient subsequently deteriorates so that they again satisfy the study entry criteria, the study treatment will be recommenced.

If at any time during the trial the treating ICU physician deems it in the patient's best interest (for example if the patient suffers significant or repeated episodes of hypoglycaemia) then, at the discretion of the treating physician, the study treatment can be withdrawn. Patients withdrawn from the randomised treatment will be followed up according to the study follow up schedule and analysed according to the intention to treat principle unless they or their legal surrogate specifically requests such follow up be ceased.

Ancillary Treatments

Other aspects of patient management are unaffected by study procedures and the treating clinicians will be free to provide whatever care is deemed appropriate and necessary.

Outcomes

The principal study outcome will be whether the patient is alive or dead at 90 days. This will be determined by the research coordinator at each participating centre. The study monitor will verify the source documentation at each monitoring visit. As death is such a robust outcome, unintended bias in outcome assessment is unlikely. Intentional bias would require collusion between the study monitor and research nurses and is considered most unlikely. Given the robustness of the outcome measure, it is unnecessary to establish a blinded outcome committee.

Secondary outcomes, also determined over the same period include:

• Death in the ICU, by 28 days and by 90 days
• Length of ICU stay
• Length of hospital stay
• The need for organ support (inotropes, renal replacement therapy and positive pressure ventilation)
• Incidence of blood stream infections
• Incidence and severity of hypoglycaemia
• In the subgroup of patients admitted with diagnosis of traumatic brain injury, a follow up to determine long term functional status as determined by Extended Glasgow Outcome Scores (GOSE) will be collected at Day 90 and Six months.

Analysis of results

The George Institute will be conducting the statistical analyses. All analyses will be performed on an intention-to-treat basis. Baseline and outcome variables will be compared using Students t test, Chi squared and the Mann-Whitney U test as appropriate. Odds ratios will be estimated using multiple logistic regression analysis. Survival analysis will be performed using Kaplan Meier and Cox's proportional hazards regression analysis.

An independent statistician will conduct two blinded interim analyses when we have primary outcome data for 2000 and 4000 patients and these will be submitted to the DSMC.

Safety and Data Monitoring Committee

An independent Safety and Data Monitoring Committee chaired by Professor Sir Richard Peto at Oxford University, comprising experts in clinical trials, biostatistics, and intensive care has been established. The committee will review unblinded data on patient characteristics, treatment compliance and study outcomes at two interim analyses (availability of primary outcome for 1500 and 3500 patients), at any other time point the committee may deem necessary to protect study participants, and at the final analysis. The committee will be charged with informing the study management committee if at any time there emerges:

1. Evidence beyond reasonable doubt of a difference between randomised groups in all cause mortality
2. Evidence likely to change the practice of many clinicians already familiar with the available evidence about the comparative effects of the two blood glucose regimens

While the definition of beyond reasonable doubt will be left to the judgement of the Safety and Data Monitoring Committee, other committees have considered that a difference in total mortality between randomised groups of three standard deviations would normally constitute such evidence. While a major focus of the Committee's brief will be to monitor total mortality, they would also be provided data on serious adverse events and would not be precluded from making recommendations based on other outcomes such as cause-specific death or serious non-fatal adverse events.

This study will provide reliable evidence about the comparative effects of different targets for blood glucose concentration in patients treated in the Australasian and Canadian intensive care setting. This evidence will have direct relevance to decisions about the care of critically ill patients admitted to ICUs in Australia and New Zealand, Canada and the rest of the world. If the study confirms the treatment effect reported in Van den Berghe's study, maintaining normoglycaemia would likely become a treatment standard worldwide.