Control of Hyperglycaemia in Paediatric intensive care trial (the CHIP trial)

Design: Primary Research: Pragmatic randomised controlled clinical trial. Setting: 10 Regional Paediatric Intensive Care Units in the UK Target population: Children aged to 16 years undergoing intensive care treatment who are receiving mechanical ventilation and inotropic support drugs and in whom it is anticipated that such treatment will continue for at least 12 hours. Children born preterm (< 36 weeks gestation), those with diabetes mellitus (Type 1) or for whom treatment withdrawal or limitation of intensive care treatment is being considered will be excluded. Health technology being assessed: Elevated blood glucose levels frequently occur following injury, major surgery or in association with major illness in children. Conventional management is to intervene to lower glucose levels with an intravenous infusion of insulin only if the glucose level exceeds approximately 12 mmol/l (Personal communication, K. Morris, UK Survey, 2006). An alternative strategy, tight glycaemic control using insulin to maintain blood glucose strictly within normal limits of 5-7 mmol/l may improve outcomes in critically ill and stressed children, as it has been shown to do in adult patients. It is not possible to blind the administration of insulin, as its impact on blood glucose levels must be closely monitored and adjustments made. Measurement of costs and outcomes: Following the influential ARDSNET study (NEJM 2000: 342;1301) we will use as the primary outcome the number of days alive and free from mechanical ventilation (VFDs) 30 days after trial entry. Death is an important outcome and is unlikely to be influenced by investigator bias. The clinical investigators consider that mechanical ventilation 'defines' need for complex intensive care and that its discontinuation will be the result of application of standard clinical care protocols rather than trial-driven bias. Secondary measurement of outcomes are: death in hospital, no. days in ICU; no. patients >4 days in ICU ; duration of mechanical ventilation; duration of inotropic drug usage; renal impairment (creatinine > 100 mol/l; need for renal replacement therapy; blood stream infection; use of antibiotics >10 days; hypoglycaemic episodes; seizures; and a measure of multi-organ dysfunction using PELOD. A full cost and cost-effectiveness analysis will be undertaken to assess whether the costs of achieving tight blood glucose control are justified by subsequent reductions in hospitalisation costs and/or by improvements in patient outcomes. Information will be collected on the resources required under both policies. Unit costs for hospital services and community care costs will be taken from appropriate reference costs databases, with some more detailed unit costs if required. Total costs will be estimated by valuing each resource use item by the appropriate unit cost. The analysis will also combine the clinical and cost data in a cost-effectiveness analysis, and report cost per death averted and cost per adverse event averted. Sensitivity analysis will be conducted to test how robust the conclusions are to assumptions made in the analysis. Based on information from PICANet, the target sample size is 1500 patients. This has over 90% power to detect an effect size of 2 VFDs within the first 30 days post-randomisation. This size not only allows for random error in the estimate of the standard deviation (assumed as 7 VFds), and for non-compliance (some patients allocated to tight control not receiving this, and some allocated to usual care being managed with tight control), but also crucially has 80% power to test for interaction between the effect of the tight glucose control and key prognostic factors (such as whether the child is a cardiac or non-cardiac patient).