Glucose-potassium-insulin infusions in the management of post-stroke hyperglycaemia: the UK Glucose Insulin in Stroke Trial (GIST-UK)

Summary Background Hyperglycaemia after acute stroke is a common finding that has been associated with an increased risk of death. We sought to determine whether treatment with glucose-potassium-insulin (GKI) infusions to maintain euglycaemia immediately after the acute event reduces death at 90 days. Methods Patients presenting within 24 h of stroke onset and with admission plasma glucose concentration between 6·0–17·0 mmol/L were randomly assigned to receive variable-dose-insulin GKI (intervention) or saline (control) as a continuous intravenous infusion for 24 h. The purpose of GKI infusion was to maintain capillary glucose at 4–7 mmol/L, with no glucose intervention in the control group. The primary outcome was death at 90 days, and the secondary endpoint was avoidance of death or severe disability at 90 days. Additional planned analyses were done to determine any differences in residual disability or neurological and functional recovery. The trial was powered to detect a mortality difference of 6% (sample size 2355), with 83% power, at the 5% two-sided significance level. This study is registered as an International Standard Randomised Controlled Trial (number ISRCTN 31118803) Findings The trial was stopped due to slow enrolment after 933 patients were recruited. For the intention-to-treat data, there was no significant reduction in mortality at 90 days (GKI vs control: odds ratio 1·14, 95% CI 0·86–1·51, p=0·37). There were no significant differences for secondary outcomes. In the GKI group, overall mean plasma glucose and mean systolic blood pressure were significantly lower than in the control group (mean difference in glucose 0·57 mmol/L, p Interpretation GKI infusions significantly reduced plasma glucose concentrations and blood pressure. Treatment within the trial protocol was not associated with significant clinical benefit, although the study was underpowered and alternative results cannot be excluded.