Thirty-five RCTs (n=8,478) were included in the review.
The method of allocation generation was clearly reported and considered appropriate in 13 studies. Blinding was reported in 10 studies, 5 of which were double-blind. Only 2 studies met strict criteria such as appropriate randomisation, double-blind status and clearly stated statistical methods.
Based on all 35 trials, patients receiving insulin therapy showed a statistically significant 15% reduction in mortality compared with controls (RR 0.85, 95% CI: 0.75, 0.97).
In studies in which the goal was to achieve glucose control, patients receiving insulin showed a 29% reduction in mortality relative to the controls (RR 0.71, 95% CI: 0.54, 0.93). Studies in which insulin was administered without aiming to achieve glucose control found no benefit of insulin on mortality.
Studies that included patients with diabetes mellitus showed a significant benefit of insulin on mortality (RR 0.73, 95% CI: 0.58, 0.90). In studies that excluded patients with insulin-requiring diabetes mellitus, the benefit of insulin was smaller, while in trials that excluded patients with a history of diabetes mellitus, there was no benefit.
Studies in which insulin was administered as a GIK solution showed no statistically significant benefit of insulin on mortality. Studies that administered insulin in a way other than GIK (n=5) showed a statistically significant 27% reduction in mortality in patients receiving insulin compared with controls (RR 0.73, 95% CI: 0.56, 0.95).
Patients who received insulin when admitted for acute myocardial infarction or cardiac surgery showed no statistically significant benefit of insulin therapy. One large trial showed that patients receiving insulin on a surgical intensive care unit had a significant 42% reduction in mortality. Studies of patients with myocardial infarction who were not treated with reperfusion showed a reduction in mortality for insulin treatment (pooled RR 0.84, 95% CI: 0.71, 1.00), but no significant benefit was found in the studies where reperfusion was used.
Based on 10 studies, patients receiving insulin were three times more likely to develop hypoglycaemia than controls (RR 3.4, 95% CI: 1.9, 6.3).