Six RCTs were identified (n=28,065 participants, range 110 to 11,140). There were three very large studies with more than 4,500 participants. The four largest studies were of high quality (adequate randomisation, allocation and blinding). The smallest study did not meet any of the three quality criteria and the next smallest study did not meet the criterion for allocation. Mean follow-up ranged from 7.5 to 17.7 years. One large study provided two different sets of data. At the end of treatment, mean HbA1C ranged from 6.5% to 7.2% in the intervention group and 7.3% to 9.4% in the control group.
There was a significantly lower incidence of macrovascular events in the intensive glucose control group compared to controls (RR 0.92, 95% CI 0.87 to 0.98, I2=0%, fixed-effect model; seven study groups). The effect did not change when the two smallest studies were removed from the analysis.
Intensive glucose control was not associated with a significant decrease in mortality (RR 0.95, 95% CI 0.80 to 1.12, I2=66.7%, random-effects model; seven study groups) and had no significant effect on incidence of cardiovascular death (RR 1.10, 95% CI 0.79 to 1.53, I2=65.2%, random-effects model; five studies).
A funnel plot for macrovascular events showed no evidence of publication bias.