Eighteen RCTs with 4,535 participants were included in the review. Nine trials reported adequate randomisation and none were stopped early, but 15 did not report whether data collection and outcome assessment were blinded. Trial duration (follow-up) ranged from 24 to 52 weeks.
All classes of anti-hyperglycaemic agents significantly reduced glycated haemoglobin (HbA1c) levels compared with placebo (weighted mean difference (WMD) -0.96%, 95% confidence interval (CI) -1.11 to -0.81; I2=63.7%; nine trials). Non-insulin agents significantly increased HbA1c compared with insulin (WMD 0.29%, 95% CI 0.06 to 0.51; I2=73.4%; 10 trials). Weight change varied by drug class. Insulins doubled the risk of severe hypoglycaemic episodes compared with non-insulin agents (15 out of 566 patients versus 6 out of 553 patients; 10 trials). None of the factors investigated by meta-regression were significantly associated with changes in HbA1c levels.
In the network meta-analysis, drug classes did not differ in their effect on HbA1c levels compared with placebo. Weight increase was seen with insulins (2.84 kg, 95% credible interval (CrI) 1.76 to 3.90) and thiazolidinediones (4.25 kg, 95% CrI 2.76 to 5.66) compared with placebo. Glucagon-like peptide-1 agonists were associated with weight loss compared with placebo (-1.63 kg, 95% CrI -2.71 to -0.60). Other significant differences between agents were reported.
There was no evidence of publication bias for HbA1c level outcome.