Twenty one RCTs (n=8482 patients) were identified. Twelve RCTs were placebo-controlled; six were active comparator studies; the remaining three RCTs had both a placebo group and an active drug control group. Nine trials investigated liraglutide and twelve studies, exenatide. Randomisation was classed as adequate in ten RCTs; blinding was adequate in nine studies and reported as ‘open label’ in five RCTs; loss to follow-up was reported as adequate in 19 RCTs; and an intention-to-treat analysis was performed in all but one RCT.
Efficacy: Glucagon-like peptide-1 receptor agonists gave a significant improvement (decrease) in glycosylated haemoglobin (HbA1c) levels compared to placebo (WMD -1.0 mmol/L, 95% CI: -1.1 to -0.8). Similar significant results were reported for subgroup analyses of liraglutide and exenatide. The effect was similar in both shorter term (less than 26 weeks) and longer term trials (more than 26 weeks) and in unpublished trials and published trials. When glucagon-like peptide-1 receptor agonists were compared with other active drugs, there was no significant difference in HbA1c level when compared with insulin (five RCTs), sulphonylurea (glimepiride) (three RCTs of liraglutide), and metformin (one RCT of liraglutide). However, for four RCTs comparing exenatide with insulin, exenatide significantly reduced self-monitored postprandial glucose after breakfast (WMD -0.67 mmol/L, 95% CI: -0.56 to -0.78) and after dinner (WMD -0.66 mmol/L, 95% CI: -1.14 to -2.73), but had no significant effect on postprandial glucose after lunch. There was a high level of heterogeneity for HbA1c, and both placebo-controlled (I2=83.6%) and active comparator-controlled (I2= 83.2%) trials (both p<0.001). For Hb1Ac and publication bias, Kendall’s tau was -0.25 (p=0.11) for placebo-controlled trials and 0.14 (p=0.36) for active comparator-controlled trials. For published trials alone, the corresponding values were -0.28 (p=0.15) or placebo-controlled trials and 0.07 (p=0.43) for active comparator-controlled trials, indicating an absence of significant publication bias.
Body mass index: Glucagon-like peptide-1 receptor agonists gave a significant reduction in body mass index compared to placebo (WMD -0.44 kg/m2, 95% CI: -0.78 to -0.10; 11 RCTs), which was also significant for exenatide compared to placebo (WMD -0.62 kg/m2 (95% CI: -1.14 to -0.10) but not significant for liraglutide compared to placebo. When compared to insulin, glucagon-like peptide-1 receptor agonists also gave a significant reduction in body mass index (WMD -1.57 kg/m2, 95% CI: -1.98 to -1.15; five RCTs).
Safety and adverse events: In placebo-controlled trials, exenatide significantly increased the risk of hypoglycaemic episodes (OR 2.92, 95% CI: 1.49 to 5.75; 10 RCTs), but the risk was only increased where the drug was combined with sulphonylurea (OR 4.62, 95% CI: 1.89 to 11.21; 10 RCTs). In RCTs which compared exenatide and insulin, exenatide was not associated with an increased risk of hypoglycaemic episodes. Liraglutide was not significantly associated with increased risk of hypoglycaemic episodes compare to placebo (five RCTs). Glucagon-like peptide-1 receptor agonists significantly increased the risk of nausea (OR 3.88, 95% CI: 2.79 to 5.42), vomiting (OR 4.23, 95% CI: 2.67 to 6.13), and diarrhoea (OR 2.36, 95% CI: 1.67 to 3.33). Similar significant results were reported for subgroup analyses of liraglutide and exenatide. There was no significant risk of severe glycaemic episodes, increased mortality, major cardiovascular events, pancreatitis, or angioedema associated with the use of either exenatide or liraglutide.