Comparison of glucose‐lowering agents after dual therapy failure in type 2 diabetes: A systematic review and network meta‐analysis of randomized controlled trials ZaccardiFrancesco DhalwaniNafeesa N. DalesJolyon ManiHamid KhuntiKamlesh DaviesMelanie J. WebbDavid R. 2018 Aims: To assess the evidence supporting the choice of third-line agents in adults with inadequately controlled type 2 diabetes. Materials and Methods: We searched randomized controlled trials (RCTs) published between January 2000 and July 2017 that reported data on cardiometabolic outcomes and hypoglycaemia for glucose-lowering agents added to metformin-based dual treatments. Data were stratified by background therapy and RCT duration, and synthesized, when possible, with network meta-analyses. Results: A total of 43 RCTs (16 590 participants) were included, with metformin combined with: sulphonylureas (SUs) in 20 RCTs; thiazolidinediones (TZDs) in 10; basal or rapid-acting insulin in 6; dipeptidyl peptidase-4 (DPP-4) inhibitors in 3; glucagon-like peptide-1 receptor agonists (GLP-1RAs) in 2; and sodium-glucose co-transporter-2 (SGLT-2) inhibitors in 2. When added to metformin and SUs, after 24 to 36 weeks, rapid-acting insulin resulted in the largest reduction in glycated haemoglobin (HbA1c; 1.6% vs placebo), followed by GLP-1RAs (1.0%), basal insulin (0.8%) and SGLT-2 inhibitors (0.7%), with no difference between GLP-1RAs and SGLT-2 inhibitors; body weight increased with insulin treatment (~3 kg vs placebo), while the greatest reduction was observed for SGLT-2 inhibitors compared with all other therapies. Limited data for hypoglycaemia indicated a similar risk for SGLT-2 inhibitors and GLP-1RAs. Results for third-line agents added to metformin and TZDs were comparable, showing similar HbA1c reduction and risk of hypoglycaemia between SGLT-2 inhibitors and GLP-1RAs, and a slightly greater reduction in body weight with SGLT-2 inhibitors vs GLP-1RAs. Data for 52 to 54 weeks were more limited: added to metformin and a SU, TZDs, GLP-1RAs or SGLT-2 inhibitors reduced HbA1c to a similar extent but had different effects on body weight (7 kg and 5 kg more with TZDs vs SGLT-2 inhibitors and GLP-1RAs, respectively; 2 kg less when comparing SGLT-2 inhibitors with GLP-1RAs). Formal analyses could not be performed for any other dual therapy failure combinations because of the small number of available RCTs. Conclusions: Moderate-quality evidence supports the choice of a third-line agent only in patients on metformin combined with a SU or a TZD, with SGLT-2 inhibitors performing generally better than other drugs. In suggesting third-line agents, future guidelines should recognize the widely differing evidence on the various dual therapy failures.