OBJECTIVES To demonstrate the use and interpretation of alternative approaches to evaluate and understand inconsistency within network meta-analyses (NMA). METHODS A network meta-analysis was performed to compare the efficacy of once weekly dulaglutide 1.5mg to licensed doses of liraglutide, exenatide, lixisenatide, and albiglutide for the treatment of type 2 diabetes. The primary endpoint was reduction in HbARESULTS For the 37-56 week add-on to MET network there were no trials linking to dulaglutide 1.5mg; therefore, a NMA could not be conducted. For the 37-56 week combination network there were no comparisons for which there were both direct and indirect evidence. For the 16-36 week add-on to MET network, there were no comparisons with statistically significant inconsistency in HbA reduction in the node splitting model and no significant design by treatment interactions. For the 16-36 week add-on to Met±SU±TZD network there was statistically significant variation between the direct and indirect estimates for the comparison between glargine and placebo and the comparison between glargine with liraglutide 1.80mg. Further sensitivity analyses were conducted removing the LEAD 5 trial (comparing placebo, glargine and liraglutide 1.80mg) from the network. CONCLUSIONS The node-splitting and treatment-by-design interaction models showed no evidence of inconsistency in the 16-36 week monotherapy network. Following removal of the LEAD 5 trial, the 16-36 week combination network also showed no evidence of inconsistency. The analyses were useful in investigating potential inconsistency across the network.