Second-Line Therapy for Patients with Diabetes Inadequately Controlled on Metformin: Addendum to Project Protocol

Section 5.1.3 Additional antidiabetes agents to be studied:  liraglutide (a GLP-1 analogue)  saxagliptin (a DPP-4 inhibitor). Section 6.1.2 (a) Selection criteria Clarification:  Primary studies will be included if second-line agents are compared as add-ons to metformin monotherapy regardless of treatment history before metformin monotherapy. This includes studies that employ a metformin monotherapy run-in period before the addition of second-line agents. Section 6.1.2 (i) Data synthesis and analysis Changes to the planned analysis:  All data from head-to-head, direct treatment comparisons that form a closed network will be synthesized at the drug class level (e.g., sulfonylureas, thiazolidinediones). For glycosylated hemoglobin (A1C), body weight, and hypoglycemia outcomes, separate mixed treatment comparison (MTC) analyses will be conducted in which sulfonylureas and thiazolidinediones will be disaggregated to the level of individual agents.  Weight loss agents (i.e., orlistat and sibutramine) are primarily used to lower body weight rather than to manage hyperglycemia. The populations enrolled in studies involving these agents are, therefore, likely to differ from studies of other agents. Hence, these agents will be excluded from MTC analysis, although any direct comparisons of weight loss agents versus placebo or other antidiabetes treatments will be reported.  Study arms that use fixed, low doses of second-line therapies will be removed from the reference case MTC meta-analysis. Doses lower than the defined daily doses from the World Health Organization will be considered low doses. Where the defined daily dose is not available, doses lower than the lowest recommended maintenance doses in product monographs will be defined as low doses.  The reference case analysis will exclude data from extension phases for all outcomes other than safety and long-term complications.  The deviance information criterion statistic will be used to assess the goodness-of-fit of all models, including subgroup and sensitivity analyses. The deviance information criterion statistic provides a measure of model fit that penalizes model complexity.