OBJECTIVES: Most patients with type 2 diabetes mellitus (T2DM) treated with insulin need intensification of their insulin regimen to achieve glycemic target goals. While there are options available for adjunctive treatments by adding non-insulin agents, in patients who were poorly controlled with the insulin therapy, the optimal therapy is unclear. We evaluate their relative treatment effects by a comprehensive analysis that incorporates evidence from a network of studies. METHODS:  A systematic search of the literature was conducted in MEDLINE, EMBASE and the Cochrane Central Register of Controlled Trials through May 2016. Selected studies were randomized controlled trials (RCTs) that evaluated treatments as add-on to insulin in T2DM patients who were uncontrolled with insulin therapy. The therapies considered were DPP4i (DPP4i/INS), GLP-1 agonist (GLP-1 agonist/INS), SGLT2i (SGLT2i/INS) or TZD (TZD/INS), and the reference comparator was placebo plus insulin (PCB/INS). The primary outcome was the change in HbA1c from baseline to the end of the intervention period. The network meta-analysis adjusted by study-level covariates in a Bayesian framework was performed on WinBUGS14. RESULTS:  We included 42 eligible RCTs. Performing the random-effects network meta-analysis including age, sex, body mass index, baseline insulin dose and the change of HbA1c of Control group as covariates, the mean reduction in HbA1c compared to PCB/INS during the intervention was -0.82% (95% Credible Interval (CI) -0.97 to -0.66%) for GLP-1 agonist/INS, -0.66 (95% CI -0.89 to -0.43%) for TZD/INS, -0.63% (95% CI -0.80 to -0.45%) for SGLT2i/INS and -0.54% (95% CI -0.67 to -0.40%) for DPP4i/INS. The probability that GLP-1 agonist/INS is the best on glycemic control was 84.47%, followed by TZD/INS (12.62%), SGLT2i/INS (3.72%) and DPP4i/INS (0.19%). CONCLUSIONS:  Our results suggest that GLP-1 agonist/INS is likely to achieve the best glycemic control with respect to reduction of HbA1c in patients with inadequately controlled T2DM.