We aimed to evaluate the comparative efficacy and safety of dipeptidyl peptidase-4 inhibitors (DPP4i), glucagon-like peptide-1 receptor agonists (GLP-1RA), sodium-glucose co-transporter 2 inhibitors (SGLT2i), or thiazolidinedione (TZD) as an adjunctive treatment in patients with poorly controlled type 2 diabetes mellitus (T2DM) on insulin therapy. We searched Medline, Embase, the Cochrane Library, and ClinicalTrials.gov through April 2016. Bayesian network meta-analyses were performed with covariate adjustment. The primary outcome was the change in glycated hemoglobin A1c (HbA1c) from baseline. Fifty randomized controlled trials covering 15,494 patients were included. GLP-1RA showed the greatest HbA1c-lowering effect compared to the control (-0.84%; 95% credible interval, -1.00% to -0.69%), followed by TZD (-0.73%; -0.93 to -0.52%), SGLT2i (-0.66%; -0.84% to -0.48%), and DPP4i (-0.54%; -0.68% to -0.39%). SGLT2i showed the greatest fasting plasma glucose reduction. GLP-1RA and SGLT2i showed greater body weight reduction, whereas TZD increased body weight. TZD was ranked the highest in terms of insulin dose reduction. The risk of hypoglycemia was increased with TZD or GLP-1RA. The study provides the best available evidence on the comparative efficacy and safety of non-insulin anti-diabetic agents on top of pre-existing insulin therapy for inadequately controlled T2DM patients.