OBJECTIVES: To compare time to discontinuation with canagliflozin versus DPP-4 inhibitors and GLP-1 agonists in patients with T2DM using retrospective claims data. METHODS: Patients with T2DM who received a first prescription for a DPP-4 inhibitor (sitagliptin, saxagliptin, linagliptin), GLP-1 agonist (liraglutide, exenatide, exenatide long-acting), or an agent that inhibits SGLT2 (canagliflozin) in 2013 were extracted from two US claims databases of commercially insured patients (Truven, Optum).  The analytical sample included only patients with ≥6 months of retrospective data prior to their first paid claim.  Discontinuation was defined as an observed refill gap ≥90 days (sensitivity analysis for 30/60 days) between two subsequent prescriptions.  Time to discontinuation was analyzed using Kaplan-Meier and Cox proportional hazards regression, including demographics, treatment background, and diabetes-related complications/comorbidities as covariates. RESULTS: P <0.0001 for all comparisons).  The adjusted hazard ratio (HR) for time to discontinuation for canagliflozin 100 mg (reference) and 300 mg (HR=0.92 [0.86;0.99]) was significantly lower versus DPP-4 inhibitors and GLP-1 agonists: sitagliptin (n=29,426; HR=1.28 [1.22;1.34]); saxagliptin (n=1,566; HR=2.01 [1.86;2.16]); linagliptin (n=1,432; HR=2.08 [1.92;2.24]); exenatide (n=2,376; HR=2.59 [2.41;2.77]); exenatide long-acting (n=5,922; HR=1.46 [1.40;1.52]); liraglutide (n=17,690; HR=1.23 [1.20;1.27]).  Being younger, male, and being on monotherapy were associated with higher discontinuation risk.  HRs were stable across sensitivity analyses using alternative discontinuation definitions.  Analyses from Optum were generally consistent with these results. CONCLUSIONS: These analyses indicate that patients who received canagliflozin versus DPP-4 inhibitors or GLP-1 agonists remained on their therapy longer, which may reflect better effectiveness and/or tolerability.