BACKGROUND: Agonists of the glucagon-like peptide-1 (GLP-1) receptor provide pharmacological levels of GLP-1 activity, whereas dipeptidyl peptidase-4 (DPP-4) inhibitors increase concentrations of endogenous GLP-1 and glucose-dependent insulinotropic polypeptide. We aimed to assess the efficacy and safety of the human GLP-1 analogue liraglutide versus the DPP-4 inhibitor sitagliptin, as adjunct treatments to metformin, in individuals with type 2 diabetes who did not achieve adequate glycaemic control with metformin alone. METHODS: In this parallel-group, open-label trial, participants (aged 18-80 years) with type 2 diabetes mellitus who had inadequate glycaemic control (glycosylated haemoglobin [HbA(1c)] 7.5-10.0%) on metformin (>or=1500 mg daily for >or=3 months) were enrolled and treated at office-based sites in Europe, the USA, and Canada. Participants were randomly allocated to receive 26 weeks' treatment with 1.2 mg (n=225) or 1.8 mg (n=221) subcutaneous liraglutide once daily, or 100 mg oral sitagliptin once daily (n=219). The primary endpoint was change in HbA(1c) from baseline to week 26. The efficacy of liraglutide versus sitagliptin was assessed hierarchically by a non-inferiority comparison, with a margin of 0.4%, followed by a superiority comparison. Analyses were done on the full analysis set with missing values imputed by last observation carried forward; seven patients assigned to liraglutide did not receive treatment and thus did not meet criteria for inclusion in the full analysis set. This trial is registered with ClinicalTrials.gov, number NCT00700817. FINDINGS: Greater lowering of mean HbA(1c) (8.5% at baseline) was achieved with 1.8 mg liraglutide (-1.50%, 95% CI -1.63 to -1.37, n=218) and 1.2 mg liraglutide (-1.24%, -1.37 to -1.11, n=221) than with sitagliptin (-0.90%, -1.03 to -0.77, n=219). Estimated mean treatment differences for liraglutide versus sitagliptin were -0.60% (95% CI -0.77 to -0.43, p<0.0001) for 1.8 mg and -0.34% (-0.51 to -0.16, p<0.0001) for 1.2 mg liraglutide. Nausea was more common with liraglutide (59 [27%] patients on 1.8 mg; 46 [21%] on 1.2 mg) than with sitagliptin (10 [5%]). Minor hypoglycaemia was recorded in about 5% of participants in each treatment group. INTERPRETATION: Liraglutide was superior to sitagliptin for reduction of HbA(1c), and was well tolerated with minimum risk of hypoglycaemia. These findings support the use of liraglutide as an effective GLP-1 agent to add to metformin. FUNDING: Novo Nordisk.