BACKGROUND: Most patients with type 2 diabetes begin pharmacotherapy with metformin, but eventually need additional treatment. We assessed the safety and efficacy of once weekly exenatide, a glucagon-like peptide 1 receptor agonist, versus maximum approved doses of the dipeptidyl peptidase-4 inhibitor, sitagliptin, or the thiazolidinedione, pioglitazone, in patients treated with metformin. METHODS: In this 26-week randomised, double-blind, double-dummy, superiority trial, patients with type 2 diabetes who had been treated with metformin, and at baseline had mean glycosylated haemoglobin (HbA(1c)) of 8.5% (SD 1.1), fasting plasma glucose of 9.1 mmol/L (2.6), and weight of 88.0 kg (20.1), were enrolled and treated at 72 sites in the USA, India, and Mexico. Patients were randomly assigned to receive: 2 mg injected exenatide once weekly plus oral placebo once daily; 100 mg oral sitagliptin once daily plus injected placebo once weekly; or 45 mg oral pioglitazone once daily plus injected placebo once weekly. Primary endpoint was change in HbA(1c) between baseline and week 26. Analysis was by intention to treat, for all patients who received at least one dose of study drug. This trial is registered with ClinicalTrials.gov, number NCT00637273. FINDINGS: 170 patients were assigned to receive once weekly exenatide, 172 to receive sitagliptin, and 172 to receive pioglitazone. 491 patients received at least one dose of study drug and were included in the intention-to-treat analysis (160 on exenatide, 166 on sitagliptin, and 165 on pioglitazone). Treatment with exenatide reduced HbA(1c) (least square mean -1.5%, 95% CI -1.7 to -1.4) significantly more than did sitagliptin (-0.9%, -1.1 to -0.7) or pioglitazone (-1.2%, -1.4 to -1.0). Treatment differences were -0.6% (95% CI -0.9 to -0.4, p<0.0001) for exenatide versus sitagliptin, and -0.3% (-0.6 to -0.1, p=0.0165) for exenatide versus pioglitazone. Weight loss with exenatide (-2.3 kg, 95% CI-2.9 to -1.7) was significantly greater than with sitagliptin (difference -1.5 kg, 95% CI -2.4 to -0.7, p=0.0002) or pioglitazone (difference -5.1 kg, -5.9 to -4.3, p<0.0001). No episodes of major hypoglycaemia occurred. The most frequent adverse events with exenatide and sitagliptin were nausea (n=38, 24%, and n=16, 10%, respectively) and diarrhoea (n=29, 18%, and n=16, 10%, respectively); upper-respiratory-tract infection (n=17, 10%) and peripheral oedema (n=13, 8%) were the most frequent events with pioglitazone. INTERPRETATION: The goal of many clinicians who manage diabetes is to achieve optimum glucose control alongside weight loss and a minimum number of hypoglycaemic episodes. Addition of exenatide once weekly to metformin achieved this goal more often than did addition of maximum daily doses of either sitagliptin or pioglitazone. FUNDING: Amylin Pharmaceuticals and Eli Lilly.