OBJECTIVE: To compare the efficacy and safety profile of adding inhaled human insulin (INH; Exubera) or metformin to sulfonylurea monotherapy in patients with poorly controlled type 2 diabetes. RESEARCH DESIGN AND METHODS: We performed an open-label, parallel, 24-week, multicenter trial. At week -1, patients uncontrolled on sulfonylurea monotherapy were divided into two HbA(1c) (A1C) arms: > or =8 to < or =9.5% (moderately high) and >9.5 to < or =12% (very high). Patients were randomized to adjunctive premeal INH (n = 225) or metformin (n = 202). The primary efficacy end point was change in A1C from baseline. RESULTS: In the A1C >9.5% arm, INH demonstrated a significantly greater reduction in A1C than metformin. Mean adjusted changes from baseline were -2.17 and -1.79%, respectively; between-treatment difference was -0.38% (95% CI -0.63 to -0.14, P = 0.002). In the A1C < or =9.5% arm, mean adjusted A1C changes were -1.94 and -1.87%, respectively (-0.07% [-0.33 to 0.19], P = 0.610), consistent with the noninferiority criterion. Hypoglycemia (events/subject-month) was greater in the INH (0.33) than in the metformin (0.15) group (risk ratio 2.16 [95% CI 1.67-2.78]), but there were no associated discontinuations. Other adverse events, except increased cough in the INH group, were similar. At week 24, changes in pulmonary function parameters were small and comparable between groups. Insulin antibody binding increased more with INH but did not have any associated clinical manifestations. CONCLUSIONS: In patients with type 2 diabetes poorly controlled on a sulfonylurea (A1C >9.5%), the addition of premeal INH significantly improves glycemic control compared with adjunctive metformin and is well tolerated.