The short-term (2-12 weeks) antihyperglycaemic efficacy of metformin (M), glibenclamide (G), and their primary combination (MG) was assessed in a double-blind study including 165 unselected patients with Type 2 diabetes. Patients with diet failure were randomized to M, G or MG. The dose was titrated with a fasting blood glucose concentration (FBG) of < 6.7 mmol l-1 as the target, using at most six dose levels, the first three comprising increasing monotherapy (M or G) or low-dose primary combination (MGL), and the second three add-on therapies (M/G and G/M) and primary combination therapy escalated to high dose (MGH). Success rates were higher on MGL than on monotherapy. The difference in achieving acceptable control (FBG < or = 7.8 mmol 1(-1)) was 70% versus 51% (95% confidence interval 3-36%, p = 0.032). When the drugs were combined, a slightly greater FBG reduction (p = 0.026) was observed, at lower dosage (p = 0.013). The response could not be predicted from body weight, but depended upon initial FBG (p = 0.019) and meal-stimulated C-peptide (p = 0.007). FBG declined progressively with increasing doses of metformin, whereas glibenclamide exerted most of its effect at low dose. Primary combination therapy with metformin and sulphonylurea may be clinically useful.