The aim of this study was to investigate the incremental and proportional effect of a sulfonylurea on insulin secretion rates at low, elevated and high blood glucose, using parallel groups with ascending or descending glucose steps to minimise potential biases of a single stepped clamp order. Following 14 days on placebo or glibenclamide (2.5 mg) tablets twice daily, separated by 14 days washout, 19 type 2 diabetic patients had ascending or descending three-step hyperinsulinaemic glucose clamps at 4, 8 and 12 mmol/l. C-peptide secretion was estimated by two-compartment C-peptide deconvolution. Patients in the ascending glucose steps group (n = 10) had mean (SD) age of 60.3 (6.5) years, BMI of 29.8 (4.9) kg/m(2) and fasting glucose on diet alone of 10.6 (2.9) mmol/l; while those in the descending glucose steps group (n = 9) had mean age of 58.2 (8.0) years, BMI of 30.5 (5.4) kg/m(2) and fasting glucose on diet alone of 9.8 (2.2) mmol/l. The geometric means (95% CI) of C-peptide secretion rates on placebo for glucose at 4.0, 8.0 and 12.0 mmol/l were 63 (46, 86), 143 (105, 195) and 205 (149, 281) pmol/min, respectively. On glibenclamide, this increased by 140 (99, 181), 126 (85, 167) and 158 (117, 199) pmol/min, respectively (p < 0.001 vs placebo). The absolute increment was significant (p < 0.001) and independent of clamp glucose concentration (p = 0.54). The proportional increase was greater at 4 mmol/l: 2.8-fold (2.4, 3.2), compared with 1.8-fold (1.5, 2.0) and 1.7-fold (1.4, 1.9) at 8 and 12 mmol/l, respectively (p < 0.001). At low-normal glucose, glibenclamide exerted a disproportionate effect on insulin secretion. This study highlights the risks of hypoglycaemia when aiming for tight glucose control on this agent.