Bedtime insulin (BI)/daytime sulfonylurea (DSU) therapy was studied double-blind in 30 non-insulin-dependent diabetes mellitus subjects in whom sulfonylurea (SU) therapy had failed. Subjects were switched to glipizide for 2 months (phase I) to confirm failure (fasting plasma glucose [FPG] 12.0 ± 0.4 mmol/1) and then randomly assigned into three groups: BI-DSU; BI-no DSU; and DSU-no BI. During phase II (3 months), the BI dose was fixed (20 U/1.73 m2, low-dose). In phase III (3 months), BI was titrated up (high-dose) to achieve good control or until hypoglycemie symptoms prevented further dose increases. In phase IV (6 months), 25 of the 30 original subjects received open-labeled, high-dose BI-DSU. Low-dose BI-DSU markedly reduced FPG (13.6 ± 0.8 to 8.0 ± 0.6 mmol/1, P < 0.001), mean 24-h glucose (P < 0.001), HbA1c (8.9 ± 0.7 to 7.6 ± 0.3%, P = 0.07), and basal hepatic glucose production (HGP) (P < 0.005). A positive correlation (r = 0.69, P < 0.05) between the declines in FPG and HGP was observed. Neither low-dose BI alone nor DSU alone reduced FPG, mean 24-h glucose, HbA1c, or basal HGP. High-dose (40 ± 5 U/day) BI plus DSU further reduced the FPG (6.3 ± 0.6 mmoM), HbA1c (7.1 ± 0.3%), mean 24-h plasma glucose, and basal HGP (all P < 0.05 vs. phase II). High-dose BI alone (38 ± 4 U/day) improved FPG (P < 0.01 vs. phase II), mean 24-h plasma glucose (P < 0.005 vs. phase II), HbA1c (P < 0.05 vs. phase II), and basal HGP (P < 0.05 vs. phase I) to a degree similar to low-dose BI-DSU but less than high-dose BI-DSU. Regardless of prior treatment, FPG and HbA1c were controlled during phase IV. Weight gain during BI-DSU was largely attributable to reduced glucosuria. In conclusion, 1) both BI-DSU and BI alone improve glycemia in SU failure patients; 2) BI-DSU is superior to BI alone during both low- and high-dose BI; 3) the efficacy of BI is related to its suppression of basal HGP; and 4) good long-term glycemie control (up to 12 months) can be achieved with combined BI-DSU therapy.