To compare the pharmacokinetics/dynamics of the long-acting insulin analog glargine with NPH, ultralente, and continuous subcutaneous (SC) infusion of insulin lispro (continuous subcutaneous insulin infusion [CSII]), 20 C-peptide-negative type 1 diabetic patients were studied on four occasions during an isoglycemic 24-h clamp. Patients received SC injection of either 0.3 U/kg glargine or NPH insulin (random sequence, crossover design). On two subsequent occasions, they received either an SC injection of ultralente (0.3 U/kg) or CSII (0.3 U x kg(-1) x 24 h(-1)) (random sequence, crossover design). After SC insulin injection or CSII, intravenous (IV) insulin was tapered, and glucose was infused to clamp plasma glucose at 130 mg/dl for 24 h. Onset of action (defined as reduction of IV insulin >50%) was earlier with NPH (0.8 +/- 0.2 h), CSII (0.5 +/- 0.1 h), and ultralente (1 +/- 0.2 h) versus glargine (1.5 +/- 0.3 h) (P < 0.05) (mean +/- SE). End of action (defined as an increase in plasma glucose >150 mg/dl) occurred later with glargine (22 +/- 4 h) than with NPH (14 +/- 3 h) (P < 0.05) but was similar with ultralente (20 +/- 6 h). NPH and ultralente exhibited a peak concentration and action (at 4.5 +/- 0.5 and 10.1 +/- 1 h, respectively) followed by waning, whereas glargine had no peak but had a flat concentration/action profile mimicking CSII. Interindividual variability (calculated as differences in SD of plasma insulin concentrations and glucose infusion rates in different treatments) was lower with glargine than with NPH and ultralente (P < 0.05) but was similar with glargine and CSII (NS). In conclusion, NPH and ultralente are both peak insulins. Duration of action of ultralente is greater, but intersubject variability is also greater than that of NPH. Glargine is a peakless insulin, it lasts nearly 24 h, it has lower intersubject variability than NPH and ultralente, and it closely mimics CSII, the gold standard of basal insulin replacement.