OBJECTIVE To provide distinct definitions and quantify the establishment of onset, peak, and duration of action for insulins. METHODS We administered single subcutaneous doses of 10 U regular insulin to 10 volunteer subjects and 25 U NPH insulin to 6 healthy male volunteer subjects on separate occasions. Each dose was given after an overnight fast during a glucose clamp to maintain an euglycemic state. We measured serum insulin concentrations and glucose infusion rates (GIR) from frequent blood sampling after each treatment. Serum insulin concentrations were related to GIR values at each collection time and a counter-clockwise hysteresis resulted. An effect compartment model was used to simultaneously describe the pharmacokinetics and pharmacodynamics of each insulin and to resolve the hysteresis. RESULTS From the resulting relationship, GIR could then be predicted, with onset and duration of action reflecting the time when effect compartment concentrations initially exceeded then declined below a 10% maximum possible effect (Emax) level. Ninety-five percent confidence intervals were constructed allowing a predictive range of values. For regular insulin, a mean onset of 0.75 h, peak of 2 h, and duration of 6 h was estimated. Mean values were also produced with NPH, with an onset of 3 h, peak of 6-7 h, and a duration of 13 h estimated. CONCLUSIONS This method estimates the onset, peak, and duration of insulin action. Although these estimates were from single doses, we believe they can provide good estimations of insulin activity.