The conventional isotope dilution technique was compared with the more accurate constant specific activity (SA) method at six different insulin levels. Paired euglycemic clamp studies were performed in 30 normal subjects (4-hour insulin infusion: 5, 10, 20, 40, 80, and 160 mU . m-2 . min-1) using primed-constant 3-3H-glucose infusion and either conventional unlabeled glucose infusates (Cold-GINF) or labeled glucose infusates (Hot-GINF) to maintain constant SA. At all insulin levels, both glucose disappearance (Rd) and hepatic glucose production (HGP) were underestimated by the conventional technique, and errors during the first 2 hours correlated with glucose infusion rates (GIRs) (r = .93, P < .00001). During the second hour, mean underestimation of HGP varied from 20% +/- 9% to 84% +/- 16% of basal rates from low-dose to high-dose insulin infusion studies. During prolonged equilibration (3 to 4 hours), errors decreased but were still significant in the two low-dose insulin infusion protocols during the fourth hour. In conclusion, using the conventional isotope dilution technique, suppression of glucose production was overestimated and stimulation of glucose Rd was underestimated, and these errors were greater the higher the GIR. Thus, artifactually greater hepatic and smaller peripheral effects may have been assumed for factors or therapies that influence insulin sensitivity in previous studies using a conventional isotope dilution technique, and therefore, reevaluation of these issues may be relevant in future studies.