Recent experimental evidence suggests that estimates of glucose effectiveness (S(G)) from the minimal model of unlabeled glucose disappearance (Cold-MM) are in error. The single-compartment glucose distribution assumption embedded in the model has been indicated as a possible source of error. In this study, to directly examine the single-compartment assumption, we measured plasma and interstitial glucose concentrations after intravenous glucose injection. Additionally, we compared the accuracy of the estimates of glucose effectiveness from the Cold-MM and the single-compartment tracer minimal model (Hot-MM). Paired labeled intravenous glucose tolerance tests (IVGTTs) were performed in each of six C-peptide-negative type 1 diabetic subjects. Two different insulin infusion protocols were used: an infusion at constant basal rates and an infusion at variable rates to mimic a normal insulin response. During the labeled IVGTT with basal insulin infusion, the microperfusion technique was employed to sample adipose tissue interstitial fluid. Marked differences between the plasma and interstitial dynamics of (cold) glucose were observed during the first 22 min after glucose injection. These results suggest that the requirements for a single-compartment representation of glucose kinetics are not satisfied during at least the first 22 min of an IVGTT. Data from the labeled IVGTT with normal insulin response were used to identify the minimal-model parameters. The measure of S(G) derived using the Cold-MM was 3.44-fold higher than the direct measure obtained from the labeled IVGTT with basal insulin infusion (0.0179+/-0.0027 vs. 0.0052+/-0.0010 min(-1), P<0.01). The measure of glucose effectiveness (S(G)*) derived by the Hot-MM was 1.36-fold higher than the direct measure available from the labeled IVGTT with basal insulin infusion (0.0079+/-0.0013 vs. 0.0058+/-0.0004 min(-1), P>0.26). These results suggest that the Hot-MM is more appropriate for the evaluation of glucose effectiveness than the Cold-MM.