Modeling Error and Apparent Isotope Discrimination Confound Estimation of Endogenous Glucose Production During Euglycemic Glucose Clamps

We previously demonstrated that conventional tracer methods applied to euglycemic-hyperinsulinemic glucose clamps result in substantially negative estimates for the rate of endogenous glucose production, particularly during the first half of 180-min clamps. We also showed that addition of tracer to the exogenous glucose infusate resulted in nonnegative endogenous glucose production (R_a) estimates. In this study, we investigated the underlying cause of negative estimates of R_a from conventional clamp/tracer methods and the reason for the difference in estimates when tracer is added to the exogenous glucose infusate. We performed euglycemic-hyperinsulinemic (300-μU/ml) clamps in normal dogs without (cold GINF protocol, n = 6) or with (hot GINF protocol, n = 6) tracer (D-[3-³H]glucose) added to the exogenous glucose infusate. In the hot GINF protocol, sufficient tracer was added to the exogenous glucose infusate such that arterial plasma specific activity (SA_a) did not change from basal through the clamp period (P > .05). In the cold GINF studies, plasma SA_a fell 81 ± 2% from the basal level by the 3rd h of clamping. We observed a significant, transient, positive venous-arterial difference in specific activity (SA_v-SA_a difference) during the cold GINF studies. The SA_v-SA_a difference reached a peak of 27 ± 6% at 30 min and diminished to a plateau of 7 ± 1% between 70 and 180 min. We also observed a positive but constant SA_v-SA_a difference (4.6 ± 0.2% between 10 and 180 min) during the hot GINF studies. The observations of a difference between hot and cold GINF endogenous R_a estimates and a positive but transient SA_v-SA_a difference during the cold GINF studies are consistent with the interpretation that a portion of the underestimation of R_a is due to insufficient mixing of endogenous and exogenous glucose for the one-compartment, fixed-pool volume model to be applicable. Alternatively, our results suggest that the one-compartment, fixed-pool volume model of glucose kinetics is insufficient to account for the complex dynamics of labeled and unlabeled glucose during euglycemic-hyperinsulinemic clamps. Improved mixing through addition of tracer to the exogenous glucose infusate or improved modeling by allowing for a variable-pool volume appears to improve the accuracy of the tracer methods; however, these approaches remain to be validated. The constant positive SA_v-SA_a difference observed during the hot GINF studies is consistent with the interpretation that an additional contributor to underestimation of endogenous R_a is apparent isotope discrimination. Whether this apparent discrimination is due to a true isotope effect or contamination of the tracer infusate remains to be determined.