Quantification of regional cerebral glucose metabolism (CMRglc) using positron emission tomography and 18F-fluorodeoxyglucose (PET-FDG) requires knowledge of the correction factor between FDG and glucose net clearance, the FDG lumped constant (LC). Because diverging values for LC have been obtained, the authors reevaluated LC by measuring the ratio of the cerebral net extraction fractions of FDG (E*) and glucose (E) from arteriovenous cerebral measurements. Thirty subjects were studied (mean age = 25 +/- 4 years): 12 during a programed infusion of FDG and 18 after a bolus injection of FDG. In the infusion study, LC was calculated as the ratio E*/E. In the bolus study, E* was calculated from the slope of a Patlak-Gjedde plot. Lumped constant was significantly smaller in the infusion study as compared with the bolus study (0.48 +/- 0.16 vs. 0.81 +/- 0.27, P < 0.001). In 4 subjects studied during continuous FDG infusion for 2.5 hours, LC decreased to 0.36 +/- 0.11. These results suggest that the "steady-state" method underestimates LC because E* continues to decline because of significant labeled product. Further, the authors provide evidence for resetting of LC toward a greater value. The subsequent resetting of CMRglc provides a physiologically more meaningful estimate and allows for comparison of CMRglc values between different methodologies.