This study was aimed at validating the in vitro estimated response characteristics of implanted glucose oxidase/H2O2 electrodes with respect to their in vivo function. Monoexponential non-linear regression analysis of sensor current vs. time curves in response to square alterations in glucose concentration gave response times T95 of between 1 and 5 min. Non-primed glucose infusions were applied to dogs with these electrodes implanted subcutaneously. The simultaneously monitored in vivo data were subjected to non-linear regression analysis. The time constants T of increases or decreases after starting or ending the glucose load were (mean +/- SEM) 53 +/- 10 and 26 +/- 4 min (significant difference, p less than 0.05) in sensor current, 28 +/- 8 and 15 +/- 2 min (NS) in whole blood, and 26 +/- 5 and 18 +/- 2 min (NS) in plasma. The in vivo kinetic patterns of sensors were not related to their in vitro response times. Non-linear regression analysis of in vitro responses of glucose sensors under clearly defined conditions is recommended as a basis for further studies. The physiological delay in the subcutaneous glucose system needs more attention in this field of research.