The calibration of the glucose sensor is a critical issue for the subcutaneous implantation of such devices, since it allows from the sensor output the estimation of the prevailing subcutaneous glucose concentration. This study has compared estimations of subcutaneous glucose concentration in rats, calculated from the sensor's signal by using calibration parameters (sensitivity coefficient and background current) determined either under in vitro or in vivo conditions. The in vitro calibration was performed in phosphate buffer at 37 degrees C. The in vivo parameters were calculated by comparing changes in the sensor output current to concomitant changes in blood glucose concentration, induced by intravenous insulin injection and glucose infusion. The apparent subcutaneous glucose concentration, calculated with in vitro parameters, remained much lower than the concomitant blood glucose levels throughout the experiments. By contrast, when it was calculated with the calibration parameters determined in vivo, either in the hypoglycemia or hyperglycemia range, the apparent subcutaneous glucose concentration well reflected blood glucose levels with a five min time lag. We conclude that it is the two-point in vivo calibration procedure which allows the calculation of an apparent subcutaneous glucose concentration best reflecting blood glucose levels.