We have developed a miniaturized glucose sensor which has been shown previously to function adequately when implanted in the subcutaneous tissue of rats and dogs. Following a glucose load, the sensor output increases, making it possible to calculate a sensitivity coefficient to glucose in vivo, and an extrapolated background current in the absence of glucose. These parameters are used for estimating at any time the apparent subcutaneous glucose concentration from the current. In the previous studies, this calibration was performed a posteriori, on the basis of the retrospective analysis of the changes in blood glucose and in the current generated by the sensor. However, for clinical application of the system, an on line estimation of glucose concentration would be necessary. Thus, this study was undertaken in order to assess the possibility of calibrating the sensor in real time, using a novel calibration procedure and a monitoring unit which was specifically designed for this purpose. This electronic device is able to measure, to filter and to store the current. During an oral glucose challenge, when a stable current is reached, it is possible to feed the unit with two different values of blood glucose and their corresponding times. The unit calculates the in vivo parameters, transforms every single value of current into an estimation of the glucose concentration, and then displays this estimation. In this study, 11 sensors were investigated of which two did not respond to glucose. In the other nine trials, the volunteers were asked to record every 30 s what appeared on the display during the secondary decrease in blood glucose. The results were analysed by comparing these readings (approximately 220 measurements per trial) to the changes in plasma glucose, measured every 15 min. The Error Grid Analysis indicated that 84.1±3.6% of the measurements were in zone A (accurate) and 15±3.6% were in zone B (acceptable). Considering each individual trial, the differences between the displayed value and the concomitant plasma glucose concentration ranged between −1.7 and 0.69 mmol/l. These excellent results were due to the absence of any significant lag between the changes in plasma glucose concentration and the changes in the result on the display. We conclude that this glucose monitoring system, based on subcutaneous sensing of glucose, is able to provide a direct on line estimation of blood glucose concentration.