Glucose monitoring is of importance for success of complex therapeutic interventions in diabetic patients. Its impact on treatment and glycemic control is demonstrated in large clinical trials. Up to eight blood glucose measurements per day are recommended. Notwithstanding, a substantial number of diabetic patients cannot or will not monitor their blood glucose appropriately. Considerable progress in control of disturbed metabolism in diabetic patients can be expected by continuous glucose monitoring. The aim of the study was to evaluate the performance of a new amperometric glucose oxidase-based glucose sensor in vitro and in vivo after subcutaneous implantation into rats. For in vitro testing current output of sensors was measured by exposure to increasing and decreasing glucose concentrations up to 472 mg dL(-1) over a time period of 7 days. After subcutaneous implantation of sensors into interscapular region of male rats glucose in interstitial fluid was evaluated and compared to glucose in arterial blood up to 7 days. Hyper- and hypoglycaemia were induced by intravenous application of glucose and insulin, respectively. Current of each implanted sensor was converted into glucose concentration using the first blood glucose measurement only. A change of current with glucose of 0.35 nA mg(-1)dL(-1) indicates high sensitivity of the sensor in vitro. The response time (90% of steady state) was calculated by approximately 60s. Test strips for blood glucose measurement as reference for sensor readings was found as an appropriate and rapidly available method in rats by comparison with established hexokinase method in an automated lab analyzer with limits of agreement of +32.8 and -25.7 mg dL(-1) in Bland-Altman analysis. In normo- and hypoglycaemic range sensor readings in interstitial fluid correlated well with blood glucose measurements whereas hyperglycaemia was not reflected by the sensor completely when blood glucose was changing rapidly. The data given characterize a sensor with high sensitivity, long term stability and short response time. A single calibration of the sensor is required only in measurement periods up to 7 days. The findings demonstrate that the sensor is a highly promising candidate for assessment in humans.