Portable glucose sensors with an accurate and longterm stable in vivo function have been in demand for several decades because such devices are of strategic importance in the diagnosis and treatment of diabetes mellitus [1±3]. Despite remarkable progress, the development of new and improvement of existing prototypes is desirable.Efforts have focused on transdermal glucose sensors because the glucose concentrations of capillary blood and interstitial liquid of the subcutaneous tissue are closely related [4±8]. These sensors measure the glucose concentration either after electroosmotic transfer of the interstitial liquid to the skin surface [9] or by a microsensor implanted into the subcutaneous tissue. Most work has been carried out using needle-like transdermal amperometric sensors with immobilised glucose oxidase (GOD) [10±18]. To render the highly sensitive and oxygen-dependent enzyme reaction glucose-limited, the immobilised enzyme is usually covered by a membrane that is less Diabetologia (2001) Abstract Aims/hypothesis. To provide a nonenzymatic sensor for glucose monitoring in subcutaneous tissue. Methods. A continuously working affinity sensor based on the glucose-dependent viscosity of a sensitive liquid containing dextran and concanavalin A has been designed by arranging a microdialysis probe, two flow-resisting capillaries and two pressure transducers in a linear flow system. It allows synchronous processing of the viscosity of the sensitive liquid at the standard glucose concentration and the glucose concentration to be measured. In preliminary human trials the sensor was implanted into the subcutaneous tissue of the forearm and its read-out was compared with capillary blood concentrations.Results. In vitro, the viscometric sensor shows a linear and long-term stable dependence on the glucose concentration without detectable drift. At the applied flow rate of the sensitive liquid (about 5 ml/h) the technical delay is 5 to 10 min. The read-out of the implanted sensor followed the dynamics of the capillary blood glucose concentrations with a time-shift of 10 to 15 min but showed a systematic error when based on precalibration with polymer-free glucose solutions. After appropriate in vivo calibration, the readout was in good or acceptable coincidence with capillary blood concentrations according to the error grid method and did not show any detectable reduction of sensitivity during the periods of measurement (up to 44 h). Conclusion/interpretion. The viscometric-affinity sensor is an efficient tool for current research on glucose monitoring in the subcutaneous tissue and can potentially be further developed for routine clinical use.[ Diabetologia (2001) 44: 416±423]