For fifty years, tremendous efforts have been directed towards the development of glucose sensors for tight glycemic control of diabetic patients. Today, millions of diabetics test their blood glucose level daily, making glucose the most commonly tested analyte. Recently, subcutaneous implantable needle-type sensors became commercially available for continuous glucose monitoring. However, these devices require frequent calibrations and are lacking accuracy and reliability. They are based on electrochemical detection, which is strongly affected by the biological environment in which the sensor is placed. In addition, an accurate and reliable continuous glucose sensor would also be of great interest for tight glycemic control in intensive care units of hospitals. However, despite the many impressive breakthroughs, the development of clinically accurate continuous glucose sensors remains a challenge. In this context, alternative approaches to overcome the limitation of electrochemical methods have been actively investigated. Among these, affinity sensing should offer several intrinsic advantages for in vivo monitoring. In this thesis, we investigate a novel viscosity-dependent affinity sensor for continuous monitoring of glucose in biological fluids such as blood and plasma. The sensing principle relies upon the viscosity variation of a sensitive fluid with glucose concentration. The sensitive fluid is based on the competitive binding of glucose and dextran with a glucose-specific binding protein, Concanavalin A. Basically, the sensor is filled with the sensitive fluid, and includes both an actuating and a sensing piezoelectric diaphragm as well as a flow-resistive microchannel. In addition, a nanoporous alumina membrane completely retains the sensitive fluid within the sensor whilst allowing glucose permeation through the membrane. The sensor was extensively tested in isotonic saline solution for physiological blood glucose concentrations between 2 and 20 mM, demonstrating an excellent accuracy, reversibility and stability for up to 3 days. In addition, the response time was close to the 10 minutes required for medical applications. However, despite the excellent short term stability, a progressive loss of sensitivity was observed for long term measurements. Concanavalin A retention by the alumina nanoporous membrane was assessed by ultraviolet absorbance spectrometry. Small leakage through the membrane was detected, which at least partly explains the sensitivity reduction over several days. Finally, the adequacy of the sensor for measurement in human blood serum and plasma was checked. Physiological glucose levels were successfully monitored, meaning that the chemical stability of the sensitive fluid and biofouling of the nanoporous alumina membrane were not an issue for short term applications. Moreover, interferences from biomolecules were limited and the sensitivity was still high enough for glucose monitoring. These results suggest that the combination of the ConA-based sensitive fluid and the microviscometer is a promising sensing principle for continuous glucose monitoring in blood.