A b b re v i a t i o n s : FBC, foreign body capsule; PC, IBM-compatible computer. A table elsewhere in this issue shows conventional and Système International (SI) units and conversion factors for many substances.

The feasibility of continuous long-term glucose monitoring in humans has not yet been demonstrated. Enzyme-based electrochemical glucose sensors with telemetric output were subcutaneously implanted and evaluated in five human subjects with type I diabetes. Subject-worn radio-receiver data-loggers stored sensor outputs. Every 1-4 weeks the subject's glucose levels were manipulated through the full clinical range of interest using standard protocols. Reference blood glucose samples were obtained every 5-10 min and analyzed in our hospital clinical laboratory and/or on glucose meters. The sensor data were evaluated versus the reference data by linear least squares regression and by the Clarke Error Grid method. After surgical explantation and device inspection, the tissue-sensor interface was evaluated histologically. The remaining sensor-membranes were also recalibrated for comparison with preimplant performance. Four of the five glucose sensors tracked glucose in vivo. One sensor responded to manipulated glucose changes for 6.2 months with clinically useful performance (>/=90% of sensor glucose values within the A and B regions of the Clarke Error Grid). For this sensor, recalibration was required every 1-4 weeks. The other three transiently responding sensors had electronic problems associated with packaging failure. The remaining sensor never tracked glucose because of failure to form any sustained connection to adjacent subcutaneous tissue. Thus, stable, clinically useful sensor performance was demonstrated in one of five subjects with diabetes for a sustained interval of greater than 6 months. While this glucose sensor implant technology shows promise in humans, it needs to be made more reliable and robust with respect to device packaging and sensor-tissue connection.

Further development of this type of sensor may provide diabetic patients with a better means of monitoring blood glucose.

An implantable potentiostat-radiotelemetry system for in vivo sensing of glucose is described. An enzyme electrode sensor measures the oxidation current of hydrogen peroxide formed by the stoichiometric conversion of glucose substrate and oxygen cofactor in an immobilized glucose oxidase layer. The sensor current is converted to a frequency and transmitted at programmable intervals (4, 32, 256 s) to a remote receiver. Low power CMOS circuitry is employed and device operation for up to 1.5 years is predicted using two series connected 250 mAh lithium cells. Crystal controlled RF frequencies uniquely identify each sensor allowing over 10 sensors within the same 10 m radius. A custom interface card allows a PC to program the receiver and handle the transmitted sensor data using software written in Microsoft C and QuickBasic. Software control allows on-the-fly sensor addition or subtraction to the sensor group being monitored. Over 10 sensors can be tracked long-term using the longest transmit interval, or four sensors can be tracked during short-term infusion studies when the transmit interval is reduced to 4 s. The design, construction, operation, and performance of the system hardware and software are described and evaluated.

An implantable glucose sensor is needed before a reliable artificial pancreas can be realized. The principles and current status of one such device, the glucose enzyme electrode, is presented and discussed. While monitoring glucose this enzyme sensor consumes enough oxygen to become oxygen-limited. This problem has been solved by developing hydrophobic membranes that are more permeable to oxygen than to glucose. Two types of membranes with this property made from (1) cross-linked albumin and (2) sebacyl chloride (nylon) are described. Placing these membranes over the glucose enzyme electrode solves the problem of oxygen limitation. Furthermore, the addition of this type of membrane increases the linear response range of the electrode to glucose to include the entire clinical range of interest (0-400 mg/dl). Other problems in developing an implantable glucose sensor are discussed. Competing strategies to achieve an implantable artificial pancreas without using electronic or mechanical components are presented and evaluated.