The aim of this article is to critically discuss the technical and clinical aspects of glucose sensors and to briefly review current technical developments. This includes sensors for spot glucose measurements as well as those used for continuous glucose monitoring. Continuous glucose monitoring in particular should supply the diabetic patient with all the information required to optimize insulin therapy and metabolic control. Such systems should also allow hypo- and hyperglycemic episodes to be avoided. During the last 30 years numerous attempts have been made to develop glucose sensors, and new major breakthroughs have been announced repeatedly. However, up until now no glucose sensor has been available that can be used by diabetic patients in daily life conditions. Also one type of glucose sensor, a glucose electrode, recently received approval by the Food and Drug Administration (USA) and is commercially available. Other glucose sensors employing the transdermal, microdialysis or open tissue microperfusion technique are currently under clinical development and may also become available in the near future. The types of glucose sensors referred to so far are not truly non-invasive, but only minimally invasive. They measure glucose concentration in the interstitial fluid of the skin or the subcutis. Non-invasive optical glucose sensors are designed to monitor glucose changes in the skin by directing light through it. They measure the characteristics of the reflected light that are changed as the result of an interaction with glucose. However, none of the attempts with optical glucose sensors have resulted thus far in the development of a sensor that allows monitoring of glucose with sufficient accuracy and precision within the clinically relevant glucose range in daily life conditions. Nevertheless, more minimal-invasive glucose sensors systems will become available for practical use in the near future, whereas it is still uncertain if this can be said for any non-invasive glucose sensor.

Diabetics would benefit greatly from a device capable of providing continuous noninvasive monitoring of their blood glucose levels. The optical scattering coefficient of tissue depends on the concentration of glucose in the extracellular fluid. A feasibility study was performed to evaluate the sensitivity of the tissue reduced scattering coefficient in response to step changes in the blood glucose levels of diabetic volunteers. Estimates of the scattering coefficient were based on measurements of the diffuse reflectance on the skin at distances of 1-10 mm from a point source. A correlation was observed between step changes in blood glucose concentration and tissue reduced scattering coefficient in 30 out of 41 subjects measured.

Research into noninvasive devices for self-monitoring of blood glucose is mainly based on near-infrared spectroscopy. Such a device is particularly desirable in the intensive therapy of patients with diabetes mellitus to achieve optimal metabolic control through frequent glucose testing. The state of noninvasive assay technology is presented. Using diffuse reflectance spectra of mucous lip tissue has advantages and drawbacks compared with tissue transmittance experiments. Different approaches have been proposed in the patent literature; however, current technology requires further significant improvements, particularly within the lower normal and hypoglycemic glucose concentration ranges.

Near-infrared (NIR) spectra of the human inner lip were obtained by using a special optimized accessory for diffuse reflectance measurements. The partial-least squares (PLS) multivariate calibration algorithm was applied for linear regression of the spectral data between 9000 and 5500 cm−1 (Λ = 1.1–1.8 μm) against blood glucose concentrations determined by a standard clinical enzymatic method. Calibration experiments with a single person were carried out under varying conditions, as well as with a population of 133 different patients, with capillary and venous blood glucose concentration values provided. A genuine correlation between the blood glucose concentrations and the NIR-spectra can be proven. A time lag of about 10 min for the glucose concentration in the spectroscopically probed tissue volume vs. the capillary concentration can be estimated. Mean-square prediction errors obtained by cross-validation were in the range of 45 to 55 mg/dL. An analysis of different variance factors showed that the major contribution to the average prediction uncertainty was due to the reduced measurement reproducibility, i.e., variations in lip position and contact pressure. The results demonstrate the feasibility of using diffuse reflectance NIR-spectroscopy for the noninvasive measurement of blood glucose.

An analytical multicomponent method for the blood substrates total protein, glucose, total cholesterol, triglycerides, and urea in human EDTA plasma by FT-IR spectroscopy is described. The spectra were obtained with the use of the attenuated total reflection technique. Partial least-squares was applied for multivariate calibration over optimized spectral ranges. The mean-square prediction errors for the population of 126 plasma samples of different patients calculated by cross-validation are in the range of clinical acceptance. Within an error variance analysis, the contributions of the reference method and the spectrometric measurement to the average (root mean square) prediction error have been estimated for each substrate, giving evidence of the limitations of the spectrometric method. The problem of the biocompatibility of the plasma has been investigated, and the protein adsorption onto the ATR crystal can be reduced to a constant and tolerable level by appropriate cleaning and rinsing. The potential for further improvement is discussed.

No abstract

A multicomponent assay for the blood substrates of total protein, glucose, total cholesterol, triglycerides and urea in human EDTA-plasma by FT-IR spectroscopy is described. Transmission near infrared spectra of plasma were recorded using a 1 mm cell. Partial least-squares was applied for multivariate calibration taking into account optimising the spectral ranges of absorbance or logarithmised single beam spectra. The standard errors of prediction for the population of 124 plasma samples of different patients calculated by cross-validation are in the range of clinical acceptance for protein, cholesterol and triglycerides, whereas glucose assay performance is slightly above the recommended level. The relative standard error of prediction for urea is 12.7% based on the mean concentration of the population studied. Within an error variance analysis, the contributions of the reference method and the spectrometric measurement to the average (RMS) prediction error have been estimated for each substrate giving evidence of the spectrometric method limitations. The potential for further improvement is discussed.

No abstract