Summary:
Modern strategies in the treatment of Diabetes mellitus require frequent blood glucose tests. Near-infrared spectroscopy [NIR] is considered to be a method for non-invasive blood glucose testing. The present study evaluated the long-term accuracy and stability of the method. Spectra of 10 patients were recorded. An individual calibration models were calculated for each patient from spectra which were recorded at the beginning of the investigation. These models were then applied to calculate blood glucose values from spectra which were recorded 84 to 169 days later. The long-term accuracy and stability of the calibration models expressed as root mean squared error of prediction [RMSP] varied from 3.1 to 35.9 mmol/l and the error grid analysis [EGA] from 13 to 92% clinically acceptable values in zones A and B. The results of the investigation indicate that long-term monitoring of blood glocose by means of NIR-spectroscopy and individual calibration models over a period of 3 to 4 months is basically possible once long-term stability of the method will be improved. Further technological progress and a better understanding of the underlying physiological processes is necessary to achieve a level of accuracy, reliability and stability of the method as required for clinical application.
Abbreviations: RMSP root mean squared error of prediction; EGA error grid analysis; LOO-CV leaving-one-out cross-validation
Key words:
Glucose self monitoring - diabetes mellitus - NIR-spectroscopy - non-invasive blood glucose determination - multivariate calibration
References
ua.mueller@med.uni-jena.de
