Continuous glucose monitoring with glucose sensors: calibration and assessment criteria.

Continuous glucose monitoring (CM) by means of minimally invasive or noninvasive glucose sensors can help to further optimize metabolic control in patients with diabetes without need for frequent capillary blood glucose measurements. Most glucose sensors measure glucose concentration in the interstitial fluid (ISF). Because of the varying conditions in this compartment, a general in vitro calibration ( = factory calibration) by the manufacturer appears not to be possible. Therefore, calibration of the sensor signal must be performed by the patient himself repeatedly. The calibration procedure can be performed by means of conventional capillary blood glucose measurements in order to transform the sensor signals obtained from the specific compartment (e.g., ISF) into "blood" glucose values. A number of aspects can influence the validity of this procedure. The relationship between changes in blood glucose and in ISF glucose, in both time and concentration dimensions, is not well understood, especially during dynamic changes. The physical lag time, which critically depends on the glucose sensor technology used, can also introduce a systematic calibration error. After the first calibration, usually performed some hours after the application of a given glucose sensor, recalibration at certain intervals is necessary. Therefore, patients should critically assess the values displayed by a CM system. In the case of implausible glucose values they should verify the results by means of a conventional capillary glucose measurement. Up to now there is no consensus on assessment criteria to be used for evaluation of CM system performance and calibration quality. Existing methods of displaying CM values against corresponding reference values, including linear regression analysis and error grid analysis, as well as numeric criteria such as System Error, Predicted Error Sum of Squares (in %), and Mean Absolute Deviation are not generally applicable to CM. It appears as if they do not allow sufficient description of data obtained with CM systems. There is a pressing need to develop novel adequate assessment criteria enabling a better characterization of CM system performance. If these were used uniformly by all manufacturers and scientists assessing CM systems, this would allow a fair comparison of the performance of different systems.